1 This file contains information about GCC releases which has been generated 2 automatically from the online release notes. It covers releases of GCC 3 (and the former EGCS project) since EGCS 1.0, on the line of development 4 that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2, 5 see ONEWS. 6 7 ====================================================================== 8 http://gcc.gnu.org/gcc-14/index.html 9 GCC 14 Release Series 10 11 June 26, 2026 12 13 The GCC developers are pleased to announce the release of GCC 14.4. 14 15 This release is a bug-fix release, containing fixes for regressions in 16 GCC 14.3 relative to previous releases of GCC. 17 18 Release History 19 20 GCC 14.4 21 Jun 23, 2026 ([1]changes, [2]documentation) 22 23 GCC 14.3 24 May 23, 2025 ([3]changes, [4]documentation) 25 26 GCC 14.2 27 Aug 1, 2024 ([5]changes, [6]documentation) 28 29 GCC 14.1 30 May 7, 2024 ([7]changes, [8]documentation) 31 32 References and Acknowledgements 33 34 GCC used to stand for the GNU C Compiler, but since the compiler 35 supports several other languages aside from C, it now stands for the 36 GNU Compiler Collection. 37 38 The GCC developers would like to thank the numerous people that have 39 contributed new features, improvements, bug fixes, and other changes as 40 well as test results to GCC. This [9]amazing group of volunteers is 41 what makes GCC successful. 42 43 For additional information about GCC please refer to the [10]GCC 44 project web site or contact the [11]GCC development mailing list. 45 46 To obtain GCC please use [12]our mirror sites or [13]our version 47 control system. 48 49 50 For questions related to the use of GCC, please consult these web 51 pages and the [14]GCC manuals. If that fails, the 52 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 53 web pages and the development of GCC are welcome on our developer 54 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 55 archives. 56 57 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 58 distribution of this entire article is permitted in any medium, 59 provided this notice is preserved. 60 61 These pages are [19]maintained by the GCC team. Last modified 62 2026-06-26. 63 64 References 65 66 1. https://gcc.gnu.org/gcc-14/changes.html 67 2. https://gcc.gnu.org/onlinedocs/14.4.0/ 68 3. https://gcc.gnu.org/gcc-14/changes.html 69 4. https://gcc.gnu.org/onlinedocs/14.3.0/ 70 5. https://gcc.gnu.org/gcc-14/changes.html 71 6. https://gcc.gnu.org/onlinedocs/14.2.0/ 72 7. https://gcc.gnu.org/gcc-14/changes.html 73 8. https://gcc.gnu.org/onlinedocs/14.1.0/ 74 9. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Contributors.html 75 10. https://gcc.gnu.org/index.html 76 11. mailto:gcc (a] gcc.gnu.org 77 12. https://gcc.gnu.org/mirrors.html 78 13. https://gcc.gnu.org/git.html 79 14. https://gcc.gnu.org/onlinedocs/ 80 15. mailto:gcc-help (a] gcc.gnu.org 81 16. mailto:gcc (a] gcc.gnu.org 82 17. https://gcc.gnu.org/lists.html 83 18. https://www.fsf.org/ 84 19. https://gcc.gnu.org/about.html 85 ====================================================================== 86 http://gcc.gnu.org/gcc-14/changes.html 87 GCC 14 Release Series 88 Changes, New Features, and Fixes 89 90 This page is a "brief" summary of some of the huge number of 91 improvements in GCC 14. You may also want to check out our [1]Porting 92 to GCC 14 page and the [2]full GCC documentation. 93 94 Caveats 95 96 * C: Support for the GCC extension, a structure containing a C99 97 flexible array member, or a union containing such a structure, is 98 not the last field of another structure, is deprecated. Refer to 99 [3]Zero Length Arrays. Any code relying on this extension should be 100 modified to ensure that C99 flexible array members only end up at 101 the ends of structures. Please use the warning option 102 [4]-Wflex-array-member-not-at-end to identify all such cases in the 103 source code and modify them. 104 * C: Certain warnings about are now errors, see [5]Porting to GCC 14 105 for details. 106 * [6]-fcf-protection=[full|branch|return|none|check] has been 107 refactored. To override -fcf-protection with a more specific 108 setting, add -fcf-protection=none followed by a specific 109 -fcf-protection=xxx. 110 * Support for the ia64*-*- target ports which have been unmaintained 111 for quite a while has been declared obsolete in GCC 14. The next 112 release of GCC will have their sources permanently removed. 113 * Support for the nios2*-*- target ports has also been declared 114 obsolete in GCC 14, and the sources will also be removed in the 115 next release of GCC. 116 * [7]-fanalyzer is still only suitable for analyzing C code. In 117 particular, using it on C++ is unlikely to give meaningful output. 118 119 [8]General Improvements 120 121 * For offload-device code generated via OpenMP and OpenACC, the math 122 and the Fortran runtime libraries will now automatically be linked, 123 when the user or compiler links them on the host side. Thus, it is 124 no longer required to explicitly pass -lm and/or -lgfortran to the 125 offload-device linker using the [9]-foffload-options= flag. 126 * New configure options: --enable-host-pie, to build the compiler 127 executables as PIE; and --enable-host-bind-now, to link the 128 compiler executables with -Wl,-z,now in order to enable additional 129 hardening. 130 * New option [10]-fhardened, an umbrella option that enables a set of 131 hardening flags. The options it enables can be displayed using the 132 --help=hardened option. 133 * New option [11]-fharden-control-flow-redundancy, to verify, at the 134 end of functions, that the visited basic blocks correspond to a 135 legitimate execution path, so as to detect and prevent attacks that 136 transfer control into the middle of functions. 137 * New type attribute [12]hardbool, for C and Ada. Hardened booleans 138 take user-specified representations for true and false, presumably 139 with higher hamming distance than standard booleans, and get 140 verified at every use, detecting memory corruption and some 141 malicious attacks. 142 * New type attribute [13]strub to control stack scrubbing properties 143 of functions and variables. The stack frame used by functions 144 marked with the attribute gets zeroed-out upon returning or 145 exception escaping. Scalar variables marked with the attribute 146 cause functions containing or accessing them to get stack scrubbing 147 enabled implicitly. 148 * New option [14]-finline-stringops, to force inline expansion of 149 memcmp, memcpy, memmove and memset, even when that is not an 150 optimization, to avoid relying on library implementations. 151 * New option [15]-ftrampoline-impl=, to choose the placement of 152 generated trampolines. The default remains the stack for almost all 153 targets. To override the default, -ftrampoline-impl=heap can be 154 passed on supported ABIs (x86_64 Darwin, x86_64 and aarch64 Linux). 155 * New function attribute [16]null_terminated_string_arg(PARAM_IDX) 156 for indicating parameters that are expected to be null-terminated 157 strings. 158 * The vectorizer now supports vectorizing loops which contain any 159 number of early breaks. This means loops such as: 160 int z[100], y[100], x[100]; 161 int foo (int n) 162 { 163 int res = 0; 164 for (int i = 0; i < n; i++) 165 { 166 y[i] = x[i] * 2; 167 res += x[i] + y[i]; 168 169 if (x[i] > 5) 170 break; 171 172 if (z[i] > 5) 173 break; 174 175 } 176 return res; 177 } 178 179 can now be vectorized on a number of targets. In this first version 180 any input data sources must either have a statically known size at 181 compile time or the vectorizer must be able to determine based on 182 auxiliary information that the accesses are aligned. 183 184 [17]New Languages and Language specific improvements 185 186 * [18]OpenMP 187 + The [19]GNU Offloading and Multi Processing Runtime Library 188 Manual has been updated and extended, improving especially the 189 description of ICVs, memory allocation, environment variables 190 and OpenMP routines. 191 + The requires directive's unified_address requirement is now 192 fulfilled by both AMD GCN and nvptx devices. AMD GCN and nvptx 193 devices now support low-latency allocators as [20]detailed in 194 the manual. Initial support for pinned-memory allocators has 195 been added and, on Linux, [21]libnuma is now used for 196 allocators requesting the nearest-partition trait (both is 197 described in the [22]memory allocation section of the manual). 198 + OpenMP 5.0: The allocate directive is now supported for stack 199 variables in C and Fortran, including the OpenMP 5.1 align 200 modifier. In C and C++ the map clause now accepts lvalue 201 expressions. For Fortran, OpenMP allocators can now be used 202 for allocatables and pointers using the allocate directive and 203 its OpenMP 5.2 replacement, the allocators directive; files 204 using this allocator and all files that might directly or 205 indirectly (intrinsic assignment, intent(out), ...) de- or 206 reallocate such-allocated variables must be compiled with the 207 [23]-fopenmp-allocators option. 208 + OpenMP 5.1: Support was added for collapsing imperfectly 209 nested loops and using present as map-type modifier and in 210 defaultmap. The indirect clause is now supported. The 211 performance of copying strided data from or to nvptx and AMD 212 GPU devices using the OpenMP 5.1 routine 213 omp_target_memcpy_rect has been improved. 214 + OpenMP 5.2: The OMP_TARGET_OFFLOAD=mandatory handling has been 215 updated for the clarifications and changes of the 5.2 216 specification. For Fortran, the list of directives permitted 217 in Fortran pure procedures was extended. Additionally, the 218 spec change has been implemented for default implicit mapping 219 of C/C++ pointers pointing to unmapped storage. The destroy 220 clause now optionally accepts the depend object as argument. 221 + OpenMP 6.0 preview (TR11/TR12): The decl attribute is now 222 supported in C++ 11 and the directive, sequence and decl 223 attributes are now supported in C 23. 224 * [24]OpenACC 225 + OpenACC 2.7: The self clause was added to be used on compute 226 constructs and the default clause for data constructs. 227 Additionally, the readonly modifier is now handled in the 228 copyin clause and cache directive. 229 + OpenACC 3.2: The following API routines are now available in 230 Fortran using the openacc module or the openacc_lib.h header 231 file: acc_malloc, acc_free, acc_map_data, acc_unmap_data, 232 acc_deviceptr, acc_hostptr, acc_memcpy_to_device, 233 acc_memcpy_to_device_async, acc_memcpy_from_device, and 234 acc_memcpy_from_device_async. 235 236 [25]Ada 237 238 * New implementation-defined aspects and pragmas: 239 + [26]Local_Restrictions, which specifies that a particular 240 subprogram does not violate one or more local restrictions, 241 nor can it call a subprogram that is not subject to the same 242 requirements. 243 + [27]User_Aspect_Definition and [28]User_Aspect, which provide 244 a mechanism for avoiding textual duplication if some set of 245 aspect specifications is needed in multiple places. 246 * New implementation-defined aspects and pragmas for verification of 247 the SPARK 2014 subset of Ada: 248 + [29]Always_Terminates, which provides a condition for a 249 subprogram to necessarily complete (either return normally or 250 raise an exception). 251 + [30]Ghost_Predicate, which introduces a subtype predicate that 252 can reference Ghost entities. 253 + [31]Exceptional_Cases, which lists exceptions that might be 254 propagated by the subprogram with side effects in the context 255 of its precondition and associates them with a specific 256 postcondition. 257 + [32]Side_Effects, which indicates that a function should be 258 handled like a procedure with respect to parameter modes, 259 Global contract, exceptional contract and termination: it may 260 have output parameters, write global variables, raise 261 exceptions and not terminate. 262 * The new attributes and contracts have been applied to the relevant 263 parts of the Ada runtime library, which has been subsequently 264 proven to be correct with SPARK 2014. 265 * Support for the LoongArch architecture. 266 * Support for VxWorks 7 Cert RTP has been removed. 267 * Additional hardening improvements. For more information related to 268 hardening options, refer to the [33]GCC Instrumentation Options and 269 the [34]GNAT Reference Manual, Security and Hardening Features. 270 * Improve style checking for redundant parentheses with [35]-gnatyz 271 * New switch [36]-gnateH to force reverse Bit_Order threshold to 64. 272 * Experimental features: 273 + [37]Storage Model: this feature proposes to redesign the 274 concepts of Storage Pools into a more efficient model allowing 275 higher performance and easier integration with low footprint 276 embedded runtimes. 277 + [38]String Interpolation: allows for easier string formatting. 278 * [39]Fixed lower bound [40][RFC] : Unconstrained array types and 279 subtypes can be specified with a lower bound that is fixed to a 280 certain value, by writing an index range that uses the syntax 281 <lower-bound-expression> .. <>. This guarantees that all objects of 282 the type or subtype will have the specified lower bound. 283 * Further clean up and improvements to the GNAT code. 284 285 [41]C family 286 287 * The Clang language extensions __has_feature and __has_extension 288 have been implemented in GCC. These are available from C, C++, and 289 Objective-C(++). This is primarily intended to aid the portability 290 of code written against Clang. 291 * GCC supports a new pragma #pragma GCC novector to indicate to the 292 vectorizer not to vectorize the loop annotated with the pragma. 293 294 [42]C 295 296 * Some more C23 features have been implemented: 297 + Bit-precise integer types (_BitInt (N) and unsigned _BitInt 298 (N)): integer types with a specified number of bits. These are 299 only supported on IA-32, x86-64 and AArch64 (little-endian) at 300 present. 301 + Structure, union and enumeration types may be defined more 302 than once in the same scope with the same contents and the 303 same tag; if such types are defined with the same contents and 304 the same tag in different scopes, the types are compatible. 305 + The <stdckdint.h> header for checked integer arithmetic. 306 * In addition to those C23 features, there are new command-line 307 options -std=c23, -std=gnu23 and -Wc11-c23-compat. These are 308 equivalent to the previous options -std=c2x, -std=gnu2x and 309 -Wc11-c2x-compat, which are deprecated but remain supported. 310 311 [43]C++ 312 313 * Several C++26 features have been implemented: 314 + [44]P1854R4, Making non-encodable string literals ill-formed 315 ([45]PR110341) 316 + [46]P2752R3, Static storage for braced initializers 317 ([47]PR110346) 318 + [48]P2361R6, Unevaluated strings ([49]PR110342) 319 + [50]P2738R1, constexpr cast from void* ([51]PR110344) 320 + [52]P2741R3, User-generated static_assert messages 321 ([53]PR110348) 322 + [54]P2169R4, Placeholder variables with no name ([55]PR110349) 323 + [56]P2864R2, Removing deprecated arithmetic conversion on 324 enumerations 325 + [57]P2748R5, Disallow binding a returned reference to a 326 temporary ([58]PR114455) 327 + [59]P2809R3, Trivial infinite loops are not undefined behavior 328 ([60]PR114462) 329 * Several C++23 features have been implemented: 330 + [61]P0847R7, Deducing this ([62]PR102609) 331 + [63]P2280R4, Using unknown references in constant expressions 332 ([64]PR106650) 333 + [65]P2564R3, consteval needs to propagate up ([66]PR107687) 334 + [67]P2582R1, Class template argument deduction from inherited 335 constructors ([68]PR106653) 336 * Several C++ Defect Reports have been resolved, e.g.: 337 + [69]DR 532, Member/nonmember operator template partial 338 ordering 339 + [70]DR 976, Deduction for const T& conversion operators 340 + [71]DR 2262, Attributes for asm-definition 341 + [72]DR 2359, Unintended copy initialization with designated 342 initializers 343 + [73]DR 2386, tuple_size requirements for structured binding 344 + [74]DR 2406, [[fallthrough]] attribute and iteration 345 statements 346 + [75]DR 2543, constinit and optimized dynamic initialization 347 + [76]DR 2586, Explicit object parameter for assignment and 348 comparison 349 + [77]DR 2735, List-initialization and conversions in overload 350 resolution 351 + [78]DR 2799, Inheriting default constructors 352 * When a diagnostic occurs involving a C++ template, GCC will now 353 quote the source code of the context at which the template is 354 instantiated ("required from here"), rather than just print 355 filename and line/column numbers. 356 * New built-in __type_pack_element to speed up traits such as 357 std::tuple_element ([79]PR100157) 358 * goto can cross the initialization of a trivially initialized object 359 with a non-trivial destructor ([80]DR 2256) 360 * -Wdangling-reference false positives have been reduced. The warning 361 does not warn about std::span-like classes; there is also a new 362 attribute gnu::no_dangling to suppress the warning. See [81]the 363 manual for more info. 364 * noexcept(expr) is now mangled as per the Itanium ABI 365 * the named return value optimization can now be performed even for 366 variables declared in an inner block of a function, see the 367 [82]test 368 * New -Wnrvo warning, to warn if the named return value optimization 369 is not performed although it is allowed by [class.copy.elision]. 370 See [83]the manual for more info. 371 * The backing array for std::initializer_list has been made static, 372 allowing combining multiple equivalent initializer-lists ([84]git) 373 * New -Welaborated-enum-base warning, to warn if an additional 374 enum-base is used in an elaborated-type-specifier 375 * Better #include hints for missing headers ([85]PR110164) 376 * The arguments of a variable template-id are coerced earlier than 377 before, so various problems are detected earlier ([86]PR89442) 378 * -Wmissing-field-initializers is no longer emitted for empty classes 379 ([87]PR110064) 380 * The constexpr code now tracks lifetimes in constant evaluation; 381 this change helps to detect bugs such as accessing a variable whose 382 lifetime has ended ([88]PR70331, [89]PR96630, [90]PR98675) 383 * Array destruction can now be devirtualized 384 * In-class member variable template partial specializations are now 385 accepted ([91]PR71954) 386 * Improved diagnostic for explicit conversion functions: when a 387 conversion doesn't work out only because the conversion function 388 necessary to do the conversion couldn't be used because it was 389 marked explicit, explain that to the user ([92]git) 390 * Corrected mangling of static/thread_local structured bindings at 391 function/block scope ([93]PR111069) 392 * [basic.scope.block]/2 violations are detected even in compound-stmt 393 of function-try-block and for block-scope external variables 394 ([94]PR52953) 395 * Improved "not a constant expression" diagnostic when taking the 396 address of a non-static constexpr variable ([95]PR91483) 397 * Non-dependent simple assignments are checked even in templates 398 ([96]PR18474) 399 * Attributes hot and cold can be applied to classes as well. See 400 [97]the manual for more info. 401 * Function template constraints, as well as CTAD placeholders, are 402 now mangled 403 * Various decltype fixes: [98]PR79620, [99]PR79378, [100]PR83167, 404 [101]PR96917 405 * New option -fdiagnostics-all-candidates to note all candidates 406 during overload resolution failure 407 * -Walloc-size and -Wcalloc-transposed-args warnings are enabled for 408 C++ as well 409 * The DR 2237 code no longer gives an error, it emits a 410 -Wtemplate-id-cdtor warning instead 411 * C++ module scanning for named modules is now available, based on 412 the format described in [102]P1689R5, Format for describing 413 dependencies of source files. The -fdeps-format=, -fdeps-file=, and 414 -fdeps-target= flags may be used to generate dependency 415 information. In GCC 14 p1689r5 is the only valid argument for 416 -fdeps-format=. 417 418 [103]Runtime Library (libstdc++) 419 420 * The libstdc++exp.a library now includes all the Filesystem TS 421 symbols from the libstdc++fs.a library. The experimental symbols 422 for the C++23 std::stacktrace class are also in libstdc++exp.a, 423 replacing the libstdc++_libbacktrace.a library that GCC 13 424 provides. This means that -lstdc++exp is the only library needed 425 for all experimental libstdc++ features. 426 * Improved experimental support for C++20, including: 427 + std::chrono::parse. 428 + Unicode-aware string handling in std::format. 429 * Improved experimental support for C++23, including: 430 + The std::ranges::to function for converting ranges to 431 containers. 432 + The std::generator view for getting results from coroutines. 433 + The <stacktrace> header is supported by default. 434 + std::print and std::println (requires linking with -lstdc++exp 435 on Windows). 436 + Formatters for std::thread::id and std::stacktrace. 437 + Smart pointer adaptors, std::out_ptr and std::inout_ptr. 438 + Some range adaptors now support move-only types. 439 * Experimental support for C++26, including: 440 + Native handles for filebuf, fstream, etc. 441 + Functions for saturation arithmetic on integers. 442 + std::to_string now uses std::format. 443 + Enhanced formatting of pointers with std::format. 444 + The std::runtime_format function to allow using non-literal 445 format strings with std::format. 446 + Testable result types for <charconv> functions. 447 + The std::text_encoding class for identifying character sets 448 (requires linking with -lstdc++exp for some member functions). 449 * Faster numeric conversions using std::to_string and 450 std::to_wstring. 451 * Updated parallel algorithms that are compatible with oneTBB. 452 * std::numeric_limits<_Float32> and std::numeric_limits<_Float64> are 453 now defined for all standard modes, not only for C++23. 454 * Added missing functions for float and long double to <cmath>. 455 * Using the std::setfill manipulator with std::istream is deprecated. 456 457 [104]D 458 459 * Support for the D programming language has been updated to version 460 2.108.1 of the language and run-time library. Full changelog for 461 this release and previous releases can be found on the 462 [105]dlang.org website. 463 464 [106]Fortran 465 466 * The compiler now accepts the -std=f2023 option, which has been 467 added in preparation of support of Fortran 2023. This option 468 increases the line-length limit for source in free-form to 10000, 469 and statements may have up to 1 million characters. 470 * With the -save-temps option, preprocessed files with the .fii 471 extension will be generated from free-form source files such as 472 .F90 and .fi from fixed-form files such as .F. 473 474 [107]Modula-2 475 476 * The automatic dependency generation options: -M, -MD, -MF, -MMD, 477 -MP, -MQ and -MT have been implemented in the compiler. 478 * The -Wcase-enum and -Wuninit-variable-checking= options have been 479 implemented to provide compile-time warnings against missing case 480 clauses and uninitialized variables respectively. 481 482 [108]Rust 483 484 * Experimental support for the Rust programming language has been 485 added. The compiler is incomplete, but already supports a subset of 486 the Rust programming language. The frontend does not support 487 compiling the Rust standard library, so it cannot be used for most 488 real-world Rust code yet. However, you can experiment with the 489 compiler to run Rust code on #[no_core] targets. 490 491 [109]libgccjit 492 493 * The libgccjit API gained 6 new entry points: 494 + [110]gcc_jit_type_get_restrict for adding restrict to types 495 ([111]LIBGCCJIT_ABI_25). 496 + 4 functions for setting attributes on functions and variables 497 ([112]LIBGCCJIT_ABI_26): 498 o gcc_jit_function_add_attribute 499 o gcc_jit_function_add_string_attribute 500 o gcc_jit_function_add_integer_array_attribute 501 o gcc_jit_lvalue_add_string_attribute 502 + [113]gcc_jit_context_new_sizeof for accessing the size of a 503 type ([114]LIBGCCJIT_ABI_27). 504 505 [115]New Targets and Target Specific Improvements 506 507 [116]AArch64 508 509 * A number of new CPUs are supported through the -mcpu and -mtune 510 options (GCC identifiers in parentheses). 511 + Ampere-1B (ampere1b). 512 + Arm Cortex-A520 (cortex-a520). 513 + Arm Cortex-A720 (cortex-a720). 514 + Arm Cortex-X4 (cortex-x4). 515 + Microsoft Cobalt-100 (cobalt-100). 516 * Additionally, the identifiers generic, generic-armv8-a and 517 generic-armv9-a can be used to optimize code generation for a good 518 blend of CPUs of a particular architecture version. These tunings 519 are also used as the default optimization targets when compiling 520 with the -march=armv8-a or -march=armv9-a options and their point 521 releases e.g. -march=armv8.2-a or -march=armv9.3-a. 522 * New features in the Arm architecture are supported in a number of 523 ways: 524 + Support is added for the Arm Streaming Matrix Extensions SME 525 and SME2 through the +sme and +sme2 extensions to -march=. In 526 particular, this includes support for the Beta state of the 527 [117]SME ACLE in the form of a new intrinsics arm_sme.h 528 intrinsics header and a number of new keyword attributes to 529 manage use of the new Streaming SVE state. For more 530 information please refer to the ACLE documentation. 531 + Libatomic is updated to implement 128-bit atomic operations 532 locklessly on systems with FEAT_LSE2. 533 + Support for FEAT_LRCPC3 is added through ACLE intrinsics in 534 arm_neon.h header and enabled through the +rcpc3 extension to 535 -march=. 536 * As well as numerous AArch64 code generation improvements, the 537 following optimization enhancements are noteworthy: 538 + A new AArch64-specific register allocation pass is added. It 539 runs in addition to standard register allocation. The pass's 540 main purpose is to make use of strided vector register 541 operands in SME instructions. However, it can also remove 542 redundant moves in normal Advanced SIMD and SVE code. The pass 543 is controlled by the new option -mearly-ra= that takes the 544 arguments all, strided, none. -mearly-ra=all is enabled by 545 default at optimization levels -O2 and above. 546 + A new optimization pass to fuse loads and stores to adjacent 547 memory locations into load and store-pair AArch64 548 instructions. The pass is enabled by default when compiling 549 with optimization and runs twice in the optimization pipeline: 550 before and after register allocation. This can be controlled 551 with the options -mearly-ldp-fusion and -mlate-ldp-fusion. 552 * Conformance with the ACLE specification is improved and a number of 553 features aimed at helping developers deploy Arm architecture 554 features are added: 555 + Support for the Beta version of the [118]Function 556 Multiversioning Specification. This feature provides 557 facilities to annotate functions with attributes that allow 558 the compiler to generate multiple versions of the function, 559 selected at runtime based on the architecture features 560 available in the system. Please refer to the ACLE 561 specification for more details. 562 + Support for more ACLE intrinsics in the arm_acle.h header, 563 including the [119]Memory prefetch intrinsics and the 564 [120]Special register intrinsics. This also includes 565 intrinsics for the extension to 128-bit system registers, 566 enabled through the +d128 extension to -march=. 567 + Intrinsics enabled by the +dotprod, +fp16, +fp16fml, +i8mm, 568 +sha3 and +sm4 extensions to -march= no longer require 569 -march=armv8.2-a or higher to be specified. Likewise, the 570 intrinsics enabled by +memtag no longer require 571 -march=armv8.5-a. 572 + Support for the [121]NEON-SVE Bridge intrinsics. These are 573 intrinsics that allow conversions between NEON and SVE 574 vectors, enabled through the inclusion of the 575 arm_neon_sve_bridge.h header. 576 * The option -mtp= is now supported for changing the TPIDR register 577 used for TLS accesses. For more details please refer to the 578 [122]documentation. 579 580 [123]AMD Radeon (GCN) 581 582 * Initial support for the AMD Radeon gfx90c (GCN5), gfx1030, gfx1036 583 (RDNA2), gfx1100 and gfx1103 (RDNA3) devices has been added. LLVM 584 15+ (assembler and linker) is [124]required to support GFX11. 585 * Improved register usage and performance on CDNA Instinct MI100 and 586 MI200 series devices. 587 * The default device architecture is now gfx900 (Vega). 588 * Fiji (gfx803) device support is now deprecated and will be removed 589 from a future release. The default compiler configuration no longer 590 uses Fiji as the default device, and no longer includes the Fiji 591 libraries. Both can be restored by configuring with 592 [125]--with-arch=fiji . 593 594 [126]arm 595 596 * The Cortex-M52 CPU is now supported through the cortex-m52 argument 597 to the -mcpu and -mtune options. 598 599 [127]AVR 600 601 * On AVR64* and AVR128* devices, read-only data is now located in 602 program memory per default and no longer in RAM. 603 + Only a 32 KiB block of program memory can be used to store and 604 access .rodata in that way. Which block is used can be 605 selected by defining the symbol __flmap. As an alternative, 606 the byte address of the block can be specified by the symbol 607 __RODATA_FLASH_START__ which takes precedence over __flmap. 608 For example, linking with 609 -Wl,--defsym,__RODATA_FLASH_START__=32k chooses the second 610 32 KiB block. 611 + The default uses the last 32 KiB block, which is also the 612 hardware default for bit-field NVMCTRL_CTRLB.FLMAP. 613 + When a non-default block is used, then NVMCTRL_CTRLB.FLMAP 614 must be initialized accordingly by hand, or AVR-LibC v2.2 that 615 implements [128]#931 can be used. The latter initializes 616 NVMCTRL_CTRLB.FLMAP in the startup code and according to the 617 value of __flmap or __RODATA_FLASH_START__. 618 + When AVR-LibC with #931 is used, then defining the symbol 619 __flmap_lock to a non-zero value will set bit 620 NVMCTRL_CTRLB.FLMAPLOCK. This will protect NVMCTRL_CTRLB.FLMAP 621 from any further changes which would be Undefined Behaviour 622 in C/C++. If you prefer to define the symbol in a C/C++ file, 623 an asm statement can be used: 624 __asm (".global __flmap_lock" "\n\t" 625 "__flmap_lock = 1"); 626 + When you do not want the code from #931, then define a global 627 symbol __do_flmap_init and the linker will not pull in that 628 code from libmcu.a any more. 629 + In order to return to the old placement of read-only data in 630 RAM, the new compiler option -mrodata-in-ram can be used. This 631 is required on devices where the hardware revision is affected 632 by a silicon bug concerning the FLMAP functionality. 633 + Read-only data is located in output section .rodata, whereas 634 it is part of .text when located in RAM. 635 + The feature is only available when the compiler is configured 636 with a version of Binutils that implements [129]PR31124, which 637 is the case for Binutils v2.42 and up. 638 + The implementation consists of two parts: 639 1. Binutils supports new emulations avrxmega2_flmap and 640 avrxmega4_flmap. The sole purpose of these emulations is 641 to provide adjusted default linker description files. 642 Apart from that, these emulations behave exactly the same 643 like avrxmega2 resp. avrxmega4. 644 2. The compiler uses a [130]device-specs file which links 645 the program with -mavrxmega2_flmap or -mavrxmega2 646 depending on -m[no-]rodata-in-ram; and similar for 647 -mavrxmega4[_flmap]. 648 This means the feature can be used with older compiler or 649 Binutils versions; all what's needed is an adjusted linker 650 script and a custom device-specs file. 651 * A new compiler option -m[no-]rodata-in-ram has been added. The 652 default is to locate read-only data in program memory for devices 653 that support it, e.g. for AVR64* and AVR128* devices as explained 654 above, and for devices from the [131]avrxmega3 and [132]avrtiny 655 families. 656 * The new built-in macro __AVR_RODATA_IN_RAM__ is supported on all 657 devices. It's defined to 0 or 1. 658 * A new optimization tries to improve code generation for indirect 659 memory accesses on [133]Reduced Tiny devices. It can be controlled 660 by the new compiler option -mfuse-add=level where level may be 0, 1 661 or 2. 662 * On the Reduced Tiny devices, the meaning of register constraint "w" 663 has been changed. It now constrains the registers R24R31 as is the 664 case for all the other devices. 665 * Support for the following devices has been added in GCC 14.3: 666 + AVR32SD20, AVR32SD28, AVR32SD32, AVR64SD28, AVR64SD32, 667 AVR64SD48. 668 * Support for the following devices has been added in GCC 14.4: 669 + AVR32DA28S, AVR32DA32S, AVR32DA48S, AVR64DA28S, AVR64DA32S, 670 AVR64DA48S AVR64DA64S, AVR128DA28S, AVR128DA32S, AVR128DA48S, 671 AVR128DA64S, AVR32EB14, AVR32EB20, AVR32EB28, AVR32EB32, 672 AVR16LA14, AVR16LA20, AVR16LA28, AVR16LA32, AVR32LA14, 673 AVR32LA20, AVR32LA28, AVR32LA32. 674 675 [134]IA-32/x86-64 676 677 * New compiler option -m[no-]evex512 was added. The compiler switch 678 enables/disables 512-bit vector. It will be default on if AVX512F 679 is enabled. 680 * Part of new feature support for Intel APX was added, including 681 EGPR, NDD, PPX and PUSH2POP2. APX support is available via the 682 -mapxf compiler switch. 683 * For inline asm support with APX, by default the EGPR feature was 684 disabled to prevent potential illegal instruction with EGPR occurs. 685 To invoke egpr usage in inline asm, use new compiler option 686 -mapx-inline-asm-use-gpr32 and user should ensure the instruction 687 supports EGPR. 688 * New ISA extension support for Intel AVX10.1 was added. AVX10.1 689 intrinsics are available via the -mavx10.1-256 compiler switch with 690 256-bit vector support. 512-bit vector support for AVX10.1 691 intrinsics are available via the -mavx10.1-512 compiler switch. 692 -mavx10.1 enables AVX10.1 intrinsics with 256-bit vector support in 693 GCC 14.1 and GCC 14.2. Since GCC 14.3, it enables AVX10.1 694 intrinsics with 512-bit vector support (and emits a warning due to 695 this behavior change). 696 * New ISA extension support for Intel AVX-VNNI-INT16 was added. 697 AVX-VNNI-INT16 intrinsics are available via the -mavxvnniint16 698 compiler switch. 699 * New ISA extension support for Intel SHA512 was added. SHA512 700 intrinsics are available via the -msha512 compiler switch. 701 * New ISA extension support for Intel SM3 was added. SM3 intrinsics 702 are available via the -msm3 compiler switch. 703 * New ISA extension support for Intel SM4 was added. SM4 intrinsics 704 are available via the -msm4 compiler switch. 705 * New ISA extension support for Intel USER_MSR was added. USER_MSR 706 intrinsics are available via the -muser_msr compiler switch. 707 * GCC now supports the Intel CPU named Clearwater Forest through 708 -march=clearwaterforest. Based on Sierra Forest, the switch in 709 addition enables the AVX-VNNI-INT16, PREFETCHI, SHA512, SM3, SM4 710 and USER_MSR ISA extensions. Since GCC 14.4, KL and WIDEKL are not 711 enabled any longer. 712 * GCC now supports the Intel CPU named Gracemont through 713 -march=gracemont. Gracemont is based on Alder Lake. Since GCC 14.4, 714 CLDEMOTE is not enabled any longer. 715 * GCC now supports the Intel CPU named Arrow Lake through 716 -march=arrowlake. Based on Alder Lake, the switch in addition 717 enables the AVX-IFMA, AVX-NE-CONVERT, AVX-VNNI-INT8 and CMPccXADD 718 ISA extensions. Since GCC 14.4, CLDEMOTE is not enabled any longer. 719 * GCC now supports the Intel CPU named Arrow Lake S through 720 -march=arrowlake-s. Based on Arrow Lake, the switch in addition 721 enables the AVX-VNNI-INT16, SHA512, SM3 and SM4 ISA extensions. 722 Since GCC 14.4, CLDEMOTE is not enabled any longer. 723 * GCC now supports the Intel CPU named Lunar Lake through 724 -march=lunarlake. Lunar Lake is based on Arrow Lake S. Since GCC 725 14.4, CLDEMOTE is not enabled any longer. 726 * GCC now supports the Intel CPU named Panther Lake through 727 -march=pantherlake. Based on Arrow Lake S, the switch in addition 728 further enables the PREFETCHI ISA extensions. Since GCC 14.4, 729 CLDEMOTE, KL, PREFETCHI and WIDEKL are not enabled any longer. 730 * Since GCC 14.4, CLDEMOTE is not enabled through the compiler 731 switches -march=alderlake, -march=meteorlake and -march=raptorlake 732 any longer. 733 * Xeon Phi CPUs support (a.k.a. Knight Landing and Knight Mill) are 734 marked as deprecated. GCC will emit a warning when using the 735 -mavx5124fmaps, -mavx5124vnniw, -mavx512er, -mavx512pf, 736 -mprefetchwt1, -march=knl, -march=knm, -mtune=knl or -mtune=knm 737 compiler switches. Support will be removed in GCC 15. 738 * [135]Hardware-assisted AddressSanitizer now works for the x86-64 739 target with LAM_U57. -fsanitize=hwaddress will enable -mlam=u57 by 740 default. 741 * GCC now supports AMD CPUs based on the znver5 core via 742 -march=znver5. In addition to the ISA extensions enabled on a 743 znver4 core, this switch further enables the AVX512VP2INTERSECT, 744 AVXVNNI, MOVDIR64B, MOVDIRI, and PREFETCHI ISA extensions. 745 * The _Float16 and __bf16 type are supported independent of SSE2. W/o 746 SSE2, these types are storage-only, compiler will issue an error 747 when they're used in conversion, unary operation, binary operation, 748 parameter passing or value return. Please use __SSE2__ to detect 749 arithmetic support of these types instead of __FLT16_MAX__(or other 750 similar Macros). 751 752 [136]MCore 753 754 * Bitfields are now signed by default per GCC policy. If you need 755 bitfields to be unsigned, use -funsigned-bitfields. 756 757 [137]LoongArch 758 759 * Support for the following [138]-march parameters has been added: 760 + la64v1.0 761 + la64v1.1 762 + la664 763 It is now recommended to use -march=la64v1.0 as the only compiler 764 option to describe the target ISA when building binaries for 765 distribution. For more information on LoongArch ISA versions, see 766 [139]Toolchain Conventions of the LoongArch Architecture. 767 * Support for the following [140]-mtune parameters has been added: 768 + generic 769 + la664 770 * New ISA Extension 771 + LSX (Loongson SIMD Extension): Support 128-bit vector 772 instructions and the intrinsics. 773 + LASX (Loongson Advanced SIMD Extension): Support 256-bit 774 vector instructions and the intrinsics. 775 + FRECIPE: Support frecipe.{s/d} and frsqrte.{s/d} instructions 776 and the intrinsics. 777 + DIV32: Support div.w[u] and mod.w[u] instructions with inputs 778 not sign-extended. 779 + LAM_BH: Support am{swap/add}[_db].{b/h} instructions. 780 + LAMCAS: Support amcas[_db].{b/h/w/d} instructions. 781 * New Built-in Macros 782 + __loongarch_arch: Target ISA preset as specified by -march=. 783 For example, compiling with -march=la64v1.0, the value of 784 __loongarch_arch is "la64v1.0". 785 + __loongarch_tune: Processor model as specified by -mtune or 786 its default value. 787 + __loongarch_{simd,sx,asx}: These macros are not defined, or 788 defined as 1. 789 + __loongarch_simd_width: The maximum SIMD bit-width enabled by 790 the compiler. (128 for lsx, and 256 for lasx). 791 + __loongarch_frecipe: It's defined to 1 or undefined. 792 + __loongarch_div32: It's defined to 1 or undefined. 793 + __loongarch_lam_bh: It's defined to 1 or undefined. 794 + __loongarch_lamcas: It's defined to 1 or undefined. 795 + __loongarch_ld_seq_sa: It's defined to 1 or undefined. 796 + __loongarch_version_major: The minimally required LoongArch 797 ISA version (major) to run the compiled program, defined to 1 798 or undefined (iff no such version is known to the compiler). 799 + __loongarch_version_minor: The minimally required LoongArch 800 ISA version (minor) to run the compiled program, defined to 0 801 1 or undefined (iff __loongarch_version_major is undefined). 802 + __FLOAT128_TYPE: It's defined to 1. 803 * New Intrinsics 804 + [141]__builtin_thread_pointer 805 + [142]__lsx_* 806 + [143]__lasx_* 807 + [144]__frecipe_{s/d} and __frsqrte_{s/d} 808 * New Compiler Option 809 + [145]-m[no-]lsx 810 + [146]-m[no-]lasx 811 + [147]-m[no-]frecipe 812 + [148]-m[no-]div32 813 + [149]-m[no-]lam-bh 814 + [150]-m[no-]lamcas 815 + [151]-m[no-]ld-seq-sa 816 + [152]-mrecip= 817 + [153]-m[no-]recip 818 + [154]-mexplicit-relocs={none,always,auto} 819 + [155]-m[no-]relax 820 + [156]-m[no-]pass-mrelax-to-as 821 + [157]-mtls-dialect={trad,desc} 822 * Support for Ada and D. 823 * Support for libffi. 824 * Enable -free by default at -O2 or higher. 825 * Enable -fsched-pressure by default at -O1 or higher. 826 * Support the extreme code model using macro instructions (under 827 -mno-explicit-relocs). 828 * Support call36. 829 * Optimizing built-in functions for memory-model-aware atomic 830 operations using hierarchical dbar instructions. 831 * TLS descriptors support. It is not enabled by default, and can be 832 enabled with [158]-mtls-dialect=desc. The default behavior can be 833 configured with --with-tls=[trad|desc]. 834 835 [159]RISC-V 836 837 * The SLP and loop vectorizer are now enabled for RISC-V when the 838 vector extension is enabled, thanks to Ju-Zhe Zhong from 839 [160]RiVAI, Pan Li from Intel, and Robin Dapp from [161]Ventana 840 Micro for contributing most of the implementation. 841 * The -mrvv-max-lmul= option has been introduced for performance 842 tuning of the loop vectorizer. The default value is 843 -mrvv-max-lmul=m1, which limits the maximum LMUL to 1. The 844 -mrvv-max-lmul=dynamic setting can dynamically select the maximum 845 LMUL value based on register pressure. 846 * Atomic code generation has been improved and is now in conformance 847 with the latest psABI specification, thanks to Patrick O'Neill from 848 [162]Rivos. 849 * Support for the vector intrinsics as specified in [163]version 1.0 850 of the RISC-V vector intrinsic specification. 851 * Support for the experimental vector crypto intrinsics as specified 852 in [164]RISC-V vector intrinsic specification, thanks to Feng Wang 853 et al. from [165]ESWIN Computing 854 * Support for the T-head vector intrinsics. 855 * Support for the scalar bitmanip and scalar crypto intrinsics, 856 thanks to Liao Shihua from [166]PLCT. 857 * Support for the large code model via option -mcmodel=large, thanks 858 to Kuan-Lin Chen from [167]Andes Technology. 859 * Support for the standard vector calling convention variant, thanks 860 to Lehua Ding from [168]RiVAI. 861 * Supports the target attribute, which allows users to compile a 862 function with specific extensions. 863 * The -march= option no longer requires the architecture string to be 864 in canonical order, with only a few constraints remaining: the 865 architecture string must start with rv[32|64][i|g|e], and must use 866 an underscore as the separator after a multi-letter extension. 867 * -march=help has been introduced to list all supported extensions. 868 * Added experimental support for the -mrvv-vector-bits=zvl option and 869 the riscv_rvv_vector_bits attribute, which specify a fixed length 870 for scalable vector types. This option is optimized for specific 871 vector core implementations; however, the code generated with this 872 option is NOT portable between the core with different VLEN, thanks 873 to Pan Li from Intel. 874 * Support for TLS descriptors has been introduced. It can be enabled 875 by the -mtls-dialect=desc option. The default behavior can be 876 configured with --with-tls=[trad|desc]. This feature requires glibc 877 2.40. Thanks to Tatsuyuki Ishi from [169]Blue Whale Systems. 878 * Support for the following standard extensions has been added: 879 + Vector crypto extensions: 880 o Zvbb 881 o Zvkb 882 o Zvbc 883 o Zvkg 884 o Zvkned 885 o Zvkhna 886 o Zvkhnb 887 o Zvksed 888 o Zvksh 889 o Zvkn 890 o Zvknc 891 o Zvkng 892 o Zvks 893 o Zvksc 894 o Zvksg 895 o Zvkt 896 + Code size reduction extensions: 897 o Zca 898 o Zcb 899 o Zce 900 o Zcf 901 o Zcd 902 o Zcmp 903 o Zcmt 904 + Zicond 905 + Zfa 906 + Ztso 907 + Zvfbfmin 908 + Zvfhmin 909 + Zvfh 910 + Za64rs 911 + Za128rs 912 + Ziccif 913 + Ziccrse 914 + Ziccamoa 915 + Zicclsm 916 + Zic64b 917 + Smaia 918 + Smepmp 919 + Smstateen 920 + Ssaia 921 + Sscofpmf 922 + Ssstateen 923 + Sstc 924 + Svinval 925 + Svnapot 926 + Svpbmt 927 * Support for the following vendor extensions has been added: 928 + T-Head: 929 o XTheadVector 930 + CORE-V: 931 o XCVmac 932 o XCValu 933 o XCVelw 934 o XCVsimd 935 o XCVbi 936 + Ventana Micro: 937 o XVentanaCondops 938 * The following new CPUs are supported through the -mcpu option (GCC 939 identifiers in parentheses). 940 + SiFive's X280 (sifive-x280). 941 + SiFive's P450 (sifive-p450). 942 + SiFive's P670 (sifive-p670). 943 * The following new CPUs are supported through the -mtune option (GCC 944 identifiers in parentheses). 945 + Generic out-of-order core (generic-ooo). 946 + SiFive's P400 series (sifive-p400-series). 947 + SiFive's P600 series (sifive-p600-series). 948 + XiangShan's Nanhu microarchitecture (xiangshan-nanhu). 949 950 [170]SPARC 951 952 * The implementation of calling conventions for small structures 953 containing arrays of floating-point components has been changed in 954 64-bit mode for the Solaris port to match the implementation of the 955 vendor compiler (and the ABI). As a result, the code generated will 956 not be binary compatible with earlier releases in these cases. 957 958 Documentation improvements 959 960 * GCC's ability to provide clickable hyperlinks to the documentation 961 has been extended, so that whenever GCC refers to a command-line 962 option in quotes in a diagnostic message, the option is a clickable 963 hyperlink (assuming a suitably capable terminal). 964 965 [171]Improvements to Static Analyzer 966 967 * New warnings: 968 + [172]-Wanalyzer-infinite-loop warns about paths through the 969 code which appear to lead to an infinite loop. 970 + [173]-Wanalyzer-overlapping-buffers warns for paths through 971 the code in which overlapping buffers are passed to an API for 972 which the behavior on such buffers is undefined. 973 + [174]-Wanalyzer-undefined-behavior-strtok warns for paths 974 through the code in which a call is made to strtok with 975 undefined behavior. 976 * Previously, the analyzer's "taint" tracking to be explicitly 977 enabled via [175]-fanalyzer-checker=taint (along with 978 [176]-fanalyzer). This is now enabled by default when 979 [177]-fanalyzer is selected, thus also enabling the 6 taint-based 980 warnings: 981 + [178]-Wanalyzer-tainted-allocation-size 982 + [179]-Wanalyzer-tainted-array-index 983 + [180]-Wanalyzer-tainted-assertion 984 + [181]-Wanalyzer-tainted-divisor 985 + [182]-Wanalyzer-tainted-offset 986 + [183]-Wanalyzer-tainted-size 987 * The analyzer will now simulate API calls that expect 988 null-terminated string arguments, and will warn about code paths in 989 which such a call is made with a buffer that isn't properly 990 terminated, either due to a read of an uninitialized byte or an 991 out-of-range accesses seen before any zero byte is seen. This 992 applies to functions that use the new 993 [184]null_terminated_string_arg(PARAM_IDX) attribute, functions 994 that use the [185]format attribute, and to the library functions 995 error (parameter 3), error_at_line (parameter 5), putenv, strchr 996 (parameter 1), and strcpy (parameter 2). 997 * The analyzer now makes use of the function attribute 998 [186]alloc_size allowing [187]-fanalyzer to emit 999 [188]-Wanalyzer-allocation-size, [189]-Wanalyzer-out-of-bounds, and 1000 [190]-Wanalyzer-tainted-allocation-size on execution paths 1001 involving allocations using such functions. 1002 * The analyzer's knowledge about the behavior of the standard library 1003 has been extended to cover fopen, strcat, strncpy, and strstr. The 1004 analyzer will also more precisely model the behavior of memcpy, 1005 memmove, strcpy, strdup, strlen, and of various atomic built-in 1006 functions. 1007 * The warning [191]-Wanalyzer-out-of-bounds has been extended so 1008 that, where possible, it will emit a text-based diagram visualizing 1009 the spatial relationship between 1010 1. the memory region that the analyzer predicts would be 1011 accessed, versus 1012 2. the range of memory that is valid to access 1013 whether they overlap, are touching, are close or far apart; which 1014 one is before or after in memory, the relative sizes involved, the 1015 direction of the access (read vs write), and, in some cases, the 1016 values of data involved. 1017 Such "text art" diagrams can be controlled (or suppressed) via a 1018 new [192]-fdiagnostics-text-art-charset= option. 1019 For example, given the out-of-bounds write in strcat in: 1020 void test (void) 1021 { 1022 char buf[10]; 1023 strcpy (buf, "hello"); 1024 strcat (buf, " world!"); 1025 } 1026 1027 it emits: 1028 1029 [0] [1] [2] [3] [4] [5] [6] [7] 1030 1031 ' ' 'w' 'o' 'r' 'l' 'd' '!' NUL 1032 1033 string literal (type: 'char[8]') 1034 1035 1036 1037 v v v v v v v v 1038 1039 [0] ... [5] ... [9] 1040 1041 'h' 'e' 'l' 'l' 'o' NUL after valid range 1042 1043 'buf' (type: 'char[10]') 1044 1045 1046 1047 1048 1049 capacity: 10 bytes overflow of 3 bytes 1050 1051 1052 1053 1054 showing that the overflow occurs partway through the second string 1055 fragment. 1056 * The analyzer will now attempt to track execution paths involving 1057 computed gotos, whereas previously it gave up on such paths. 1058 1059 [193]Improvements to SARIF support 1060 1061 * The SARIF output from [194]-fdiagnostics-format= now adds 1062 indentation and newlines to reflect the logical JSON structure of 1063 the data. The previous compact behavior can be restored via the new 1064 option [195]-fno-diagnostics-json-formatting. This also applies to 1065 the older output format named "json". 1066 * If profiling information about the compiler itself is requested via 1067 [196]-ftime-report, and a SARIF output format is requested via 1068 [197]-fdiagnostics-format=, then the timing and memory usage data 1069 is now written in JSON form into the SARIF output, rather than as 1070 plain text to stderr. 1071 1072 [198]Improvements for plugin authors 1073 1074 * GCC diagnostics have been able to have execution paths associated 1075 with them since GCC 10, but previously these were required to be 1076 single-threaded. As of GCC 14, these execution paths can have 1077 multiple named threads associated with them, with each event being 1078 associated with one of the threads. No existing GCC diagnostics 1079 take advantage of this, but GCC plugins may find this useful for 1080 their own diagnostics; an [199]example is provided in the 1081 testsuite. 1082 * GCC's diagnostics can now optionally add per-diagnostic property 1083 bags to the SARIF output, allowing plugins to capture custom data 1084 as needed with their diagnostics. 1085 1086 [200]GCC 14.1 1087 1088 This is the [201]list of problem reports (PRs) from GCC's bug tracking 1089 system that are known to be fixed in the 14.1 release. This list might 1090 not be complete (that is, it is possible that some PRs that have been 1091 fixed are not listed here). 1092 1093 [202]GCC 14.2 1094 1095 This is the [203]list of problem reports (PRs) from GCC's bug tracking 1096 system that are known to be fixed in the 14.2 release. This list might 1097 not be complete (that is, it is possible that some PRs that have been 1098 fixed are not listed here). 1099 1100 [204]GCC 14.3 1101 1102 This is the [205]list of problem reports (PRs) from GCC's bug tracking 1103 system that are known to be fixed in the 14.3 release. This list might 1104 not be complete (that is, it is possible that some PRs that have been 1105 fixed are not listed here). 1106 1107 [206]GCC 14.4 1108 1109 This is the [207]list of problem reports (PRs) from GCC's bug tracking 1110 system that are known to be fixed in the 14.4 release. This list might 1111 not be complete (that is, it is possible that some PRs that have been 1112 fixed are not listed here). 1113 1114 1115 For questions related to the use of GCC, please consult these web 1116 pages and the [208]GCC manuals. If that fails, the 1117 [209]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1118 web pages and the development of GCC are welcome on our developer 1119 list at [210]gcc (a] gcc.gnu.org. All of [211]our lists have public 1120 archives. 1121 1122 Copyright (C) [212]Free Software Foundation, Inc. Verbatim copying and 1123 distribution of this entire article is permitted in any medium, 1124 provided this notice is preserved. 1125 1126 These pages are [213]maintained by the GCC team. Last modified 1127 2026-06-26. 1128 1129 References 1130 1131 1. https://gcc.gnu.org/gcc-14/porting_to.html 1132 2. https://gcc.gnu.org/onlinedocs/index.html#current 1133 3. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Zero-Length.html 1134 4. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Warning-Options.html#index-Wflex-array-member-not-at-end 1135 5. https://gcc.gnu.org/gcc-14/porting_to.html#warnings-as-errors 1136 6. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html 1137 7. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html 1138 8. https://gcc.gnu.org/gcc-14/changes.html#general 1139 9. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C-Dialect-Options.html#index-foffload-options 1140 10. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fhardened 1141 11. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fharden-control-flow-redundancy 1142 12. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Type-Attributes.html#index-hardbool-type-attribute 1143 13. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Type-Attributes.html#index-strub-type-attribute 1144 14. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Optimize-Options.html#index-finline-stringops 1145 15. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Code-Gen-Options.html#index-ftrampoline-impl 1146 16. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-null_005fterminated_005fstring_005farg-function-attribute 1147 17. https://gcc.gnu.org/gcc-14/changes.html#languages 1148 18. https://gcc.gnu.org/projects/gomp/ 1149 19. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/ 1150 20. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/Offload-Target-Specifics.html 1151 21. https://github.com/numactl/numactl 1152 22. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/Memory-allocation.html 1153 23. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gfortran/Fortran-Dialect-Options.html#index-fopenmp-allocators 1154 24. https://gcc.gnu.org/wiki/OpenACC 1155 25. https://gcc.gnu.org/gcc-14/changes.html#ada 1156 26. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Local_005fRestrictions.html 1157 27. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Pragma-User_005fAspect_005fDefinition.html 1158 28. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-User_005fAspect.html 1159 29. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Always_005fTerminates.html 1160 30. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Ghost_005fPredicate.html 1161 31. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Exceptional_005fCases.html 1162 32. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Side_005fEffects.html 1163 33. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fharden-compares 1164 34. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Security-Hardening-Features.html 1165 35. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_ugn/Style-Checking.html#index--gnatyz-_0028gcc_0029 1166 36. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_ugn/Alphabetical-List-of-All-Switches.html#index--gnateH-_0028gcc_0029 1167 37. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Pragma-Storage_005fModel.html 1168 38. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/String-interpolation.html 1169 39. https://gcc.gnu.org/onlinedocs/gnat_rm/Fixed-lower-bounds-for-array-types-and-subtypes.html 1170 40. https://github.com/AdaCore/ada-spark-rfcs/blob/master/features/rfc-fixed-lower-bound.rst 1171 41. https://gcc.gnu.org/gcc-14/changes.html#c-family 1172 42. https://gcc.gnu.org/gcc-14/changes.html#c 1173 43. https://gcc.gnu.org/gcc-14/changes.html#cxx 1174 44. https://wg21.link/P1854R4 1175 45. https://gcc.gnu.org/PR110341 1176 46. https://wg21.link/P2752R3 1177 47. https://gcc.gnu.org/PR110346 1178 48. https://wg21.link/P2361R6 1179 49. https://gcc.gnu.org/PR110342 1180 50. https://wg21.link/P2738R1 1181 51. https://gcc.gnu.org/PR110344 1182 52. https://wg21.link/P2741R3 1183 53. https://gcc.gnu.org/PR110348 1184 54. https://wg21.link/P2169R4 1185 55. https://gcc.gnu.org/PR110349 1186 56. https://wg21.link/P2864R2 1187 57. https://wg21.link/P2748R5 1188 58. https://gcc.gnu.org/PR114455 1189 59. https://wg21.link/P2809R3 1190 60. https://gcc.gnu.org/PR114462 1191 61. https://wg21.link/P0847R7 1192 62. https://gcc.gnu.org/PR102609 1193 63. https://wg21.link/P2280R4 1194 64. https://gcc.gnu.org/PR106650 1195 65. https://wg21.link/P2564R3 1196 66. https://gcc.gnu.org/PR107687 1197 67. https://wg21.link/P2582R1 1198 68. https://gcc.gnu.org/PR106653 1199 69. https://wg21.link/cwg532 1200 70. https://wg21.link/cwg976 1201 71. https://wg21.link/cwg2262 1202 72. https://wg21.link/cwg2359 1203 73. https://wg21.link/cwg2386 1204 74. https://wg21.link/cwg2406 1205 75. https://wg21.link/cwg2543 1206 76. https://wg21.link/cwg2586 1207 77. https://wg21.link/cwg2735 1208 78. https://wg21.link/cwg2799 1209 79. https://gcc.gnu.org/PR100157 1210 80. https://cplusplus.github.io/CWG/issues/2256.html 1211 81. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdangling-reference 1212 82. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/testsuite/g++.dg/opt/nrv23.C;h=9e1253cd830a84ad4de5ff3076a07c543afe344f;hb=7e0b65b239c3a0d68ce94896b236b03de666ffd6 1213 83. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Warning-Options.html#index-Wnrvo 1214 84. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=4d935f52b0d5c00fcc154461b87415ebd8791a94 1215 85. https://gcc.gnu.org/PR110164 1216 86. https://gcc.gnu.org/PR89442 1217 87. https://gcc.gnu.org/PR110064 1218 88. https://gcc.gnu.org/PR70331 1219 89. https://gcc.gnu.org/PR96630 1220 90. https://gcc.gnu.org/PR98675 1221 91. https://gcc.gnu.org/PR71954 1222 92. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=85ad41a494e31311f8a6b2dbe930a128c5e85840 1223 93. https://gcc.gnu.org/PR111069 1224 94. https://gcc.gnu.org/PR52953 1225 95. https://gcc.gnu.org/PR91483 1226 96. https://gcc.gnu.org/PR18474 1227 97. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C_002b_002b-Attributes.html#index-cold-type-attribute 1228 98. https://gcc.gnu.org/PR79620 1229 99. https://gcc.gnu.org/PR79378 1230 100. https://gcc.gnu.org/PR83167 1231 101. https://gcc.gnu.org/PR96917 1232 102. https://wg21.link/P1689R5 1233 103. https://gcc.gnu.org/gcc-14/changes.html#libstdcxx 1234 104. https://gcc.gnu.org/gcc-14/changes.html#d 1235 105. https://dlang.org/changelog/2.108.1.html 1236 106. https://gcc.gnu.org/gcc-14/changes.html#fortran 1237 107. https://gcc.gnu.org/gcc-14/changes.html#modula2 1238 108. https://gcc.gnu.org/gcc-14/changes.html#rust 1239 109. https://gcc.gnu.org/gcc-14/changes.html#jit 1240 110. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/types.html#c.gcc_jit_type_get_restrict 1241 111. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-25 1242 112. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-26 1243 113. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_sizeof 1244 114. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-27 1245 115. https://gcc.gnu.org/gcc-14/changes.html#targets 1246 116. https://gcc.gnu.org/gcc-14/changes.html#aarch64 1247 117. https://github.com/ARM-software/acle/blob/main/main/acle.md#sme-language-extensions-and-intrinsics 1248 118. https://github.com/ARM-software/acle/blob/main/main/acle.md#function-multi-versioning 1249 119. https://github.com/ARM-software/acle/blob/main/main/acle.md#memory-prefetch-intrinsics 1250 120. https://github.com/ARM-software/acle/blob/main/main/acle.md#special-register-intrinsics 1251 121. https://github.com/ARM-software/acle/blob/main/main/acle.md#neon-sve-bridge 1252 122. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AArch64-Options.html 1253 123. https://gcc.gnu.org/gcc-14/changes.html#amdgcn 1254 124. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 1255 125. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 1256 126. https://gcc.gnu.org/gcc-14/changes.html#arm 1257 127. https://gcc.gnu.org/gcc-14/changes.html#avr 1258 128. https://github.com/avrdudes/avr-libc/issues/931 1259 129. https://sourceware.org/PR31124 1260 130. https://gcc.gnu.org/wiki/avr-gcc#spec-files 1261 131. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrxmega3 1262 132. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrtiny 1263 133. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrtiny 1264 134. https://gcc.gnu.org/gcc-14/changes.html#x86 1265 135. https://gcc.gnu.org/gcc-11/changes.html 1266 136. https://gcc.gnu.org/gcc-14/changes.html#mcore 1267 137. https://gcc.gnu.org/gcc-14/changes.html#loongarch 1268 138. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-march-7 1269 139. https://github.com/loongson/la-toolchain-conventions/ 1270 140. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtune-8 1271 141. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Base-Built-in-Functions.html 1272 142. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-SX-Vector-Intrinsics.html 1273 143. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-ASX-Vector-Intrinsics.html 1274 144. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Base-Built-in-Functions.html 1275 145. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-ml_005ba_005dsx 1276 146. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-ml_005ba_005dsx 1277 147. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mfrecipe 1278 148. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mdiv32 1279 149. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mlam-bh 1280 150. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mlamcas 1281 151. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mld-seq-sa 1282 152. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrecip_003dopt 1283 153. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrecip 1284 154. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mexplicit-relocs-1 1285 155. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrelax-2 1286 156. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mpass-mrelax-to-as 1287 157. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtls-dialect-1 1288 158. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtls-dialect-1 1289 159. https://gcc.gnu.org/gcc-14/changes.html#riscv 1290 160. https://rivai-ic.com.cn/ 1291 161. https://www.ventanamicro.com/ 1292 162. https://www.rivosinc.com/ 1293 163. https://github.com/riscv-non-isa/riscv-rvv-intrinsic-doc/tree/v0.11.x 1294 164. https://github.com/riscv-non-isa/riscv-rvv-intrinsic-doc/pull/234 1295 165. https://eswincomputing.com/ 1296 166. https://plctlab.org/ 1297 167. https://www.andestech.com/ 1298 168. https://rivai-ic.com.cn/ 1299 169. https://bluewhale.systems/ 1300 170. https://gcc.gnu.org/gcc-14/changes.html#sparc 1301 171. https://gcc.gnu.org/gcc-14/changes.html#analyzer 1302 172. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-infinite-loop 1303 173. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-overlapping-buffers 1304 174. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-undefined-behavior-strtok 1305 175. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 1306 176. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1307 177. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1308 178. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 1309 179. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 1310 180. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-assertion 1311 181. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 1312 182. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 1313 183. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 1314 184. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-null_005fterminated_005fstring_005farg-function-attribute 1315 185. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-format-function-attribute 1316 186. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-alloc_005fsize-function-attribute 1317 187. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1318 188. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-allocation-size 1319 189. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 1320 190. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 1321 191. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 1322 192. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-text-art-charset 1323 193. https://gcc.gnu.org/gcc-14/changes.html#sarif 1324 194. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1325 195. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-json-formatting 1326 196. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Developer-Options.html#index-ftime-report 1327 197. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1328 198. https://gcc.gnu.org/gcc-14/changes.html#plugins 1329 199. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a1e9f3ed7aa49adad02190ace0614e0b37fc089 1330 200. https://gcc.gnu.org/gcc-14/changes.html#14.1 1331 201. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.0 1332 202. https://gcc.gnu.org/gcc-14/changes.html#14.2 1333 203. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.2 1334 204. https://gcc.gnu.org/gcc-14/changes.html#14.3 1335 205. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.3 1336 206. https://gcc.gnu.org/gcc-14/changes.html#14.4 1337 207. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.4 1338 208. https://gcc.gnu.org/onlinedocs/ 1339 209. mailto:gcc-help (a] gcc.gnu.org 1340 210. mailto:gcc (a] gcc.gnu.org 1341 211. https://gcc.gnu.org/lists.html 1342 212. https://www.fsf.org/ 1343 213. https://gcc.gnu.org/about.html 1344 ====================================================================== 1345 http://gcc.gnu.org/gcc-13/index.html 1346 GCC 13 Release Series 1347 1348 June 5, 2025 1349 1350 The GCC developers are pleased to announce the release of GCC 13.4. 1351 1352 This release is a bug-fix release, containing fixes for regressions in 1353 GCC 13.3 relative to previous releases of GCC. 1354 1355 Release History 1356 1357 GCC 13.4 1358 June 5, 2025 ([1]changes, [2]documentation) 1359 1360 GCC 13.3 1361 May 21, 2024 ([3]changes, [4]documentation) 1362 1363 GCC 13.2 1364 July 27, 2023 ([5]changes, [6]documentation) 1365 1366 GCC 13.1 1367 April 26, 2023 ([7]changes, [8]documentation) 1368 1369 References and Acknowledgements 1370 1371 GCC used to stand for the GNU C Compiler, but since the compiler 1372 supports several other languages aside from C, it now stands for the 1373 GNU Compiler Collection. 1374 1375 The GCC developers would like to thank the numerous people that have 1376 contributed new features, improvements, bug fixes, and other changes as 1377 well as test results to GCC. This [9]amazing group of volunteers is 1378 what makes GCC successful. 1379 1380 For additional information about GCC please refer to the [10]GCC 1381 project web site or contact the [11]GCC development mailing list. 1382 1383 To obtain GCC please use [12]our mirror sites or [13]our version 1384 control system. 1385 1386 1387 For questions related to the use of GCC, please consult these web 1388 pages and the [14]GCC manuals. If that fails, the 1389 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1390 web pages and the development of GCC are welcome on our developer 1391 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 1392 archives. 1393 1394 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 1395 distribution of this entire article is permitted in any medium, 1396 provided this notice is preserved. 1397 1398 These pages are [19]maintained by the GCC team. Last modified 1399 2025-06-05. 1400 1401 References 1402 1403 1. https://gcc.gnu.org/gcc-13/changes.html 1404 2. https://gcc.gnu.org/onlinedocs/13.4.0/ 1405 3. https://gcc.gnu.org/gcc-13/changes.html 1406 4. https://gcc.gnu.org/onlinedocs/13.3.0/ 1407 5. https://gcc.gnu.org/gcc-13/changes.html 1408 6. https://gcc.gnu.org/onlinedocs/13.2.0/ 1409 7. https://gcc.gnu.org/gcc-13/changes.html 1410 8. https://gcc.gnu.org/onlinedocs/13.1.0/ 1411 9. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Contributors.html 1412 10. https://gcc.gnu.org/index.html 1413 11. mailto:gcc (a] gcc.gnu.org 1414 12. https://gcc.gnu.org/mirrors.html 1415 13. https://gcc.gnu.org/git.html 1416 14. https://gcc.gnu.org/onlinedocs/ 1417 15. mailto:gcc-help (a] gcc.gnu.org 1418 16. mailto:gcc (a] gcc.gnu.org 1419 17. https://gcc.gnu.org/lists.html 1420 18. https://www.fsf.org/ 1421 19. https://gcc.gnu.org/about.html 1422 ====================================================================== 1423 http://gcc.gnu.org/gcc-13/changes.html 1424 GCC 13 Release Series 1425 Changes, New Features, and Fixes 1426 1427 This page is a "brief" summary of some of the huge number of 1428 improvements in GCC 13. You may also want to check out our [1]Porting 1429 to GCC 13 page and the [2]full GCC documentation. 1430 1431 Caveats 1432 1433 * OpenMP offloading to Intel MIC has been removed. 1434 * The support for the cr16-elf, tilegx*-linux, tilepro*-linux, 1435 hppa[12]*-*-hpux10*, hppa[12]*-*-hpux11* and m32c-rtems 1436 configurations has been removed. 1437 * Support for Solaris 11.3 (*-*-solaris2.11.3) has been declared 1438 obsolete. The next release of GCC will have corresponding code 1439 permanently removed. Details can be found in the [3]announcement. 1440 * Support for emitting the STABS debugging format (including the 1441 -gstabs and -gxcoff options) has been removed. (This means the dbx 1442 debugger is no longer supported, either.) 1443 * Legacy debug info compression option -gz=zlib-gnu was removed and 1444 the option is ignored right now. 1445 * [4]-Warray-bounds=2 will no longer issue warnings for out of bounds 1446 accesses to trailing struct members of one-element array type 1447 anymore. Instead it diagnoses accesses to trailing arrays according 1448 to [5]-fstrict-flex-arrays. 1449 * [6]-fanalyzer is still only suitable for analyzing C code. In 1450 particular, using it on C++ is unlikely to give meaningful output. 1451 * In the arm port, support for the iWMMXt extensions, enabled through 1452 -mcpu=iwmmxt, has been deprecated and will be removed in a future 1453 release. This includes support for the [7]iWMMXt built-in 1454 functions. 1455 * For C++, construction of the global iostream objects std::cout, 1456 std::cin, etc. is now done inside the standard library, instead of 1457 in every source file that includes the <iostream> header. This 1458 change improves the start-up performance of C++ programs, but it 1459 means that code compiled with GCC 13.1 will crash if the correct 1460 version of libstdc++.so is not used at run time. See the 1461 [8]documentation about using the right libstdc++.so at run time. 1462 Future GCC releases will mitigate the problem so that the program 1463 cannot be run at all with an older libstdc++.so. 1464 1465 [9]General Improvements 1466 1467 * [10]OpenMP 1468 + Reverse offload is now supported with AMD GCN and nvptx 1469 devices. Additionally, the requires handling has been improved 1470 and all clauses are now accepted. If a requirement cannot be 1471 fulfilled for an accessible device, this device is excluded 1472 from the list of available devices. This may imply that the 1473 only device left is the host (the initial device). In 1474 particular, unified_address and unified_shared_memory are 1475 unsupported by all non-host devices. 1476 + OpenMP 5.0: Fortran now supports some non-rectangular loop 1477 nests; for C/C++, the support was added in GCC 11. 1478 + The following OpenMP 5.1 features have been added: the 1479 omp_all_memory reserved locator, the inoutset modifier to the 1480 depend clause, the nowait clause for the taskwait directive 1481 and the omp_target_is_accessible, omp_target_memcpy_async, 1482 omp_target_memcpy_rect_async and omp_get_mapped_ptr API 1483 routines. The assume and assumes directives, the begin/end 1484 declare target syntax in C/C++ and device-specific ICV 1485 settings with environment variables are now supported. 1486 + Initial support for OpenMP 5.2 features has been added: 1487 firstprivate and allocate clauses on the scope construct; the 1488 OpenMP 5.2 syntax of the linear clause; new enum/constants 1489 omp_initial_device and omp_invalid_device; and optionally 1490 omitting the map-type in target enter/exit data. The enter 1491 clause (as alias for to) has been added to the declare target 1492 directive. Also added have been the omp_in_explicit_task 1493 routine and the doacross clause as alias for depend with 1494 source/sink modifier. 1495 + The _ALL suffix to the device-scope environment variables 1496 added in Technical Report (TR11) is already handled. 1497 + For user defined allocators requesting high bandwidth or large 1498 capacity memspaces or interleaved partitioning, the 1499 [11]memkind library is used, if available at run time. 1500 * AddressSanitizer defaults to detect_stack_use_after_return=1 on 1501 GNU/Linux targets. For compatibility, it can be disabled with env 1502 ASAN_OPTIONS=detect_stack_use_after_return=0. 1503 * New debug info compression option value -gz=zstd has been added. 1504 * Link-time optimization improvements: 1505 + LTO supports the newly added jobserver of GNU make jobserver 1506 that uses named pipes (--jobserver-style=fifo) by default. 1507 + If make's jobserver is active, parallel LTO WPA streaming 1508 communicates with it and thus avoids system overcommitting. 1509 * -Ofast, -ffast-math and -funsafe-math-optimizations will no longer 1510 add startup code to alter the floating-point environment when 1511 producing a shared object with -shared. 1512 * GCC can now emit its diagnostics using [12]SARIF. This is a 1513 JSON-based format suited for capturing the results of static 1514 analysis tools (like GCC's [13]-fanalyzer), but it can also be used 1515 to capture other GCC warnings and errors in a machine-readable 1516 format. Specifically, the [14]-fdiagnostics-format= option has been 1517 extended to support these new values: 1518 + -fdiagnostics-format=sarif-stderr 1519 + -fdiagnostics-format=sarif-file 1520 + -fdiagnostics-format=json-stderr, a synonym for the existing 1521 -fdiagnostics-format=json 1522 + -fdiagnostics-format=json-file 1523 where the json-prefixed variants refer to GCC's own JSON diagnostic 1524 format. 1525 * Support for profiling and test coverage in freestanding 1526 environments has been added, see also [15]Profiling and Test 1527 Coverage in Freestanding Environments. 1528 * New options -fharden-compares and -fharden-conditional-branches to 1529 verify compares and conditional branches, to detect some 1530 power-deprivation hardware attacks, using reversed conditions. 1531 1532 [16]New Languages and Language specific improvements 1533 1534 [17]Ada 1535 1536 * Traceback support added in RTEMS for the PPC ELF and ARM 1537 architectures. 1538 * Support for versions older than VxWorks 7 has been removed. 1539 * General improvements to the contracts in the standard libraries. 1540 * Addition of GNAT.Binary_Search. 1541 * Further additions and fixes for the Ada 2022 specification. 1542 * The Pragma SPARK_Mode=>Auto is now accepted. Contract analysis has 1543 been further improved. 1544 * Documentation improvements. 1545 1546 [18]C family 1547 1548 * New warnings: 1549 + [19]-Wxor-used-as-pow warns about uses of ^, the exclusive or 1550 operator, where it appears the user meant exponentiation 1551 ([20]PR90885) 1552 * Three new function attributes for documenting int arguments that 1553 are file descriptors: 1554 + [21]__attribute__((fd_arg(N))) 1555 + [22]__attribute__((fd_arg_read(N))) 1556 + [23]__attribute__((fd_arg_write(N))) 1557 These are used by [24]-fanalyzer to detect misuses of file 1558 descriptors. 1559 * A new statement attribute for C++23 [25]P1774R8 Portable 1560 assumptions support also in C or older C++: 1561 [26]__attribute__((assume(EXPR))); 1562 * GCC can now control when to treat the trailing array of a structure 1563 as a flexible array member for the purpose of accessing the 1564 elements of such an array. By default, all trailing arrays in 1565 aggregates are treated as flexible array members. Use the new 1566 command-line option [27]-fstrict-flex-arrays to control which array 1567 members are treated as flexible arrays. 1568 1569 [28]C 1570 1571 * Several C23 features have been implemented: 1572 + [29]N3042, Introduce the nullptr constant 1573 + [30]N2963, Enhanced Enumerations (fixed underlying types) 1574 + [31]N2975, Relax requirements for variadic parameter lists 1575 + [32]N3007, Type inference for object definitions (auto) 1576 + [33]N3018, The constexpr specifier for object definitions 1577 + [34]N3038, Introduce storage-class specifiers for compound 1578 literals 1579 + typeof (previously supported as an extension) and 1580 typeof_unqual 1581 + New keywords alignas, alignof, bool, false, static_assert, 1582 thread_local, true 1583 + [35]N2764, The noreturn attribute 1584 + Support for empty initializer braces 1585 + __STDC_VERSION_*_H__ header version macros 1586 + Removal of ATOMIC_VAR_INIT 1587 + unreachable macro in <stddef.h> 1588 + Removal of trigraphs 1589 + Removal of unprototyped functions 1590 + printf and scanf format checking with [36]-Wformat for %wN and 1591 %wfN format length modifiers 1592 + [37]N2836, Identifier Syntax using Unicode Standard Annex 31 1593 * In addition to those C23 features, existing features adopted in C23 1594 have been adjusted to follow C23 requirements and are not diagnosed 1595 with -std=c2x -Wpedantic. 1596 * New warnings: 1597 + [38]-Wenum-int-mismatch warns about mismatches between an 1598 enumerated type and an integer type ([39]PR105131) 1599 1600 [40]C++ 1601 1602 * Excess precision support (which has been available in C since GCC 1603 4.5) has been implemented for C++ as well. It is enabled by default 1604 in strict standard modes like -std=c++17, where it defaults to 1605 -fexcess-precision=standard, while in GNU standard modes like 1606 -std=gnu++20 it defaults to -fexcess-precision=fast. The option 1607 mainly affects IA-32/x86-64 using x87 math and in some cases on 1608 Motorola 68000, where float and double expressions are evaluated in 1609 long double precision and S/390, System z, IBM z Systems where 1610 float expressions are evaluated in double precision. Also, on 1611 several architectures where std::float16_t or std::bfloat16_t types 1612 are supported those are evaluated in float precision. 1613 -fexcess-precision=fast restores previous behavior. 1614 * Several C++23 features have been implemented: 1615 + [41]P2324R1, Labels at the end of compound statements 1616 ([42]PR103539) 1617 + [43]P2255R2, A type trait to detect reference binding to 1618 temporary ([44]PR104477) 1619 + [45]P2327R1, De-deprecating volatile compound operations 1620 + [46]P2437R1, Support for #warning ([47]PR106646) 1621 + [48]P2290R3, Delimited escape sequences ([49]PR106645) 1622 + [50]P2071R2, Named universal character escapes ([51]PR106648) 1623 + [52]P2513R3, char8_t Compatibility and Portability Fix 1624 ([53]PR106656) 1625 + [54]P1169R4, static operator() ([55]PR106651) 1626 + [56]P2266R3, Simpler implicit move ([57]PR101165) 1627 + [58]P2468R2, The Equality Operator You Are Looking For 1628 ([59]PR106644) 1629 + [60]P2362R3, Remove non-encodable wide character literals and 1630 multicharacter wide character literals ([61]PR106647) 1631 + [62]P2448R2, Relaxing some constexpr restrictions 1632 ([63]PR106649) 1633 + [64]P1467R9, Extended floating-point types and standard names 1634 ([65]PR106652) 1635 + [66]P1774R8, Portable assumptions ([67]PR106654) 1636 + [68]P2295R6, Support for UTF-8 as a portable source file 1637 encoding ([69]PR106655) 1638 + [70]P2589R1, static operator[] ([71]PR107684) 1639 * New warnings: 1640 + [72]-Wself-move warns when a value is moved to itself with 1641 std::move ([73]PR81159) 1642 + [74]-Wdangling-reference warns when a reference is bound to a 1643 temporary whose lifetime has ended ([75]PR106393) 1644 * The [76]-Wpessimizing-move and [77]-Wredundant-move warnings have 1645 been extended to warn in more contexts. 1646 * The [78]-nostdlib++ option has been added, to enable linking with 1647 g++ without implicitly linking in the C++ standard library. 1648 1649 [79]Runtime Library (libstdc++) 1650 1651 * Improved experimental support for C++20, including: 1652 + <format> header and std::format. 1653 + std::chrono::utc_clock and other clocks, time zones, and 1654 std::format support in the <chrono> header. 1655 * Improved experimental support for C++23, including: 1656 + Additions to the <ranges> header: views::zip, 1657 views::zip_transform, views::adjacent, 1658 views::adjacent_transform views::pairwise, views::slide, 1659 views::chunk, views::chunk_by, views::repeat, views::chunk_by, 1660 views::cartesian_product, views::as_rvalue, views::enumerate, 1661 views::as_const. 1662 + Additions to the <algorithm> header: ranges::contains, 1663 ranges::contains_subrange, ranges::iota, ranges::find_last, 1664 ranges::find_last_if, ranges::find_last_if_not, 1665 ranges::fold_left, ranges::fold_left_first, 1666 ranges::fold_right, ranges::fold_right_last, 1667 ranges::fold_left_with_iter, 1668 ranges::fold_left_first_with_iter. 1669 + Monadic operations for std::expected. 1670 + Constexpr std::bitset, std::to_chars and std::from_chars. 1671 + Library support for extended floating-point types. 1672 * Support for the <experimental/scope> header from v3 of the Library 1673 Fundamentals Technical Specification. 1674 * Support for the <experimental/synchronized_value> header from v2 of 1675 the Concurrency Technical Specification. 1676 * Support for many previously unavailable features in freestanding 1677 mode, thanks to Arsen Arsenovi. For example, std::tuple is now 1678 available for freestanding compilation. The freestanding subset 1679 contains all the components made freestanding by [80]P1642, and 1680 libstdc++ adds more components such as std::array and 1681 std::string_view. Additionally, libstdc++ now respects the 1682 -ffreestanding compiler option and so it is not necessary to build 1683 a separate freestanding installation of libstdc++. Compiling with 1684 -ffreestanding restricts the available features to the freestanding 1685 subset, even if libstdc++ was built as a full, hosted 1686 implementation. 1687 1688 [81]D 1689 1690 * Support for the D programming language has been updated to version 1691 2.103.1 of the language and run-time library. Full changelog for 1692 this release and previous releases can be found on the 1693 [82]dlang.org website. 1694 * The following GCC attributes are now recognized and available from 1695 the gcc.attributes module with short-hand aliases for convenience: 1696 + @attribute("no_sanitize", arguments) or 1697 @no_sanitize(arguments). 1698 + @attribute("register") or @register. 1699 + @attribute("simd") or @simd. 1700 + @attribute("simd_clones", mask) or @simd_clones(mask). 1701 + @attribute("visibility", arguments) or @visibility(arguments). 1702 * New aliases have been added to gcc.attributes for compatibility 1703 with ldc.attributes. 1704 + The @hidden attribute is an alias for @attribute("visibility", 1705 "hidden"). 1706 + The @noSanitize attribute is an alias for 1707 @attribute("no_sanitize"). 1708 * Vector operation intrinsics prefetch, loadUnaligned, 1709 storeUnaligned, shuffle, shufflevector, extractelement, 1710 insertelement, convertvector, and blendvector have been added to 1711 the gcc.simd module. 1712 * New warnings: 1713 + [83]-Wbuiltin-declaration-mismatch= warns when a built-in 1714 function is declared with the wrong signature. 1715 + [84]-Wmismatched-special-enum warns when a special enum is 1716 declared with the wrong base type. 1717 * New version identifier D_Optimized is now predefined when the -O 1718 option, or any higher optimization level is used. 1719 * The predefinition of version D_Exceptions can now by controlled by 1720 the option -fexception. 1721 * The predefinition of version D_TypeInfo can now by controlled by 1722 the option -frtti. 1723 * The -fdebug= and -fversion= compiler switches no longer accept an 1724 integer argument. 1725 1726 [85]Fortran 1727 1728 * Finalization is now fully supported. 1729 1730 [86]Go 1731 1732 * GCC 13, like GCC 12, provides a complete implementation of the Go 1733 1.18 user packages. 1734 * Although Go 1.18 includes support for generic programming, that 1735 support is not yet available in GCC. 1736 1737 [87]Modula-2 1738 1739 * Support for the language Modula-2 has been added. This includes 1740 support for the ISO/IEC 10514-1, PIM2, PIM3, PIM4 dialects together 1741 with a complete set of ISO/IEC 10514-1 and PIM libraries. 1742 * The <* noreturn *> attribute is supported with the -Wreturn-type 1743 [88]option. 1744 1745 [89]New Targets and Target Specific Improvements 1746 1747 [90]AArch64 1748 1749 * The AArch64 target now supports Decimal Floating-point in the BID 1750 format through the libbid library. 1751 * A number of new CPUs are supported through the -mcpu and -mtune 1752 options (GCC identifiers in parentheses). 1753 + Ampere-1A (ampere1a). 1754 + Arm Cortex-A715 (cortex-a715). 1755 + Arm Cortex-X1C (cortex-x1c). 1756 + Arm Cortex-X3 (cortex-x3). 1757 + Arm Neoverse V2 (neoverse-v2). 1758 * Support has been added for the armv9.1-a, armv9.2-a and armv9.3-a 1759 arguments to the -march= option. 1760 * The FEAT_LRCPC feature is now supported by generating the LDAPR 1761 instructions for C and C++ atomic loads with an acquire memory 1762 model. This is enabled when compiling with the +rcpc extension to 1763 -march or a CPU target that supports this feature. 1764 * The FEAT_CSSC feature from the 2022 Arm Architecture extensions is 1765 supported through the +cssc extension option. When enabled, scalar 1766 operations like integer minimum, maximum, absolute value, count 1767 trailing zeroes (__builtin_ctz), population count 1768 (__builtin_popcount) can be implemented in a single instruction. 1769 * The FEAT_LSE2 feature is now supported through libatomic and 1770 provides lockless 16-byte atomics on systems that implement it. 1771 1772 [91]AMD Radeon (GCN) 1773 1774 * Support for the Instinct MI200 series devices ([92] gfx90a) has 1775 been added. 1776 * SIMD vectorization support has been improved; this and 1777 stack-handling changes [93]require newlib 4.3.0 (or newer). 1778 1779 [94]arm 1780 1781 * A number of new CPUs are supported through the -mcpu and -mtune 1782 options (GCC identifiers in parentheses). 1783 + STAR-MC1 (star-mc1). 1784 + Arm Cortex-X1C (cortex-x1c). 1785 + Arm Cortex-M85 (cortex-m85). 1786 * Support has been added for the M-profile PACBTI extension that can 1787 help harden the generated code against return-oriented and 1788 jump-oriented attacks. It can be enabled through the 1789 -mbranch-protection= option. 1790 1791 [95]AVR 1792 1793 * Support for the following devices has been added in GCC 13.3: 1794 + ATtiny102, ATtiny104, ATtiny424, ATtiny426, ATtiny427, 1795 ATtiny824, ATtiny826, ATtiny827, ATtiny1624, ATtiny1626, 1796 ATtiny1627, ATtiny3224, ATtiny3226, ATtiny3227, AVR32DD14, 1797 AVR32DD20, AVR32DD28, AVR32DD32, AVR32DU14, AVR32DU20, 1798 AVR32DU28, AVR32DU32, AVR32EA28, AVR32EA32, AVR32EA48, 1799 AVR64DD14, AVR64DD20, AVR64DD28, AVR64DD32, AVR64DU28, 1800 AVR64DU32, AVR64EA28, AVR64EA32, AVR64EA48, ATA5787, ATA5835, 1801 ATA5700M322. 1802 * Support for the following devices has been added in GCC 13.4: 1803 + AVR32SD20, AVR32SD28, AVR32SD32, AVR64SD28, AVR64SD32, 1804 AVR64SD48. 1805 * Support for the following devices has been added in GCC 13.5: 1806 + AVR32DA28S, AVR32DA32S, AVR32DA48S, AVR64DA28S, AVR64DA32S, 1807 AVR64DA48S AVR64DA64S, AVR128DA28S, AVR128DA32S, AVR128DA48S, 1808 AVR128DA64S, AVR32EB14, AVR32EB20, AVR32EB28, AVR32EB32, 1809 AVR16LA14, AVR16LA20, AVR16LA28, AVR16LA32, AVR32LA14, 1810 AVR32LA20, AVR32LA28, AVR32LA32. 1811 1812 [96]IA-32/x86-64 1813 1814 * For both C and C++ the __bf16 type is supported on x86 systems with 1815 SSE2 and above enabled. 1816 * Use this __bf16 type for AVX512BF16 intrinsics instead of 1817 __bfloat16 which is typedef for short. __bf16 is now part of the 1818 x86 psABI. Users need to adjust their AVX512BF16-related source 1819 code when upgrading to GCC 13. 1820 * New ISA extension support for Intel AMX-COMPLEX was added. 1821 AMX-COMPLEX intrinsics are available via the -mamx-complex compiler 1822 switch. 1823 * New ISA extension support for Intel AMX-FP16 was added. AMX-FP16 1824 intrinsics are available via the -mamx-fp16 compiler switch. 1825 * New ISA extension support for Intel AVX-IFMA was added. AVX-IFMA 1826 intrinsics are available via the -mavxifma compiler switch. 1827 * New ISA extension support for Intel AVX-NE-CONVERT was added. 1828 AVX-NE-CONVERT intrinsics are available via the -mavxneconvert 1829 compiler switch. 1830 * New ISA extension support for Intel AVX-VNNI-INT8 was added. 1831 AVX-VNNI-INT8 intrinsics are available via the -mavxvnniint8 1832 compiler switch. 1833 * New ISA extension support for Intel CMPccXADD was added. CMPccXADD 1834 intrinsics are available via the -mcmpccxadd compiler switch. 1835 * New ISA extension support for Intel PREFETCHI was added. PREFETCHI 1836 intrinsics are available via the -mprefetchi compiler switch. 1837 * New ISA extension support for Intel RAO-INT was added. RAO-INT 1838 intrinsics are available via the -mraoint compiler switch. 1839 * GCC now supports the Intel CPU named Raptor Lake through 1840 -march=raptorlake. Raptor Lake is based on Alder Lake. Since GCC 1841 13.5, CLDEMOTE is not enabled any longer. 1842 * GCC now supports the Intel CPU named Meteor Lake through 1843 -march=meteorlake. Meteor Lake is based on Alder Lake. Since GCC 1844 13.5, CLDEMOTE is not enabled any longer. 1845 * GCC now supports the Intel CPU named Sierra Forest through 1846 -march=sierraforest. Based on ISA extensions enabled on Alder Lake, 1847 the switch in addition enables the AVX-IFMA, AVX-NE-CONVERT, 1848 AVX-VNNI-INT8 and CMPccXADD ISA extensions. Since GCC 13.2, ENQCMD 1849 and UINTR are further enabled. 1850 * GCC now supports the Intel CPU named Grand Ridge through 1851 -march=grandridge. Based on Sierra Forest, the switch in addition 1852 enables the RAO-INT ISA extensions. Since GCC 13.2, ENQCMD and 1853 UINTR are further enabled. Since GCC 13.3, RAO-INT is not enabled 1854 any longer. 1855 * GCC now supports the Intel CPU named Emerald Rapids through 1856 -march=emeraldrapids. Emerald Rapids is based on Sapphire Rapids. 1857 * GCC now supports the Intel CPU named Granite Rapids through 1858 -march=graniterapids. Based on Sapphire Rapids, the switch in 1859 addition enables the AMX-FP16 and PREFETCHI ISA extensions. 1860 * GCC now supports the Intel CPU named Granite Rapids D through 1861 -march=graniterapids-d. Based on Granite Rapids, the switch in 1862 addition enables the AMX-COMPLEX ISA extensions. 1863 * Since GCC 13.5, CLDEMOTE is not enabled through the compiler switch 1864 -march=alderlake any longer. 1865 * GCC now supports AMD CPUs based on the znver4 core via 1866 -march=znver4. The switch makes GCC consider using 512-bit vectors 1867 when auto-vectorizing. 1868 1869 [97]LoongArch 1870 1871 * New features 1872 + The new command-line option -mexplicit-relocs decides whether 1873 to use the assembler relocation operator when dealing with 1874 symbolic addresses. It is enabled by default if a compatible 1875 assembler (binutils 2.40 or later) is present at GCC build 1876 time. 1877 + The new command-line option -mdirect-extern-access can be used 1878 to prevent accessing external symbols through GOT. 1879 + The new variable attribute [98]model has been added. 1880 * Built-in functions 1881 + The rint and copysign mathematical builtins (and their float 1882 variants) are now implemented as inline LoongArch intrinsics. 1883 + The lrint, logb, scalbln, scalbn and ldexp mathematical 1884 builtins (and their float variants) are now implemented as 1885 inline LoongArch intrinsics when using -fno-math-errno. 1886 + The lceil and lfloor mathematical builtins (and their float 1887 variants) are now implemented as inline LoongArch intrinsics 1888 when using -ffp-int-builtin-inexact. 1889 * Subprojects Support 1890 + libvtv now supports LoongArch. 1891 + libitm now supports LoongArch. 1892 + Address sanitizers other than HWASan and TSan are now 1893 supported on LoongArch. 1894 1895 [99]NVPTX 1896 1897 * The default value for the [100]-march option can be now changed 1898 when [101]building GCC using the [102]--with-arch= configure 1899 option. GCC's target libraries are then build both with sm_30 and 1900 the specified target architecture. If not specified, GCC defaults 1901 to sm_30. 1902 1903 [103]RISC-V 1904 1905 * Support for vector intrinsics as specified in [104]version 0.11 of 1906 the RISC-V vector intrinsic specification, thanks Ju-Zhe Zhong from 1907 [105]RiVAI for contributing most of implementation. 1908 * Support for the following standard extensions has been added: 1909 + Zawrs 1910 + Zbkb 1911 + Zbkc 1912 + Zbkx 1913 + Zdinx 1914 + Zfinx 1915 + Zfh 1916 + Zfhmin 1917 + Zhinx 1918 + Zhinxmin 1919 + Zicbom 1920 + Zicbop 1921 + Zicboz 1922 + Zknd 1923 + Zkne 1924 + Zksed 1925 + Zksh 1926 + Zmmul 1927 * Support for the following vendor extensions has been added: 1928 + XTheadBa 1929 + XTheadBb 1930 + XTheadBs 1931 + XTheadCmo 1932 + XTheadCondMov 1933 + XTheadFMemIdx 1934 + XTheadFmv 1935 + XTheadInt 1936 + XTheadMac 1937 + XTheadMemIdx 1938 + XTheadMemPair 1939 + XTheadSync 1940 * The following new CPUs are supported through the -mcpu option (GCC 1941 identifiers in parentheses). 1942 + T-Head's XuanTie C906 (thead-c906). 1943 * Improves the multi-lib selection mechanism for the bare-metal 1944 toolchain (riscv*-elf*). GCC will now automatically select the 1945 best-fit multi-lib candidate instead of requiring all possible 1946 reuse rules to be listed at build time. 1947 1948 [106]Operating Systems 1949 1950 [107]Windows 1951 1952 * The GNU threads library used by the win32 thread model has been 1953 reimplemented using direct Win32 API calls, except for the 1954 Objective-C specific subset. It requires Windows XP/Server 2003 or 1955 later. The new implementation also adds the support needed for the 1956 C++11 threads, using again direct Win32 API calls; this additional 1957 layer requires Windows Vista/Server 2008 or later. It is 1958 recommended to use a recent version of MinGW-W64 in conjunction 1959 with the win32 thread model. 1960 1961 [108]Improvements to Static Analyzer 1962 1963 * The analyzer has gained 20 new warnings: 1964 + [109]-Wanalyzer-allocation-size 1965 + [110]-Wanalyzer-deref-before-check 1966 + [111]-Wanalyzer-exposure-through-uninit-copy 1967 + Seven new warnings relating to misuse of file descriptors: 1968 o [112]-Wanalyzer-fd-access-mode-mismatch 1969 o [113]-Wanalyzer-fd-double-close 1970 o [114]-Wanalyzer-fd-leak 1971 o [115]-Wanalyzer-fd-phase-mismatch (e.g. calling accept on 1972 a socket before calling listen on it) 1973 o [116]-Wanalyzer-fd-type-mismatch (e.g. using a stream 1974 socket operation on a datagram socket) 1975 o [117]-Wanalyzer-fd-use-after-close 1976 o [118]-Wanalyzer-fd-use-without-check 1977 along with special-casing handling of the behavior of open, 1978 close, creat, dup, dup2, dup3, pipe, pipe2, read, and write. 1979 + [119]-Wanalyzer-imprecise-fp-arithmetic 1980 + [120]-Wanalyzer-infinite-recursion 1981 + [121]-Wanalyzer-jump-through-null 1982 + [122]-Wanalyzer-out-of-bounds 1983 + [123]-Wanalyzer-putenv-of-auto-var 1984 + [124]-Wanalyzer-tainted-assertion 1985 + Four new warnings for misuses of <stdarg.h>: 1986 o [125]-Wanalyzer-va-list-leak for complaining about 1987 missing va_end after a va_start or va_copy 1988 o [126]-Wanalyzer-va-list-use-after-va-end for complaining 1989 about va_arg or va_copy used on a va_list that's had 1990 va_end called on it 1991 o [127]-Wanalyzer-va-arg-type-mismatch for type-checking of 1992 va_arg usage in interprocedural execution paths against 1993 the types of the parameters that were actually passed to 1994 the variadic call 1995 o [128]-Wanalyzer-va-list-exhausted for complaining in 1996 interprocedural execution paths if va_arg is used too 1997 many times on a va_list 1998 along with numerous other improvements. 1999 2000 [129]Improvements for plugin authors 2001 2002 * GCC diagnostics can now be [130]associated with rules such as from 2003 coding standards documents, or specifications. Such rules have a 2004 code name and can have a URL, which GCC can print in text form or 2005 capture in its [131]SARIF output when emitting diagnostics. 2006 2007 Other significant improvements 2008 2009 [132]GCC 13.1 2010 2011 This is the [133]list of problem reports (PRs) from GCC's bug tracking 2012 system that are known to be fixed in the 13.1 release. This list might 2013 not be complete (that is, it is possible that some PRs that have been 2014 fixed are not listed here). 2015 2016 [134]GCC 13.2 2017 2018 This is the [135]list of problem reports (PRs) from GCC's bug tracking 2019 system that are known to be fixed in the 13.2 release. This list might 2020 not be complete (that is, it is possible that some PRs that have been 2021 fixed are not listed here). 2022 2023 [136]GCC 13.3 2024 2025 This is the [137]list of problem reports (PRs) from GCC's bug tracking 2026 system that are known to be fixed in the 13.3 release. This list might 2027 not be complete (that is, it is possible that some PRs that have been 2028 fixed are not listed here). 2029 2030 Language Specific Changes 2031 2032 C++ 2033 2034 * The libstdc++exp.a library now includes all the Filesystem TS 2035 symbols from the libstdc++fs.a library, and the experimental 2036 symbols for the C++23 std::stacktrace class from the 2037 libstdc++_libbacktrace.a library. This means that -lstdc++exp is 2038 the only library needed for all experimental libstdc++ features. 2039 2040 [138]GCC 13.4 2041 2042 This is the [139]list of problem reports (PRs) from GCC's bug tracking 2043 system that are known to be fixed in the 13.4 release. This list might 2044 not be complete (that is, it is possible that some PRs that have been 2045 fixed are not listed here). 2046 2047 2048 For questions related to the use of GCC, please consult these web 2049 pages and the [140]GCC manuals. If that fails, the 2050 [141]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2051 web pages and the development of GCC are welcome on our developer 2052 list at [142]gcc (a] gcc.gnu.org. All of [143]our lists have public 2053 archives. 2054 2055 Copyright (C) [144]Free Software Foundation, Inc. Verbatim copying and 2056 distribution of this entire article is permitted in any medium, 2057 provided this notice is preserved. 2058 2059 These pages are [145]maintained by the GCC team. Last modified 2060 2026-03-15. 2061 2062 References 2063 2064 1. https://gcc.gnu.org/gcc-13/porting_to.html 2065 2. https://gcc.gnu.org/onlinedocs/index.html#current 2066 3. https://gcc.gnu.org/pipermail/gcc/2022-December/240322.html 2067 4. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Warray-bounds 2068 5. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C-Dialect-Options.html#index-fstrict-flex-arrays 2069 6. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2070 7. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/ARM-iWMMXt-Built-in-Functions.html 2071 8. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dynamic_or_shared.html#manual.intro.using.linkage.dynamic 2072 9. https://gcc.gnu.org/gcc-13/changes.html#general 2073 10. https://gcc.gnu.org/projects/gomp/ 2074 11. http://memkind.github.io/memkind/ 2075 12. https://sarifweb.azurewebsites.net/ 2076 13. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2077 14. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2078 15. https://gcc.gnu.org/onlinedocs/gcc/Freestanding-Environments.html 2079 16. https://gcc.gnu.org/gcc-13/changes.html#languages 2080 17. https://gcc.gnu.org/gcc-13/changes.html#ada 2081 18. https://gcc.gnu.org/gcc-13/changes.html#c-family 2082 19. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wxor-used-as-pow 2083 20. https://gcc.gnu.org/PR90885 2084 21. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg-function-attribute 2085 22. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg_005fread-function-attribute 2086 23. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg_005fwrite-function-attribute 2087 24. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2088 25. https://wg21.link/p1774r8 2089 26. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Statement-Attributes.html#index-assume-statement-attribute 2090 27. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C-Dialect-Options.html#index-fstrict-flex-arrays 2091 28. https://gcc.gnu.org/gcc-13/changes.html#c 2092 29. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3042.htm 2093 30. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2963.htm 2094 31. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2975.pdf 2095 32. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3007.htm 2096 33. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3018.htm 2097 34. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3038.htm 2098 35. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2764.pdf 2099 36. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wformat 2100 37. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2836.pdf 2101 38. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wenum-int-mismatch 2102 39. https://gcc.gnu.org/PR105131 2103 40. https://gcc.gnu.org/gcc-13/changes.html#cxx 2104 41. https://wg21.link/p2324 2105 42. https://gcc.gnu.org/PR103539 2106 43. https://wg21.link/p2255 2107 44. https://gcc.gnu.org/PR104477 2108 45. https://wg21.link/p2327 2109 46. https://wg21.link/p2437 2110 47. https://gcc.gnu.org/PR106646 2111 48. https://wg21.link/p2290 2112 49. https://gcc.gnu.org/PR106645 2113 50. https://wg21.link/p2071 2114 51. https://gcc.gnu.org/PR106648 2115 52. https://wg21.link/p2513 2116 53. https://gcc.gnu.org/PR106656 2117 54. https://wg21.link/p1169r4 2118 55. https://gcc.gnu.org/PR106651 2119 56. https://wg21.link/p2266r3 2120 57. https://gcc.gnu.org/PR101165 2121 58. https://wg21.link/p2468r2 2122 59. https://gcc.gnu.org/PR106644 2123 60. https://wg21.link/p2362r3 2124 61. https://gcc.gnu.org/PR106647 2125 62. https://wg21.link/p2448r2 2126 63. https://gcc.gnu.org/PR106649 2127 64. https://wg21.link/p1467r9 2128 65. https://gcc.gnu.org/PR106652 2129 66. https://wg21.link/p1774r8 2130 67. https://gcc.gnu.org/PR106654 2131 68. https://wg21.link/p2295r6 2132 69. https://gcc.gnu.org/PR106655 2133 70. https://wg21.link/p2589r1 2134 71. https://gcc.gnu.org/PR107684 2135 72. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wself-move 2136 73. https://gcc.gnu.org/PR81159 2137 74. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdangling-reference 2138 75. https://gcc.gnu.org/PR106393 2139 76. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 2140 77. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 2141 78. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Link-Options.html#index-nostdlib_002b_002b 2142 79. https://gcc.gnu.org/gcc-13/changes.html#libstdcxx 2143 80. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p1642r11.html 2144 81. https://gcc.gnu.org/gcc-13/changes.html#d 2145 82. https://dlang.org/changelog/2.103.1.html 2146 83. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gdc/Warnings.html#index-Wno-builtin-declaration-mismatch 2147 84. https://gcc.gnu.org/onlinedocs/gcc-13.2.0/gdc/Warnings.html#index-Wmismatched-special-enum 2148 85. https://gcc.gnu.org/gcc-13/changes.html#fortran 2149 86. https://gcc.gnu.org/gcc-13/changes.html#go 2150 87. https://gcc.gnu.org/gcc-13/changes.html#modula2 2151 88. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gm2/Compiler-options.html 2152 89. https://gcc.gnu.org/gcc-13/changes.html#targets 2153 90. https://gcc.gnu.org/gcc-13/changes.html#aarch64 2154 91. https://gcc.gnu.org/gcc-13/changes.html#amdgcn 2155 92. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/AMD-GCN-Options.html 2156 93. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 2157 94. https://gcc.gnu.org/gcc-13/changes.html#arm 2158 95. https://gcc.gnu.org/gcc-13/changes.html#avr 2159 96. https://gcc.gnu.org/gcc-13/changes.html#x86 2160 97. https://gcc.gnu.org/gcc-13/changes.html#loongarch 2161 98. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/LoongArch-Variable-Attributes.html#LoongArch-Variable-Attributes 2162 99. https://gcc.gnu.org/gcc-13/changes.html#nvptx 2163 100. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Nvidia-PTX-Options.html 2164 101. https://gcc.gnu.org/install/ 2165 102. https://gcc.gnu.org/install/specific.html#nvptx-x-none 2166 103. https://gcc.gnu.org/gcc-13/changes.html#riscv 2167 104. https://github.com/riscv-non-isa/riscv-rvv-intrinsic-doc/tree/v0.11.x 2168 105. https://rivai-ic.com.cn/ 2169 106. https://gcc.gnu.org/gcc-13/changes.html#os 2170 107. https://gcc.gnu.org/gcc-13/changes.html#windows 2171 108. https://gcc.gnu.org/gcc-13/changes.html#analyzer 2172 109. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-allocation-size 2173 110. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-deref-before-check 2174 111. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-exposure-through-uninit-copy 2175 112. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-access-mode-mismatch 2176 113. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-double-close 2177 114. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-leak 2178 115. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-phase-mismatch 2179 116. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-type-mismatch 2180 117. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-use-after-close 2181 118. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-use-without-check 2182 119. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-imprecise-fp-arithmetic 2183 120. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-infinite-recursion 2184 121. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-jump-through-null 2185 122. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 2186 123. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-putenv-of-auto-var 2187 124. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-assertion 2188 125. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-leak 2189 126. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-use-after-va-end 2190 127. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-arg-type-mismatch 2191 128. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-exhausted 2192 129. https://gcc.gnu.org/gcc-13/changes.html#plugins 2193 130. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=0b14f590e3e9d95b8211b77d992589d5ab4c25f0 2194 131. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2195 132. https://gcc.gnu.org/gcc-13/changes.html#13.1 2196 133. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.0 2197 134. https://gcc.gnu.org/gcc-13/changes.html#13.2 2198 135. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.2 2199 136. https://gcc.gnu.org/gcc-13/changes.html#13.3 2200 137. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.3 2201 138. https://gcc.gnu.org/gcc-13/changes.html#13.4 2202 139. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.4 2203 140. https://gcc.gnu.org/onlinedocs/ 2204 141. mailto:gcc-help (a] gcc.gnu.org 2205 142. mailto:gcc (a] gcc.gnu.org 2206 143. https://gcc.gnu.org/lists.html 2207 144. https://www.fsf.org/ 2208 145. https://gcc.gnu.org/about.html 2209 ====================================================================== 2210 http://gcc.gnu.org/gcc-12/index.html 2211 GCC 12 Release Series 2212 2213 (This release series is no longer supported.) 2214 2215 July 11, 2025 2216 2217 The GCC developers are pleased to announce the release of GCC 12.5. 2218 2219 This release is a bug-fix release, containing fixes for regressions in 2220 GCC 12.4 relative to previous releases of GCC. 2221 2222 Release History 2223 2224 GCC 12.5 2225 July 11, 2025 ([1]changes, [2]documentation) 2226 2227 GCC 12.4 2228 Jun 20, 2024 ([3]changes, [4]documentation) 2229 2230 GCC 12.3 2231 May 8, 2023 ([5]changes, [6]documentation) 2232 2233 GCC 12.2 2234 Aug 19, 2022 ([7]changes, [8]documentation) 2235 2236 GCC 12.1 2237 May 6, 2022 ([9]changes, [10]documentation) 2238 2239 References and Acknowledgements 2240 2241 GCC used to stand for the GNU C Compiler, but since the compiler 2242 supports several other languages aside from C, it now stands for the 2243 GNU Compiler Collection. 2244 2245 The GCC developers would like to thank the numerous people that have 2246 contributed new features, improvements, bug fixes, and other changes as 2247 well as test results to GCC. This [11]amazing group of volunteers is 2248 what makes GCC successful. 2249 2250 For additional information about GCC please refer to the [12]GCC 2251 project web site or contact the [13]GCC development mailing list. 2252 2253 To obtain GCC please use [14]our mirror sites or [15]our version 2254 control system. 2255 2256 2257 For questions related to the use of GCC, please consult these web 2258 pages and the [16]GCC manuals. If that fails, the 2259 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2260 web pages and the development of GCC are welcome on our developer 2261 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 2262 archives. 2263 2264 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 2265 distribution of this entire article is permitted in any medium, 2266 provided this notice is preserved. 2267 2268 These pages are [21]maintained by the GCC team. Last modified 2269 2025-08-04. 2270 2271 References 2272 2273 1. https://gcc.gnu.org/gcc-12/changes.html 2274 2. https://gcc.gnu.org/onlinedocs/12.5.0/ 2275 3. https://gcc.gnu.org/gcc-12/changes.html 2276 4. https://gcc.gnu.org/onlinedocs/12.4.0/ 2277 5. https://gcc.gnu.org/gcc-12/changes.html 2278 6. https://gcc.gnu.org/onlinedocs/12.3.0/ 2279 7. https://gcc.gnu.org/gcc-12/changes.html 2280 8. https://gcc.gnu.org/onlinedocs/12.2.0/ 2281 9. https://gcc.gnu.org/gcc-12/changes.html 2282 10. https://gcc.gnu.org/onlinedocs/12.1.0/ 2283 11. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Contributors.html 2284 12. https://gcc.gnu.org/index.html 2285 13. mailto:gcc (a] gcc.gnu.org 2286 14. https://gcc.gnu.org/mirrors.html 2287 15. https://gcc.gnu.org/git.html 2288 16. https://gcc.gnu.org/onlinedocs/ 2289 17. mailto:gcc-help (a] gcc.gnu.org 2290 18. mailto:gcc (a] gcc.gnu.org 2291 19. https://gcc.gnu.org/lists.html 2292 20. https://www.fsf.org/ 2293 21. https://gcc.gnu.org/about.html 2294 ====================================================================== 2295 http://gcc.gnu.org/gcc-12/changes.html 2296 GCC 12 Release Series 2297 Changes, New Features, and Fixes 2298 2299 This page is a "brief" summary of some of the huge number of 2300 improvements in GCC 12. You may also want to check out our [1]Porting 2301 to GCC 12 page and the [2]full GCC documentation. 2302 2303 Caveats 2304 2305 * An ABI incompatibility between C and C++ when passing or returning 2306 by value certain aggregates containing zero width bit-fields has 2307 been discovered on various targets. As mentioned in [3]PR102024, 2308 since the [4]PR42217 fix in GCC 4.5 the C++ front-end has been 2309 removing zero width bit-fields from the internal representation of 2310 the aggregates after the layout of those aggregates, but the C 2311 front-end kept them, so passing e.g. struct S { float a; int : 0; 2312 float b; } or struct T { float c; int : 0; } by value could differ 2313 between C and C++. Starting with GCC 12 the C++ front-end no longer 2314 removes those bit-fields from the internal representation and per 2315 clarified psABI some targets have been changed, so that they either 2316 ignore those bit-fields in the argument passing by value decisions 2317 in both C and C++, or they always take them into account. x86-64, 2318 ARM and AArch64 will always ignore them (so there is a C ABI 2319 incompatibility between GCC 11 and earlier with GCC 12 or later), 2320 PowerPC64 ELFv2 and S/390 always take them into account (so there 2321 is a C++ ABI incompatibility, GCC 4.4 and earlier compatible with 2322 GCC 12 or later, incompatible with GCC 4.5 through GCC 11). RISC-V 2323 has changed the handling of these already starting with GCC 10. As 2324 the ABI requires, MIPS takes them into account handling function 2325 return values so there is a C++ ABI incompatibility with GCC 4.5 2326 through 11. For function arguments on MIPS, refer to [5]the MIPS 2327 specific entry. GCC 12 on the above targets will report such 2328 incompatibilities as warnings or other diagnostics unless 2329 -Wno-psabi is used. 2330 * C: Computed gotos require a pointer type now. 2331 * C++: Two non-standard std::pair constructors have been deprecated. 2332 These allowed the use of an rvalue and a literal 0 to construct a 2333 pair containing a move-only type and a pointer. The nullptr keyword 2334 should be used to initialize the pointer member instead of a 2335 literal 0, as this is portable to other C++ implementations. 2336 * The configuration option --enable-libstdcxx-allocator no longer 2337 supports the bitmap, mt, and pool arguments. Those configurations 2338 had been broken for some time. 2339 * D: Building and bootstrapping GDC, the D compiler, now requires a 2340 working GDC (GCC version 9.1 or later) and D runtime library, 2341 libphobos, as the D front end is written in D. On some targets, 2342 libphobos isn't enabled by default, but compiles and works if 2343 --enable-libphobos is used. Other targets may require a more recent 2344 version of GCC to bootstrap. Specifics are documented for affected 2345 targets in the [6]Installing GCC manual. 2346 * Fortran: OpenMP code using the omp_lib.h include file can no longer 2347 be compiled with -std=f95 but now requires at least -std=f2003. 2348 Alternatively, use the omp_lib module, which still supports 2349 -std=f95 and is recommended to be used instead in general. 2350 * OpenMP offloading to Intel MIC has been deprecated and will be 2351 removed in a future release. 2352 * The cr16 target with the cr16-*-* configuration has been obsoleted 2353 and will be removed in a future release. 2354 * The hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11* configurations 2355 targeting 32-bit PA-RISC with HP-UX have been obsoleted and will be 2356 removed in a future release. 2357 * The m32c*-*-rtems* configuration has been obsoleted and will be 2358 removed in a future release. 2359 * The support for the m32r-*-linux*, m32rle-*-linux*, 2360 m68k*-*-openbsd* and vax-*-openbsd* configurations has been 2361 removed. 2362 * STABS: Support for emitting the STABS debugging format is 2363 deprecated and will be removed in the next release. All ports now 2364 default to emit DWARF (version 2 or later) debugging info or are 2365 obsoleted. 2366 * The optimization level -Ofast now implies 2367 -fno-semantic-interposition. 2368 2369 [7]General Improvements 2370 2371 * Vectorization is enabled at -O2 which is now equivalent to what 2372 would have been -O2 -ftree-vectorize -fvect-cost-model=very-cheap 2373 in the past. Note that the default vectorizer cost model has been 2374 changed; it used to behave as if -fvect-cost-model=cheap had been 2375 specified. 2376 * GCC now supports the [8]ShadowCallStack sanitizer, which can be 2377 enabled using the command-line option 2378 [9]-fsanitize=shadow-call-stack. This sanitizer currently only 2379 works on AArch64 targets and it requires an environment in which 2380 all code has been compiled with -ffixed-r18. Its primary initial 2381 user is the Linux kernel. 2382 2383 [10]New Languages and Language specific improvements 2384 2385 * OpenMP 2386 + OpenMP 5.0 support has been extended: The close map modifier 2387 and the affinity clause are now supported. In addition, 2388 Fortran gained the following features which were available in 2389 C and C++ before: declare variant is now available, depobj, 2390 mutexinoutset and iterator can now also be used with the 2391 depend clause, defaultmap has been updated for OpenMP 5.0, and 2392 the loop directive and combined directives involving the 2393 master directive have been added. 2394 + The following OpenMP 5.1 features have been added: support for 2395 expressing OpenMP directives as C++ 11 attributes, the masked 2396 and scope constructs, the nothing and error directives, and 2397 using primary with the proc_bind clause and OMP_PROC_BIND 2398 environment variable, the reproducible and unconstrained 2399 modifiers to the order clause, and, for C/C++ only, the align 2400 and allocator modifiers to the allocate clause and the atomic 2401 extensions are now available. The OMP_PLACE environment 2402 variable supports the OpenMP 5.1 features. In addition, the 2403 OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT environment variables 2404 and their associated API routines are now supported as well as 2405 the memory-allocation routines added for Fortran and extended 2406 for C/C++ in OpenMP 5.1. In Fortran code, strictly structured 2407 blocks can be used. 2408 + The [11]OpenMP Implementation Status can be found in the 2409 libgomp manual. 2410 * Version 2.6 of the [12]OpenACC specification continues to be 2411 maintained and improved in the C, C++ and Fortran compilers. See 2412 the [13]implementation status section on the OpenACC wiki page and 2413 the [14]run-time library documentation for further information. In 2414 addition to general performance tuning and bug fixing, new features 2415 include: 2416 + OpenACC worker parallelism for [15]AMD GPUs (already for a 2417 long time supported for [16]Nvidia GPUs). 2418 + Data privatization/sharing at the OpenACC gang level. 2419 + Considerable improvements for the experimental OpenACC 2420 'kernels' decomposition ([17]--param 2421 openacc-kernels=decompose). 2422 + A new warning flag [18]-Wopenacc-parallelism to warn about 2423 potentially suboptimal choices related to OpenACC parallelism. 2424 * The offload target code generation for OpenMP and OpenACC can now 2425 be better adjusted using the new [19]-foffload-options= flag and 2426 the pre-existing but now documented [20]-foffload= flag. 2427 2428 [21]Ada 2429 2430 * Ada 2022 2431 + Added the -gnat2022 flag to indicate strict Ada 2022 2432 compliance. The old -gnat2020 flag is now deprecated. 2433 + Support for Big Numbers (Annex G) has seen continuous 2434 improvements. It is now considered complete. It is compatible 2435 with SPARK, i.e. can be used from SPARK code. 2436 + Continuous improvements to the Ada 2022 standard since GCC 11. 2437 + Greatly improved compile time support. More functions can now 2438 have the with Static aspect and can be used in more contexts. 2439 * Ada 2022 extensions. The use of the -gnatX flag is necessary to 2440 access these features as they are not considered stable or 2441 standard. 2442 + Fixed lower bound for unconstrained arrays. 2443 o type Matrix is array (Natural range 0 .. <>, Natural 2444 range 0 .. <>) of Integer; is now valid. 2445 o Subtypes can also specify a lower bound: subtype String_1 2446 is String (1 .. <>);. Boundaries from slices will "slide" 2447 to the correct lower bound of the subtype. 2448 + Generalized Object.Operand notation. The following code is now 2449 valid V.Add_Element(42);, with V being a vector, for example. 2450 + Additional when constructs. Keywords return, goto and raise 2451 can now use when in addition to the existing exit when. The 2452 following expression is therefore now valid raise 2453 Constraint_Error with "Element is null" when Element = null; 2454 + Pattern matching 2455 o The case statement has been extended to cover records and 2456 arrays as well as finer grained casing on scalar types. 2457 In the future it is expected to provide more compile time 2458 guarantees when accessing discriminated fields. Case 2459 exhaustion is supported for pattern matching. An example 2460 would be 2461 type Sign is (Neg, Zero, Pos); 2462 2463 function Multiply (S1, S2 : Sign) return Sign is 2464 (case (S1, S2) is 2465 when (Neg, Neg) | (Pos, Pos) => Pos, 2466 when (Zero, <>) | (<>, Zero) => Zero, 2467 when (Neg, Pos) | (Pos, Neg) => Neg); 2468 2469 * gnatfind and gnatxref, which were already deprecated, have been 2470 removed. 2471 * Greatly expanded code covered by contracts. Thanks to this work, 2472 there are now several Ada standard libraries fully proven in SPARK 2473 which means they have no runtime nor logical errors. They are 2474 mostly numeric and string handling libraries. 2475 * Enable return-slot optimization for Pure functions. 2476 * General optimizations, improvements and additions to the standard 2477 library. Performance, correctness and in some cases stability was 2478 improved. Memory pools have also seen some minor enhancements. 2479 * Improvements to embedded-RTOS targets such as RTEMS, VxWorks and 2480 QNX. Older targets were removed or cleaned. 2481 * Added some [22]hardening features. 2482 2483 [23]C family 2484 2485 * Support for __builtin_shufflevector compatible with the clang 2486 language extension was added. 2487 * Support for attribute unavailable was added. 2488 * A new built-in function, __builtin_assoc_barrier, was added. It can 2489 be used to inhibit re-association of floating-point expressions. 2490 * Support for __builtin_dynamic_object_size compatible with the clang 2491 language extension was added. 2492 * New warnings: 2493 + [24]-Wbidi-chars warns about potentially misleading UTF-8 2494 bidirectional control characters. The default is 2495 -Wbidi-chars=unpaired ([25]PR103026) 2496 + [26]-Warray-compare warns about comparisons between two 2497 operands of array type ([27]PR97573) 2498 * Enhancements to existing warnings: 2499 + [28]-Wattributes has been extended so that it's possible to 2500 use -Wno-attributes=ns::attr or -Wno-attributes=ns:: to 2501 suppress warnings about unknown scoped attributes (in C++11 2502 and C2X). Similarly, #pragma GCC diagnostic ignored_attributes 2503 "vendor::attr" can be used to achieve the same effect 2504 ([29]PR101940) 2505 2506 [30]C 2507 2508 * Some new features from the upcoming C2X revision of the ISO C 2509 standard are supported with -std=c2x and -std=gnu2x. Some of these 2510 features are also supported as extensions when compiling for older 2511 language versions. In addition to the features listed, some 2512 features previously supported as extensions and now added to the C 2513 standard are enabled by default in C2X mode and not diagnosed with 2514 -std=c2x -Wpedantic. 2515 + Digit separators (as in C++) are supported for C2X. 2516 + The #elifdef and #elifndef preprocessing directives are now 2517 supported. 2518 + The printf and scanf format checking with [31]-Wformat now 2519 supports the %b format specified by C2X for binary integers, 2520 and the %B format recommended by C2X for printf. 2521 2522 [32]C++ 2523 2524 * Several C++23 features have been implemented: 2525 + [33]P1938R3, if consteval ([34]PR100974) 2526 + [35]P0849R8, auto(x): decay-copy in the language 2527 ([36]PR103049) 2528 + [37]P2242R3, Non-literal variables (and labels and gotos) in 2529 constexpr functions ([38]PR102612) 2530 + [39]P2334R1, Support for preprocessing directives elifdef and 2531 elifndef ([40]PR102616) 2532 + [41]P2360R0, Extend init-statement to allow alias-declaration 2533 ([42]PR102617) 2534 + [43]P2128R6, Multidimensional subscript operator 2535 + [44]DR 2397, auto specifier for pointers and references to 2536 arrays ([45]PR100975) 2537 * Several C++ Defect Reports have been resolved, e.g.: 2538 + [46]DR 960, Covariant functions and lvalue/rvalue references 2539 + [47]DR 1227, Mixing immediate and non-immediate contexts in 2540 deduction failure 2541 + [48]DR 1315, Restrictions on non-type template arguments in 2542 partial specializations 2543 + [49]DR 2082, Referring to parameters in unevaluated operands 2544 of default arguments 2545 + [50]DR 2351, void{} 2546 + [51]DR 2374, Overly permissive specification of enum 2547 direct-list-initialization 2548 + [52]DR 2397, auto specifier for pointers and references to 2549 arrays 2550 + [53]DR 2446, Questionable type-dependency of concept-ids 2551 * New command-line option -fimplicit-constexpr can be used to make 2552 inline functions implicitly constexpr ([54]git) 2553 * New command-line option -ffold-simple-inlines can be used to fold 2554 calls to certain trivial inline functions (currently std::move, 2555 std::forward, std::addressof and std::as_const). In contrast to 2556 inlining such calls, folding means that no intermediate code or 2557 debug information will be generated for them; this minimizes the 2558 abstraction penalty incurred for using these functions versus using 2559 the fundamental operations from which they're defined (e.g. 2560 std::move versus static_cast). This flag is enabled by default when 2561 -fno-inline is not active. 2562 * Deduction guides can be declared at class scope ([55]PR79501) 2563 * [56]-Wuninitialized warns about using uninitialized variables in 2564 member initializer lists ([57]PR19808) 2565 * [58]-Wint-in-bool-context is now disabled when instantiating a 2566 template ([59]git) 2567 * Stricter checking of attributes on friend declarations: if a friend 2568 declaration has an attribute, that declaration must be a 2569 definition. Moreover, a C++11 attribute cannot appear in the middle 2570 of the decl-specifier-seq. ([60]PR99032) 2571 * New warning options for C++ language mismatches: 2572 -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions, 2573 -Wc++20-extensions, and -Wc++23-extensions. They are enabled by 2574 default and can be used to control existing pedwarns about 2575 occurrences of new C++ constructs in code using an old C++ standard 2576 dialect. 2577 * New warning [61]-Wmissing-requires warns about missing requires 2578 ([62]git) 2579 * The existing std::is_constant_evaluated in if warning was extended 2580 to warn in more cases ([63]PR100995) 2581 * [64]-Waddress has been enhanced so that it now warns about, for 2582 instance, comparing the address of a nonstatic member function to 2583 null ([65]PR102103) 2584 * Errors about narrowing are no longer hidden if they occur in system 2585 headers 2586 * Ordered comparison of null pointers is now rejected ([66]PR99701) 2587 * Anonymous structs with bases are now rejected ([67]git) 2588 * The compiler rejects taking the address of an immediate member 2589 function ([68]PR102753) 2590 * The compiler has support for C++20 2591 __cpp_lib_is_pointer_interconvertible and 2592 __cpp_lib_is_layout_compatible to help the C++ library implement 2593 [69]P0466, Layout-compatibility and Pointer-interconvertibility 2594 Traits ([70]PR101539) 2595 * Memory usage of constraint subsumption has been improved 2596 ([71]PR100828) 2597 * constinit thread_local variables are optimized better 2598 ([72]PR101786) 2599 * Support for C++17 std::hardware_destructive_interference_size was 2600 added, along with the [73]-Winterference-size warning ([74]git) 2601 * Many bugs in the CTAD handling have been fixed ([75]PR101344, 2602 [76]PR101883, [77]PR89062, [78]PR101233, [79]PR88252, [80]PR86439, 2603 [81]PR98832, [82]PR102933 ...) 2604 * Two-stage name lookup for dependent operator expressions has been 2605 corrected ([83]PR51577) 2606 * Several issues with constrained variable templates have been fixed 2607 ([84]PR98486) 2608 * The compiler performs less instantiating when doing speculative 2609 constant evaluation ([85]git) 2610 * Various diagnostic improvements; e.g., a more precise caret 2611 location for pointer-to-member expressions 2612 * The new -fconstexpr-fp-except flag allows IEC559 floating point 2613 exceptions in constant-expressions. 2614 2615 [86]Runtime Library (libstdc++) 2616 2617 * Improved experimental C++20 support, including: 2618 + std::vector, std::basic_string, std::optional, and 2619 std::variant can be used in constexpr functions. 2620 + std::make_shared for arrays with default initialization, and 2621 std::atomic<std::shared_ptr<T>>. 2622 + Layout-compatibility and pointer-interconvertibility traits. 2623 * Improved experimental C++23 support, including: 2624 + Monadic operations for std::optional. 2625 + std::expected 2626 + std::move_only_function 2627 + <spanstream> 2628 + std::basic_string::resize_and_overwrite 2629 + std::unique_ptr can be used in constexpr functions. 2630 + <stacktrace> (not built by default, requires linking to an 2631 extra library). 2632 + <stdatomic.h> 2633 + std::invoke_r 2634 + constexpr std::type_info::operator== 2635 2636 [87]D 2637 2638 * New features: 2639 + Support for the D programming language has been updated to 2640 version 2.100.1 of the language and run-time library. Full 2641 changelog for this release and previous releases can be found 2642 on the [88]dlang.org website. 2643 + On supported targets, the __traits(compiles) expression can 2644 now be used to determine whether a target-specific built-in is 2645 available without error during CTFE ([89]PR101127). 2646 + Functions annotated with pragma(inline, true) are now compiled 2647 into every module where they are used from ([90]PR106563). 2648 + Partial support for directly importing C99 sources into a D 2649 compilation ([91]ImportC) has been added to the language. A 2650 notable missing feature is support for preprocessing C 2651 imports, which can be worked around by preprocessing all C 2652 sources used for importing ahead of time. 2653 * New language options: 2654 + -fcheck=, enables or disables the code generation of specific 2655 run-time contract checks. 2656 + -fcheckaction=, controls the run-time behavior on an assert, 2657 array bounds check, or final switch contract failure. The 2658 default is -fcheckaction=throw. 2659 + -fdump-c++-spec=, dumps all compiled extern(C++) declarations 2660 as C++ code to the given file. The supplementary option 2661 -fdump-c++-spec-verbose turns on emission of comments for 2662 ignored declarations in the generated spec. 2663 + -fextern-std=, controls which C++ standard extern(C++) 2664 declarations are compiled to be compatible with. The default 2665 is -fextern-std=c++17. 2666 + -fpreview=, added to enable upcoming D language features in 2667 the compiler. 2668 + -frevert=, added to revert D language changes to support older 2669 D codebases that need more time to transition. 2670 + -fsave-mixins=, saves mixins expanded at compile-time to a 2671 file. 2672 * Deprecated and removed features: 2673 + The -Wtemplates compiler switch has been removed, as it had 2674 been superceded by -ftransition=templates, which more 2675 accurately reports on which templates have been instantiated. 2676 + The -ftransition=dip25 and -ftransition=dip1000 compiler 2677 switches have been renamed to -fpreview=dip25 and 2678 -fpreview=dip1000. 2679 2680 [92]Fortran 2681 2682 * WG5/N1942, "TS 29113 Further Interoperability of Fortran with C", 2683 is now fully supported. In addition to implementing previously 2684 missing functionality, such as support for character arguments of 2685 length greater than one in functions marked bind(c) and gaps in the 2686 handling for assumed-rank arrays, numerous other bugs have been 2687 fixed, and an extensive set of new conformance test cases has been 2688 added. 2689 * GCC 12 now uses OPERATION as the name of the function to the 2690 CO_REDUCE intrinsic for the pairwise reduction, thus conforming to 2691 the Fortran 2018 standard. Previous versions used OPERATOR which 2692 conforms to TS 18508. 2693 * On POWER systems which support it, the -mabi=ieeelongdouble option 2694 now selects the IEEE 128-bit floating point format for 2695 REAL(KIND=16). R16_IBM and R16_IEEE have been added to the 2696 -fconvert option, the CONVERT specifier of the OPEN statement and 2697 the GFORTRAN_CONVERT_UNIT environment variable. 2698 2699 [93]Go 2700 2701 * GCC 12 provides a complete implementation of the Go 1.18 user 2702 packages. 2703 * Although Go 1.18 includes support for generic programming, that 2704 support is not yet available in GCC. 2705 2706 [94]libgccjit 2707 2708 * The libgccjit API gained 30 new entry points: 2709 + 17 new "reflection" entry points for querying functions and 2710 types ([95]LIBGCCJIT_ABI_16) 2711 + [96]gcc_jit_lvalue_set_tls_model for supporting thread-local 2712 variables ([97]LIBGCCJIT_ABI_17) 2713 + [98]gcc_jit_lvalue_set_link_section for setting the link 2714 section of global variables, analogous to 2715 [99]__attribute__((section(".section"))) 2716 ([100]LIBGCCJIT_ABI_18) 2717 + 4 new entry points for initializing global variables and 2718 creating constructors for rvalues ([101]LIBGCCJIT_ABI_19) 2719 + Support for sized integer types, including 128-bit integers 2720 and helper functions for such types ([102]LIBGCCJIT_ABI_20) 2721 + [103]gcc_jit_context_new_bitcast for reinterpreting the bits 2722 of an rvalue as a different type ([104]LIBGCCJIT_ABI_21) 2723 + [105]gcc_jit_lvalue_set_register_name for setting a specific 2724 register for a variable ([106]LIBGCCJIT_ABI_22) 2725 + [107]gcc_jit_context_set_bool_print_errors_to_stderr 2726 ([108]LIBGCCJIT_ABI_23) 2727 + 2 new entry points for setting the alignment of a variable 2728 ([109]LIBGCCJIT_ABI_24) 2729 * libgccjit has gained support for the use of various atomic builtins 2730 ([110]PR96066, [111]PR96067) 2731 * [112]gcc_jit_context_new_cast is now able to handle truncation and 2732 extension between different integer types ([113]PR95498) 2733 2734 [114]New Targets and Target Specific Improvements 2735 2736 [115]AArch64 & arm 2737 2738 * Newer revisions of the Arm Architecture are supported as arguments 2739 to the -march option: armv8.7-a, armv8.8-a, armv9-a. 2740 * The Arm Cortex-A510 CPU is now supported through the cortex-a510 2741 argument to the -mcpu and -mtune options. 2742 * GCC can now auto-vectorize operations performing sign-differing 2743 dot-product operations, taking advantage of instructions in the 2744 Advanced SIMD (AArch64/AArch32) and SVE (AArch64) instruction sets. 2745 2746 [116]AArch64 2747 2748 * A number of new CPUs are supported through the -mcpu and -mtune 2749 options (GCC identifiers in parentheses). 2750 + Ampere-1 (ampere1). 2751 + Arm Cortex-A710 (cortex-a710). 2752 + Arm Cortex-X2 (cortex-x2). 2753 * The 64-byte atomic load/store intrinsics to accelerator memory from 2754 the [117]2020 Arm Architecture extensions are supported through the 2755 +ls64 option extension. 2756 * Initial code generation support is supported for hardware 2757 instructions used to accelerate the memcpy,memmove and memset 2758 standard functions. These instructions can be generated when 2759 compiling with the +mopsoption extension. 2760 * The ACLE Advanced SIMD intrinsics accessible through the arm_neon.h 2761 header have been significantly reimplemented and generate 2762 higher-performing code than previous GCC versions. 2763 * The option -mtune=neoverse-512tvb is added to tune for Arm Neoverse 2764 cores that have a total vector bandwidth of 512 bits. Please refer 2765 to the documentation for more details. 2766 2767 [118]AMD Radeon (GCN) 2768 2769 * Debug experience with ROCGDB has been improved. 2770 * Support for the type __int128_t/integer(kind=16) was added. 2771 * For offloading, the limitation of using only one wavefront per 2772 compute unit (CU) has been lifted. Up to 40 workgroups per CU and 2773 16 wavefronts per workgroup are supported (up to a limit of 40 2774 wavefronts in total, per CU). Additionally, the number of used 2775 wavefronts and workgroups was tuned for performance. 2776 2777 [119]arm 2778 2779 * Support is added for accessing the stack canary value via the TLS 2780 register through the -fstack-protector-guard=tls and 2781 -mstack-protector-guard-offset= options. This intended for use in 2782 Linux kernel development. Please refer to the documentation for 2783 more details. 2784 2785 [120]BPF 2786 2787 * Support for CO-RE (compile-once, run-everywhere) has been added to 2788 the BPF back end. CO-RE allows to compile portable BPF programs 2789 that are able to run among different versions of the Linux kernel. 2790 2791 [121]IA-32/x86-64 2792 2793 * New ISA extension support for Intel AVX512-FP16 was added. 2794 AVX512-FP16 intrinsics are available via the -mavx512fp16 compiler 2795 switch. 2796 * Since GCC 12.5, CLDEMOTE is not enabled through the compiler switch 2797 -march=alderlake any longer. 2798 * For both C and C++ the _Float16 type is supported on x86 systems 2799 with SSE2 enabled. Without {-mavx512fp16}, all operations will be 2800 emulated in software and float instructions. 2801 * Mitigation against straight line speculation (SLS) for function 2802 return and indirect jump is supported via 2803 -mharden-sls=[none|all|return|indirect-jmp]. 2804 * Add CS prefix to call and jmp to indirect thunk with branch target 2805 in r8-r15 registers via -mindirect-branch-cs-prefix. 2806 * Always use global offset table (GOT) to access external data and 2807 function symbols when the new -mno-direct-extern-access 2808 command-line option is specified. 2809 2810 [122]LoongArch 2811 2812 * Support for the LoongArch architecture instruction set has been 2813 added. 2814 * The Loongson CPU codename LA464 and LoongArch 64-bit generic CPU 2815 codename loongarch64 are supported through the -march= and -mtune= 2816 options (GCC identifiers in parentheses). 2817 + Loongson LA464 core (la464). 2818 + LoongArch 64-bit generic core (loongarch64). 2819 2820 [123]MIPS 2821 2822 * The ABI passing arguments containing zero-width fields (for 2823 example, C/C++ zero-width bit-fields, GNU C/C++ zero-length arrays, 2824 and GNU C empty structs) has changed. Now a zero-width field will 2825 not prevent an aligned 64-bit floating-point field next to it from 2826 being passed through FPR. This is compatible with LLVM, but 2827 incompatible with previous GCC releases. GCC 12 on MIPS will report 2828 such incompatibilities as an inform unless -Wno-psabi is used. 2829 * The ABI returning values containing C++17 empty bases has changed. 2830 Now an empty base will not prevent an aggregate containing only one 2831 or two floating-point fields from being returned through FPR. This 2832 is compatible with GCC 6 and earlier, but incompatible with GCC 7 2833 through 11. GCC 12 on MIPS will report such incompatibilities as an 2834 inform unless -Wno-psabi is used. 2835 2836 [124]NVPTX 2837 2838 * The -march flag has been added. The -misa flag is now considered an 2839 alias of the -march flag. 2840 * Support for PTX ISA target architectures sm_53, sm_70, sm_75 and 2841 sm_80 has been added. These can be specified using the -march flag. 2842 * The default PTX ISA target architecture has been set back to sm_30, 2843 to fix support for sm_30 boards. 2844 * The -march-map flag has been added. The -march-map value will be 2845 mapped to an valid -march flag value. For instance, 2846 -march-map=sm_50 maps to -march=sm_35. This can be used to specify 2847 that generated code is to be executed on a board with at least some 2848 specific compute capability, without having to know the valid 2849 values for the -march flag. 2850 * The -mptx flag has been added to specify the PTX ISA version for 2851 the generated code; permitted values are 3.1 (matches previous GCC 2852 versions), 6.0, 6.3, and 7.0. If not specified, the used version is 2853 the minimal version required for -march but at least 6.0. 2854 * An mptx-3.1 multilib was added. This allows using older drivers 2855 which do not support PTX ISA version 6.0. 2856 * The new __PTX_SM__ predefined macro allows code to check the PTX 2857 ISA target architecture being targeted by the compiler. 2858 * The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__ 2859 predefined macros allows code to check the PTX ISA version being 2860 targeted by the compiler. 2861 2862 [125]PowerPC / PowerPC64 / RS6000 2863 2864 * The internal implementation of Power's target-specific built-in 2865 functions has been rewritten to be easier and less error-prone to 2866 maintain. Every attempt has been made to ensure that the new 2867 behavior matches the old behavior, but inevitably some bugs can be 2868 expected. Please [126]report any problems. 2869 * The built-in functions __builtin_get_texasr, __builtin_get_texasru, 2870 __builtin_get_tfhar, __builtin_get_tfiar, __builtin_set_texasr, 2871 __builtin_set_texasru, __builtin_set_tfhar, and __builtin_set_tfiar 2872 now behave as documented in all supported configurations. On prior 2873 releases, the arguments and return values of these functions were 2874 treated as unsigned long long instead of as unsigned long, when the 2875 options -m32 -mpowerpc64 were in effect. 2876 * The overloaded built-in functions vec_cntlz_lsbb and vec_cnttz_lsbb 2877 now behave as documented. On prior releases, these built-in 2878 functions had incorrect semantics on little-endian targets. 2879 2880 [127]PRU 2881 2882 * The [128]__regio_symbol variable qualifier has been added. It 2883 allows easier access in C programs to the __R30 and __R31 CPU I/O 2884 registers. 2885 2886 [129]RISC-V 2887 2888 * The default ISA spec version was bumped to 20191213; see [130]this 2889 announcement for details. 2890 * New ISA extension support for zba, zbb, zbc, zbs was added. 2891 * New ISA extension support for vector and scalar crypto was added, 2892 only support architecture testing macro and -march= parsing. 2893 * The option -mtune=thead-c906 is added to tune for T-HEAD c906 2894 cores. 2895 * libstdc++ no longer attempts to detect built-in atomics. 2896 Distributions that have out-of-tree workarounds for -latomic should 2897 check their ABIs again. 2898 2899 [131]Operating Systems 2900 2901 [132]Improvements to Static Analyzer 2902 2903 * The analyzer has gained a 2904 [133]-Wanalyzer-use-of-uninitialized-value warning, similar to 2905 [134]-Wuninitialized and [135]-Wmaybe-uninitialized, but based on 2906 an interprocedural path-sensitive analysis ([136]PR95006). 2907 Such warnings are not disabled by the new 2908 [137]-ftrivial-auto-var-init (see below), as the latter is 2909 considered a mitigation option. 2910 * [138]-Wanalyzer-write-to-const and 2911 [139]-Wanalyzer-write-to-string-literal will now check for 2912 [140]__attribute__ ((access, ....)) on calls to externally-defined 2913 functions, and complain about read-only regions pointed to by 2914 arguments marked with a write_only or read_write attribute 2915 ([141]PR104793). 2916 * The analyzer's "taint" mode, activated by 2917 [142]-fanalyzer-checker=taint (in addition to [143]-fanalyzer), has 2918 gained four new taint-based warnings: 2919 + [144]-Wanalyzer-tainted-allocation-size for e.g. 2920 attacker-controlled malloc and alloca, 2921 + [145]-Wanalyzer-tainted-divisor for detecting where an 2922 attacker can inject a divide-by-zero, 2923 + [146]-Wanalyzer-tainted-offset for attacker-controlled pointer 2924 offsets, 2925 + [147]-Wanalyzer-tainted-size for attacker-controlled values 2926 being used as a size parameter to calls to memset or to 2927 functions marked with [148]__attribute__ ((access, ....)). 2928 The existing [149]-Wanalyzer-tainted-array-index has been reworded 2929 to talk about "attacker-controlled" rather than "tainted" values, 2930 for consistency with the new warnings. 2931 A new [150]__attribute__ ((tainted_args)) has been added to the C 2932 and C++ front ends, usable on functions, and on function pointer 2933 callback fields in structs. The analyzer's taint mode will treat 2934 all parameters and buffers pointed to by parameters of such 2935 functions as being attacker-controlled, such as for annotating 2936 system calls in an operating system kernel as being an "attack 2937 surface". 2938 * The analyzer now respects [151]__attribute__((const)): it will 2939 treat such functions as returning the same value when given the 2940 same inputs ([152]PR104434), and as having no side effects 2941 ([153]PR104576). 2942 * The analyzer is now able to split its analysis into multiple 2943 execution paths in places where there isn't a split in the control 2944 flow graph. For example, it now handles realloc calls by splitting 2945 the execution path into three possible outcomes for the call: 2946 + failure, returning NULL 2947 + success, growing the buffer in-place without moving it 2948 + success, allocating a new buffer, copying the content of the 2949 old buffer to it, and freeing the old buffer 2950 * The analyzer's interprocedural path exploration logic is now able 2951 to track calls through function pointers. 2952 * The analyzer now makes the assumption that if we know PTR is 2953 non-NULL, then (PTR + OFFSET) is also non-NULL. This isn't strictly 2954 true, but eliminates false positives in practice ([154]PR101962). 2955 * The analyzer has gained some initial support for inline assembler 2956 code. This is extremely limited, and is purely to help suppress 2957 false positives when analyzing the Linux kernel, which makes heavy 2958 use of inline assembler ([155]PR101570). 2959 * The way the analyzer tracks the state of memory along an execution 2960 path has been improved in various ways for GCC 12: 2961 + An optimization for representing bulk updates to memory (e.g. 2962 zero fills) has been removed as it never worked well. In GCC 2963 12 it has been replaced with a simpler and more accurate 2964 approach, eliminating many false positives ([156]PR95006). 2965 + Various optimizations have been added, speeding up the 2966 analysis on a particularly problematic source file from 4 2967 minutes down to 17 seconds ([157]PR104943, [158]PR104954, and 2968 [159]PR104955). 2969 + The analyzer now tracks the sizes of dynamically-allocated 2970 regions, both on the heap (via malloc etc) and stack (via 2971 alloca), though none of the analyzer warnings make use of this 2972 yet in GCC 12. 2973 * The analyzer's handling of switch statements has been rewritten, 2974 fixing various bugs. 2975 2976 Other significant improvements 2977 2978 [160]Eliminating uninitialized variables 2979 2980 * GCC can now [161]initialize all stack variables implicitly, 2981 including padding. This is intended to eliminate all classes of 2982 uninitialized stack variable flaws. Lack of explicit initialization 2983 will still warn when [162]-Wuninitialized is active. For best 2984 debugging, use of the new command-line option 2985 [163]-ftrivial-auto-var-init=pattern can be used to fill variables 2986 with a repeated 0xFE pattern, which tends to illuminate many bugs 2987 (e.g. pointers receive invalid addresses, sizes and indices are 2988 very large). For best production results, the new command-line 2989 option [164]-ftrivial-auto-var-init=zero can be used to fill 2990 variables with 0x00, which tends to provide a safer state for bugs 2991 (e.g. pointers are NULL, strings are NUL filled, and sizes and 2992 indices are 0). 2993 2994 [165]Debugging formats 2995 2996 * GCC can now generate debugging information in CTF, a lightweight 2997 debugging format that provides information about C types and the 2998 association between functions and data symbols and types. This 2999 format is designed to be embedded in ELF files and to be very 3000 compact and simple. A new command-line option -gctf enables the 3001 generation of CTF. 3002 * GCC can now generate debugging information in BTF. This is a 3003 debugging format mainly used in BPF programs and the Linux kernel. 3004 The compiler can generate BTF for any target, when enabled with the 3005 command-line option -gbtf 3006 3007 [166]GCC 12.1 3008 3009 This is the [167]list of problem reports (PRs) from GCC's bug tracking 3010 system that are known to be fixed in the 12.1 release. This list might 3011 not be complete (that is, it is possible that some PRs that have been 3012 fixed are not listed here). 3013 3014 [168]GCC 12.2 3015 3016 This is the [169]list of problem reports (PRs) from GCC's bug tracking 3017 system that are known to be fixed in the 12.2 release. This list might 3018 not be complete (that is, it is possible that some PRs that have been 3019 fixed are not listed here). 3020 3021 Target Specific Changes 3022 3023 LoongArch 3024 3025 * The default setting of -m[check|no-check]-zero-division is changed 3026 for optimized code. Now -mno-check-zero-division is the default for 3027 all optimization levels but -O0 and -Og. The old behavior can be 3028 obtained by explicitly passing -mcheck-zero-division to GCC. 3029 3030 [170]GCC 12.3 3031 3032 Target Specific Changes 3033 3034 x86-64 3035 3036 * GCC now supports AMD CPUs based on the znver4 core via 3037 -march=znver4. The switch makes GCC consider using 512-bit vectors 3038 when auto-vectorizing. 3039 3040 This is the [171]list of problem reports (PRs) from GCC's bug tracking 3041 system that are known to be fixed in the 12.3 release. This list might 3042 not be complete (that is, it is possible that some PRs that have been 3043 fixed are not listed here). 3044 3045 [172]GCC 12.4 3046 3047 This is the [173]list of problem reports (PRs) from GCC's bug tracking 3048 system that are known to be fixed in the 12.4 release. This list might 3049 not be complete (that is, it is possible that some PRs that have been 3050 fixed are not listed here). 3051 3052 [174]GCC 12.5 3053 3054 This is the [175]list of problem reports (PRs) from GCC's bug tracking 3055 system that are known to be fixed in the 12.5 release. This list might 3056 not be complete (that is, it is possible that some PRs that have been 3057 fixed are not listed here). 3058 3059 3060 For questions related to the use of GCC, please consult these web 3061 pages and the [176]GCC manuals. If that fails, the 3062 [177]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3063 web pages and the development of GCC are welcome on our developer 3064 list at [178]gcc (a] gcc.gnu.org. All of [179]our lists have public 3065 archives. 3066 3067 Copyright (C) [180]Free Software Foundation, Inc. Verbatim copying and 3068 distribution of this entire article is permitted in any medium, 3069 provided this notice is preserved. 3070 3071 These pages are [181]maintained by the GCC team. Last modified 3072 2025-12-29. 3073 3074 References 3075 3076 1. https://gcc.gnu.org/gcc-12/porting_to.html 3077 2. https://gcc.gnu.org/onlinedocs/index.html#current 3078 3. https://gcc.gnu.org/PR102024 3079 4. https://gcc.gnu.org/PR42217 3080 5. https://gcc.gnu.org/gcc-12/changes.html#mips_zero_width_fields 3081 6. https://gcc.gnu.org/install/specific.html 3082 7. https://gcc.gnu.org/gcc-12/changes.html#general 3083 8. https://clang.llvm.org/docs/ShadowCallStack.html 3084 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Instrumentation-Options.html#index-fsanitize_003dshadow-call-stack 3085 10. https://gcc.gnu.org/gcc-12/changes.html#languages 3086 11. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/OpenMP-Implementation-Status.html 3087 12. https://www.openacc.org/ 3088 13. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-12 3089 14. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/Enabling-OpenACC.html 3090 15. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 3091 16. https://gcc.gnu.org/gcc-12/changes.html#nvptx 3092 17. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-param 3093 18. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wopenacc-parallelism 3094 19. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload-options 3095 20. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload 3096 21. https://gcc.gnu.org/gcc-12/changes.html#ada 3097 22. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gnat_rm/Security-Hardening-Features.html#Security-Hardening-Features 3098 23. https://gcc.gnu.org/gcc-12/changes.html#c-family 3099 24. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wbidi-chars 3100 25. https://gcc.gnu.org/PR103026 3101 26. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Warray-compare 3102 27. https://gcc.gnu.org/PR97573 3103 28. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wattributes 3104 29. https://gcc.gnu.org/PR101940 3105 30. https://gcc.gnu.org/gcc-12/changes.html#c 3106 31. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wformat 3107 32. https://gcc.gnu.org/gcc-12/changes.html#cxx 3108 33. https://wg21.link/p1938 3109 34. https://gcc.gnu.org/PR100974 3110 35. https://wg21.link/p0849 3111 36. https://gcc.gnu.org/PR103049 3112 37. https://wg21.link/p2242 3113 38. https://gcc.gnu.org/PR102612 3114 39. https://wg21.link/p2334 3115 40. https://gcc.gnu.org/PR102616 3116 41. https://wg21.link/p2360 3117 42. https://gcc.gnu.org/PR102617 3118 43. https://wg21.link/p2128 3119 44. https://wg21.link/cwg2397 3120 45. https://gcc.gnu.org/PR100975 3121 46. https://wg21.link/cwg960 3122 47. https://wg21.link/cwg1227 3123 48. https://wg21.link/cwg1315 3124 49. https://wg21.link/cwg2082 3125 50. https://wg21.link/cwg2351 3126 51. https://wg21.link/cwg2374 3127 52. https://wg21.link/cwg2397 3128 53. https://wg21.link/cwg2446 3129 54. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=87c2080b 3130 55. https://gcc.gnu.org/PR79501 3131 56. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3132 57. https://gcc.gnu.org/PR19808 3133 58. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wint-in-bool-context 3134 59. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a2b12bc 3135 60. https://gcc.gnu.org/PR99032 3136 61. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmissing-requires 3137 62. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=e18e56c7 3138 63. https://gcc.gnu.org/PR100995 3139 64. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Waddress 3140 65. https://gcc.gnu.org/PR102103 3141 66. https://gcc.gnu.org/PR99701 3142 67. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3ead06c1 3143 68. https://gcc.gnu.org/PR102753 3144 69. https://wg21.link/p0466 3145 70. https://gcc.gnu.org/PR101539 3146 71. https://gcc.gnu.org/PR100828 3147 72. https://gcc.gnu.org/PR101786 3148 73. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Winterference-size 3149 74. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=76b75018 3150 75. https://gcc.gnu.org/PR101344 3151 76. https://gcc.gnu.org/PR101883 3152 77. https://gcc.gnu.org/PR89062 3153 78. https://gcc.gnu.org/PR101233 3154 79. https://gcc.gnu.org/PR88252 3155 80. https://gcc.gnu.org/PR86439 3156 81. https://gcc.gnu.org/PR98832 3157 82. https://gcc.gnu.org/PR102933 3158 83. https://gcc.gnu.org/PR51577 3159 84. https://gcc.gnu.org/PR98486 3160 85. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=1595fe44 3161 86. https://gcc.gnu.org/gcc-12/changes.html#libstdcxx 3162 87. https://gcc.gnu.org/gcc-12/changes.html#d 3163 88. https://dlang.org/changelog/2.100.1.html 3164 89. https://gcc.gnu.org/PR101127 3165 90. https://gcc.gnu.org/PR106563 3166 91. https://dlang.org/spec/importc.html 3167 92. https://gcc.gnu.org/gcc-12/changes.html#fortran 3168 93. https://gcc.gnu.org/gcc-12/changes.html#go 3169 94. https://gcc.gnu.org/gcc-12/changes.html#jit 3170 95. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-16 3171 96. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_tls_model 3172 97. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-17 3173 98. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_link_section 3174 99. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Variable-Attributes.html#index-section-variable-attribute 3175 100. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-18 3176 101. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-19 3177 102. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-20 3178 103. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_bitcast 3179 104. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-21 3180 105. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_register_name 3181 106. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-22 3182 107. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/contexts.html#c.gcc_jit_context_set_bool_print_errors_to_stderr 3183 108. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-23 3184 109. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-24 3185 110. https://gcc.gnu.org/PR96066 3186 111. https://gcc.gnu.org/PR96067 3187 112. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_cast 3188 113. https://gcc.gnu.org/PR95498 3189 114. https://gcc.gnu.org/gcc-12/changes.html#targets 3190 115. https://gcc.gnu.org/gcc-12/changes.html#arm-targets 3191 116. https://gcc.gnu.org/gcc-12/changes.html#aarch64 3192 117. https://developer.arm.com/community/arm-community-blogs/b/architectures-and-processors-blog/posts/arm-a-profile-architecture-developments-2020 3193 118. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 3194 119. https://gcc.gnu.org/gcc-12/changes.html#arm 3195 120. https://gcc.gnu.org/gcc-12/changes.html#bpf 3196 121. https://gcc.gnu.org/gcc-12/changes.html#x86 3197 122. https://gcc.gnu.org/gcc-12/changes.html#loongarch 3198 123. https://gcc.gnu.org/gcc-12/changes.html#mips 3199 124. https://gcc.gnu.org/gcc-12/changes.html#nvptx 3200 125. https://gcc.gnu.org/gcc-12/changes.html#powerpc 3201 126. https://gcc.gnu.org/bugs/ 3202 127. https://gcc.gnu.org/gcc-12/changes.html#pru 3203 128. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Named-Address-Spaces.html#PRU-Named-Address-Spaces 3204 129. https://gcc.gnu.org/gcc-12/changes.html#riscv 3205 130. https://groups.google.com/a/groups.riscv.org/g/sw-dev/c/aE1ZeHHCYf4 3206 131. https://gcc.gnu.org/gcc-12/changes.html#os 3207 132. https://gcc.gnu.org/gcc-12/changes.html#analyzer 3208 133. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-use-of-uninitialized-value 3209 134. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3210 135. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 3211 136. https://gcc.gnu.org/PR95006 3212 137. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3213 138. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 3214 139. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 3215 140. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 3216 141. https://gcc.gnu.org/PR104793 3217 142. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 3218 143. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 3219 144. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 3220 145. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 3221 146. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 3222 147. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 3223 148. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 3224 149. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 3225 150. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-tainted_005fargs-function-attribute 3226 151. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-const-function-attribute 3227 152. https://gcc.gnu.org/PR104434 3228 153. https://gcc.gnu.org/PR104576 3229 154. https://gcc.gnu.org/PR101962 3230 155. https://gcc.gnu.org/PR101570 3231 156. https://gcc.gnu.org/PR95006 3232 157. https://gcc.gnu.org/PR104943 3233 158. https://gcc.gnu.org/PR104954 3234 159. https://gcc.gnu.org/PR104955 3235 160. https://gcc.gnu.org/gcc-12/changes.html#uninitialized 3236 161. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3237 162. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3238 163. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3239 164. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3240 165. https://gcc.gnu.org/gcc-12/changes.html#debug 3241 166. https://gcc.gnu.org/gcc-12/changes.html#12.1 3242 167. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.0 3243 168. https://gcc.gnu.org/gcc-12/changes.html#12.2 3244 169. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.2 3245 170. https://gcc.gnu.org/gcc-12/changes.html#12.3 3246 171. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.3 3247 172. https://gcc.gnu.org/gcc-12/changes.html#12.4 3248 173. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.4 3249 174. https://gcc.gnu.org/gcc-12/changes.html#12.5 3250 175. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.5 3251 176. https://gcc.gnu.org/onlinedocs/ 3252 177. mailto:gcc-help (a] gcc.gnu.org 3253 178. mailto:gcc (a] gcc.gnu.org 3254 179. https://gcc.gnu.org/lists.html 3255 180. https://www.fsf.org/ 3256 181. https://gcc.gnu.org/about.html 3257 ====================================================================== 3258 http://gcc.gnu.org/gcc-11/index.html 3259 GCC 11 Release Series 3260 3261 (This release series is no longer supported.) 3262 3263 July 19, 2024 3264 3265 The GCC developers are pleased to announce the release of GCC 11.5. 3266 3267 This release is a bug-fix release, containing fixes for regressions in 3268 GCC 11.4 relative to previous releases of GCC. 3269 3270 Release History 3271 3272 GCC 11.5 3273 July 19, 2024 ([1]changes, [2]documentation) 3274 3275 GCC 11.4 3276 May 29, 2023 ([3]changes, [4]documentation) 3277 3278 GCC 11.3 3279 April 21, 2022 ([5]changes, [6]documentation) 3280 3281 GCC 11.2 3282 July 28, 2021 ([7]changes, [8]documentation) 3283 3284 GCC 11.1 3285 April 27, 2021 ([9]changes, [10]documentation) 3286 3287 References and Acknowledgements 3288 3289 GCC used to stand for the GNU C Compiler, but since the compiler 3290 supports several other languages aside from C, it now stands for the 3291 GNU Compiler Collection. 3292 3293 The GCC developers would like to thank the numerous people that have 3294 contributed new features, improvements, bug fixes, and other changes as 3295 well as test results to GCC. This [11]amazing group of volunteers is 3296 what makes GCC successful. 3297 3298 For additional information about GCC please refer to the [12]GCC 3299 project web site or contact the [13]GCC development mailing list. 3300 3301 To obtain GCC please use [14]our mirror sites or [15]our version 3302 control system. 3303 3304 3305 For questions related to the use of GCC, please consult these web 3306 pages and the [16]GCC manuals. If that fails, the 3307 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3308 web pages and the development of GCC are welcome on our developer 3309 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 3310 archives. 3311 3312 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 3313 distribution of this entire article is permitted in any medium, 3314 provided this notice is preserved. 3315 3316 These pages are [21]maintained by the GCC team. Last modified 3317 2025-08-04. 3318 3319 References 3320 3321 1. https://gcc.gnu.org/gcc-11/changes.html 3322 2. https://gcc.gnu.org/onlinedocs/11.5.0/ 3323 3. https://gcc.gnu.org/gcc-11/changes.html 3324 4. https://gcc.gnu.org/onlinedocs/11.4.0/ 3325 5. https://gcc.gnu.org/gcc-11/changes.html 3326 6. https://gcc.gnu.org/onlinedocs/11.3.0/ 3327 7. https://gcc.gnu.org/gcc-11/changes.html 3328 8. https://gcc.gnu.org/onlinedocs/11.2.0/ 3329 9. https://gcc.gnu.org/gcc-11/changes.html 3330 10. https://gcc.gnu.org/onlinedocs/11.1.0/ 3331 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html 3332 12. https://gcc.gnu.org/index.html 3333 13. mailto:gcc (a] gcc.gnu.org 3334 14. https://gcc.gnu.org/mirrors.html 3335 15. https://gcc.gnu.org/git.html 3336 16. https://gcc.gnu.org/onlinedocs/ 3337 17. mailto:gcc-help (a] gcc.gnu.org 3338 18. mailto:gcc (a] gcc.gnu.org 3339 19. https://gcc.gnu.org/lists.html 3340 20. https://www.fsf.org/ 3341 21. https://gcc.gnu.org/about.html 3342 ====================================================================== 3343 http://gcc.gnu.org/gcc-11/changes.html 3344 GCC 11 Release Series 3345 Changes, New Features, and Fixes 3346 3347 This page is a "brief" summary of some of the huge number of 3348 improvements in GCC 11. You may also want to check out our [1]Porting 3349 to GCC 11 page and the [2]full GCC documentation. 3350 3351 Caveats 3352 3353 * The default mode for C++ is now -std=gnu++17 instead of 3354 -std=gnu++14. Note that [3]C++17 changes to template template 3355 parameter matching can be disabled independently of other features 3356 with -fno-new-ttp-matching. 3357 * When building GCC itself, the host compiler must now support C++11, 3358 rather than C++98. In particular bootstrapping GCC 11 using an 3359 older version of GCC requires a binary of GCC 4.8 or later, rather 3360 than of GCC 3.4 or later as was the case for bootstrapping GCC 10. 3361 * Naming and location of auxiliary and dump output files changed. If 3362 you compile multiple input files in a single command, if you enable 3363 Link Time Optimization, or if you use -dumpbase, -dumpdir, 3364 -save-temps=*, and you expect any file other than the primary 3365 output file(s) to be created as a side effect, watch out for 3366 improvements and a few surprises. See [4]the patch, particularly 3367 its textual description, for more details about the changes. 3368 * -gsplit-dwarf no longer enables debug info generation on its own 3369 but requires a separate -g for this. 3370 * The libstdc++ configure option --enable-cheaders=c_std is 3371 deprecated and will be removed in a future release. It should be 3372 possible to use --enable-cheaders=c_global (the default) with no 3373 change in behavior. 3374 * The front end for compiling BRIG format of Heterogeneous System 3375 Architecture Intermediate Language (HSAIL) has been deprecated and 3376 will likely be removed in a future release. 3377 * Some short options of the gcov tool have been renamed: -i to -j and 3378 -j to -H. 3379 3380 [5]General Improvements 3381 3382 * [6]ThreadSanitizer improvements to support alternative runtimes and 3383 environments. The [7]Linux Kernel Concurrency Sanitizer (KCSAN) is 3384 now supported. 3385 + Add --param tsan-distinguish-volatile to optionally emit 3386 instrumentation distinguishing volatile accesses. 3387 + Add --param tsan-instrument-func-entry-exit to optionally 3388 control if function entries and exits should be instrumented. 3389 * In previous releases of GCC, the "column numbers" emitted in 3390 diagnostics were actually a count of bytes from the start of the 3391 source line. This could be problematic, both because of: 3392 + multibyte characters (requiring more than one byte to encode), 3393 and 3394 + multicolumn characters (requiring more than one column to 3395 display in a monospace font) 3396 For example, the character ("GREEK SMALL LETTER PI (U+03C0)") 3397 occupies one column, and its UTF-8 encoding requires two bytes; the 3398 character ("SLIGHTLY SMILING FACE (U+1F642)") occupies two 3399 columns, and its UTF-8 encoding requires four bytes. 3400 In GCC 11 the column numbers default to being column numbers, 3401 respecting multi-column characters. The old behavior can be 3402 restored using a new option [8]-fdiagnostics-column-unit=byte. 3403 There is also a new option [9]-fdiagnostics-column-origin=, 3404 allowing the pre-existing default of the left-hand column being 3405 column 1 to be overridden if desired (e.g. for 0-based columns). 3406 The output of [10]-fdiagnostics-format=json has been extended to 3407 supply both byte counts and column numbers for all source 3408 locations. 3409 Additionally, in previous releases of GCC, tab characters in the 3410 source would be emitted verbatim when quoting source code, but be 3411 prefixed with whitespace or line number information, leading to 3412 misalignments in the resulting output when compared with the actual 3413 source. Tab characters are now printed as an appropriate number of 3414 spaces, using the [11]-ftabstop option (which defaults to 8 spaces 3415 per tab stop). 3416 * Introduce [12]Hardware-assisted AddressSanitizer support. This 3417 sanitizer currently only works for the AArch64 target. It helps 3418 debug address problems similarly to [13]AddressSanitizer but is 3419 based on partial hardware assistance and provides probabilistic 3420 protection to use less RAM at run time. [14]Hardware-assisted 3421 AddressSanitizer is not production-ready for user space, and is 3422 provided mainly for use compiling the Linux Kernel. 3423 To use this sanitizer the command line arguments are: 3424 + -fsanitize=hwaddress to instrument userspace code. 3425 + -fsanitize=kernel-hwaddress to instrument kernel code. 3426 * For targets that produce DWARF debugging information GCC now 3427 defaults to [15]DWARF version 5 (with the exception of VxWorks and 3428 Darwin/Mac OS X which default to version 2 and AIX which defaults 3429 to version 4). This can produce up to 25% more compact debug 3430 information compared to earlier versions. 3431 To take full advantage of DWARF version 5 GCC needs to be built 3432 against binutils version 2.35.2 or higher. When GCC is built 3433 against earlier versions of binutils GCC will still emit DWARF 3434 version 5 for most debuginfo data, but will generate version 4 3435 debug line tables (even when explicitly given -gdwarf-5). 3436 The following debug information consumers can process DWARF version 3437 5: 3438 + GDB 8.0, or higher 3439 + valgrind 3.17.0 3440 + elfutils 0.172, or higher (for use with systemtap, 3441 dwarves/pahole, perf and libabigail) 3442 + dwz 0.14 3443 Programs embedding libbacktrace are urged to upgrade to the version 3444 shipping with GCC 11. 3445 To make GCC 11 generate an older DWARF version use -g together with 3446 -gdwarf-2, -gdwarf-3 or -gdwarf-4. 3447 * Vectorizer improvements: 3448 + The straight-line code vectorizer now considers the whole 3449 function when vectorizing and can handle opportunities 3450 crossing CFG merges and backedges. 3451 * A series of conditional expressions that compare the same variable 3452 can be transformed into a switch statement if each of them contains 3453 a comparison expression. Example: 3454 int IsHTMLWhitespace(int aChar) { 3455 return aChar == 0x0009 || aChar == 0x000A || 3456 aChar == 0x000C || aChar == 0x000D || 3457 aChar == 0x0020; 3458 } 3459 3460 This statement can be transformed into a switch statement and then 3461 expanded into a bit-test. 3462 * New command-line options: 3463 + [16]-fbit-tests, enabled by default, can be used to enable or 3464 disable switch expansion using bit-tests. 3465 * Inter-procedural optimization improvements: 3466 + A new IPA-modref pass was added to track side effects of 3467 function calls and improve precision of points-to-analysis. 3468 The pass can be controlled by the [17]-fipa-modref option. 3469 + The identical code folding pass (controlled by [18]-fipa-icf) 3470 was significantly improved to increase the number of unified 3471 functions and to reduce compile-time memory use. 3472 + IPA-CP (Interprocedural constant propagation) heuristics 3473 improved its estimation of potential usefulness of known loop 3474 bounds and strides by taking the estimated frequency of these 3475 loops into account. 3476 * Link-time optimization improvements: 3477 + The LTO bytecode format was optimized for smaller object files 3478 and faster streaming. 3479 + Memory allocation of the linking stage was improved to reduce 3480 peak memory use. 3481 * Profile driven optimization improvements: 3482 + Using [19]-fprofile-values, was improved by tracking more 3483 target values for e.g. indirect calls. 3484 + GCOV data file format outputs smaller files by representing 3485 zero counters in a more compact way. 3486 3487 [20]New Languages and Language specific improvements 3488 3489 * GCC 11 adds support for non-rectangular loop nests in OpenMP 3490 constructs and the allocator routines of [21]OpenMP 5.0, including 3491 initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD 3492 environment variable and the active-levels routines are now 3493 supported. For C/C++, the declare variant and map support has been 3494 extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP 3495 5.0 support has been extended, including the following features 3496 which were before only available in C and C++: order(concurrent), 3497 device_type, memorder-clauses for flush, lastprivate with 3498 conditional modifier, atomic construct and reduction clause 3499 extensions of OpenMP 5.0, if clause with simd and cancel modifiers, 3500 target data without map clause, and limited support for the 3501 requires construct. 3502 * Version 2.6 of the [22]OpenACC specification continues to be 3503 maintained and improved in the C, C++ and Fortran compilers. See 3504 the [23]implementation status section on the OpenACC wiki page and 3505 the [24]run-time library documentation for further information. 3506 3507 [25]C family 3508 3509 * New attributes: 3510 + The [26]no_stack_protector attribute has been added to mark 3511 functions which should not be instrumented with stack 3512 protection (-fstack-protector). 3513 + The existing [27]malloc attribute has been extended so that it 3514 can be used to identify allocator/deallocator API pairs. A 3515 pair of new [28]-Wmismatched-dealloc and 3516 [29]-Wmismatched-new-delete warnings will complain about 3517 mismatched calls, and [30]-Wfree-nonheap-object about 3518 deallocation calls with pointers not obtained from allocation 3519 functions. Additionally, the static analyzer will use these 3520 attributes when checking for leaks, double-frees, 3521 use-after-frees, and similar issues. 3522 * New warnings: 3523 + [31]-Wmismatched-dealloc, enabled by default, warns about 3524 calls to deallocation functions with pointers returned from 3525 mismatched allocation functions. 3526 + [32]-Wsizeof-array-div, enabled by -Wall, warns about 3527 divisions of two sizeof operators when the first one is 3528 applied to an array and the divisor does not equal the size of 3529 the array element. 3530 + [33]-Wstringop-overread, enabled by default, warns about calls 3531 to string functions reading past the end of the arrays passed 3532 to them as arguments. In prior GCC releases most instances of 3533 his warning are diagnosed by -Wstringop-overflow. 3534 + [34]-Wtsan, enabled by default, warns about unsupported 3535 features in ThreadSanitizer (currently 3536 std::atomic_thread_fence). 3537 * Enhancements to existing warnings: 3538 + [35]-Wfree-nonheap-object detects many more instances of calls 3539 to deallocation functions with pointers not returned from a 3540 dynamic memory allocation function. 3541 + [36]-Wmaybe-uninitialized diagnoses passing pointers or 3542 references to uninitialized memory to functions taking 3543 const-qualified arguments. 3544 + [37]-Wuninitialized detects reads from uninitialized 3545 dynamically allocated memory. 3546 * For ELF targets that support the GNU or FreeBSD OSABIs, the used 3547 attribute will now save the symbol declaration it is applied to 3548 from linker garbage collection. 3549 To support this behavior, used symbols that have not been placed in 3550 specific sections (e.g. with the section attribute, or the 3551 -f{function,data}-sections options) will be placed in new, unique 3552 sections. 3553 This functionality requires Binutils version 2.36 or later. 3554 3555 [38]C 3556 3557 * Several new features from the upcoming C2X revision of the ISO C 3558 standard are supported with -std=c2x and -std=gnu2x. Some of these 3559 features are also supported as extensions when compiling for older 3560 language versions. In addition to the features listed, some 3561 features previously supported as extensions and now added to the C 3562 standard are enabled by default in C2X mode and not diagnosed with 3563 -std=c2x -Wpedantic. 3564 + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>. 3565 + As in C++, function definitions no longer need to give names 3566 for unused function parameters. 3567 + The expansions of the true and false macros in <stdbool.h> 3568 have changed so that they have type bool. 3569 + The [[nodiscard]] standard attribute is now supported. 3570 + The __has_c_attribute preprocessor operator is now supported. 3571 + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN, 3572 DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros 3573 for _FloatN, _FloatNx and _DecimalN types, are provided in 3574 <float.h>. There are also corresponding built-in functions 3575 __builtin_nansdN for decimal signaling NaNs. 3576 + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and 3577 LDBL_IS_IEC_60559 are provided in <float.h>. 3578 + The feature test macro __STDC_WANT_IEC_60559_EXT__ is 3579 supported by <float.h>. 3580 + Labels may appear before declarations and at the end of a 3581 compound statement. 3582 * New warnings: 3583 + [39]-Warray-parameter, enabled by -Wall, warns about 3584 redeclarations of functions with ordinary array arguments 3585 declared using inconsistent forms. The warning also enables 3586 the detection of the likely out of bounds accesses in calls to 3587 such functions with smaller arrays. 3588 + [40]-Wvla-parameter, enabled by -Wall, warns redeclarations of 3589 functions with variable length array arguments declared using 3590 inconsistent forms or with mismatched bounds. The warning also 3591 enables the detection of the likely out of bounds accesses in 3592 calls to such functions with smaller arrays. 3593 3594 [41]C++ 3595 3596 * The default mode has been changed to -std=gnu++17. 3597 * Several C++20 features have been implemented: 3598 + the compiler now supports consteval virtual functions 3599 + P2082R1, Fixing CTAD for aggregates 3600 + P0593R6, Pseudo-destructors end object lifetimes 3601 + P1907R1, Inconsistencies with non-type template parameters 3602 (complete implementation) 3603 + P1975R0, Fixing the wording of parenthesized 3604 aggregate-initialization 3605 + P1009R2, Array size deduction in new-expressions 3606 + P1099R5, using enum 3607 + Modules, Requires -fmodules-ts and some aspects are 3608 incomplete. Refer to [42]C++ 20 Status 3609 * The C++ front end has experimental support for some of the upcoming 3610 C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b 3611 or -std=gnu++2b flags, including 3612 + P0330R8, Literal Suffix for (signed) size_t. 3613 For a full list of new features, see [43]the C++ status page. 3614 * Several C++ Defect Reports have been resolved, e.g.: 3615 + DR 625, Use of auto as a template-argument 3616 + DR 1512, Pointer comparison vs qualification conversions 3617 + DR 1722, Should lambda to function pointer conversion function 3618 be noexcept? 3619 + DR 1914, Duplicate standard attributes 3620 + DR 2032, Default template-arguments of variable templates 3621 + DR 2289, Uniqueness of decomposition declaration names 3622 + DR 2237, Can a template-id name a constructor? 3623 + DR 2303, Partial ordering and recursive variadic inheritance 3624 + DR 2369, Ordering between constraints and substitution 3625 + DR 2450, braced-init-list as a template-argument 3626 * G++ now performs better access checking in templates ([44]PR41437). 3627 * reinterpret_casts in constexpr evaluation are now checked more 3628 completely ([45]PR95307). 3629 * The diagnostic for static_assert has been improved: the compiler 3630 now shows the expression including its template arguments (if there 3631 were any), and can point to the failing clause if the condition 3632 comprised of any logical AND operators ([46]PR97518). 3633 * New warnings: 3634 + [47]-Wctad-maybe-unsupported, disabled by default, warns about 3635 performing class template argument deduction on a type with no 3636 deduction guides. 3637 + [48]-Wrange-loop-construct, enabled by -Wall, warns when a 3638 range-based for-loop is creating unnecessary and expensive 3639 copies. 3640 + [49]-Wdeprecated-enum-enum-conversion, enabled by default in 3641 C++20, warns about deprecated arithmetic conversions on 3642 operands of enumeration types, as outlined in 3643 [depr.arith.conv.enum]. 3644 + [50]-Wdeprecated-enum-float-conversion, enabled by default in 3645 C++20, warns about deprecated arithmetic conversions on 3646 operands where one is of enumeration type and the other is of 3647 a floating-point type, as outlined in [depr.arith.conv.enum]. 3648 + [51]-Wmismatched-new-delete, enabled by -Wall, warns about 3649 calls to C++ operator delete with pointers returned from 3650 mismatched forms of operator new or from other mismatched 3651 allocation functions. 3652 + [52]-Wvexing-parse, enabled by default, warns about the most 3653 vexing parse rule: the cases when a declaration looks like a 3654 variable definition, but the C++ language requires it to be 3655 interpreted as a function declaration. 3656 * Enhancements to existing warnings: 3657 + [53]-Wnonnull considers the implicit this argument of every 3658 C++ nonstatic member function to have been implicitly declared 3659 with attribute nonnull and triggers warnings for calls where 3660 the pointer is null. 3661 3662 [54]Runtime Library (libstdc++) 3663 3664 * Improved C++17 support, including: 3665 + std::from_chars and std::to_chars for floating-point types. 3666 * Improved experimental C++20 support, including: 3667 + Calendar additions to <chrono>. Thanks to Cassio Neri for 3668 optimizations. 3669 + std::bit_cast 3670 + std::source_location 3671 + Atomic wait and notify operations. 3672 + <barrier>, <latch>, and <semaphore> 3673 + <syncstream> 3674 + Efficient access to basic_stringbuf's buffer. 3675 + Heterogeneous lookup in unordered containers. 3676 * Experimental C++23 support, including: 3677 + contains member functions for strings, thanks to Paul Fee. 3678 + std::to_underlying, std::is_scoped_enum 3679 * Experimental support for Data-Parallel Types (simd) from the 3680 Parallelism 2 TS, thanks to Matthias Kretz. 3681 * Faster std::uniform_int_distribution, thanks to Daniel Lemire. 3682 3683 [55]D 3684 3685 * New features: 3686 + A new bottom type typeof(*null) has been added to represent 3687 run-time errors and non-terminating functions. This also 3688 introduces a new standard alias for the type named noreturn, 3689 and is implicitly imported into every module. 3690 + Printf-like and scanf-like functions are now detected by 3691 prefixing them with pragma(printf) for printf-like functions 3692 or pragma(scanf) for scanf-like functions. 3693 + The __traits() expression now supports the extensions 3694 isDeprecated, isDisabled, isFuture, isModule, isPackage, 3695 child, isReturnOnStack, isZeroInit, getTargetInfo, 3696 getLocation, hasPostblit, isCopyable, getVisibility, and 3697 totype. 3698 + An expression-based contract syntax has been added to the 3699 language. 3700 + Function literals can now return a value by reference with the 3701 ref keyword. 3702 + A new syntax is available to declare aliases to function types 3703 using the alias syntax based on the assignment operator. 3704 + New types __c_complex_float, __c_complex_double, 3705 __c_complex_real, and __c_wchar_t have been added for 3706 interfacing with C and C++ code, and are available from the 3707 core.stdc.config module. 3708 + User-defined attributes can now be used to annotate enum 3709 members, alias declarations, and function parameters. 3710 + Template alias parameters can now be instantiated with basic 3711 types such as int or void function(). 3712 + The mixin construct can now be used as types in the form 3713 mixin(string) var. 3714 + The mixin construct can now take an argument list, same as 3715 pragma(msg). 3716 * New intrinsics: 3717 + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror 3718 have been added. 3719 + Byte swap intrinsic core.bitop.byteswap for swapping bytes in 3720 a 2-byte ushort has been added. 3721 + Math intrinsics available from core.math now have overloads 3722 for float and double types. 3723 + Volatile intrinsics core.volatile.volatileLoad and 3724 core.volatile.volatileStore have been moved from the 3725 core.bitop module. 3726 * New attributes: 3727 + The following GCC attributes are now recognized and available 3728 from the gcc.attributes module with short-hand aliases for 3729 convenience: 3730 o @attribute("alloc_size", arguments) or 3731 @alloc_size(arguments). 3732 o @attribute("always_inline") or @always_inline. 3733 o @attribute("used") or @used. 3734 o @attribute("optimize", arguments) or 3735 @optimize(arguments). 3736 o @attribute("cold") or @cold. 3737 o @attribute("noplt") or @noplt. 3738 o @attribute("target_clones", arguments) or 3739 @target_clones(arguments). 3740 o @attribute("no_icf") or @no_icf. 3741 o @attribute("noipa") or @noipa. 3742 o @attribute("symver", arguments) or @symver(arguments). 3743 + New aliases have been added to gcc.attributes for 3744 compatibility with ldc.attributes. 3745 o The @allocSize(arguments) attribute is the same as 3746 @alloc_size(arguments), but uses a 0-based index for 3747 function arguments. 3748 o The @assumeUsed attribute is an alias for 3749 @attribute("used"). 3750 o The @fastmath attribute is an alias for 3751 @optimize("Ofast"). 3752 o The @naked attribute is an alias for @attribute("naked"). 3753 This attribute may not be available on all targets. 3754 o The @restrict attribute has been added to specify that a 3755 function parameter is to be restrict-qualified in the C99 3756 sense of the term. 3757 o The @optStrategy(strategy) attribute is an alias for 3758 @optimize("O0") when the strategy is "none", otherwise 3759 @optimize("Os") for the "optsize" and "minsize" 3760 strategies. 3761 o The @polly attribute is an alias for 3762 @optimize("loop-parallelize-all"). 3763 o The @section(name) attribute is an alias for 3764 @attribute("section", name). 3765 o The @target(arguments) attribute is an alias for 3766 attribute("target", arguments). 3767 o The @weak attribute is an alias for @attribute("weak"). 3768 * New language options: 3769 + -fweak-templates, added to control whether declarations that 3770 can be defined in multiple objects should be emitted as weak 3771 symbols. The default is to emit all symbols with extern 3772 linkage as weak, unless the target lacks support for weak 3773 symbols. 3774 + -Wdeprecated, this option is now enabled by default. 3775 + -Wextra, this option now turns on all warnings that are not 3776 part of the core D language front-end - -Waddress, 3777 -Wcast-result, -Wunknown-pragmas. 3778 + -Wvarargs, added to turn on warnings about questionable usage 3779 of the va_start intrinsic. 3780 * Deprecated and removed features: 3781 + Compiler-recognized attributes are now made available from the 3782 gcc.attributes module, the former module gcc.attribute has 3783 been deprecated and will be removed in a future release. 3784 + The @attribute("alias") attribute has been removed, as it had 3785 been superseded by pragma(mangle). 3786 + The @attribute("forceinline") attribute has been removed and 3787 renamed to @always_inline. 3788 + __vector types that are not supported in hardware are now 3789 rejected at compile-time. Previously all vector types were 3790 accepted by the compiler and emulated when target support was 3791 absent. 3792 + The extern(Pascal) linkage attribute has been removed. 3793 + The deprecation phase for -ftransition=import and 3794 -ftransition=checkimports is finished. These switches no 3795 longer have an effect and are now removed. Symbols that are 3796 not visible in a particular scope will no longer be found by 3797 the compiler. 3798 + It is now an error to use private variables selectively 3799 imported from other modules. Due to a bug, some imported 3800 private members were visible from other modules, violating the 3801 specification. 3802 + The -fweak compiler switch has been removed, as it existed 3803 only for testing. 3804 3805 [56]Fortran 3806 3807 * Added DEPRECATED to !GCC$'s attributes directive. 3808 3809 [57]Go 3810 3811 * GCC 11 provides a complete implementation of the Go 1.16.3 user 3812 packages. 3813 3814 [58]libgccjit 3815 3816 * libgccjit was marked as merely "Alpha" quality when [59]originally 3817 added in GCC 5. Given that we have maintained [60]API and ABI 3818 compatibility since then and it is in use by various projects, we 3819 have removed that caveat. 3820 * libgccjit can now be built for MinGW 3821 * The libgccjit API gained 10 new entry points: 3822 + [61]gcc_jit_global_set_initializer 3823 + 9 entrypoints for [62]directly embedding asm statements into a 3824 compile, analogous to inline asm in the C front end 3825 3826 [63]New Targets and Target Specific Improvements 3827 3828 [64]AArch64 & arm 3829 3830 * A number of new CPUs are supported through arguments to the -mcpu 3831 and -mtune options in both the arm and aarch64 back ends (GCC 3832 identifiers in parentheses): 3833 + Arm Cortex-A78 (cortex-a78). 3834 + Arm Cortex-A78AE (cortex-a78ae). 3835 + Arm Cortex-A78C (cortex-a78c). 3836 + Arm Cortex-X1 (cortex-x1). 3837 + Arm Neoverse V1 (neoverse-v1). 3838 + Arm Neoverse N2 (neoverse-n2). 3839 * GCC can now auto-vectorize operations performing addition, 3840 subtraction, multiplication and the accumulate/subtract variants on 3841 complex numbers, taking advantage of the Advanced SIMD instructions 3842 in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64) 3843 and MVE (AArch32 M-profile) instruction sets. 3844 3845 [65]AArch64 3846 3847 * In addition to the above, the following AArch64-only CPUs are now 3848 supported: 3849 + Fujitsu A64FX (a64fx). 3850 + Arm Cortex-R82 (cortex-r82). 3851 * The AArch64 Armv8-R architecture is now supported through the 3852 -march=armv8-r option. 3853 * Mitigation against the [66]Straight-line Speculation vulnerability 3854 is supported with the -mharden-sls= option. Please refer to the 3855 documentation for usage instructions. 3856 * The availability of Advanced SIMD intrinsics available through the 3857 arm_neon.h header is improved and GCC 11 supports the full set of 3858 intrinsics defined by ACLE Q3 2020. 3859 3860 [67]AMD Radeon (GCN) 3861 3862 * Initial support for gfx908 GPUs has been added. 3863 3864 [68]arm 3865 3866 * Initial auto-vectorization is now available when targeting the MVE 3867 instruction set. 3868 * GCC can now make use of the Low Overhead Branch instruction in 3869 Armv8.1-M to optimize loop counters and checks. 3870 * The -mcpu=cortex-m55 option now supports the extensions +nomve and 3871 +nomve.fp to control generation of MVE and MVE floating-point 3872 instructions. 3873 3874 [69]IA-32/x86-64 3875 3876 * New ISA extension support for Intel TSXLDTRK was added to GCC. 3877 TSXLDTRK intrinsics are available via the -mtsxldtrk compiler 3878 switch. 3879 * New ISA extension support for Intel SERIALIZE was added to GCC. 3880 SERIALIZE intrinsics are available via the -mserialize compiler 3881 switch. 3882 * New ISA extension support for Intel HRESET was added to GCC. HRESET 3883 intrinsics are available via the -mhreset compiler switch. 3884 * New ISA extension support for Intel UINTR was added to GCC. UINTR 3885 intrinsics are available via the -muintr compiler switch. 3886 * New ISA extension support for Intel KEYLOCKER was added to GCC. 3887 KEYLOCKER intrinsics are available via the -mkeylocker compiler 3888 switch. 3889 * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16 3890 was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are 3891 available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler 3892 switches. 3893 * New ISA extension support for Intel AVX-VNNI was added to GCC. 3894 AVX-VNNI intrinsics are available via the -mavxvnni compiler 3895 switch. 3896 * GCC now supports the Intel CPU named Sapphire Rapids through 3897 -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B, 3898 AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG, 3899 TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA 3900 extensions. 3901 * GCC now supports the Intel CPU named Alderlake through 3902 -march=alderlake. The switch enables the CLDEMOTE, PTWRITE, 3903 WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions. 3904 * GCC now supports the Intel CPU named Rocketlake through 3905 -march=rocketlake. Rocket Lake is based on Icelake client and minus 3906 SGX. 3907 * GCC now supports AMD CPUs based on the znver3 core via 3908 -march=znver3. 3909 * GCC now supports micro-architecture levels defined in the x86-64 3910 psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4. 3911 3912 [70]Nios II 3913 3914 * The options -mcustom-insn=N no longer produce compiler warnings if 3915 the custom instruction is not generated due to missing optimization 3916 options such as -fno-math-errno, -ffinite-math-only, or 3917 -funsafe-math-optimizations. These warnings were not consistently 3918 emitted for all custom instructions. 3919 * The -mcustom-fpu-cfg=fph2 option has been added to enable the 3920 custom instructions supported by the Nios II Floating Point 3921 Hardware 2 Component. 3922 3923 [71]NVPTX 3924 3925 * The -misa default has changed from sm_30 to sm_35. 3926 * The -m32 compiler switch has been removed. 3927 * The -msoft-stack-reserve-local format has been fixed. Previously, 3928 it accepted -msoft-stack-reserve-local<n>. It now accepts 3929 -msoft-stack-reserve-local=<n>. 3930 3931 [72]S/390, System z, IBM Z Systems 3932 3933 * The behavior when compiling with -fexcess-precision=standard (e.g., 3934 implied by -std=c99) on s390(x) targets can now be controlled at 3935 configure time with the flag --enable-s390-excess-float-precision. 3936 When enabled, GCC will maintain previous behavior and evaluate 3937 float expressions in double precision, which aligns with the 3938 definition of float_t as double. With the flag disabled, GCC will 3939 always evaluate float expressions in single precision. In native 3940 builds and cross compiles that have target libc headers, GCC will 3941 by default match the definition of float_t in the installed glibc. 3942 3943 [73]RISC-V 3944 3945 * Support address sanitizer for RISC-V. 3946 * Support big-endian for RISC-V, thanks to Marcus Comstedt. 3947 * Implement new style of architecture extension test macros: each 3948 architecture extension has a corresponding feature test macro, 3949 which can be used to test its existence and version information. 3950 * Legacy architecture extension test macros like __riscv_atomic are 3951 deprecated, but will still be supported for at least 2 release 3952 cycles. 3953 * Support IFUNC for riscv*-*-linux*. 3954 * Add new option -misa-spec=* to control ISA spec version. This 3955 controls the default version of each extensions. It defaults to 3956 2.2. 3957 * Introduce the --with-multilib-generator configure time option. This 3958 allows for flexible config multi-lib settings. Its syntax is the 3959 same as RISC-V's multilib-generator. 3960 * Extend the sytax for multilib-generator, support expansion operator 3961 * to reduce the complexity of complicated multi-lib re-use rules. 3962 * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the 3963 pipeline model and architecture extensions, like -mtune=* plus 3964 -march=*. 3965 * Support for TLS stack protector canary access, thanks to Cooper Qu. 3966 * Support __builtin_thread_pointer for RISC-V. 3967 * Introduce shorten_memrefs optimization, which reduces the code size 3968 for memory access, thanks to Craig Blackmore. 3969 3970 [74]Operating Systems 3971 3972 [75]AIX 3973 3974 * GCC for AIX can be built as a 64-bit application and the runtime is 3975 built as FAT libraries containing both 32-bit and 64-bit objects. 3976 * Support AIX Vector Extended ABI with -mabi=vec-extabi. 3977 * Thread-Local uninitiated data placed in local common section. 3978 * Use thread-safe access in ctype. 3979 * Link with libc128.a when long-double-128 enabled. 3980 3981 [76]Improvements to Static Analyzer 3982 3983 * The implementation of how program state is tracked within 3984 [77]-fanalyzer has been completely rewritten for GCC 11, fixing 3985 numerous bugs, and allowing for the analyzer to scale to larger C 3986 source files. 3987 * The analysis of allocations and deallocations has been generalized 3988 beyond malloc and free. 3989 + As preliminary work towards eventually supporting C++, the 3990 malloc/free checking will also check new/delete and 3991 new[]/delete[]. However, C++ is not yet properly supported by 3992 [78]-fanalyzer (for example, exception-handling is 3993 unimplemented). 3994 + As noted above, the existing [79]malloc attribute has been 3995 extended so that it can be used to identify 3996 allocator/deallocator API pairs. The analyzer will use these 3997 attributes when checking for leaks, double-frees, 3998 use-after-frees, and similar issues. 3999 + A new [80]-Wanalyzer-mismatching-deallocation warning has been 4000 added, covering such mismatches as using scalar delete rather 4001 vector delete[]. 4002 * The analyzer has gained warnings 4003 [81]-Wanalyzer-shift-count-negative, 4004 [82]-Wanalyzer-shift-count-overflow, [83]-Wanalyzer-write-to-const, 4005 and [84]-Wanalyzer-write-to-string-literal, all enabled by default 4006 when [85]-fanalyzer is enabled. 4007 * The analyzer can now be extended by GCC plugins, allowing for 4008 domain-specific path-sensitive warnings. An example of using a 4009 [86]GCC plugin to check for misuses of CPython's global interpreter 4010 lock can be seen in the test suite 4011 * The analyzer has gained new debugging options 4012 [87]-fdump-analyzer-json and [88]-fno-analyzer-feasibility. 4013 4014 Other significant improvements 4015 4016 * GCC has gained a new environment variable 4017 [89]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to 4018 request machine-readable fix-it hints without needing to adjust 4019 build flags. 4020 4021 [90]GCC 11.1 4022 4023 This is the [91]list of problem reports (PRs) from GCC's bug tracking 4024 system that are known to be fixed in the 11.1 release. This list might 4025 not be complete (that is, it is possible that some PRs that have been 4026 fixed are not listed here). 4027 4028 [92]GCC 11.2 4029 4030 This is the [93]list of problem reports (PRs) from GCC's bug tracking 4031 system that are known to be fixed in the 11.2 release. This list might 4032 not be complete (that is, it is possible that some PRs that have been 4033 fixed are not listed here). 4034 4035 [94]GCC 11.3 4036 4037 This is the [95]list of problem reports (PRs) from GCC's bug tracking 4038 system that are known to be fixed in the 11.3 release. This list might 4039 not be complete (that is, it is possible that some PRs that have been 4040 fixed are not listed here). 4041 4042 [96]GCC 11.4 4043 4044 This is the [97]list of problem reports (PRs) from GCC's bug tracking 4045 system that are known to be fixed in the 11.4 release. This list might 4046 not be complete (that is, it is possible that some PRs that have been 4047 fixed are not listed here). 4048 4049 Target Specific Changes 4050 4051 x86-64 4052 4053 * The x86-64 ABI of passing and returning structure with a 64-bit 4054 single precision vector changed in GCC 11.1 when 3DNOW is disabled. 4055 Disabling 3DNOW no longer changes how they are passed nor returned. 4056 This ABI change is now diagnosed with -Wpsabi. 4057 * Mitigation against straight line speculation (SLS) for function 4058 return and indirect jump is supported via 4059 -mharden-sls=[none|all|return|indirect-jmp]. 4060 * Add CS prefix to call and jmp to indirect thunk with branch target 4061 in r8-r15 registers via -mindirect-branch-cs-prefix. 4062 4063 [98]GCC 11.5 4064 4065 This is the [99]list of problem reports (PRs) from GCC's bug tracking 4066 system that are known to be fixed in the 11.5 release. This list might 4067 not be complete (that is, it is possible that some PRs that have been 4068 fixed are not listed here). 4069 4070 Caveats 4071 4072 aarch64 4073 4074 * Due to a bug introduced a few weeks before the 11.5.0 release if 4075 the compiler is configured without explicit --with-arch=, 4076 --with=cpu= and/or --with-tune= configure options the compiler 4077 without explicit -march= etc. options might act as if asked for 4078 cortex-a34. This can be fixed by applying manually the 4079 [100]r12-8060 commit patch on top of GCC 11.5.0. See [101]PR116029 4080 for more details. GCC 11.4.0 or earlier releases are not affected. 4081 4082 4083 For questions related to the use of GCC, please consult these web 4084 pages and the [102]GCC manuals. If that fails, the 4085 [103]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4086 web pages and the development of GCC are welcome on our developer 4087 list at [104]gcc (a] gcc.gnu.org. All of [105]our lists have public 4088 archives. 4089 4090 Copyright (C) [106]Free Software Foundation, Inc. Verbatim copying and 4091 distribution of this entire article is permitted in any medium, 4092 provided this notice is preserved. 4093 4094 These pages are [107]maintained by the GCC team. Last modified 4095 2026-06-07. 4096 4097 References 4098 4099 1. https://gcc.gnu.org/gcc-11/porting_to.html 4100 2. https://gcc.gnu.org/onlinedocs/index.html#current 4101 3. https://wg21.link/p0522r0 4102 4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html 4103 5. https://gcc.gnu.org/gcc-11/changes.html#general 4104 6. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual 4105 7. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html 4106 8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit 4107 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin 4108 10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 4109 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop 4110 12. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 4111 13. https://github.com/google/sanitizers/wiki/AddressSanitizer 4112 14. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 4113 15. https://dwarfstd.org/doc/DWARF5.pdf 4114 16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests 4115 17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref 4116 18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf 4117 19. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values 4118 20. https://gcc.gnu.org/gcc-11/changes.html#languages 4119 21. https://www.openmp.org/specifications/ 4120 22. https://www.openacc.org/ 4121 23. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-11 4122 24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html 4123 25. https://gcc.gnu.org/gcc-11/changes.html#c-family 4124 26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute 4125 27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 4126 28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 4127 29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 4128 30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 4129 31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 4130 32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div 4131 33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread 4132 34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan 4133 35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 4134 36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 4135 37. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized 4136 38. https://gcc.gnu.org/gcc-11/changes.html#c 4137 39. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter 4138 40. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter 4139 41. https://gcc.gnu.org/gcc-11/changes.html#cxx 4140 42. https://gcc.gnu.org/projects/cxx-status.html#cxx20 4141 43. https://gcc.gnu.org/projects/cxx-status.html#cxx23 4142 44. https://gcc.gnu.org/PR41437 4143 45. https://gcc.gnu.org/PR95307 4144 46. https://gcc.gnu.org/PR97518 4145 47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported 4146 48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct 4147 49. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion 4148 50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion 4149 51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 4150 52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse 4151 53. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull 4152 54. https://gcc.gnu.org/gcc-11/changes.html#libstdcxx 4153 55. https://gcc.gnu.org/gcc-11/changes.html#d 4154 56. https://gcc.gnu.org/gcc-11/changes.html#fortran 4155 57. https://gcc.gnu.org/gcc-11/changes.html#go 4156 58. https://gcc.gnu.org/gcc-11/changes.html#jit 4157 59. https://gcc.gnu.org/gcc-5/changes.html#jit 4158 60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html 4159 61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer 4160 62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html 4161 63. https://gcc.gnu.org/gcc-11/changes.html#targets 4162 64. https://gcc.gnu.org/gcc-11/changes.html#arm-targets 4163 65. https://gcc.gnu.org/gcc-11/changes.html#aarch64 4164 66. https://developer.arm.com/documentation/110280/latest 4165 67. https://gcc.gnu.org/gcc-11/changes.html#amdgcn 4166 68. https://gcc.gnu.org/gcc-11/changes.html#arm 4167 69. https://gcc.gnu.org/gcc-11/changes.html#x86 4168 70. https://gcc.gnu.org/gcc-11/changes.html#nios2 4169 71. https://gcc.gnu.org/gcc-11/changes.html#nvptx 4170 72. https://gcc.gnu.org/gcc-11/changes.html#s390 4171 73. https://gcc.gnu.org/gcc-11/changes.html#riscv 4172 74. https://gcc.gnu.org/gcc-11/changes.html#os 4173 75. https://gcc.gnu.org/gcc-11/changes.html#aix 4174 76. https://gcc.gnu.org/gcc-11/changes.html#analyzer 4175 77. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4176 78. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4177 79. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 4178 80. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation 4179 81. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative 4180 82. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow 4181 83. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 4182 84. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 4183 85. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4184 86. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825 4185 87. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json 4186 88. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility 4187 89. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT 4188 90. https://gcc.gnu.org/gcc-11/changes.html#11.1 4189 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0 4190 92. https://gcc.gnu.org/gcc-11/changes.html#11.2 4191 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2 4192 94. https://gcc.gnu.org/gcc-11/changes.html#11.3 4193 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.3 4194 96. https://gcc.gnu.org/gcc-11/changes.html#11.4 4195 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.4 4196 98. https://gcc.gnu.org/gcc-11/changes.html#11.5 4197 99. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.5 4198 100. https://gcc.gnu.org/r12-8060 4199 101. https://gcc.gnu.org/PR116029 4200 102. https://gcc.gnu.org/onlinedocs/ 4201 103. mailto:gcc-help (a] gcc.gnu.org 4202 104. mailto:gcc (a] gcc.gnu.org 4203 105. https://gcc.gnu.org/lists.html 4204 106. https://www.fsf.org/ 4205 107. https://gcc.gnu.org/about.html 4206 ====================================================================== 4207 http://gcc.gnu.org/gcc-10/index.html 4208 GCC 10 Release Series 4209 4210 (This release series is no longer supported.) 4211 4212 July, 7, 2023 4213 4214 The [1]GNU project and the GCC developers are pleased to announce the 4215 release of GCC 10.5. 4216 4217 This release is a bug-fix release, containing fixes for regressions in 4218 GCC 10.4 relative to previous releases of GCC. 4219 4220 Release History 4221 4222 GCC 10.5 4223 July 7, 2023 ([2]changes, [3]documentation) 4224 4225 GCC 10.4 4226 June 28, 2022 ([4]changes, [5]documentation) 4227 4228 GCC 10.3 4229 April 8, 2021 ([6]changes, [7]documentation) 4230 4231 GCC 10.2 4232 July 23, 2020 ([8]changes, [9]documentation) 4233 4234 GCC 10.1 4235 May 7, 2020 ([10]changes, [11]documentation) 4236 4237 References and Acknowledgements 4238 4239 GCC used to stand for the GNU C Compiler, but since the compiler 4240 supports several other languages aside from C, it now stands for the 4241 GNU Compiler Collection. 4242 4243 The GCC developers would like to thank the numerous people that have 4244 contributed new features, improvements, bug fixes, and other changes as 4245 well as test results to GCC. This [12]amazing group of volunteers is 4246 what makes GCC successful. 4247 4248 For additional information about GCC please refer to the [13]GCC 4249 project web site or contact the [14]GCC development mailing list. 4250 4251 To obtain GCC please use [15]our mirror sites or [16]our version 4252 control system. 4253 4254 4255 For questions related to the use of GCC, please consult these web 4256 pages and the [17]GCC manuals. If that fails, the 4257 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4258 web pages and the development of GCC are welcome on our developer 4259 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 4260 archives. 4261 4262 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 4263 distribution of this entire article is permitted in any medium, 4264 provided this notice is preserved. 4265 4266 These pages are [22]maintained by the GCC team. Last modified 4267 2024-05-30. 4268 4269 References 4270 4271 1. http://www.gnu.org/ 4272 2. https://gcc.gnu.org/gcc-10/changes.html 4273 3. https://gcc.gnu.org/onlinedocs/10.5.0/ 4274 4. https://gcc.gnu.org/gcc-10/changes.html 4275 5. https://gcc.gnu.org/onlinedocs/10.4.0/ 4276 6. https://gcc.gnu.org/gcc-10/changes.html 4277 7. https://gcc.gnu.org/onlinedocs/10.3.0/ 4278 8. https://gcc.gnu.org/gcc-10/changes.html 4279 9. https://gcc.gnu.org/onlinedocs/10.2.0/ 4280 10. https://gcc.gnu.org/gcc-10/changes.html 4281 11. https://gcc.gnu.org/onlinedocs/10.1.0/ 4282 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html 4283 13. https://gcc.gnu.org/index.html 4284 14. mailto:gcc (a] gcc.gnu.org 4285 15. https://gcc.gnu.org/mirrors.html 4286 16. https://gcc.gnu.org/git.html 4287 17. https://gcc.gnu.org/onlinedocs/ 4288 18. mailto:gcc-help (a] gcc.gnu.org 4289 19. mailto:gcc (a] gcc.gnu.org 4290 20. https://gcc.gnu.org/lists.html 4291 21. https://www.fsf.org/ 4292 22. https://gcc.gnu.org/about.html 4293 ====================================================================== 4294 http://gcc.gnu.org/gcc-10/changes.html 4295 GCC 10 Release Series 4296 Changes, New Features, and Fixes 4297 4298 This page is a "brief" summary of some of the huge number of 4299 improvements in GCC 10. You may also want to check out our [1]Porting 4300 to GCC 10 page and the [2]full GCC documentation. 4301 4302 Caveats 4303 4304 * An ABI incompatibility between C++14 and C++17 has been fixed. On 4305 some targets a class with a zero-sized subobject would be passed 4306 incorrectly when compiled as C++17 or C++20. See the [3]C++ notes 4307 below for more details. 4308 * The deprecated Profile Mode and array_allocator extensions have 4309 been removed from libstdc++. 4310 * The non-standard std::__is_nullptr_t type trait is deprecated and 4311 will be removed from libstdc++ in a future release. The standard 4312 trait std::is_null_pointer should be instead. 4313 * The minimum version of the [4]MPFR library required for building 4314 GCC has been increased to version 3.1.0 (released 2011-10-03). 4315 * The automatic template instantiation at link time (-frepo) has been 4316 removed. 4317 * The --param allow-store-data-races internal parameter has been 4318 removed in favor of a new official option -fallow-store-data-races. 4319 While default behavior is unchanged and the new option allows to 4320 correctly maintain a per compilation unit setting across link-time 4321 optimization, alteration of the default via --param 4322 allow-store-data-races will now be diagnosed and build systems have 4323 to be adjusted accordingly. 4324 * Offloading to Heterogeneous System Architecture Intermediate 4325 Language (HSAIL) has been deprecated and will likely be removed in 4326 a future release. 4327 * The type of the std::iterator base class of 4328 std::istreambuf_iterator was changed in C++98 mode to be consistent 4329 with C++11 and later standards. See the [5]libstdc++ notes below 4330 for more details. 4331 * GCC 10.5 does not bootstrap with a C++98 compiler; if you need to 4332 start from C++98, you should build 10.4 or 9.5 instead. 4333 4334 [6]General Improvements 4335 4336 * New built-in functions: 4337 + The [7]__has_builtin built-in preprocessor operator can be 4338 used to query support for built-in functions provided by GCC 4339 and other compilers that support it. 4340 + __builtin_roundeven for the corresponding function from 4341 ISO/IEC TS 18661. 4342 * New command-line options: 4343 + [8]-fallocation-dce removes unneeded pairs of new and delete 4344 operators. 4345 + [9]-fprofile-partial-training can now be used to inform the 4346 compiler that code paths not covered by the training run 4347 should not be optimized for size. 4348 + [10]-fprofile-reproducible controls level of reproducibility 4349 of profile gathered by [11]-fprofile-generate. This makes it 4350 possible to rebuild program with same outcome which is useful, 4351 for example, for distribution packages. 4352 + [12]-fprofile-prefix-path can be used in combination with 4353 -fprofile-generate=profile_dir and -fprofile-use=profile_dir 4354 to inform GCC where the base directory of build source tree is 4355 in case it differs between instrumentation and optimized 4356 builds. 4357 + [13]-fanalyzer enables a new static analysis pass and 4358 associated warnings. This pass performs a time-consuming 4359 exploration of paths through the code in the hope of detecting 4360 various common errors, such as double-free bugs. This option 4361 should be regarded as experimental in this release. In 4362 particular, analysis of non-C code is unlikely to work. 4363 * Inter-procedural optimization improvements: 4364 + The inter-procedural scalar replacement of aggregates 4365 (IPA-SRA) pass was re-implemented to work at link-time and can 4366 now also remove computing and returning unused return values. 4367 + [14]-finline-functions is now enabled at -O2 and was retuned 4368 for better code size versus runtime performance trade-offs. 4369 Inliner heuristics was also significantly sped up to avoid 4370 negative impact to -flto -O2 compile times. 4371 + Inliner heuristics and function cloning can now use 4372 value-range information to predict effectivity of individual 4373 transformations. 4374 + During link-time optimization the C++ One Definition Rule is 4375 used to increase precision of type based alias analysis. 4376 * Link-time optimization improvements: 4377 + A new binary [15]lto-dump has been added. It dumps various 4378 information about LTO bytecode object files. 4379 + The parallel phase of the LTO can automatically detect a 4380 running make's jobserver or fall back to number of available 4381 cores. 4382 + The LTO bytecode can be compressed with the [16]zstd 4383 algorithm. The configure script automatically detects zstd 4384 support. 4385 + Most --param values can now be specified at translation unit 4386 granularity. This includes all parameters controlling the 4387 inliner and other inter-procedural optimizations. Unlike 4388 earlier releases, GCC 10 will ignore parameters controlling 4389 optimizations specified at link-time and apply parameters 4390 specified at compile-time in the same manner as done for 4391 optimization flags. 4392 * Profile driven optimization improvements: 4393 + Profile maintenance during compilation and hot/cold code 4394 partitioning have been improved. 4395 + Using [17]-fprofile-values, an instrumented binary can track 4396 multiple values (up to 4) for e.g. indirect calls and provide 4397 more precise profile information. 4398 4399 [18]New Languages and Language-Specific Improvements 4400 4401 * Version 2.6 of the [19]OpenACC specification is now supported by 4402 the C, C++ and Fortran compilers. See the [20]implementation status 4403 section on the OpenACC wiki page and the [21]run-time library 4404 documentation for further information. 4405 * GCC 10 adds a number of newly implemented [22]OpenMP 5.0 features 4406 such as conditional lastprivate clause, scan and loop directives, 4407 order(concurrent) and use_device_addr clauses support, if clause on 4408 simd construct, and partial support for the declare variant 4409 directive, getting closer to full support of the OpenMP 5.0 4410 standard. 4411 * OpenMP and OpenACC now support [23]offloading to AMD Radeon (GCN) 4412 GPUs; supported are the third-generation Fiji (fiji) and the 4413 fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906). 4414 4415 [24]C family 4416 4417 * New attributes: 4418 + The access function and type attribute has been added to 4419 describe how a function accesses objects passed to it by 4420 pointer or reference, and to associate such arguments with 4421 integer arguments denoting the objects' sizes. The attribute 4422 is used to enable the detection of invalid accesses by 4423 user-defined functions, such as those diagnosed by 4424 -Wstringop-overflow. 4425 + The symver attribute can be used to bind symbols to specific 4426 version nodes on ELF platforms. This is preferred to using 4427 inline assembly with GNU as symver directive because the 4428 latter is not compatible with link-time optimizations. 4429 * New warnings: 4430 + [25]-Wstring-compare, enabled by -Wextra, warns about equality 4431 and inequality expressions between zero and the result of a 4432 call to either strcmp and strncmp that evaluate to a constant 4433 as a result of the length of one argument being greater than 4434 the size of the array pointed to by the other. 4435 + [26]-Wzero-length-bounds, enabled by -Warray-bounds, warns 4436 about accesses to elements of zero-length arrays that might 4437 overlap other members of the same object. 4438 * Enhancements to existing warnings: 4439 + [27]-Warray-bounds detects more out-of-bounds accesses to 4440 member arrays as well as accesses to elements of zero-length 4441 arrays. 4442 + [28]-Wformat-overflow makes full use of string length 4443 information computed by the strlen optimization pass. 4444 + [29]-Wrestrict detects overlapping accesses to dynamically 4445 allocated objects. 4446 + [30]-Wreturn-local-addr diagnoses more instances of return 4447 statements returning addresses of automatic variables. 4448 + [31]-Wstringop-overflow detects more out-of-bounds stores to 4449 member arrays including zero-length arrays, dynamically 4450 allocated objects and variable length arrays, as well as more 4451 instances of reads of unterminated character arrays by string 4452 built-in functions. The warning also detects out-of-bounds 4453 accesses by calls to user-defined functions declared with the 4454 new attribute access. 4455 + [32]-Warith-conversion re-enables warnings from -Wconversion, 4456 -Wfloat-conversion, and -Wsign-conversion that are now off by 4457 default for an expression where the result of an arithmetic 4458 operation will not fit in the target type due to promotion, 4459 but the operands of the expression do fit in the target type. 4460 * Extended characters in identifiers may now be specified directly in 4461 the input encoding (UTF-8, by default), in addition to the UCN 4462 syntax (\uNNNN or \UNNNNNNNN) that is already supported: 4463 4464 static const int = 3; 4465 int get_nave_pi() { 4466 return ; 4467 } 4468 4469 [33]C 4470 4471 * Several new features from the upcoming C2X revision of the ISO C 4472 standard are supported with -std=c2x and -std=gnu2x. Some of these 4473 features are also supported as extensions when compiling for older 4474 language versions. In addition to the features listed, some 4475 features previously supported as extensions and now added to the C 4476 standard are enabled by default in C2X mode and not diagnosed with 4477 -std=c2x -Wpedantic. 4478 + The [[]] attribute syntax is supported, as in C++. Existing 4479 attributes can be used with this syntax in forms such as 4480 [[gnu::const]]. The standard attributes [[deprecated]], 4481 [[fallthrough]] and [[maybe_unused]] are supported. 4482 + UTF-8 character constants using the u8'' syntax are supported. 4483 + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and 4484 LDBL_NORM_MAX. 4485 + When decimal floating-point arithmetic is supported, <float.h> 4486 defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and 4487 DEC128_TRUE_MIN, in addition to the macros that were 4488 previously only defined if __STDC_WANT_DEC_FP__ was defined 4489 before including <float.h>. 4490 + In C2X mode, empty parentheses in a function definition give 4491 that function a type with a prototype for subsequent calls; 4492 other old-style function definitions are diagnosed by default 4493 in C2X mode. 4494 + The strftime format checking supports the %OB and %Ob formats. 4495 + In C2X mode, -fno-fp-int-builtin-inexact is enabled by 4496 default. 4497 * GCC now defaults to -fno-common. As a result, global variable 4498 accesses are more efficient on various targets. In C, global 4499 variables with multiple tentative definitions now result in linker 4500 errors. With -fcommon such definitions are silently merged during 4501 linking. 4502 4503 [34]C++ 4504 4505 * Several C++20 features have been implemented: 4506 + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, 4507 P0848R3, P1616R1, P1452R2 4508 + P1668R1, Permitting Unevaluated inline-assembly in constexpr 4509 Functions 4510 + P1161R3, Deprecate a[b,c] 4511 + P0848R3, Conditionally Trivial Special Member Functions 4512 + P1091R3, Extending structured bindings 4513 + P1143R2, Adding the constinit keyword 4514 + P1152R4, Deprecating volatile 4515 + P0388R4, Permit conversions to arrays of unknown bound 4516 + P0784R7, constexpr new 4517 + P1301R4, [[nodiscard("with reason")]] 4518 + P1814R0, class template argument deduction for alias templates 4519 + P1816R0, class template argument deduction for aggregates 4520 + P0960R3, Parenthesized initialization of aggregates 4521 + P1331R2, Allow trivial default initialization in constexpr 4522 contexts 4523 + P1327R1, Allowing dynamic_cast and polymorphic typeid in 4524 constexpr contexts 4525 + P0912R5, Coroutines (requires -fcoroutines) 4526 * Several C++ Defect Reports have been resolved, e.g.: 4527 + DR 1560, lvalue-to-rvalue conversion in ?: 4528 + DR 1813, __is_standard_layout for a class with repeated bases 4529 + DR 2094, volatile scalars are trivially copyable, 4530 + DR 2096, constraints on literal unions 4531 + DR 2413, typename in conversion-function-ids 4532 + DR 2352, Similar types and reference binding 4533 + DR 1601, Promotion of enumeration with fixed underlying type 4534 + DR 330, Qualification conversions and pointers to arrays of 4535 pointers 4536 + DR 1307, Overload resolution based on size of array 4537 initializer-list 4538 + DR 1710, Missing template keyword in class-or-decltype 4539 * New warnings: 4540 + [35]-Wmismatched-tags, disabled by default, warns about 4541 declarations of structs, classes, and class templates and 4542 their specializations with a class-key that does not match 4543 either the definition or the first declaration if no 4544 definition is provided. The option is provided to ease 4545 portability to Windows-based compilers. 4546 + [36]-Wredundant-tags, disabled by default, warns about 4547 redundant class-key and enum-key in contexts where the key can 4548 be eliminated without causing an syntactic ambiguity. 4549 * G++ can now detect modifying constant objects in constexpr 4550 evaluation (which is undefined behavior). 4551 * G++ no longer emits bogus -Wsign-conversion warnings with explicit 4552 casts. 4553 * Narrowing is now detected in more contexts (e.g., case values). 4554 * Memory consumption of the compiler has been reduced in constexpr 4555 evaluation. 4556 * The noexcept-specifier is now properly treated as a complete-class 4557 context as per [class.mem]. 4558 * The attribute deprecated can now be used on namespaces too. 4559 * The ABI of passing and returning certain C++ classes by value 4560 changed on several targets in GCC 10, including [37]AArch64, 4561 [38]ARM, [39]PowerPC ELFv2, [40]S/390 and [41]Itanium. These 4562 changes affect classes with a zero-sized subobject (an empty base 4563 class, or data member with the [[no_unique_address]] attribute) 4564 where all other non-static data members have the same type (this is 4565 called a "homogeneous aggregate" in some ABI specifications, or if 4566 there is only one such member, a "single element"). In -std=c++17 4567 and -std=c++20 modes, classes with an empty base class were not 4568 considered to have a single element or to be a homogeneous 4569 aggregate, and so could be passed differently (in the wrong 4570 registers or at the wrong stack address). This could make code 4571 compiled with -std=c++17 and -std=c++14 ABI incompatible. This has 4572 been corrected and the empty bases are ignored in those ABI 4573 decisions, so functions compiled with -std=c++14 and -std=c++17 are 4574 now ABI compatible again. Example: struct empty {}; struct S : 4575 empty { float f; }; void f(S);. Similarly, in classes containing 4576 non-static data members with empty class types using the C++20 4577 [[no_unique_address]] attribute, those members weren't ignored in 4578 the ABI argument passing decisions as they should be. Both of these 4579 ABI changes are now diagnosed with -Wpsabi. 4580 4581 [42]Runtime Library (libstdc++) 4582 4583 * Improved experimental C++2a support, including: 4584 + Library concepts in <concepts> and <iterator>. 4585 + Constrained algorithms in <ranges>, <algorithm>, and <memory> 4586 (thanks to Patrick Palka). 4587 + New algorithms shift_left and shift_right (thanks to Patrick 4588 Palka). 4589 + std::span (thanks to JeanHeyd Meneide). 4590 + Three-way comparisons in <compare> and throughout the library. 4591 + Constexpr support in <algorithm> and elsewhere (thanks to 4592 Edward Smith-Rowland). 4593 + <stop_token> and std::jthread (thanks to Thomas Rodgers). 4594 + std::atomic_ref and std::atomic<floating point>. 4595 + Integer comparison functions (cmp_equal, cmp_less etc.). 4596 + std::ssize, std::to_array. 4597 + std::construct_at, std::destroy, constexpr std::allocator. 4598 + Mathematical constants in <numbers>. 4599 * Support for RDSEED in std::random_device. 4600 * Reduced header dependencies, leading to faster compilation for some 4601 code. 4602 * The std::iterator base class of std::istreambuf_iterator was 4603 changed in C++98 mode to be consistent with C++11 and later 4604 standards. This is expected to have no noticeable effect except in 4605 the unlikely case of a class which has potentially overlapping 4606 subobjects of type std::istreambuf_iterator<C> and another iterator 4607 type with a std::iterator<input_iterator_tag, C, ...> base class. 4608 The layout of such a type might change when compiled as C++98. 4609 [43]Bug 92285 has more details and concrete examples. 4610 4611 [44]D 4612 4613 * Support for static foreach has been implemented. 4614 * Aliases can now be created directly from any __traits that return 4615 symbols or tuples. Previously, an AliasSeq was necessary in order 4616 to alias their return. 4617 * It is now possible to detect the language ABI specified for a 4618 struct, class, or interface using __traits(getLinkage, ...). 4619 * Support for core.math.toPrec intrinsics has been added. These 4620 intrinsics guarantee the rounding to specific floating-point 4621 precisions at specified points in the code. 4622 * Support for pragma(inline) has been implemented. Previously the 4623 pragma was recognized, but had no effect on the compilation. 4624 * Optional parentheses in asm operands are deprecated and will be 4625 removed in a future release. 4626 * All content imported files are now included in the make dependency 4627 list when compiling with -M. 4628 * Compiler recognized attributes provided by the gcc.attribute module 4629 will now take effect when applied to function prototypes as well as 4630 when applied to full function declarations. 4631 * Added a --enable-libphobos-checking configure option to control 4632 whether run-time checks are compiled into the D runtime library. 4633 * Added a --with-libphobos-druntime-only configure option to indicate 4634 whether to build only the core D runtime library, or both the core 4635 and standard libraries into libphobos. 4636 4637 [45]Fortran 4638 4639 * use_device_addr of version 5.0 of the [46]OpenMP specification is 4640 now supported. Note that otherwise OpenMP 4.5 is partially 4641 supported in the Fortran compiler; the largest missing item is 4642 structure element mapping. 4643 * The default buffer size for I/O using unformatted files has been 4644 increased to 1048576. The buffer size for can now be set at runtime 4645 via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and 4646 GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted 4647 files, respectively. 4648 * Mismatches between actual and dummy argument lists in a single file 4649 are now rejected with an error. Use the new option 4650 -fallow-argument-mismatch to turn these errors into warnings; this 4651 option is implied with -std=legacy. -Wargument-mismatch has been 4652 removed. 4653 * The handling of a BOZ literal constant has been reworked to provide 4654 better conformance to the Fortran 2008 and 2018 standards. In these 4655 Fortran standards, a BOZ literal constant is a typeless and 4656 kindless entity. As a part of the rework, documented and 4657 undocumented extensions to the Fortran standard now emit errors 4658 during compilation. Some of these extensions are permitted with the 4659 -fallow-invalid-boz option, which degrades the error to a warning 4660 and the code is compiled as with older gfortran. 4661 * At any optimization level except-Os, gfortran now uses inline 4662 packing for arguments instead of calling a library routine. If the 4663 source contains a large number of arguments that need to be 4664 repacked, code size or time for compilation can become excessive. 4665 If that is the case, -fno-inline-arg-packing can be used to disable 4666 inline argument packing. 4667 * Legacy extensions: 4668 + For formatted input/output, if the explicit widths after the 4669 data-edit descriptors I, F and G have been omitted, default 4670 widths are used. 4671 + A blank format item at the end of a format specification, i.e. 4672 nothing following the final comma, is allowed. Use the option 4673 -fdec-blank-format-item; this option is implied with -fdec. 4674 + The existing support for AUTOMATIC and STATIC attributes has 4675 been extended to allow variables with the AUTOMATIC attribute 4676 to be used in EQUIVALENCE statements. Use -fdec-static; this 4677 option is implied by -fdec. 4678 + Allow character literals in assignments and DATA statements 4679 for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. 4680 Use the option -fdec-char-conversions; this option is implied 4681 with -fdec. 4682 + DEC comparisons, i.e. allow Hollerith constants to be used in 4683 comparisons with INTEGER, REAL, COMPLEX and CHARACTER 4684 expressions. Use the option -fdec. 4685 * Character type names in errors and warnings now include len in 4686 addition to kind; * is used for assumed length. The kind is omitted 4687 if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4). 4688 * CO_BROADCAST now supports derived type variables including objects 4689 with allocatable components. In this case, the optional arguments 4690 STAT= and ERRMSG= are currently ignored. 4691 * The handling of module and submodule names has been reworked to 4692 allow the full 63-character length mandated by the standard. 4693 Previously symbol names were truncated if the combined length of 4694 module, submodule, and function name exceeded 126 characters. This 4695 change therefore breaks the ABI, but only for cases where this 126 4696 character limit was exceeded. 4697 4698 [47]Go 4699 4700 * GCC 10 provides a complete implementation of the Go 1.14.6 user 4701 packages. 4702 4703 [48]libgccjit 4704 4705 * The libgccjit API gained four new entry points: 4706 + [49]gcc_jit_version_major, [50]gcc_jit_version_minor, and 4707 [51]gcc_jit_version_patchlevel for programmatically checking 4708 the libgccjit version from client code, and 4709 + [52]gcc_jit_context_new_bitfield 4710 4711 [53]New Targets and Target Specific Improvements 4712 4713 [54]AArch64 & arm 4714 4715 * The AArch64 and arm ports now support condition flag output 4716 constraints in inline assembly, as indicated by the 4717 __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for 4718 A32 and T32 targets. Please refer to the documentation for more 4719 details. 4720 4721 [55]AArch64 4722 4723 * There have been several improvements related to the Scalable Vector 4724 Extension (SVE): 4725 + The SVE ACLE types and intrinsics are now supported. They can 4726 be accessed using the header file arm_sve.h. 4727 + It is now possible to create fixed-length SVE types using the 4728 arm_sve_vector_bits attribute. For example: 4729 #if __ARM_FEATURE_SVE_BITS==512 4730 typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512))); 4731 typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512))); 4732 #endif 4733 + -mlow-precision-div, -mlow-precision-sqrt and 4734 -mlow-precision-recip-sqrt now work for SVE. 4735 + -msve-vector-bits=128 now generates vector-length-specific 4736 code for little-endian targets. It continues to generate 4737 vector-length-agnostic code for big-endian targets, just as 4738 previous releases did for all targets. 4739 + The vectorizer is now able to use extending loads and 4740 truncating stores, including gather loads and scatter stores. 4741 + The vectorizer now compares the cost of vectorizing with SVE 4742 and vectorizing with Advanced SIMD and tries to pick the best 4743 one. Previously it would always use SVE if possible. 4744 + If a vector loop uses Advanced SIMD rather than SVE, the 4745 vectorizer now considers using SVE to vectorize the left-over 4746 elements (the scalar tail or epilog). 4747 + Besides these specific points, there have been many general 4748 improvements to the way that the vectorizer uses SVE. 4749 * The -mbranch-protection=pac-ret option now accepts the optional 4750 argument +b-key extension to perform return address signing with 4751 the B-key instead of the A-key. 4752 * The option -moutline-atomics has been added to aid deployment of 4753 the Large System Extensions (LSE) on GNU/Linux systems built with a 4754 baseline architecture targeting Armv8-A. When the option is 4755 specified code is emitted to detect the presence of LSE 4756 instructions at runtime and use them for standard atomic 4757 operations. For more information please refer to the documentation. 4758 * The Transactional Memory Extension is now supported through ACLE 4759 intrinsics. It can be enabled through the +tme option extension 4760 (for example, -march=armv8.5-a+tme). 4761 * A number of features from Armv8.5-A are now supported through ACLE 4762 intrinsics. These include: 4763 + The random number instructions that can be enabled through the 4764 (already present in GCC 9.1) +rng option extension. 4765 + Floating-point intrinsics to round to integer instructions 4766 from Armv8.5-A when targeting -march=armv8.5-a or later. 4767 + Memory Tagging Extension intrinsics enabled through the 4768 +memtag option extension. 4769 * Similarly, the following Armv8.6-A features are now supported 4770 through ACLE intrinsics: 4771 + The bfloat16 extension. This extension is enabled 4772 automatically when Armv8.6-A is selected (such as by 4773 -march=armv8.6-a). It can also be enabled for Armv8.2-A and 4774 later using the +bf16 option extension. 4775 + The Matrix Multiply extension. This extension is split into 4776 three parts, one for each supported data type: 4777 o Support for 8-bit integer matrix multiply instructions. 4778 This extension is enabled automatically when Armv8.6-A is 4779 selected. It can also be enabled for Armv8.2-A and later 4780 using the +i8mm option extension. 4781 o Support for 32-bit floating-point matrix multiply 4782 instructions. This extension can be enabled using the 4783 +f32mm option extension, which also has the effect of 4784 enabling SVE. 4785 o Support for 64-bit floating-point matrix multiply 4786 instructions. This extension can be enabled using the 4787 +f64mm option extension, which likewise has the effect of 4788 enabling SVE. 4789 * SVE2 is now supported through ACLE intrinsics and (to a limited 4790 extent) through autovectorization. It can be enabled through the 4791 +sve2 option extension (for example, -march=armv8.5-a+sve2). 4792 Additional extensions can be enabled through +sve2-sm4, +sve2-aes, 4793 +sve2-sha3 and +sve2-bitperm. 4794 * Support has been added for the following processors (GCC 4795 identifiers in parentheses): 4796 + Arm Cortex-A77 (cortex-a77). 4797 + Arm Cortex-A76AE (cortex-a76ae). 4798 + Arm Cortex-A65 (cortex-a65). 4799 + Arm Cortex-A65AE (cortex-a65ae). 4800 + Arm Cortex-A34 (cortex-a34). 4801 + Marvell ThunderX3 (thunderx3t110). 4802 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4803 options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as 4804 arguments to the equivalent target attributes and pragmas. 4805 4806 [56]arm 4807 4808 * Support for the FDPIC ABI has been added. It uses 64-bit function 4809 descriptors to represent pointers to functions, and enables code 4810 sharing on MMU-less systems. The corresponding target triple is 4811 arm-uclinuxfdpiceabi, and the C library is uclibc-ng. 4812 * Support has been added for the Arm EABI on NetBSD through the 4813 arm*-*-netbsdelf-*eabi* triplet. 4814 * The handling of 64-bit integer operations has been significantly 4815 reworked and improved leading to improved performance and reduced 4816 stack usage when using 64-bit integral data types. The option 4817 -mneon-for-64bits is now deprecated and will be removed in a future 4818 release. 4819 * Support has been added for the following processors (GCC 4820 identifiers in parentheses): 4821 + Arm Cortex-A77 (cortex-a77). 4822 + Arm Cortex-A76AE (cortex-a76ae). 4823 + Arm Cortex-M35P (cortex-m35p). 4824 + Arm Cortex-M55 (cortex-m55). 4825 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4826 options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p. 4827 * Support has been extended for the [57]ACLE data-processing 4828 intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit 4829 multiplication and other related intrinsics aimed at DSP algorithm 4830 optimization. 4831 * Support for -mpure-code in Thumb-1 (v6m) has been added: this 4832 M-profile feature is no longer restricted to targets with MOVT. For 4833 example, -mcpu=cortex-m0 now supports this option. 4834 * Support for the [58]Armv8.1-M Mainline Architecture has been added. 4835 + Armv8.1-M Mainline can be enabled by using the 4836 -march=armv8.1-m.main command-line option. 4837 * Support for the [59]MVE beta ACLE intrinsics has been added. These 4838 intrinsics can be enabled by including the arm_mve.h header file 4839 and passing the +mve or +mve.fp option extensions (for example: 4840 -march=armv8.1-m.main+mve). 4841 * Support for the Custom Datapath Extension beta ACLE intrinsics has 4842 been added. 4843 * Support for Armv8.1-M Mainline Security Extensions architecture has 4844 been added. The -mcmse option, when used in combination with an 4845 Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main 4846 -mcmse), now leads to the generation of improved code sequences 4847 when changing security states. 4848 4849 [60]AMD Radeon (GCN) 4850 4851 * Code generation and in particular vectorization support have been 4852 much improved. 4853 4854 [61]ARC 4855 4856 * The interrupt service routine functions save all used registers, 4857 including extension registers and auxiliary registers used by Zero 4858 Overhead Loops. 4859 * Improve code size by using multiple short instructions instead of a 4860 single long mov or ior instruction when its long immediate constant 4861 is known. 4862 * Fix usage of the accumulator register for ARC600. 4863 * Fix issues with uncached attribute. 4864 * Remove -mq-class option. 4865 * Improve 64-bit integer addition and subtraction operations. 4866 4867 [62]AVR 4868 4869 * Support for the XMEGA-like devices 4870 4871 ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, 4872 ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, 4873 ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, 4874 ATmega4808, ATmega4809 4875 has been added. 4876 * A new command-line option -nodevicespecs has been added. It allows 4877 to provide a custom device-specs file by means of 4878 4879 avr-gcc -nodevicespecs -specs=my-spec-file <options> 4880 and without the need to provide options -B and -mmcu=. See [63]AVR 4881 command-line options for details. This feature is also available in 4882 GCC 9.3+ and GCC 8.4+. 4883 * New command-line options -mdouble=[32,64] and -mlong-double=[32,64] 4884 have been added. They allow to choose the size (in bits) of the 4885 double and long double types, respectively. Whether or not the 4886 mentioned layouts are available, whether the options act as a 4887 multilib option, and the default for either option are controlled 4888 by the new [64]AVR configure options --with-double= and 4889 --with-long-double=. 4890 * A new configure option --with-libf7= has been added. It controls to 4891 which level avr-libgcc provides 64-bit floating point support by 4892 means of [65]Libf7. 4893 * A new configure option --with-double-comparison= has been added. 4894 It's unlikely you need to set this option by hand. 4895 4896 [66]IA-32/x86-64 4897 4898 * Support to expand __builtin_roundeven into the appropriate SSE 4.1 4899 instruction has been added. 4900 * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD 4901 intrinsics are available via the -menqcmd compiler switch. 4902 * GCC now supports the Intel CPU named Cooperlake through 4903 -march=cooperlake. The switch enables the AVX512BF16 ISA 4904 extensions. 4905 * GCC now supports the Intel CPU named Tigerlake through 4906 -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B 4907 AVX512VP2INTERSECT ISA extensions. 4908 4909 [67]MIPS 4910 4911 * The mips*-*-linux* targets now mark object files with appropriate 4912 GNU-stack note, facilitating use of non-executable stack hardening 4913 on GNU/Linux. The soft-float targets have this feature enabled by 4914 default, while for hard-float targets it is required for GCC to be 4915 configured with --with-glibc-version=2.31 against glibc 2.31 or 4916 later. 4917 4918 [68]PowerPC / PowerPC64 / RS6000 4919 4920 * Many vector builtins have been listed as deprecated in the 4921 [69]64-Bit ELF V2 ABI Specification for quite a number of years. 4922 The vector builtins listed in Tables A.8 through A.10 are now 4923 deprecated for GCC 10, and will likely be removed from support in 4924 GCC 11. Note that this does not result in any loss of function. 4925 These deprecated builtins generally provide somewhat nonsensical 4926 argument lists (for example, mixing signed, unsigned, and bool 4927 vector arguments arbitrarily), or are duplicate builtins that are 4928 inconsistent with the expected naming scheme. We expect that this 4929 will be unlikely to affect much if any code, and any required code 4930 changes will be trivial. 4931 4932 [70]PRU 4933 4934 * A new back end targeting TI PRU I/O processors has been contributed 4935 to GCC. 4936 4937 [71]RISC-V 4938 4939 * The riscv*-*-* targets now require GNU binutils version 2.30 or 4940 later, to support new assembly instructions produced by GCC. 4941 4942 [72]V850 4943 4944 * The ABI for V850 nested functions has been changed. Previously the 4945 V850 port used %r20 for the static chain pointer, now the port uses 4946 %r19. This corrects a long standing latent bug in the v850 port 4947 where a call to a nested function would unexpectedly change the 4948 value in %r20. 4949 4950 [73]Operating Systems 4951 4952 [74]Improvements for plugin authors 4953 4954 * GCC diagnostics can now have a chain of events associated with 4955 them, describing a path through the code that triggers the problem. 4956 These can be printed by the diagnostics subsystem in various ways, 4957 controlled by the [75]-fdiagnostics-path-format option, or captured 4958 in JSON form via [76]-fdiagnostics-format=json. 4959 * GCC diagnostics can now be associated with [77]CWE weakness 4960 identifiers, which will appear on the standard error stream, and in 4961 the JSON output from [78]-fdiagnostics-format=json. 4962 4963 Other significant improvements 4964 4965 * To allow inline expansion of both memcpy and memmove, the existing 4966 movmem instruction patterns used for non-overlapping memory copies 4967 have been renamed to cpymem. The movmem name is now used for 4968 overlapping memory moves, consistent with the library functions 4969 memcpy and memmove. 4970 * For many releases, when GCC emits a warning it prints the option 4971 controlling that warning. As of GCC 10, that option text is now a 4972 clickable hyperlink for the documentation of that option (assuming 4973 a [79]sufficiently capable terminal). This behavior can be 4974 controlled via a new [80]-fdiagnostics-urls option (along with 4975 various environment variables and heuristics documented with that 4976 option). 4977 4978 GCC 10.1 4979 4980 This is the [81]list of problem reports (PRs) from GCC's bug tracking 4981 system that are known to be fixed in the 10.1 release. This list might 4982 not be complete (that is, it is possible that some PRs that have been 4983 fixed are not listed here). 4984 4985 [82]GCC 10.2 4986 4987 This is the [83]list of problem reports (PRs) from GCC's bug tracking 4988 system that are known to be fixed in the 10.2 release. This list might 4989 not be complete (that is, it is possible that some PRs that have been 4990 fixed are not listed here). 4991 4992 [84]GCC 10.3 4993 4994 This is the [85]list of problem reports (PRs) from GCC's bug tracking 4995 system that are known to be fixed in the 10.3 release. This list might 4996 not be complete (that is, it is possible that some PRs that have been 4997 fixed are not listed here). 4998 4999 Target Specific Changes 5000 5001 AArch64 5002 5003 * A bug with the Random Number intrinsics in the arm_acle.h header 5004 that resulted in an incorrect status result being returned has been 5005 fixed. 5006 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 5007 options are -mcpu=a64fx and -mtune=a64fx respectively. In 5008 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 5009 tunes the code for the A64FX. This includes tuning the SVE code, 5010 although by default the code is still length-agnostic and so works 5011 for all SVE implementations. Adding -msve-vector-bits=512 makes the 5012 code specific to 512-bit SVE. 5013 5014 x86-64 5015 5016 * GCC 10.3 supports AMD CPUs based on the znver3 core via 5017 -march=znver3. 5018 5019 [86]GCC 10.4 5020 5021 This is the [87]list of problem reports (PRs) from GCC's bug tracking 5022 system that are known to be fixed in the 10.4 release. This list might 5023 not be complete (that is, it is possible that some PRs that have been 5024 fixed are not listed here). 5025 5026 Target Specific Changes 5027 5028 x86-64 5029 5030 * The x86-64 ABI of passing and returning structures with a 64-bit 5031 integer vector changed in GCC 10.1 when MMX is disabled. Disabling 5032 MMX no longer changes how they are passed nor returned. This ABI 5033 change is now diagnosed with -Wpsabi. 5034 5035 [88]GCC 10.5 5036 5037 This is the [89]list of problem reports (PRs) from GCC's bug tracking 5038 system that are known to be fixed in the 10.5 release. This list might 5039 not be complete (that is, it is possible that some PRs that have been 5040 fixed are not listed here). 5041 5042 5043 For questions related to the use of GCC, please consult these web 5044 pages and the [90]GCC manuals. If that fails, the 5045 [91]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5046 web pages and the development of GCC are welcome on our developer 5047 list at [92]gcc (a] gcc.gnu.org. All of [93]our lists have public 5048 archives. 5049 5050 Copyright (C) [94]Free Software Foundation, Inc. Verbatim copying and 5051 distribution of this entire article is permitted in any medium, 5052 provided this notice is preserved. 5053 5054 These pages are [95]maintained by the GCC team. Last modified 5055 2026-06-16. 5056 5057 References 5058 5059 1. https://gcc.gnu.org/gcc-10/porting_to.html 5060 2. https://gcc.gnu.org/onlinedocs/index.html#current 5061 3. https://gcc.gnu.org/gcc-10/changes.html#empty_base 5062 4. https://www.mpfr.org/ 5063 5. https://gcc.gnu.org/gcc-10/changes.html#iterator_base 5064 6. https://gcc.gnu.org/gcc-10/changes.html#general 5065 7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin 5066 8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce 5067 9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training 5068 10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible 5069 11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate 5070 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path 5071 13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html 5072 14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions 5073 15. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html 5074 16. https://facebook.github.io/zstd/ 5075 17. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values 5076 18. https://gcc.gnu.org/gcc-10/changes.html#languages 5077 19. https://www.openacc.org/ 5078 20. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-10 5079 21. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1 5080 22. https://www.openmp.org/specifications/ 5081 23. https://gcc.gnu.org/wiki/Offloading 5082 24. https://gcc.gnu.org/gcc-10/changes.html#c-family 5083 25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare 5084 26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds 5085 27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds 5086 28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow 5087 29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict 5088 30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr 5089 31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow 5090 32. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion 5091 33. https://gcc.gnu.org/gcc-10/changes.html#c 5092 34. https://gcc.gnu.org/gcc-10/changes.html#cxx 5093 35. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags 5094 36. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags 5095 37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383 5096 38. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711 5097 39. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707 5098 40. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704 5099 41. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706 5100 42. https://gcc.gnu.org/gcc-10/changes.html#libstdcxx 5101 43. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285 5102 44. https://gcc.gnu.org/gcc-10/changes.html#d 5103 45. https://gcc.gnu.org/gcc-10/changes.html#fortran 5104 46. https://www.openmp.org/specifications/ 5105 47. https://gcc.gnu.org/gcc-10/changes.html#go 5106 48. https://gcc.gnu.org/gcc-10/changes.html#jit 5107 49. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major 5108 50. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor 5109 51. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel 5110 52. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield 5111 53. https://gcc.gnu.org/gcc-10/changes.html#targets 5112 54. https://gcc.gnu.org/gcc-10/changes.html#arm-targets 5113 55. https://gcc.gnu.org/gcc-10/changes.html#aarch64 5114 56. https://gcc.gnu.org/gcc-10/changes.html#arm 5115 57. https://github.com/ARM-software/acle/releases 5116 58. https://developer.arm.com/Architectures/M-Profile Architecture 5117 59. https://developer.arm.com/architectures/instruction-sets/intrinsics/ 5118 60. https://gcc.gnu.org/gcc-10/changes.html#amdgcn 5119 61. https://gcc.gnu.org/gcc-10/changes.html#arc 5120 62. https://gcc.gnu.org/gcc-10/changes.html#avr 5121 63. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs 5122 64. https://gcc.gnu.org/install/configure.html#avr 5123 65. https://gcc.gnu.org/wiki/avr-gcc#Libf7 5124 66. https://gcc.gnu.org/gcc-10/changes.html#x86 5125 67. https://gcc.gnu.org/gcc-10/changes.html#mips 5126 68. https://gcc.gnu.org/gcc-10/changes.html#powerpc 5127 69. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture 5128 70. https://gcc.gnu.org/gcc-10/changes.html#pru 5129 71. https://gcc.gnu.org/gcc-10/changes.html#riscv 5130 72. https://gcc.gnu.org/gcc-10/changes.html#v850 5131 73. https://gcc.gnu.org/gcc-10/changes.html#os 5132 74. https://gcc.gnu.org/gcc-10/changes.html#plugins 5133 75. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format 5134 76. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 5135 77. https://cwe.mitre.org/ 5136 78. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 5137 79. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda 5138 80. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls 5139 81. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0 5140 82. https://gcc.gnu.org/gcc-10/changes.html#GCC10.2 5141 83. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2 5142 84. https://gcc.gnu.org/gcc-10/changes.html#GCC10.3 5143 85. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3 5144 86. https://gcc.gnu.org/gcc-10/changes.html#GCC10.4 5145 87. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4 5146 88. https://gcc.gnu.org/gcc-10/changes.html#GCC10.5 5147 89. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.5 5148 90. https://gcc.gnu.org/onlinedocs/ 5149 91. mailto:gcc-help (a] gcc.gnu.org 5150 92. mailto:gcc (a] gcc.gnu.org 5151 93. https://gcc.gnu.org/lists.html 5152 94. https://www.fsf.org/ 5153 95. https://gcc.gnu.org/about.html 5154 ====================================================================== 5155 http://gcc.gnu.org/gcc-9/index.html 5156 GCC 9 Release Series 5157 5158 (This release series is no longer supported.) 5159 5160 May 27, 2022 5161 5162 The [1]GNU project and the GCC developers are pleased to announce the 5163 release of GCC 9.5. 5164 5165 This release is a bug-fix release, containing fixes for regressions in 5166 GCC 9.4 relative to previous releases of GCC. 5167 5168 Release History 5169 5170 GCC 9.5 5171 May 27, 2022 ([2]changes, [3]documentation) 5172 5173 GCC 9.4 5174 June 1, 2021 ([4]changes, [5]documentation) 5175 5176 GCC 9.3 5177 Mar 12, 2020 ([6]changes, [7]documentation) 5178 5179 GCC 9.2 5180 Aug 12, 2019 ([8]changes, [9]documentation) 5181 5182 GCC 9.1 5183 May 3, 2019 ([10]changes, [11]documentation) 5184 5185 References and Acknowledgements 5186 5187 GCC used to stand for the GNU C Compiler, but since the compiler 5188 supports several other languages aside from C, it now stands for the 5189 GNU Compiler Collection. 5190 5191 The GCC developers would like to thank the numerous people that have 5192 contributed new features, improvements, bug fixes, and other changes as 5193 well as test results to GCC. This [12]amazing group of volunteers is 5194 what makes GCC successful. 5195 5196 For additional information about GCC please refer to the [13]GCC 5197 project web site or contact the [14]GCC development mailing list. 5198 5199 To obtain GCC please use [15]our mirror sites or [16]our version 5200 control system. 5201 5202 5203 For questions related to the use of GCC, please consult these web 5204 pages and the [17]GCC manuals. If that fails, the 5205 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5206 web pages and the development of GCC are welcome on our developer 5207 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 5208 archives. 5209 5210 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 5211 distribution of this entire article is permitted in any medium, 5212 provided this notice is preserved. 5213 5214 These pages are [22]maintained by the GCC team. Last modified 5215 2024-05-30. 5216 5217 References 5218 5219 1. http://www.gnu.org/ 5220 2. https://gcc.gnu.org/gcc-9/changes.html 5221 3. https://gcc.gnu.org/onlinedocs/9.5.0/ 5222 4. https://gcc.gnu.org/gcc-9/changes.html 5223 5. https://gcc.gnu.org/onlinedocs/9.4.0/ 5224 6. https://gcc.gnu.org/gcc-9/changes.html 5225 7. https://gcc.gnu.org/onlinedocs/9.3.0/ 5226 8. https://gcc.gnu.org/gcc-9/changes.html 5227 9. https://gcc.gnu.org/onlinedocs/9.2.0/ 5228 10. https://gcc.gnu.org/gcc-9/changes.html 5229 11. https://gcc.gnu.org/onlinedocs/9.1.0/ 5230 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html 5231 13. https://gcc.gnu.org/index.html 5232 14. mailto:gcc (a] gcc.gnu.org 5233 15. https://gcc.gnu.org/mirrors.html 5234 16. https://gcc.gnu.org/git.html 5235 17. https://gcc.gnu.org/onlinedocs/ 5236 18. mailto:gcc-help (a] gcc.gnu.org 5237 19. mailto:gcc (a] gcc.gnu.org 5238 20. https://gcc.gnu.org/lists.html 5239 21. https://www.fsf.org/ 5240 22. https://gcc.gnu.org/about.html 5241 ====================================================================== 5242 http://gcc.gnu.org/gcc-9/changes.html 5243 GCC 9 Release Series 5244 Changes, New Features, and Fixes 5245 5246 This page is a "brief" summary of some of the huge number of 5247 improvements in GCC 9. You may also want to check out our [1]Porting to 5248 GCC 9 page and the [2]full GCC documentation. 5249 5250 Caveats 5251 5252 * On Arm targets (arm*-*-*), [3]a bug in the implementation of the 5253 procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has 5254 been fixed: a structure containing a bit-field based on a 64-bit 5255 integral type and where no other element in a structure required 5256 64-bit alignment could be passed incorrectly to functions. This is 5257 an ABI change. If the option -Wpsabi is enabled (on by default) the 5258 compiler will emit a diagnostic note for code that might be 5259 affected. 5260 * Support for a number of older systems and recently unmaintained or 5261 untested target ports of GCC has been declared obsolete in GCC 9. 5262 Unless there is activity to revive them, the next release of GCC 5263 will have their sources permanently removed. 5264 The following ports for individual systems on particular 5265 architectures have been obsoleted: 5266 + Solaris 10 (*-*-solaris2.10). Details can be found in the 5267 [4]announcement. 5268 + Cell Broadband Engine SPU (spu*-*-*). Details can be found in 5269 the [5]announcement. 5270 * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause 5271 ABI incompatibilities with object files compiled with other 5272 versions of GCC. If the std::rotate algorithm is called with an 5273 empty range then it might cause a divide-by-zero error (as a SIGFPE 5274 signal) and crash. The change has been reverted for GCC 9.2.0 and 5275 future releases. For more details see [6]Bug 90920. The problem can 5276 be avoided by recompiling any objects that might call std::rotate 5277 with an empty range, so that the GCC 9.1.0 definition of 5278 std::rotate is not used. 5279 * The automatic template instantiation at link time ([7]-frepo) has 5280 been deprecated and will be removed in a future release. 5281 * The --with-default-libstdcxx-abi=gcc4-compatible configure option 5282 is broken in the 9.1 and 9.2 releases, producing a shared library 5283 with missing symbols (see [8]Bug 90361). As a workaround, configure 5284 without that option and build GCC as normal, then edit the 5285 installed <bits/c++config.h> headers to define the 5286 _GLIBCXX_USE_CXX11_ABI macro to 0. 5287 5288 [9]General Improvements 5289 5290 The following GCC command-line options have been introduced or 5291 improved. 5292 * All command-line options that take a byte-size argument accept 5293 64-bit integers as well as standard SI and IEC suffixes such as kb 5294 and KiB, MB and MiB, or GB and GiB denoting the corresponding 5295 multiples of bytes. See [10]Invoking GCC for more. 5296 * A new option [11]-flive-patching=[inline-only-static|inline-clone] 5297 generates code suitable for live patching. At the same time it 5298 provides multiple-level control over IPA optimizations. See the 5299 user guide for more details. 5300 * A new option, --completion, has been added to provide more fine 5301 option completion in a shell. It is intended to be used by 5302 Bash-completion. 5303 * GCC's diagnostics now print source code with a left margin showing 5304 line numbers, configurable with 5305 [12]-fno-diagnostics-show-line-numbers. 5306 GCC's diagnostics can also now label regions of the source code to 5307 show pertinent information, such as the types within an expression. 5308 $ g++ t.cc 5309 t.cc: In function 'int test(const shape&, const shape&)': 5310 t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl 5311 e>' and 'boxed_value<double>') 5312 14 | return (width(s1) * height(s1) 5313 | ~~~~~~~~~~~~~~~~~~~~~~ 5314 | | 5315 | boxed_value<[...]> 5316 15 | + width(s2) * height(s2)); 5317 | ^ ~~~~~~~~~~~~~~~~~~~~~~ 5318 | | 5319 | boxed_value<[...]> 5320 5321 These labels can be disabled via [13]-fno-diagnostics-show-labels. 5322 * A new option [14]-fdiagnostics-format=json has been introduced for 5323 emitting diagnostics in a machine-readable format. 5324 * The alignment-related options [15]-falign-functions, 5325 [16]-falign-labels, [17]-falign-loops, and [18]-falign-jumps 5326 received support for a secondary alignment (e.g. 5327 -falign-loops=n:m:n2:m2). 5328 * New pair of profiling options ([19]-fprofile-filter-files and 5329 [20]-fprofile-exclude-files) has been added. The options help to 5330 filter which source files are instrumented. 5331 * AddressSanitizer generates more compact redzones for automatic 5332 variables. That helps to reduce memory footprint of a sanitized 5333 binary. 5334 * Numerous improvements have been made to the output of 5335 [21]-fopt-info. 5336 Messages are now prefixed with optimized, missed, or note, rather 5337 than the old behavior of all being prefixed with note. 5338 The output from -fopt-info can now contain information on inlining 5339 decisions: 5340 $ g++ -c inline.cc -O2 -fopt-info-inline-all 5341 inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)( 5342 E)) [with T = char**; E = char*]/2. 5343 inline.cc:24:11: optimized: Inlining void foreach(T, T, void (*)(E)) [with T = 5344 char**; E = char*]/2 into int main(int, char**)/1. 5345 inline.cc:19:12: missed: not inlinable: void inline_me(char*)/0 -> int std::pu 5346 ts(const char*)/3, function body not available 5347 inline.cc:13:8: optimized: Inlined void inline_me(char*)/4 into int main(int, c 5348 har**)/1 which now has time 127.363637 and size 11, net change of +0. 5349 Unit growth for small function inlining: 16->16 (0%) 5350 5351 Inlined 2 calls, eliminated 1 functions 5352 5353 5354 The output from the vectorizer has been rationalized so that failed 5355 attempts to vectorize a loop are displayed in the form 5356 [LOOP-LOCATION]: couldn't vectorize this loop 5357 [PROBLEM-LOCATION]: because of [REASON] 5358 5359 rather than an exhaustive log of all decisions made by the 5360 vectorizer. For example: 5361 $ gcc -c v.c -O3 -fopt-info-all-vec 5362 v.c:7:3: missed: couldn't vectorize loop 5363 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 5364 mory"); 5365 v.c:3:6: note: vectorized 0 loops in function. 5366 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 5367 mory"); 5368 5369 The old behavior can be obtained via a new -internals suboption of 5370 -fopt-info. 5371 * A new option, [22]-fsave-optimization-record has been added, which 5372 writes a SRCFILE.opt-record.json.gz file describing the 5373 optimization decisions made by GCC. This is similar to the output 5374 of -fopt-info, but with additional metadata such as the inlining 5375 chain, and profile information (if available). 5376 * Inter-procedural propagation of stack alignment can now be 5377 controlled by [23]-fipa-stack-alignment. 5378 * Propagation of addressability, readonly, and writeonly flags on 5379 static variables can now be controlled by 5380 [24]-fipa-reference-addressable. 5381 5382 The following built-in functions have been introduced. 5383 * [25]__builtin_expect_with_probability to provide branch prediction 5384 probability hints to the optimizer. 5385 * [26]__builtin_has_attribute determines whether a function, type, or 5386 variable has been declared with some attribute. 5387 * [27]__builtin_speculation_safe_value can be used to help mitigate 5388 against unsafe speculative execution. 5389 5390 The following attributes have been introduced. 5391 * The [28]copy function attribute has been added. The attribute can 5392 also be applied to type definitions and to variable declarations. 5393 5394 A large number of improvements to code generation have been made, 5395 including but not limited to the following. 5396 * Switch expansion has been improved by using a different strategy 5397 (jump table, bit test, decision tree) for a subset of switch cases. 5398 * A linear function expression defined as a switch statement can be 5399 transformed by [29]-ftree-switch-conversion. For example: 5400 5401 int 5402 foo (int how) 5403 { 5404 switch (how) { 5405 case 2: how = 205; break; 5406 case 3: how = 305; break; 5407 case 4: how = 405; break; 5408 case 5: how = 505; break; 5409 case 6: how = 605; break; 5410 } 5411 return how; 5412 } 5413 5414 can be transformed into 100 * how + 5 (for values defined in the 5415 switch statement). 5416 * Inter-procedural optimization improvements: 5417 + Inliner defaults were tuned to better suit modern C++ 5418 codebases, especially when built with link time-optimizations. 5419 New parameters max-inline-insns-small, max-inline-insns-size, 5420 uninlined-function-insns, uninlined-function-time, 5421 uninlined-thunk-insns, and uninlined-thunk-time were added. 5422 + Hot/cold partitioning is now more precise and aggressive. 5423 + Improved scalability for very large translation units 5424 (especially when link-time optimizing large programs). 5425 * Profile driven optimization improvements: 5426 + [30]-fprofile-use now enables [31]-fversion-loops-for-strides, 5427 [32]-floop-interchange, [33]-floop-unroll-and-jam, 5428 [34]-ftree-loop-distribution. 5429 + Streaming of counter histograms was removed, which reduces the 5430 size of profile files. Histograms are computed on the fly with 5431 link-time optimization. 5432 + The parameter hot-bb-count-ws-permille was reduced from 999 to 5433 990 to account for more precise histograms. 5434 * Link-time optimization improvements: 5435 + Types are now simplified prior to streaming resulting in 5436 significant reductions of LTO object file sizes and link-time 5437 memory use as well as improvements of link-time parallelism. 5438 + The default number of partitions (--param lto-partitions) was 5439 increased from 32 to 128 enabling effective use of CPUs with 5440 more than 32 hyperthreads. --param 5441 lto-max-streaming-parallelism can now be used to control the 5442 number of streaming processes. 5443 + Warnings on C++ One Decl Rule violations (-Wodr) are now more 5444 informative and produce fewer redundant results. 5445 Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an 5446 8-core machine was reduced by about 5% compared to GCC 8.3, and the 5447 size of LTO object files by 7%. LTO link time improves by 11% on an 5448 8-core machine and scales significantly better for more parallel 5449 build environments. The serial stage of the link-time optimization 5450 is 28% faster consuming 20% less memory. The parallel stage now 5451 scales to up to 128 partitions rather than 32 and reduces memory 5452 use for every worker by 30%. 5453 5454 The following improvements to the gcov command-line utility have been 5455 made. 5456 * The gcov tool received a new option [35]--use-hotness-colors (-q) 5457 that can provide perf-like coloring of hot functions. 5458 * The gcov tool has changed its intermediate format to a new JSON 5459 format. 5460 5461 [36]New Languages and Language specific improvements 5462 5463 [37]OpenACC support in C, C++, and Fortran continues to be maintained 5464 and improved. Most of the OpenACC 2.5 specification is implemented. See 5465 the [38]implementation status section on the OpenACC wiki page for 5466 further information. 5467 5468 [39]C family 5469 5470 * Version 5.0 of the [40]OpenMP specification is now partially 5471 supported in the C and C++ compilers. For details which features of 5472 OpenMP 5.0 are and which are not supported in the GCC 9 release see 5473 [41]this mail. 5474 * New extensions: 5475 + [42]__builtin_convertvector built-in for vector conversions 5476 has been added. 5477 * New warnings: 5478 + [43]-Waddress-of-packed-member, enabled by default, warns 5479 about an unaligned pointer value from the address of a packed 5480 member of a struct or union. 5481 * Enhancements to existing warnings: 5482 + [44]-Warray-bounds detects more instances of out-of-bounds 5483 indices. 5484 + [45]-Wattribute-alias also detects attribute mismatches 5485 between alias declarations and their targets, in addition to 5486 mismatches between their types. 5487 + [46]-Wformat-overflow and [47]-Wformat-truncation have been 5488 extended to all formatted input/output functions (where 5489 applicable) and enhanced to detect a subset of instances of 5490 reading past the end of unterminated constant character arrays 5491 in %s directives. 5492 + [48]-Wmissing-attributes detects instances of missing function 5493 attributes on declarations of aliases and weak references. 5494 + [49]-Wstringop-truncation also detects a subset of instances 5495 of reading past the end of unterminated constant character 5496 arrays, 5497 * If a macro is used with the wrong argument count, the C and C++ 5498 front ends now show the definition of that macro via a note. 5499 * The spelling corrector now considers transposed letters, and the 5500 threshold for similarity has been tightened, to avoid nonsensical 5501 suggestions. 5502 5503 [50]C 5504 5505 * There is now experimental support for -std=c2x, to select support 5506 for the upcoming C2X revision of the ISO C standard. This standard 5507 is in the early stages of development and the only feature 5508 supported in GCC 9 is _Static_assert with a single argument 5509 (support for _Static_assert with two arguments was added in C11 and 5510 GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU 5511 extensions, and -Wc11-c2x-compat, to warn for uses of features 5512 added in C2X (such warnings are also enabled by use of -Wpedantic 5513 if not using -std=c2x or -std=gnu2x). 5514 * New warnings: 5515 + [51]-Wabsolute-value warns for calls to standard functions 5516 that compute the absolute value of an argument when a more 5517 appropriate standard function is available. For example, 5518 calling abs(3.14) triggers the warning because the appropriate 5519 function to call to compute the absolute value of a double 5520 argument is fabs. The option also triggers warnings when the 5521 argument in a call to such a function has an unsigned type. 5522 This warning can be suppressed with an explicit type cast and 5523 it is also enabled by -Wextra. 5524 5525 [52]C++ 5526 5527 * New warnings: 5528 + [53]-Wdeprecated-copy, implied by -Wextra, warns about the 5529 C++11 deprecation of implicitly declared copy constructor and 5530 assignment operator if one of them is user-provided. 5531 -Wdeprecated-copy-dtor also warns if the destructor is 5532 user-provided, as specified in C++11. 5533 + [54]-Winit-list-lifetime, on by default, warns about uses of 5534 std::initializer_list that are likely to result in a dangling 5535 pointer, such as returning or assigning from a temporary list. 5536 + [55]-Wredundant-move, implied by -Wextra, warns about 5537 redundant calls to std::move. 5538 + [56]-Wpessimizing-move, implied by -Wall, warns when a call to 5539 std::move prevents copy elision. 5540 + [57]-Wclass-conversion, on by default, warns when a conversion 5541 function will never be called due to the type it converts to. 5542 * The C++ front end has experimental support for some of the upcoming 5543 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 5544 including range-based for statements with initializer, default 5545 constructible and assignable stateless lambdas, lambdas in 5546 unevaluated contexts, language support for empty data members, 5547 allowing pack expansion in lambda init-capture, likely and unlikely 5548 attributes, class types in non-type template parameters, allowing 5549 virtual function calls in constant expressions, explicit(bool), 5550 std::is_constant_evaluated, nested inline namespaces, etc. For a 5551 full list of new features, see [58]the C++ status page. 5552 * The C++ front end now preserves source locations for literals, 5553 id-expression, and mem-initializer for longer. For example it is 5554 now able to pin-point the pertinent locations for bad 5555 initializations such as these 5556 $ g++ -c bad-inits.cc 5557 bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization 5558 10 | { 3, json::object }, 5559 | ~~~~~~^~~~~~ 5560 | | 5561 | json 5562 bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f 5563 permissive] 5564 14 | char buffers[3][5] = { "red", "green", "blue" }; 5565 | ^~~~~~~ 5566 bad-inits.cc: In constructor 'X::X()': 5567 bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv 5568 e] 5569 17 | X() : one(42), two(42), three(42) 5570 | ^~ 5571 | | 5572 | int 5573 5574 rather than emitting the error at the final closing parenthesis or 5575 brace. 5576 * Error-reporting of overload resolution has been special-cased to 5577 make the case of a single failed candidate easier to read. For 5578 example: 5579 $ g++ param-type-mismatch.cc 5580 param-type-mismatch.cc: In function 'int test(int, const char*, float)': 5581 param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char* 5582 *' 5583 8 | return foo::member_1 (first, second, third); 5584 | ^~~~~~ 5585 | | 5586 | const char* 5587 param-type-mismatch.cc:3:46: note: initializing argument 2 of 'static int foo: 5588 :member_1(int, const char**, float)' 5589 3 | static int member_1 (int one, const char **two, float three); 5590 | ~~~~~~~~~~~~~^~~ 5591 5592 highlights both the problematic argument, and the parameter that it 5593 can't be converted to. 5594 * Diagnostics involving binary operators now use color to distinguish 5595 the two operands, and label them separately (as per the example of 5596 source labelling above). 5597 * Diagnostics involving function calls now highlight the pertinent 5598 parameter of the declaration in more places. 5599 $ g++ bad-conversion.cc 5600 bad-conversion.cc: In function 'void caller()': 5601 bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*' 5602 9 | callee (0, false, 2); 5603 | ^~~~~ 5604 | | 5605 | bool 5606 bad-conversion.cc:3:19: note: initializing argument 2 of 'void callee(int, voi 5607 d*, int)' 5608 3 | void callee (int, void *, int) 5609 | ^~~~~~ 5610 5611 * The C++ front end's implementation of [59]-Wformat now shows 5612 precise locations within string literals, and underlines the 5613 pertinent arguments at bogus call sites (the C front end has been 5614 doing this since GCC 7). For example: 5615 $ g++ -c bad-printf.cc -Wall 5616 bad-printf.cc: In function 'void print_field(const char*, float, long int, long 5617 int)': 5618 bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type 5619 'int', but argument 3 has type 'long int' [-Wformat=] 5620 6 | printf ("%s: %*ld ", fieldname, column - width, value); 5621 | ~^~~ ~~~~~~~~~~~~~~ 5622 | | | 5623 | int long int 5624 bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b 5625 ut argument 4 has type 'double' [-Wformat=] 5626 6 | printf ("%s: %*ld ", fieldname, column - width, value); 5627 | ~~~^ ~~~~~ 5628 | | | 5629 | long int double 5630 | %*f 5631 5632 * The C++ front end has gained new fix-it hints for forgetting the 5633 return *this; needed by various C++ operators: 5634 $ g++ -c operator.cc 5635 operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr 5636 &)': 5637 operator.cc:7:3: warning: no return statement in function returning non-void [-W 5638 return-type] 5639 6 | m_ptr = other.m_ptr; 5640 +++ |+ return *this; 5641 7 | } 5642 | ^ 5643 5644 for when the compiler needs a typename: 5645 $ g++ -c template.cc 5646 template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i 5647 s a dependent scope 5648 3 | Traits::type type; 5649 | ^~~~~~ 5650 | typename 5651 5652 when trying to use an accessor member as if it were a data member: 5653 $ g++ -c fncall.cc 5654 fncall.cc: In function 'void hangman(const mystring&)': 5655 fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length 5656 () const' (did you forget the '()' ?) 5657 12 | if (str.get_length > 0) 5658 | ~~~~^~~~~~~~~~ 5659 | () 5660 5661 for C++11's scoped enums: 5662 $ g++ -c enums.cc 5663 enums.cc: In function 'void json::test(const json::value&)': 5664 enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js 5665 on::kind::STRING'? 5666 12 | if (v.get_kind () == STRING) 5667 | ^~~~~~ 5668 | json::kind::STRING 5669 enums.cc:3:44: note: 'json::kind::STRING' declared here 5670 3 | enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ }; 5671 | ^~~~~~ 5672 5673 and a tweak to integrate the suggestions about misspelled members 5674 with that for accessors: 5675 $ g++ -c accessor-fixit.cc 5676 accessor-fixit.cc: In function 'int test(t*)': 5677 accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m 5678 ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const') 5679 17 | return ptr->ratio; 5680 | ^~~~~ 5681 | get_ratio() 5682 5683 In addition, various diagnostics in the C++ front-end have been 5684 streamlined by consolidating the suggestion into the initial error, 5685 rather than emitting a follow-up note: 5686 $ g++ typo.cc 5687 typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF 5688 _SIZE'? 5689 5 | uint8_t buf[BUFSIZE]; 5690 | ^~~~~~~ 5691 | BUF_SIZE 5692 5693 [60]Runtime Library (libstdc++) 5694 5695 * Improved support for C++17, including: 5696 + The C++17 implementation is no longer experimental. 5697 + Parallel algorithms and <execution> (requires [61]Thread 5698 Building Blocks 2018 or newer). 5699 + <memory_resource>. 5700 + Using the types and functions in <filesystem> does not require 5701 linking with -lstdc++fs now. 5702 * Improved experimental support for C++2a, including: 5703 + Type traits std::remove_cvref, std::unwrap_reference, 5704 std::unwrap_decay_ref, std::is_nothrow_convertible, and 5705 std::type_identity. 5706 + Headers <bit> and <version>. 5707 + Uniform container erasure (std::erase_if). 5708 + contains member of maps and sets. 5709 + String prefix and suffix checking (starts_with, ends_with). 5710 + Functions std::midpoint and std::lerp for interpolation. 5711 + std::bind_front. 5712 + std::visit<R>. 5713 + std::assume_aligned. 5714 + Uses-allocator construction utilities. 5715 + std::pmr::polymorphic_allocator<std::byte>. 5716 + Library support for char8_t type. 5717 + Destroying delete. 5718 + std::is_constant_evaluated() function. 5719 * Support for opening file streams with wide character paths on 5720 Windows 5721 * Incomplete support for the C++17 Filesystem library and the 5722 Filesystem TS on Windows. 5723 * Incomplete, experimental support for the Networking TS. 5724 5725 [62]D 5726 5727 * Support for the D programming language has been added to GCC, 5728 implementing version 2.076 of the language and run-time library. 5729 5730 [63]Fortran 5731 5732 * Asynchronous I/O is now fully supported. The program needs to be 5733 linked against the pthreads library to use it, otherwise the I/O is 5734 done synchronously. For systems which do not support POSIX 5735 condition variables, such as AIX, all I/O is still done 5736 synchronously. 5737 * The BACK argument for MINLOC and MAXLOC has been implemented. 5738 * The FINDLOC intrinsic function has been implemented. 5739 * The IS_CONTIGUOUS intrinsic function has been implemented. 5740 * Direct access to the real and imaginary parts of a complex variable 5741 via c%re and c%im has been implemented. 5742 * Type parameter inquiry via str%len and a%kind has been implemented. 5743 * C descriptors and the ISO_Fortran_binding.h source file have been 5744 implemented. 5745 * The MAX and MIN intrinsics are no longer guaranteed to return any 5746 particular value in case one of the arguments is a NaN. Note that 5747 this conforms to the Fortran standard and to what other Fortran 5748 compilers do. If there is a need to handle that case in some 5749 specific way, one needs to explicitly check for NaN's before 5750 calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the 5751 intrinsic module IEEE_ARITHMETIC. 5752 * A new command-line option [64]-fdec-include, set also by the 5753 [65]-fdec option, has been added to increase compatibility with 5754 legacy code. With this option, an INCLUDE directive is also parsed 5755 as a statement, which allows the directive to be spread across 5756 multiple source lines with line continuations. 5757 * A new [66]BUILTIN directive, has been added. The purpose of the 5758 directive is to provide an API between the GCC compiler and the GNU 5759 C Library which would define vector implementations of math 5760 routines. 5761 5762 [67]Go 5763 5764 * GCC 9 provides a complete implementation of the Go 1.12.2 user 5765 packages. 5766 5767 [68]libgccjit 5768 5769 * The libgccjit API gained a new entry point: 5770 [69]gcc_jit_context_add_driver_option. 5771 5772 [70]New Targets and Target Specific Improvements 5773 5774 [71]AArch64 & Arm 5775 5776 * Support has been added for the following processors (GCC 5777 identifiers in parentheses): 5778 + Arm Cortex-A76 (cortex-a76). 5779 + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE 5780 (cortex-a76.cortex-a55). 5781 + Arm Neoverse N1 (neoverse-n1). 5782 The GCC identifiers can be used as arguments to the -mcpu or -mtune 5783 options, for example: -mcpu=cortex-a76 or 5784 -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent 5785 target attributes and pragmas. 5786 * The Armv8.3-A complex number instructions are now supported via 5787 intrinsics when the option -march=armv8.3-a or equivalent is 5788 specified. For the half-precision floating-point variants of these 5789 instructions use the architecture extension flag +fp16, e.g. 5790 -march=armv8.3-a+fp16. 5791 The intrinsics are defined by the ACLE specification. 5792 * The Armv8.5-A architecture is now supported through the 5793 -march=armv8.5-a option. 5794 * The Armv8.5-A architecture also adds some security features that 5795 are optional to all older architecture versions. These are now 5796 supported and only affect the assembler. 5797 + Speculation Barrier instruction through the -march=armv8-a+sb 5798 option. 5799 + Execution and Data Prediction Restriction instructions through 5800 the -march=armv8-a+predres option. 5801 + Speculative Store Bypass Safe instruction through the 5802 -march=armv8-a+ssbs option. This does not require a compiler 5803 option for Arm and thus -march=armv8-a+ssbs is an 5804 AArch64-specific option. 5805 5806 [72]AArch64 specific 5807 5808 * Support has been added for the Arm Neoverse E1 processor 5809 (-mcpu=neoverse-e1). 5810 * The AArch64 port now has support for stack clash protection using 5811 the [73]-fstack-clash-protection option. The probing interval/guard 5812 size can be set by using --param 5813 stack-clash-protection-guard-size=12|16. The value of this 5814 parameter must be in bytes represented as a power of two. The two 5815 supported values for this parameter are 12 (for a 4KiB size, 2^12) 5816 and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and 5817 can be changed at configure time using the flag 5818 --with-stack-clash-protection-guard-size=12|16. 5819 * The option -msign-return-address= has been deprecated. This has 5820 been replaced by the new -mbranch-protection= option. This new 5821 option can now be used to enable the return address signing as well 5822 as the new Branch Target Identification feature of Armv8.5-A 5823 architecture. For more information on the arguments accepted by 5824 this option, please refer to [74]AArch64-Options. 5825 * The following optional extensions to Armv8.5-A architecture are now 5826 supported and only affect the assembler. 5827 + Random Number Generation instructions through the 5828 -march=armv8.5-a+rng option. 5829 + Memory Tagging Extension through the -march=armv8.5-a+memtag 5830 option. 5831 5832 [75]Arm specific 5833 5834 * Support for the deprecated Armv2 and Armv3 architectures and their 5835 variants has been removed. Their corresponding -march values and 5836 the -mcpu options that used these architectures have been removed. 5837 * Support for the Armv5 and Armv5E architectures (which have no known 5838 implementations) has been removed. Note that Armv5T, Armv5TE and 5839 Armv5TEJ architectures remain supported. 5840 * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using 5841 their respective -mcpu options. 5842 5843 [76]AMD GCN 5844 5845 * A new back end targeting AMD GCN GPUs has been contributed to GCC. 5846 The implementation is currently limited to compiling 5847 single-threaded, stand-alone programs. Future versions will add 5848 support for offloading multi-threaded kernels via OpenMP and 5849 OpenACC. The following devices are supported (GCC identifiers in 5850 parentheses): 5851 + Fiji (fiji). 5852 + Vega 10 (gfx900). 5853 5854 [77]ARC 5855 5856 * LRA is now on by default for the ARC target. This can be controlled 5857 by -mlra. 5858 * Add support for frame code-density and branch-and-index 5859 instructions. 5860 5861 [78]C-SKY 5862 5863 * A new back end targeting C-SKY V2 processors has been contributed 5864 to GCC. 5865 5866 [79]IA-32/x86-64 5867 5868 * Support of Intel MPX (Memory Protection Extensions) has been 5869 removed. 5870 * New ISA extension support for Intel PTWRITE was added to GCC. 5871 PTWRITE intrinsics are available via the -mptwrite compiler switch. 5872 * GCC now supports the Intel CPU named Cascade Lake with AVX512 5873 extensions through -march=cascadelake. The switch enables the 5874 following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW, 5875 AVX512DQ, AVX512VNNI. 5876 5877 [80]MIPS 5878 5879 * The Loongson loongson-mmi and loongson-ext extensions have been 5880 split from loongson3a: 5881 + loongson-mmi contains the Loongson MMI (MultiMedia extensions 5882 Instructions). 5883 + loongson-ext contains the Loongson EXT (EXTensions 5884 instructions). 5885 * The Loongson EXT2 (EXTensions R2 instructions) are now supported. 5886 + loongson-ext2 contains the Loongson EXT2 instructions. 5887 Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and 5888 -m[no-]loongson-ext2 enable or disable those extensions. 5889 * Support has been added for the following processors (GCC 5890 identifiers in parentheses): 5891 + Loongson 3A1000 (gs464) which enables loongson-mmi, 5892 loongson-ext by default. 5893 + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi, 5894 loongson-ext, loongson-ext2 by default. 5895 + Loongson 2K1000 (gs264e) which enables loongson-ext, 5896 loongson-ext2, msa by default. 5897 The GCC identifiers can be used as arguments to the -mcpu and 5898 -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments 5899 to the equivalent target attributes and pragmas. 5900 5901 [81]OpenRISC 5902 5903 * A new back end targeting OpenRISC processors has been contributed 5904 to GCC. 5905 5906 [82]S/390, System z, IBM z Systems 5907 5908 * Support for the arch13 architecture has been added. When using the 5909 -march=arch13 option, the compiler will generate code making use of 5910 the new instructions introduced with the vector enhancement 5911 facility 2 and the miscellaneous instruction extension facility 2. 5912 The -mtune=arch13 option enables arch13 specific instruction 5913 scheduling without making use of new instructions. 5914 * Builtins for the new vector instructions have been added and can be 5915 enabled using the -mzvector option. 5916 * Support for ESA architecture machines g5 and g6 is deprecated since 5917 GCC 6.1.0 and has been removed now. 5918 * When compiling with -march=z14 or higher GCC emits alignments hints 5919 on the vector load/store instructions (8 or 16 byte). 5920 * Functions now have a default alignment of 16 bytes. This helps with 5921 branch prediction effects. 5922 * -mfentry is now supported. As well as the mcount mechanism the 5923 __fentry__ is called before the function prologue. However, since 5924 just a single instruction is required to call __fentry__ the call 5925 sequence imposes a smaller overhead than mcount (4 instructions). 5926 The produced code is compatible only with newer glibc versions, 5927 which provide the __fentry__ symbol and do not clobber r0 when 5928 resolving lazily bound functions. -mfentry is only supported when 5929 generating 64-bit code and does not work with nested C functions. 5930 * The -mnop-mcount option can be used to emit NOP instructions 5931 instead of an mcount or fentry call stub. 5932 * With the -mrecord-mcount option a __mcount_loc section is generated 5933 containing pointers to each profiling call stub. This is useful for 5934 automatically patching in and out calls. 5935 5936 [83]Operating Systems 5937 5938 [84]Solaris 5939 5940 * g++ now unconditionally enables large file support when compiling 5941 32-bit code. 5942 * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has 5943 been merged from LLVM. For the moment, this only works for 32-bit 5944 code on both SPARC and x86. 5945 * An initial port of the D runtime library has been completed on 5946 Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC 5947 support is still work-in-progress. 5948 5949 [85]Windows 5950 5951 * A C++ Microsoft ABI bitfield layout bug, [86]PR87137 has been 5952 fixed. A non-field declaration could cause the current bitfield 5953 allocation unit to be completed, incorrectly placing a following 5954 bitfield into a new allocation unit. The Microsoft ABI is selected 5955 for: 5956 + Mingw targets 5957 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 5958 option is specified, or __attribute__((ms_struct)) is used 5959 + SuperH targets when the -mhitachi option is specified, or 5960 __attribute__((renesas)) is used 5961 5962 [87]Improvements for plugin authors 5963 5964 * GCC's diagnostic subsystem now has a way to logically group 5965 together related diagnostics, auto_diagnostic_group. Such 5966 diagnostics will be nested by the output of 5967 [88]-fdiagnostics-format=json. 5968 * GCC now has a set of [89]user experience guidelines for GCC, with 5969 information and advice on implementing new diagnostics. 5970 5971 Other significant improvements 5972 5973 * GCC's internal "selftest" suite now runs for C++ as well as C (in 5974 debug builds of the compiler). 5975 5976 [90]GCC 9.1 5977 5978 This is the [91]list of problem reports (PRs) from GCC's bug tracking 5979 system that are known to be fixed in the 9.1 release. This list might 5980 not be complete (that is, it is possible that some PRs that have been 5981 fixed are not listed here). 5982 5983 [92]GCC 9.2 5984 5985 This is the [93]list of problem reports (PRs) from GCC's bug tracking 5986 system that are known to be fixed in the 9.2 release. This list might 5987 not be complete (that is, it is possible that some PRs that have been 5988 fixed are not listed here). 5989 5990 [94]GCC 9.3 5991 5992 This is the [95]list of problem reports (PRs) from GCC's bug tracking 5993 system that are known to be fixed in the 9.3 release. This list might 5994 not be complete (that is, it is possible that some PRs that have been 5995 fixed are not listed here). 5996 5997 [96]GCC 9.4 5998 5999 This is the [97]list of problem reports (PRs) from GCC's bug tracking 6000 system that are known to be fixed in the 9.4 release. This list might 6001 not be complete (that is, it is possible that some PRs that have been 6002 fixed are not listed here). 6003 6004 Target Specific Changes 6005 6006 AArch64 6007 6008 * The option -moutline-atomics has been added to aid deployment of 6009 the Large System Extensions (LSE) on GNU/Linux systems built with a 6010 baseline architecture targeting Armv8-A. When the option is 6011 specified code is emitted to detect the presence of LSE 6012 instructions at run time and use them for standard atomic 6013 operations. For more information please refer to the documentation. 6014 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 6015 options are -mcpu=a64fx and -mtune=a64fx respectively. In 6016 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 6017 tunes the code for the A64FX. This includes tuning the SVE code, 6018 although by default the code is still length-agnostic and so works 6019 for all SVE implementations. Adding -msve-vector-bits=512 makes the 6020 code specific to 512-bit SVE. 6021 6022 [98]GCC 9.5 6023 6024 This is the [99]list of problem reports (PRs) from GCC's bug tracking 6025 system that are known to be fixed in the 9.5 release. This list might 6026 not be complete (that is, it is possible that some PRs that have been 6027 fixed are not listed here). 6028 6029 6030 For questions related to the use of GCC, please consult these web 6031 pages and the [100]GCC manuals. If that fails, the 6032 [101]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6033 web pages and the development of GCC are welcome on our developer 6034 list at [102]gcc (a] gcc.gnu.org. All of [103]our lists have public 6035 archives. 6036 6037 Copyright (C) [104]Free Software Foundation, Inc. Verbatim copying and 6038 distribution of this entire article is permitted in any medium, 6039 provided this notice is preserved. 6040 6041 These pages are [105]maintained by the GCC team. Last modified 6042 2026-06-10. 6043 6044 References 6045 6046 1. https://gcc.gnu.org/gcc-9/porting_to.html 6047 2. https://gcc.gnu.org/onlinedocs/index.html#current 6048 3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469 6049 4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html 6050 5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html 6051 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920 6052 7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo 6053 8. https://gcc.gnu.org/PR90361 6054 9. https://gcc.gnu.org/gcc-9/changes.html#general 6055 10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC 6056 11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching 6057 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers 6058 13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels 6059 14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 6060 15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions 6061 16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels 6062 17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops 6063 18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps 6064 19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files 6065 20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files 6066 21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info 6067 22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record 6068 23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment 6069 24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable 6070 25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability 6071 26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1 6072 27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1 6073 28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute 6074 29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion 6075 30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use 6076 31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides 6077 32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange 6078 33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam 6079 34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution 6080 35. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov 6081 36. https://gcc.gnu.org/gcc-9/changes.html#languages 6082 37. https://www.openacc.org/ 6083 38. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-9 6084 39. https://gcc.gnu.org/gcc-9/changes.html#c-family 6085 40. https://www.openmp.org/specifications/ 6086 41. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html 6087 42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector 6088 43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member 6089 44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds 6090 45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias 6091 46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow 6092 47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation 6093 48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 6094 49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 6095 50. https://gcc.gnu.org/gcc-9/changes.html#c 6096 51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value 6097 52. https://gcc.gnu.org/gcc-9/changes.html#cxx 6098 53. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy 6099 54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime 6100 55. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 6101 56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 6102 57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion 6103 58. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 6104 59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat 6105 60. https://gcc.gnu.org/gcc-9/changes.html#libstdcxx 6106 61. https://github.com/uxlfoundation/oneTBB 6107 62. https://gcc.gnu.org/gcc-9/changes.html#d 6108 63. https://gcc.gnu.org/gcc-9/changes.html#fortran 6109 64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include 6110 65. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec 6111 66. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive 6112 67. https://gcc.gnu.org/gcc-9/changes.html#go 6113 68. https://gcc.gnu.org/gcc-9/changes.html#jit 6114 69. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option 6115 70. https://gcc.gnu.org/gcc-9/changes.html#targets 6116 71. https://gcc.gnu.org/gcc-9/changes.html#arm-targets 6117 72. https://gcc.gnu.org/gcc-9/changes.html#aarch64 6118 73. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector 6119 74. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options 6120 75. https://gcc.gnu.org/gcc-9/changes.html#arm 6121 76. https://gcc.gnu.org/gcc-9/changes.html#amdgcn 6122 77. https://gcc.gnu.org/gcc-9/changes.html#arc 6123 78. https://gcc.gnu.org/gcc-9/changes.html#csky 6124 79. https://gcc.gnu.org/gcc-9/changes.html#x86 6125 80. https://gcc.gnu.org/gcc-9/changes.html#mips 6126 81. https://gcc.gnu.org/gcc-9/changes.html#or1k 6127 82. https://gcc.gnu.org/gcc-9/changes.html#s390 6128 83. https://gcc.gnu.org/gcc-9/changes.html#os 6129 84. https://gcc.gnu.org/gcc-9/changes.html#solaris 6130 85. https://gcc.gnu.org/gcc-9/changes.html#windows 6131 86. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 6132 87. https://gcc.gnu.org/gcc-9/changes.html#plugins 6133 88. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 6134 89. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html 6135 90. https://gcc.gnu.org/gcc-9/changes.html#GCC9.1 6136 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0 6137 92. https://gcc.gnu.org/gcc-9/changes.html#GCC9.2 6138 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2 6139 94. https://gcc.gnu.org/gcc-9/changes.html#GCC9.3 6140 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3 6141 96. https://gcc.gnu.org/gcc-9/changes.html#GCC9.4 6142 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4 6143 98. https://gcc.gnu.org/gcc-9/changes.html#GCC9.5 6144 99. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5 6145 100. https://gcc.gnu.org/onlinedocs/ 6146 101. mailto:gcc-help (a] gcc.gnu.org 6147 102. mailto:gcc (a] gcc.gnu.org 6148 103. https://gcc.gnu.org/lists.html 6149 104. https://www.fsf.org/ 6150 105. https://gcc.gnu.org/about.html 6151 ====================================================================== 6152 http://gcc.gnu.org/gcc-8/index.html 6153 GCC 8 Release Series 6154 6155 (This release series is no longer supported.) 6156 6157 May 14, 2021 6158 6159 The [1]GNU project and the GCC developers are pleased to announce the 6160 release of GCC 8.5. 6161 6162 This release is a bug-fix release, containing fixes for regressions in 6163 GCC 8.4 relative to previous releases of GCC. 6164 6165 Release History 6166 6167 GCC 8.5 6168 May 14, 2021 ([2]changes, [3]documentation) 6169 6170 GCC 8.4 6171 Mar 4, 2020 ([4]changes, [5]documentation) 6172 6173 GCC 8.3 6174 Feb 22, 2019 ([6]changes, [7]documentation) 6175 6176 GCC 8.2 6177 Jul 26, 2018 ([8]changes, [9]documentation) 6178 6179 GCC 8.1 6180 May 2, 2018 ([10]changes, [11]documentation) 6181 6182 References and Acknowledgements 6183 6184 GCC used to stand for the GNU C Compiler, but since the compiler 6185 supports several other languages aside from C, it now stands for the 6186 GNU Compiler Collection. 6187 6188 A list of [12]successful builds is updated as new information becomes 6189 available. 6190 6191 The GCC developers would like to thank the numerous people that have 6192 contributed new features, improvements, bug fixes, and other changes as 6193 well as test results to GCC. This [13]amazing group of volunteers is 6194 what makes GCC successful. 6195 6196 For additional information about GCC please refer to the [14]GCC 6197 project web site or contact the [15]GCC development mailing list. 6198 6199 To obtain GCC please use [16]our mirror sites or [17]our version 6200 control system. 6201 6202 6203 For questions related to the use of GCC, please consult these web 6204 pages and the [18]GCC manuals. If that fails, the 6205 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6206 web pages and the development of GCC are welcome on our developer 6207 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6208 archives. 6209 6210 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6211 distribution of this entire article is permitted in any medium, 6212 provided this notice is preserved. 6213 6214 These pages are [23]maintained by the GCC team. Last modified 6215 2024-05-30. 6216 6217 References 6218 6219 1. http://www.gnu.org/ 6220 2. https://gcc.gnu.org/gcc-8/changes.html 6221 3. https://gcc.gnu.org/onlinedocs/8.5.0/ 6222 4. https://gcc.gnu.org/gcc-8/changes.html 6223 5. https://gcc.gnu.org/onlinedocs/8.4.0/ 6224 6. https://gcc.gnu.org/gcc-8/changes.html 6225 7. https://gcc.gnu.org/onlinedocs/8.3.0/ 6226 8. https://gcc.gnu.org/gcc-8/changes.html 6227 9. https://gcc.gnu.org/onlinedocs/8.2.0/ 6228 10. https://gcc.gnu.org/gcc-8/changes.html 6229 11. https://gcc.gnu.org/onlinedocs/8.1.0/ 6230 12. https://gcc.gnu.org/gcc-8/buildstat.html 6231 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 6232 14. https://gcc.gnu.org/index.html 6233 15. mailto:gcc (a] gcc.gnu.org 6234 16. https://gcc.gnu.org/mirrors.html 6235 17. https://gcc.gnu.org/git.html 6236 18. https://gcc.gnu.org/onlinedocs/ 6237 19. mailto:gcc-help (a] gcc.gnu.org 6238 20. mailto:gcc (a] gcc.gnu.org 6239 21. https://gcc.gnu.org/lists.html 6240 22. https://www.fsf.org/ 6241 23. https://gcc.gnu.org/about.html 6242 ====================================================================== 6243 http://gcc.gnu.org/gcc-8/changes.html 6244 GCC 8 Release Series 6245 Changes, New Features, and Fixes 6246 6247 This page is a "brief" summary of some of the huge number of 6248 improvements in GCC 8. You may also want to check out our [1]Porting to 6249 GCC 8 page and the [2]full GCC documentation. 6250 6251 Caveats 6252 6253 * The default mode for C is now -std=gnu17 instead of -std=gnu11. 6254 * Support for the obsolete SDB/coff debug info format has been 6255 removed. The option -gcoff no longer does anything. 6256 * The Cilk+ extensions to the C and C++ languages have been removed. 6257 * The MPX extensions to the C and C++ languages have been deprecated 6258 and will be removed in a future release. 6259 * The extension allowing arithmetic on std::atomic<void*> and types 6260 like std::atomic<R(*)()> has been deprecated. 6261 * The non-standard C++0x std::copy_exception function was removed. 6262 std::make_exception_ptr should be used instead. 6263 * Support for the powerpc*-*-*spe* target ports which have been 6264 recently unmaintained and untested in GCC has been declared 6265 obsolete in GCC 8 as [3]announced. Unless there is activity to 6266 revive them, the next release of GCC will have their sources 6267 permanently removed. 6268 6269 [4]General Improvements 6270 6271 * Inter-procedural optimization improvements: 6272 + Reworked run-time estimation metrics leading to more realistic 6273 guesses driving inliner and cloning heuristics. 6274 + The ipa-pure-const pass is extended to propagate the malloc 6275 attribute, and the corresponding warning option 6276 -Wsuggest-attribute=malloc emits a diagnostic for functions 6277 which can be annotated with the malloc attribute. 6278 * Profile driven optimization improvements: 6279 + New infrastructure for representing profiles (both statically 6280 guessed and profile feedback) which allows propagation of 6281 additional information about the reliability of the profile. 6282 + A number of improvements in the profile updating code solving 6283 problems found by new verification code. 6284 + Static detection of code which is not executed in a valid run 6285 of the program. This includes paths which trigger undefined 6286 behavior as well as calls to functions declared with the cold 6287 attribute. Newly the noreturn attribute does not imply all 6288 effects of cold to differentiate between exit (which is 6289 noreturn) and abort (which is in addition not executed in 6290 valid runs). 6291 + -freorder-blocks-and-partition, a pass splitting function 6292 bodies into hot and cold regions, is now enabled by default at 6293 -O2 and higher for x86 and x86-64. 6294 * Link-time optimization improvements: 6295 + We have significantly improved debug information on ELF 6296 targets using DWARF by properly preserving language-specific 6297 information. This allows for example the libstdc++ 6298 pretty-printers to work with LTO optimized executables. 6299 * A new option -fcf-protection=[full|branch|return|none] is 6300 introduced to perform code instrumentation to increase program 6301 security by checking that target addresses of control-flow transfer 6302 instructions (such as indirect function call, function return, 6303 indirect jump) are valid. Currently the instrumentation is 6304 supported on x86 GNU/Linux targets only. See the user guide for 6305 further information about the option syntax and section "New 6306 Targets and Target Specific Improvements" for IA-32/x86-64 for more 6307 details. 6308 * The -gcolumn-info option is now enabled by default. It includes 6309 column information in addition to just filenames and line numbers 6310 in DWARF debugging information. 6311 * The polyhedral-based loop nest optimization pass 6312 -floop-nest-optimize has been overhauled. It's still considered 6313 experimental and may not result in any runtime improvements. 6314 * Two new classical loop nest optimization passes have been added. 6315 -floop-unroll-and-jam performs outer loop unrolling and fusing of 6316 the inner loop copies. -floop-interchange exchanges loops in a loop 6317 nest to improve data locality. Both passes are enabled by default 6318 at -O3 and above. 6319 * The classic loop nest optimization pass -ftree-loop-distribution 6320 has been improved and enabled by default at -O3 and above. It 6321 supports loop nest distribution in some restricted scenarios; it 6322 also supports cancellable innermost loop distribution with loop 6323 versioning under run-time alias checks. 6324 * The new option -fstack-clash-protection causes the compiler to 6325 insert probes whenever stack space is allocated statically or 6326 dynamically to reliably detect stack overflows and thus mitigate 6327 the attack vector that relies on jumping over a stack guard page as 6328 provided by the operating system. 6329 * A new pragma GCC unroll has been implemented in the C family of 6330 languages, as well as Fortran and Ada, so as to make it possible 6331 for the user to have a finer-grained control over the loop 6332 unrolling optimization. 6333 * GCC has been enhanced to detect more instances of meaningless or 6334 mutually exclusive attribute specifications and handle such 6335 conflicts more consistently. Mutually exclusive attribute 6336 specifications are ignored with a warning regardless of whether 6337 they appear on the same declaration or on distinct declarations of 6338 the same entity. For example, because the noreturn attribute on the 6339 second declaration below is mutually exclusive with the malloc 6340 attribute on the first, it is ignored and a warning is issued. 6341 > 6342 void* __attribute__ ((malloc)) f (unsigned); 6343 void* __attribute__ ((noreturn)) f (unsigned); 6344 6345 warning: ignoring attribute 'noreturn' because it conflicts with attribute 6346 'malloc' [-Wattributes] 6347 * The gcov tool can distinguish functions that begin on a same line 6348 in a source file. This can be a different template instantiation or 6349 a class constructor: 6350 6351 File 'ins.C' 6352 Lines executed:100.00% of 8 6353 Creating 'ins.C.gcov' 6354 6355 -: 0:Source:ins.C 6356 -: 0:Graph:ins.gcno 6357 -: 0:Data:ins.gcda 6358 -: 0:Runs:1 6359 -: 0:Programs:1 6360 -: 1:template<class T> 6361 -: 2:class Foo 6362 -: 3:{ 6363 -: 4: public: 6364 2: 5: Foo(): b (1000) {} 6365 ------------------ 6366 Foo<char>::Foo(): 6367 1: 5: Foo(): b (1000) {} 6368 ------------------ 6369 Foo<int>::Foo(): 6370 1: 5: Foo(): b (1000) {} 6371 ------------------ 6372 2: 6: void inc () { b++; } 6373 ------------------ 6374 Foo<char>::inc(): 6375 1: 6: void inc () { b++; } 6376 ------------------ 6377 Foo<int>::inc(): 6378 1: 6: void inc () { b++; } 6379 ------------------ 6380 -: 7: 6381 -: 8: private: 6382 -: 9: int b; 6383 -: 10:}; 6384 -: 11: 6385 1: 12:int main(int argc, char **argv) 6386 -: 13:{ 6387 1: 14: Foo<int> a; 6388 1: 15: Foo<char> b; 6389 -: 16: 6390 1: 17: a.inc (); 6391 1: 18: b.inc (); 6392 1: 19:} 6393 6394 * The gcov tool has more accurate numbers for execution of lines in a 6395 source file. 6396 * The gcov tool can use TERM colors to provide more readable output. 6397 * AddressSanitizer gained a new pair of sanitization options, 6398 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 6399 warn about subtraction (or comparison) of pointers that point to a 6400 different memory object: 6401 6402 int 6403 main () 6404 { 6405 /* Heap allocated memory. */ 6406 char *heap1 = (char *)__builtin_malloc (42); 6407 char *heap2 = (char *)__builtin_malloc (42); 6408 if (heap1 > heap2) 6409 return 1; 6410 6411 return 0; 6412 } 6413 6414 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 6415 00050 6416 #0 0x40070f in main /tmp/pointer-compare.c:7 6417 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6418 #2 0x400629 in _start (/tmp/a.out+0x400629) 6419 6420 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 6421 00000003a) 6422 allocated by thread T0 here: 6423 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 6424 _malloc_linux.cc:86 6425 #1 0x4006ea in main /tmp/pointer-compare.c:5 6426 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6427 6428 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 6429 00000007a) 6430 allocated by thread T0 here: 6431 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 6432 _malloc_linux.cc:86 6433 #1 0x4006f8 in main /tmp/pointer-compare.c:6 6434 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6435 6436 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 6437 6438 * The store merging pass has been enhanced to handle bit-fields and 6439 not just constant stores, but also data copying from adjacent 6440 memory locations into other adjacent memory locations, including 6441 bitwise logical operations on the data. The pass can also handle 6442 byte swapping into memory locations. 6443 * The undefined behavior sanitizer gained two new options included in 6444 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 6445 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 6446 builtins, and -fsanitize=pointer-overflow which performs cheap run 6447 time tests for pointer wrapping. 6448 * A new attribute no_sanitize can be applied to functions to instruct 6449 the compiler not to do sanitization of the options provided as 6450 arguments to the attribute. Acceptable values for no_sanitize match 6451 those acceptable by the -fsanitize command-line option. 6452 6453 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 6454 f () { /* Do something. */; } 6455 6456 [5]New Languages and Language specific improvements 6457 6458 [6]Ada 6459 6460 * For its internal exception handling used on the host for error 6461 recovery in the front-end, the compiler now relies on the native 6462 exception handling mechanism of the host platform, which should be 6463 more efficient than the former mechanism. 6464 6465 [7]BRIG (HSAIL) 6466 6467 In this release cycle, the focus for the BRIGFE was on stabilization 6468 and performance improvements. Also a couple of completely new features 6469 were added. 6470 * Improved support for function and module scope group segment 6471 variables. PRM specs define function and module scope group segment 6472 variables as an experimental feature. However, PRM test suite uses 6473 them. Now group segment is handled by separate book keeping of 6474 module scope and function (kernel) offsets. Each function has a 6475 "frame" in the group segment offset to which is given as an 6476 argument, similar to traditional call stack frame handling. 6477 * Reduce the number of type conversions due to the untyped HSAIL 6478 registers. Instead of always representing the HSAIL's untyped 6479 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 6480 code and builds the register variables as a type used the most when 6481 storing or reading data to/from each register. This reduces the 6482 number of total casts which cannot be always optimized away. 6483 * Support for BRIG_KIND_NONE directives. 6484 * Made -O3 the default optimization level for BRIGFE. 6485 * Fixed illegal addresses generated from address expressions which 6486 refer only to offset 0. 6487 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 6488 mode, the offset is treated as 32-bit unless it's in global, 6489 read-only or kernarg address space. 6490 * Fixed a crash caused sometimes by calls with more than 4 arguments. 6491 * Fixed a mis-execution issue with kernels that have both unexpanded 6492 ID functions and calls to subfunctions. 6493 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 6494 avoid illegal optimizations. 6495 * Ensure per WI copies of private variables are aligned correctly. 6496 * libhsail-rt: Assume the host runtime allocates the work group 6497 memory. 6498 6499 [8]C family 6500 6501 * New command-line options have been added for the C and C++ 6502 compilers: 6503 + [9]-Wmultistatement-macros warns about unsafe macros expanding 6504 to multiple statements used as a body of a statement such as 6505 if, else, while, switch, or for. 6506 + [10]-Wstringop-truncation warns for calls to bounded string 6507 manipulation functions such as strncat, strncpy, and stpncpy 6508 that might either truncate the copied string or leave the 6509 destination unchanged. For example, the following call to 6510 strncat is diagnosed because it appends just three of the four 6511 characters from the source string. 6512 void append (char *buf, size_t bufsize) 6513 { 6514 strncat (buf, ".txt", 3); 6515 } 6516 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 6517 Wstringop-truncation] 6518 Similarly, in the following example, the call to strncpy 6519 specifies the size of the destination buffer as the bound. If 6520 the length of the source string is equal to or greater than 6521 this size the result of the copy will not be NUL-terminated. 6522 Therefore, the call is also diagnosed. To avoid the warning, 6523 specify sizeof buf - 1 as the bound and set the last element 6524 of the buffer to NUL. 6525 void copy (const char *s) 6526 { 6527 char buf[80]; 6528 strncpy (buf, s, sizeof buf); 6529 6530 } 6531 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 6532 tion] 6533 The -Wstringop-truncation option is included in -Wall. 6534 Note that due to GCC bug [11]82944, defining strncat, strncpy, 6535 or stpncpy as a macro in a system header as some 6536 implementations do, suppresses the warning. 6537 + [12]-Wif-not-aligned controls warnings issued in response to 6538 invalid uses of objects declared with attribute 6539 [13]warn_if_not_aligned. 6540 The -Wif-not-aligned option is included in -Wall. 6541 + [14]-Wmissing-attributes warns when a declaration of a 6542 function is missing one or more attributes that a related 6543 function is declared with and whose absence may adversely 6544 affect the correctness or efficiency of generated code. For 6545 example, in C++, the warning is issued when an explicit 6546 specialization of a primary template declared with attribute 6547 alloc_align, alloc_size, assume_aligned, format, format_arg, 6548 malloc, or nonnull is declared without it. Attributes 6549 deprecated, error, and warning suppress the warning. 6550 The -Wmissing-attributes option is included in -Wall. 6551 + [15]-Wpacked-not-aligned warns when a struct or union declared 6552 with attribute packed defines a member with an explicitly 6553 specified alignment greater than 1. Such a member will wind up 6554 under-aligned. For example, a warning will be issued for the 6555 definition of struct A in the following: 6556 struct __attribute__ ((aligned (8))) 6557 S8 { char a[8]; }; 6558 6559 struct __attribute__ ((packed)) A 6560 { 6561 struct S8 s8; 6562 }; 6563 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 6564 The -Wpacked-not-aligned option is included in -Wall. 6565 + -Wcast-function-type warns when a function pointer is cast to 6566 an incompatible function pointer. This warning is enabled by 6567 -Wextra. 6568 + -Wsizeof-pointer-div warns for suspicious divisions of the 6569 size of a pointer by the size of the elements it points to, 6570 which looks like the usual way to compute the array size but 6571 won't work out correctly with pointers. This warning is 6572 enabled by -Wall. 6573 + -Wcast-align=strict warns whenever a pointer is cast such that 6574 the required alignment of the target is increased. For 6575 example, warn if a char * is cast to an int * regardless of 6576 the target machine. 6577 + -fprofile-abs-path creates absolute path names in the .gcno 6578 files. This allows gcov to find the correct sources in 6579 projects where compilations occur with different working 6580 directories. 6581 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 6582 signed integer overflow is now undefined by default at all 6583 optimization levels. Using -fsanitize=signed-integer-overflow is 6584 now the preferred way to audit code, -Wstrict-overflow is 6585 deprecated. 6586 * The [16]-Warray-bounds option has been improved to detect more 6587 instances of out-of-bounds array indices and pointer offsets. For 6588 example, negative or excessive indices into flexible array members 6589 and string literals are detected. 6590 * The [17]-Wrestrict option introduced in GCC 7 has been enhanced to 6591 detect many more instances of overlapping accesses to objects via 6592 restrict-qualified arguments to standard memory and string 6593 manipulation functions such as memcpy and strcpy. For example, the 6594 strcpy call in the function below attempts to truncate the string 6595 by replacing its initial characters with the last four. However, 6596 because the function writes the terminating NUL into a[4], the 6597 copies overlap and the call is diagnosed. 6598 void f (void) 6599 { 6600 char a[] = "abcd1234"; 6601 strcpy (a, a + 4); 6602 6603 } 6604 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 6605 4 [-Wrestrict] 6606 The -Wrestrict option is included in -Wall. 6607 * Several optimizer enhancements have enabled improvements to the 6608 [18]-Wformat-overflow and [19]-Wformat-truncation options. The 6609 warnings detect more instances of buffer overflow and truncation 6610 than in GCC 7 and are better at avoiding certain kinds of false 6611 positives. 6612 * When reporting mismatching argument types at a function call, the C 6613 and C++ compilers now underline both the argument and the pertinent 6614 parameter in the declaration. 6615 $ gcc arg-type-mismatch.cc 6616 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 6617 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 6618 [-fpermissive] 6619 return callee(first, second, third); 6620 ^~~~~~ 6621 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 6622 onst char*, float)' 6623 extern int callee(int one, const char *two, float three); 6624 ~~~~~~~~~~~~^~~ 6625 6626 * When reporting on unrecognized identifiers, the C and C++ compilers 6627 will now emit fix-it hints suggesting #include directives for 6628 various headers in the C and C++ standard libraries. 6629 $ gcc incomplete.c 6630 incomplete.c: In function 'test': 6631 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 6632 return NULL; 6633 ^~~~ 6634 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 6635 t to '#include <stddef.h>'? 6636 incomplete.c:1:1: 6637 +#include <stddef.h> 6638 const char *test(void) 6639 incomplete.c:3:10: 6640 return NULL; 6641 ^~~~ 6642 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 6643 ch function it appears in 6644 6645 $ gcc incomplete.cc 6646 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 6647 std::string s("hello world"); 6648 ^~~~~~ 6649 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 6650 forget to '#include <string>'? 6651 +#include <string> 6652 std::string s("hello world"); 6653 ^~~ 6654 6655 * The C and C++ compilers now use more intuitive locations when 6656 reporting on missing semicolons, and offer fix-it hints: 6657 $ gcc t.c 6658 t.c: In function 'test': 6659 t.c:3:12: error: expected ';' before '}' token 6660 return 42 6661 ^ 6662 ; 6663 } 6664 ~ 6665 6666 * When reporting on missing '}' and ')' tokens, the C and C++ 6667 compilers will now highlight the corresponding '{' and '(' token, 6668 issuing a 'note' if it's on a separate line: 6669 $ gcc unclosed.c 6670 unclosed.c: In function 'log_when_out_of_range': 6671 unclosed.c:12:50: error: expected ')' before '{' token 6672 && (temperature < MIN || temperature > MAX) { 6673 ^~ 6674 ) 6675 unclosed.c:11:6: note: to match this '(' 6676 if (logging_enabled && check_range () 6677 ^ 6678 6679 or highlighting it directly if it's on the same line: 6680 $ gcc unclosed-2.c 6681 unclosed-2.c: In function 'test': 6682 unclosed-2.c:8:45: error: expected ')' before '{' token 6683 if (temperature < MIN || temperature > MAX { 6684 ~ ^~ 6685 ) 6686 6687 They will also emit fix-it hints. 6688 6689 [20]C 6690 6691 * New options -std=c17, to select support for the 2018 edition of the 6692 ISO C standard (__STDC_VERSION__ == 201710L), and -std=gnu17, for 6693 C17 with GNU extensions. 6694 * The default mode has been changed to -std=gnu17. 6695 6696 [21]C++ 6697 6698 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 6699 convention, which changes the ABI for some uncommon code: 6700 + Passing an empty class as an argument now takes up no space on 6701 x86_64, as required by the psABI. 6702 + Passing or returning a class with only deleted copy and move 6703 constructors now uses the same calling convention as a class 6704 with a non-trivial copy or move constructor. This only affects 6705 C++17 mode, as in earlier standards passing or returning such 6706 a class was impossible. 6707 + WARNING: In GCC 8.1 the second change mistakenly also affects 6708 classes with a deleted copy constructor and defaulted trivial 6709 move constructor (bug [22]c++/86094). This issue is fixed in 6710 GCC 8.2 (-fabi-version=13). 6711 You can test whether these changes affect your code with -Wabi=11 6712 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 6713 problematic for your project, the GCC 7 ABI can be selected with 6714 -fabi-version=11. 6715 * The value of the C++11 alignof operator has been corrected to match 6716 C _Alignof (minimum alignment) rather than GNU __alignof__ 6717 (preferred alignment); on ia32 targets this means that 6718 alignof(double) is now 4 rather than 8. Code that wants the 6719 preferred alignment should use __alignof__ instead. 6720 * New command-line options have been added for the C++ compiler to 6721 control warnings: 6722 + [23]-Wclass-memaccess warns when objects of non-trivial class 6723 types are manipulated in potentially unsafe ways by raw memory 6724 functions such as memcpy, or realloc. The warning helps detect 6725 calls that bypass user-defined constructors or copy-assignment 6726 operators, corrupt virtual table pointers, data members of 6727 const-qualified types or references, or member pointers. The 6728 warning also detects calls that would bypass access controls 6729 to data members. For example, a call such as: 6730 memcpy (&std::cout, &std::cerr, sizeof std::cout); 6731 results in 6732 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 6733 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 6734 o trivial copy-assignment [-Wclass-memaccess] 6735 The -Wclass-memaccess option is included in -Wall. 6736 * The C++ front end has experimental support for some of the upcoming 6737 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 6738 including designated initializers, default member initializers for 6739 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 6740 lambda [=, this] captures, etc. For a full list of new features, 6741 see [24]the C++ status page. 6742 * When reporting on attempts to access private fields of a class or 6743 struct, the C++ compiler will now offer fix-it hints showing how to 6744 use an accessor function to get at the field in question, if one 6745 exists. 6746 $ gcc accessor.cc 6747 accessor.cc: In function 'void test(foo*)': 6748 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 6749 if (ptr->m_ratio >= 0.5) 6750 ^~~~~~~ 6751 accessor.cc:7:10: note: declared private here 6752 double m_ratio; 6753 ^~~~~~~ 6754 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 6755 foo::get_ratio() const' 6756 if (ptr->m_ratio >= 0.5) 6757 ^~~~~~~ 6758 get_ratio() 6759 6760 * The C++ compiler can now give you a hint if you use a macro before 6761 it was defined (e.g. if you mess up the order of your #include 6762 directives): 6763 $ gcc ordering.cc 6764 ordering.cc:2:24: error: expected ';' at end of member declaration 6765 virtual void clone() const OVERRIDE { } 6766 ^~~~~ 6767 ; 6768 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 6769 virtual void clone() const OVERRIDE { } 6770 ^~~~~~~~ 6771 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 6772 In file included from ordering.cc:5: 6773 c++11-compat.h:2: note: it was later defined here 6774 #define OVERRIDE override 6775 6776 6777 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 6778 you when you can use a static_cast, const_cast, or 6779 reinterpret_cast. 6780 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 6781 old-style-cast-fixits.cc: In function 'void test(void*)': 6782 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 6783 -Wold-style-cast] 6784 foo *f = (foo *)ptr; 6785 ^~~ 6786 ---------- 6787 static_cast<foo *> (ptr) 6788 6789 * When reporting on problems within extern "C" linkage 6790 specifications, the C++ compiler will now display the location of 6791 the start of the extern "C". 6792 $ gcc -c extern-c.cc 6793 extern-c.cc:3:1: error: template with C linkage 6794 template <typename T> void test (void); 6795 ^~~~~~~~ 6796 In file included from extern-c.cc:1: 6797 unclosed.h:1:1: note: 'extern "C"' linkage started here 6798 extern "C" { 6799 ^~~~~~~~~~ 6800 extern-c.cc:3:39: error: expected '}' at end of input 6801 template <typename T> void test (void); 6802 ^ 6803 In file included from extern-c.cc:1: 6804 unclosed.h:1:12: note: to match this '{' 6805 extern "C" { 6806 ^ 6807 6808 * When reporting on mismatching template types, the C++ compiler will 6809 now use color to highlight the mismatching parts of the template, 6810 and will elide the parameters that are common between two 6811 mismatching templates, printing [...] instead: 6812 $ gcc templates.cc 6813 templates.cc: In function 'void test()': 6814 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 6815 e>' to 'vector<int>' 6816 fn_1(vector<double> ()); 6817 ^~~~~~~~~~~~~~~~~ 6818 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 6819 ,double>' to 'map<[...],int>' 6820 fn_2(map<int, double>()); 6821 ^~~~~~~~~~~~~~~~~~ 6822 6823 Those [...] elided parameters can be seen using -fno-elide-type: 6824 $ gcc templates.cc -fno-elide-type 6825 templates.cc: In function 'void test()': 6826 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 6827 e>' to 'vector<int>' 6828 fn_1(vector<double> ()); 6829 ^~~~~~~~~~~~~~~~~ 6830 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 6831 ouble>' to 'map<int,int>' 6832 fn_2(map<int, double>()); 6833 ^~~~~~~~~~~~~~~~~~ 6834 6835 The C++ compiler has also gained an option 6836 -fdiagnostics-show-template-tree which visualizes such mismatching 6837 templates in a hierarchical form: 6838 $ gcc templates-2.cc -fdiagnostics-show-template-tree 6839 templates-2.cc: In function 'void test()': 6840 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 6841 ble>' to 'vector<int>' 6842 vector< 6843 [double != int]> 6844 fn_1(vector<double> ()); 6845 ^~~~~~~~~~~~~~~~~ 6846 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 6847 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 6848 ap<[...],vector<float>>,vector<float>>' 6849 map< 6850 map< 6851 [...], 6852 vector< 6853 [double != float]>>, 6854 vector< 6855 [double != float]>> 6856 fn_2(map<map<int, vector<double>>, vector<double>> ()); 6857 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 6858 6859 which again works with -fno-elide-type: 6860 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 6861 templates-2.cc: In function 'void test()': 6862 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 6863 ble>' to 'vector<int>' 6864 vector< 6865 [double != int]> 6866 fn_1(vector<double> ()); 6867 ^~~~~~~~~~~~~~~~~ 6868 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 6869 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 6870 <int,vector<float>>,vector<float>>' 6871 map< 6872 map< 6873 int, 6874 vector< 6875 [double != float]>>, 6876 vector< 6877 [double != float]>> 6878 fn_2(map<map<int, vector<double>>, vector<double>> ()); 6879 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 6880 6881 * Flowing off the end of a non-void function is considered 6882 unreachable and may be subject to optimization on that basis. As a 6883 result of this change, -Wreturn-type warnings are enabled by 6884 default for C++. 6885 6886 [25]Runtime Library (libstdc++) 6887 6888 * Improved experimental support for C++17, including the following 6889 features: 6890 + Deduction guides to support class template argument deduction. 6891 + std::filesystem implementation. 6892 + std::char_traits<char> and std::char_traits<wchar_t> are 6893 usable in constant expressions. 6894 + std::to_chars and std::from_chars (for integers only, not for 6895 floating point types). 6896 * Experimental support for C++2a: std::to_address (thanks to Glen 6897 Fernandes) and std::endian. 6898 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 6899 entropy count for the random device, if known (thanks to Xi 6900 Ruoyao). 6901 * Support for std::experimental::source_location. 6902 * AddressSanitizer integration for std::vector, detecting 6903 out-of-range accesses to the unused capacity of a vector. 6904 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 6905 Mathematical Special Functions. 6906 6907 [26]Fortran 6908 6909 * The main version of libfortran has been changed to 5. 6910 * Parameterized derived types, a major feature of Fortran 2003, have 6911 been implemented. 6912 * Partial support is provided for Fortran 2018 teams, which are 6913 hierarchical subsets of images that execute independently of other 6914 image subsets. 6915 * The maximum rank for arrays has been increased to 15, conforming to 6916 the Fortran 2008 standard. 6917 * Transformational intrinsics are now fully supported in 6918 initialization expressions. 6919 * New flag -fc-prototypes to write C prototypes for BIND(C) 6920 procedures and variables. 6921 * If -fmax-stack-var-size is honored if given together with -Ofast, 6922 -fstack-arrays is no longer set in that case. 6923 * New options -fdefault-real-16 and -fdefault-real-10 to control the 6924 default kind of REAL variables. 6925 * A warning is now issued if an array subscript inside a DO loop 6926 could lead to an out-of-bounds-access. The new option 6927 -Wdo-subscript, enabled by -Wextra, warns about this even if the 6928 compiler can not prove that the code will be executed. 6929 * The Fortran front end now attempts to interchange loops if it is 6930 deemed profitable. So far, this is restricted to FORALL and DO 6931 CONCURRENT statements with multiple indices. This behavior be 6932 controlled with the new flag -ffrontend-loop-interchange, which is 6933 enabled with optimization by default. The 6934 -Wfrontend-loop-interchange option warns about such occurrences. 6935 * When an actual argument contains too few elements for a dummy 6936 argument, an error is now issued. The -std=legacy option can be 6937 used to still compile such code. 6938 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 6939 integers, making records larger than 2GiB possible. 6940 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 6941 effect. The record length for preconnected units is now larger than 6942 any practical limit, same as for sequential access units opened 6943 without an explicit RECL= specifier. 6944 * Character variables longer than HUGE(0) elements are now possible 6945 on 64-bit targets. Note that this changes the procedure call ABI 6946 for all procedures with character arguments on 64-bit targets, as 6947 the type of the hidden character length argument has changed. The 6948 hidden character length argument is now of type INTEGER(C_SIZE_T). 6949 * Partial support is provided for Fortran 2018 teams, which are 6950 hierarchical subsets of images that execute independently of other 6951 image subsets. 6952 6953 [27]Go 6954 6955 * GCC 8 provides a complete implementation of the Go 1.10.1 user 6956 packages. 6957 * The garbage collector is now fully concurrent. As before, values 6958 stored on the stack are scanned conservatively, but value stored in 6959 the heap are scanned precisely. 6960 * Escape analysis is fully implemented and enabled by default in the 6961 Go front end. This significantly reduces the number of heap 6962 allocations by allocating values on the stack instead. 6963 6964 [28]libgccjit 6965 6966 The libgccjit API gained four new entry points: 6967 * [29]gcc_jit_type_get_vector and 6968 * [30]gcc_jit_context_new_rvalue_from_vector for working with 6969 vectors, 6970 * [31]gcc_jit_type_get_aligned 6971 * [32]gcc_jit_function_get_address 6972 6973 The C code generated by [33]gcc_jit_context_dump_reproducer_to_file is 6974 now easier-to-read. 6975 6976 [34]New Targets and Target Specific Improvements 6977 6978 [35]AArch64 6979 6980 * The Armv8.4-A architecture is now supported. It can be used by 6981 specifying the -march=armv8.4-a option. 6982 * The Dot Product instructions are now supported as an optional 6983 extension to the Armv8.2-A architecture and newer and are mandatory 6984 on Armv8.4-A. The extension can be used by specifying the +dotprod 6985 architecture extension. E.g. -march=armv8.2-a+dotprod. 6986 * The Armv8-A +crypto extension has now been split into two 6987 extensions for finer grained control: 6988 + +aes which contains the Armv8-A AES crytographic instructions. 6989 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 6990 instructions. 6991 Using +crypto will now enable these two extensions. 6992 * New Armv8.4-A FP16 Floating Point Multiplication Variant 6993 instructions have been added. These instructions are mandatory in 6994 Armv8.4-A but available as an optional extension to Armv8.2-A and 6995 Armv8.3-A. The new extension can be used by specifying the +fp16fml 6996 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 6997 the instructions can be enabled by specifying +fp16. 6998 * New cryptographic instructions have been added as optional 6999 extensions to Armv8.2-A and newer. These instructions can be 7000 enabled with: 7001 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 7002 implies +sha2. 7003 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 7004 * The Scalable Vector Extension (SVE) is now supported as an optional 7005 extension to the Armv8.2-A architecture and newer. This support 7006 includes automatic vectorization with SVE instructions, but it does 7007 not yet include the SVE Arm C Language Extensions (ACLE). It can be 7008 enabled by specifying the +sve architecture extension (for example, 7009 -march=armv8.2-a+sve). By default, the generated code works with 7010 all vector lengths, but it can be made specific to N-bit vectors 7011 using -msve-vector-bits=N. 7012 * Support has been added for the following processors (GCC 7013 identifiers in parentheses): 7014 + Arm Cortex-A75 (cortex-a75). 7015 + Arm Cortex-A55 (cortex-a55). 7016 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 7017 (cortex-a75.cortex-a55). 7018 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7019 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 7020 arguments to the equivalent target attributes and pragmas. 7021 7022 [36]ARC 7023 7024 * Added support for: 7025 + Fast interrupts. 7026 + Naked functions. 7027 + aux variable attributes. 7028 + uncached type qualifier. 7029 + Secure functions via sjli instruction. 7030 * New exception handling implementation. 7031 * Revamped trampoline implementation. 7032 * Refactored small data feature implementation, controlled via the -G 7033 command-line option. 7034 * New support for reduced register set ARC architecture 7035 configurations, controlled via the -mrf16 command-line option. 7036 * Refurbished and improved support for zero overhead loops. 7037 Introduced -mlpc-width command-line option to control the width of 7038 the lp_count register. 7039 7040 [37]ARM 7041 7042 * The -mfpu option now takes a new option setting of -mfpu=auto. When 7043 set to this the floating-point and SIMD settings are derived from 7044 the settings of the -mcpu or -march options. The internal CPU 7045 configurations have been updated with information about the 7046 permitted floating-point configurations supported. See the user 7047 guide for further information about the extended option syntax for 7048 controlling architectural extensions via the -march option. 7049 -mfpu=auto is now the default setting unless the compiler has been 7050 configured with an explicit --with-fpu option. 7051 * The -march and -mcpu options now accept optional extensions to the 7052 architecture or CPU option, allowing the user to enable or disable 7053 any such extensions supported by that architecture or CPU such as 7054 (but not limited to) floating-point and AdvancedSIMD. For example: 7055 the option -mcpu=cortex-a53+nofp will generate code for the 7056 Cortex-A53 processor with no floating-point support. This, in 7057 combination with the new -mfpu=auto option, provides a 7058 straightforward way of specifying a valid build target through a 7059 single -mcpu or -march option. The -mtune option accepts the same 7060 arguments as -mcpu but only the CPU name has an effect on tuning. 7061 The architecture extensions do not have any effect. For details of 7062 what extensions a particular architecture or CPU option supports 7063 please refer to the [38]documentation. 7064 * The -mstructure-size-boundary option has been deprecated and will 7065 be removed in a future release. 7066 * The default link behavior for Armv6 and Armv7-R targets has been 7067 changed to produce BE8 format when generating big-endian images. A 7068 new flag -mbe32 can be used to force the linker to produce legacy 7069 BE32 format images. There is no change of behavior for Armv6-M and 7070 other Armv7 or later targets: these already defaulted to BE8 7071 format. This change brings GCC into alignment with other compilers 7072 for the ARM architecture. 7073 * The Armv8-R architecture is now supported. It can be used by 7074 specifying the -march=armv8-r option. 7075 * The Armv8.3-A architecture is now supported. It can be used by 7076 specifying the -march=armv8.3-a option. 7077 * The Armv8.4-A architecture is now supported. It can be used by 7078 specifying the -march=armv8.4-a option. 7079 * The Dot Product instructions are now supported as an optional 7080 extension to the Armv8.2-A architecture and newer and are mandatory 7081 on Armv8.4-A. The extension can be used by specifying the +dotprod 7082 architecture extension. E.g. -march=armv8.2-a+dotprod. 7083 * Support for setting extensions and architectures using the GCC 7084 target pragma and attribute has been added. It can be used by 7085 specifying #pragma GCC target ("arch=..."), #pragma GCC target 7086 ("+extension"), __attribute__((target("arch=..."))) or 7087 __attribute__((target("+extension"))). 7088 * New Armv8.4-A FP16 Floating Point Multiplication Variant 7089 instructions have been added. These instructions are mandatory in 7090 Armv8.4-A but available as an optional extension to Armv8.2-A and 7091 Armv8.3-A. The new extension can be used by specifying the +fp16fml 7092 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 7093 the instructions can be enabled by specifying +fp16. 7094 * Support has been added for the following processors (GCC 7095 identifiers in parentheses): 7096 + Arm Cortex-A75 (cortex-a75). 7097 + Arm Cortex-A55 (cortex-a55). 7098 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 7099 (cortex-a75.cortex-a55). 7100 + Arm Cortex-R52 for Armv8-R (cortex-r52). 7101 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7102 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 7103 arguments to the equivalent target attributes and pragmas. 7104 7105 [39]AVR 7106 7107 * The AVR port now supports the following XMEGA-like devices: 7108 7109 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 7110 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 7111 ATtiny3214, ATtiny3216, ATtiny3217 7112 The new devices are listed under [40]-mmcu=avrxmega3. 7113 + These devices see flash memory in the RAM address space, so 7114 that features like PROGMEM and __flash are not needed any more 7115 (as opposed to other AVR families for which read-only data 7116 will be located in RAM except special, non-standard features 7117 are used to locate and access such data). This requires that 7118 the compiler is used with Binutils 2.29 or newer so that 7119 [41]read-only data will be located in flash memory. 7120 + A new command-line option -mshort-calls is supported. This 7121 option is used internally for multilib selection of the 7122 avrxmega3 variants. It is not an optimization option. Do not 7123 set it by hand. 7124 * The compiler now generates [42]efficient interrupt service routine 7125 (ISR) prologues and epilogues. This is achieved by using the new 7126 [43]AVR pseudo instruction __gcc_isr which is supported and 7127 resolved by the GNU assembler. 7128 + As the __gcc_isr pseudo-instruction will be resolved by the 7129 assembler, inline assembly is transparent to the process. This 7130 means that when inline assembly uses an instruction like INC 7131 that clobbers the condition code, then the assembler will 7132 detect this and generate an appropriate ISR prologue / 7133 epilogue chunk to save / restore SREG as needed. 7134 + A new command-line option -mno-gas-isr-prologues disables the 7135 generation of the __gcc_isr pseudo instruction. Any non-naked 7136 ISR will save and restore SREG, tmp_reg and zero_reg, no 7137 matter whether the respective register is clobbered or used. 7138 + The feature is turned on per default for all optimization 7139 levels except for -O0 and -Og. It is explicitly enabled by 7140 means of option -mgas-isr-prologues. 7141 + Support has been added for a new [44]AVR function attribute 7142 no_gccisr. It can be used to disable __gcc_isr pseudo 7143 instruction generation for individual ISRs. 7144 + This optimization is only available if GCC is configured with 7145 GNU Binutils 2.29 or newer; or at least with a version of 7146 Binutils that implements feature [45]PR21683. 7147 * The compiler no more saves / restores registers in main; the effect 7148 is the same as if attribute OS_task was specified for main. This 7149 optimization can be switched off by the new command-line option 7150 -mno-main-is-OS_task. 7151 7152 [46]IA-32/x86-64 7153 7154 * The x86 port now supports the naked function attribute. 7155 * Better tuning for znver1 and Intel Core based CPUs. 7156 * Vectorization cost metrics has been reworked leading to significant 7157 improvements on some benchmarks. 7158 * GCC now supports the Intel CPU named Cannonlake through 7159 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 7160 and SHA ISA extensions. 7161 * GCC now supports the Intel CPU named Icelake through 7162 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 7163 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 7164 ISA extensions. 7165 * GCC now supports the Intel Control-flow Enforcement Technology 7166 (CET) extension through -fcf-protection option. 7167 7168 [47]NDS32 7169 7170 * New command-line options -mext-perf, -mext-perf2, and -mext-string 7171 have been added for performance extension instructions. 7172 7173 [48]Nios II 7174 7175 * The Nios II back end has been improved to generate better-optimized 7176 code. Changes include switching to LRA, more accurate cost models, 7177 and more compact code for addressing static variables. 7178 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 7179 added. 7180 * The stack-smashing protection options are now enabled on Nios II. 7181 7182 [49]PA-RISC 7183 7184 * The default call ABI on 32-bit linux has been changed from callee 7185 copies to caller copies. This affects objects larger than eight 7186 bytes passed by value. The goal is to improve compatibility with 7187 x86 and resolve issues with OpenMP. 7188 * Other PA-RISC targets are unchanged. 7189 7190 [50]PowerPC / PowerPC64 / RS6000 7191 7192 * The PowerPC SPE support is split off to a separate powerpcspe port. 7193 The separate port is deprecated and might be removed in a future 7194 release. 7195 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 7196 powerpc*-*-linux*paired*) is deprecated and will be removed in a 7197 future release. 7198 * The Xilinx floating point support (-mxilinx-fpu, 7199 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 7200 release. 7201 * Support for using big-endian AltiVec intrinsics on a little-endian 7202 target (-maltivec=be) is deprecated and will be removed in a future 7203 release. 7204 7205 [51]Tile 7206 7207 * The TILE-Gx port is deprecated and will be removed in a future 7208 release. 7209 7210 [52]Operating Systems 7211 7212 [53]Windows 7213 7214 * GCC on Microsoft Windows can now be configured via 7215 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 7216 specific behavior for GCC itself with regards to supporting the 7217 wildcard character. Prior versions of GCC would follow the 7218 configuration of the MinGW runtime. This behavior can still be 7219 obtained by not using the above options or by using 7220 --enable-mingw-wildcard=platform. 7221 7222 [54]Improvements for plugin authors 7223 7224 * Plugins can now register a callback hook for when comments are 7225 encountered by the C and C++ compilers, e.g. allowing for plugins 7226 to handle documentation markup in code comments. 7227 * The gdbinit support script for debugging GCC now has a 7228 break-on-diagnostic command, providing an easy way to trigger a 7229 breakpoint whenever a diagnostic is emitted. 7230 * The API for creating fix-it hints now supports newlines, and for 7231 emitting mutually incompatible fix-it hints for one diagnostic. 7232 7233 [55]GCC 8.1 7234 7235 This is the [56]list of problem reports (PRs) from GCC's bug tracking 7236 system that are known to be fixed in the 8.1 release. This list might 7237 not be complete (that is, it is possible that some PRs that have been 7238 fixed are not listed here). 7239 7240 [57]GCC 8.2 7241 7242 This is the [58]list of problem reports (PRs) from GCC's bug tracking 7243 system that are known to be fixed in the 8.2 release. This list might 7244 not be complete (that is, it is possible that some PRs that have been 7245 fixed are not listed here). 7246 7247 General Improvements 7248 7249 * Fixed LTO link-time performance problems caused by an overflow in 7250 the partitioning algorithm while building large binaries. 7251 7252 Language Specific Changes 7253 7254 C++ 7255 7256 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 7257 returning of classes with a deleted copy constructor and defaulted 7258 trivial move constructor (bug [59]c++/86094). GCC 8.2 introduces 7259 -fabi-version=13 and makes it the default, ABI incompatibilities 7260 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [60]C++ 7261 changes for more details. 7262 7263 Target Specific Changes 7264 7265 IA-32/x86-64 7266 7267 * -mtune=native performance regression [61]PR84413 on Intel Skylake 7268 processors has been fixed. 7269 7270 [62]GCC 8.3 7271 7272 This is the [63]list of problem reports (PRs) from GCC's bug tracking 7273 system that are known to be fixed in the 8.3 release. This list might 7274 not be complete (that is, it is possible that some PRs that have been 7275 fixed are not listed here). 7276 7277 Windows 7278 7279 * A C++ Microsoft ABI bitfield layout bug, [64]PR87137 has been 7280 fixed. A non-field declaration could cause the current bitfield 7281 allocation unit to be completed, incorrectly placing a following 7282 bitfield into a new allocation unit. The Microsoft ABI is selected 7283 for: 7284 + Mingw targets 7285 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 7286 option is specified, or __attribute__((ms_struct)) is used 7287 + SuperH targets when the -mhitachi option is specified, or 7288 __attribute__((renesas)) is used 7289 GCC 8 introduced additional cases of this defect, but rather than 7290 resolve only those regressions, we decided to resolve all the cases 7291 of this defect in single change. 7292 7293 [65]GCC 8.4 7294 7295 This is the [66]list of problem reports (PRs) from GCC's bug tracking 7296 system that are known to be fixed in the 8.4 release. This list might 7297 not be complete (that is, it is possible that some PRs that have been 7298 fixed are not listed here). 7299 7300 [67]GCC 8.5 7301 7302 This is the [68]list of problem reports (PRs) from GCC's bug tracking 7303 system that are known to be fixed in the 8.5 release. This list might 7304 not be complete (that is, it is possible that some PRs that have been 7305 fixed are not listed here). 7306 7307 Target Specific Changes 7308 7309 AArch64 7310 7311 * The option -moutline-atomics has been added to aid deployment of 7312 the Large System Extensions (LSE) on GNU/Linux systems built with a 7313 baseline architecture targeting Armv8-A. When the option is 7314 specified code is emitted to detect the presence of LSE 7315 instructions at run time and use them for standard atomic 7316 operations. For more information please refer to the documentation. 7317 7318 7319 For questions related to the use of GCC, please consult these web 7320 pages and the [69]GCC manuals. If that fails, the 7321 [70]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7322 web pages and the development of GCC are welcome on our developer 7323 list at [71]gcc (a] gcc.gnu.org. All of [72]our lists have public 7324 archives. 7325 7326 Copyright (C) [73]Free Software Foundation, Inc. Verbatim copying and 7327 distribution of this entire article is permitted in any medium, 7328 provided this notice is preserved. 7329 7330 These pages are [74]maintained by the GCC team. Last modified 7331 2025-01-31. 7332 7333 References 7334 7335 1. https://gcc.gnu.org/gcc-8/porting_to.html 7336 2. https://gcc.gnu.org/onlinedocs/index.html#current 7337 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 7338 4. https://gcc.gnu.org/gcc-8/changes.html#general 7339 5. https://gcc.gnu.org/gcc-8/changes.html#languages 7340 6. https://gcc.gnu.org/gcc-8/changes.html#ada 7341 7. https://gcc.gnu.org/gcc-8/changes.html#brig 7342 8. https://gcc.gnu.org/gcc-8/changes.html#c-family 7343 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 7344 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 7345 11. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 7346 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 7347 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 7348 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 7349 15. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 7350 16. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 7351 17. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 7352 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 7353 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 7354 20. https://gcc.gnu.org/gcc-8/changes.html#c 7355 21. https://gcc.gnu.org/gcc-8/changes.html#cxx 7356 22. https://gcc.gnu.org/PR86094 7357 23. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 7358 24. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 7359 25. https://gcc.gnu.org/gcc-8/changes.html#libstdcxx 7360 26. https://gcc.gnu.org/gcc-8/changes.html#fortran 7361 27. https://gcc.gnu.org/gcc-8/changes.html#go 7362 28. https://gcc.gnu.org/gcc-8/changes.html#jit 7363 29. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 7364 30. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 7365 31. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 7366 32. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 7367 33. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 7368 34. https://gcc.gnu.org/gcc-8/changes.html#targets 7369 35. https://gcc.gnu.org/gcc-8/changes.html#aarch64 7370 36. https://gcc.gnu.org/gcc-8/changes.html#arc 7371 37. https://gcc.gnu.org/gcc-8/changes.html#arm 7372 38. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 7373 39. https://gcc.gnu.org/gcc-8/changes.html#avr 7374 40. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 7375 41. https://sourceware.org/PR21472 7376 42. https://gcc.gnu.org/PR20296 7377 43. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 7378 44. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 7379 45. https://sourceware.org/PR21683 7380 46. https://gcc.gnu.org/gcc-8/changes.html#x86 7381 47. https://gcc.gnu.org/gcc-8/changes.html#nds32 7382 48. https://gcc.gnu.org/gcc-8/changes.html#nios2 7383 49. https://gcc.gnu.org/gcc-8/changes.html#hppa 7384 50. https://gcc.gnu.org/gcc-8/changes.html#powerpc 7385 51. https://gcc.gnu.org/gcc-8/changes.html#Tile 7386 52. https://gcc.gnu.org/gcc-8/changes.html#os 7387 53. https://gcc.gnu.org/gcc-8/changes.html#windows 7388 54. https://gcc.gnu.org/gcc-8/changes.html#plugins 7389 55. https://gcc.gnu.org/gcc-8/changes.html#GCC8.1 7390 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 7391 57. https://gcc.gnu.org/gcc-8/changes.html#GCC8.2 7392 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 7393 59. https://gcc.gnu.org/PR86094 7394 60. https://gcc.gnu.org/gcc-8/changes.html#cxx 7395 61. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 7396 62. https://gcc.gnu.org/gcc-8/changes.html#GCC8.3 7397 63. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 7398 64. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 7399 65. https://gcc.gnu.org/gcc-8/changes.html#GCC8.4 7400 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 7401 67. https://gcc.gnu.org/gcc-8/changes.html#GCC8.5 7402 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5 7403 69. https://gcc.gnu.org/onlinedocs/ 7404 70. mailto:gcc-help (a] gcc.gnu.org 7405 71. mailto:gcc (a] gcc.gnu.org 7406 72. https://gcc.gnu.org/lists.html 7407 73. https://www.fsf.org/ 7408 74. https://gcc.gnu.org/about.html 7409 ====================================================================== 7410 http://gcc.gnu.org/gcc-7/index.html 7411 GCC 7 Release Series 7412 7413 (This release series is no longer supported.) 7414 7415 Nov 14, 2019 7416 7417 The [1]GNU project and the GCC developers are pleased to announce the 7418 release of GCC 7.5. 7419 7420 This release is a bug-fix release, containing fixes for regressions in 7421 GCC 7.4 relative to previous releases of GCC. 7422 7423 Release History 7424 7425 GCC 7.5 7426 Nov 14, 2019 ([2]changes, [3]documentation) 7427 7428 GCC 7.4 7429 Dec 6, 2018 ([4]changes, [5]documentation) 7430 7431 GCC 7.3 7432 Jan 25, 2018 ([6]changes, [7]documentation) 7433 7434 GCC 7.2 7435 Aug 14, 2017 ([8]changes, [9]documentation) 7436 7437 GCC 7.1 7438 May 2, 2017 ([10]changes, [11]documentation) 7439 7440 References and Acknowledgements 7441 7442 GCC used to stand for the GNU C Compiler, but since the compiler 7443 supports several other languages aside from C, it now stands for the 7444 GNU Compiler Collection. 7445 7446 A list of [12]successful builds is updated as new information becomes 7447 available. 7448 7449 The GCC developers would like to thank the numerous people that have 7450 contributed new features, improvements, bug fixes, and other changes as 7451 well as test results to GCC. This [13]amazing group of volunteers is 7452 what makes GCC successful. 7453 7454 For additional information about GCC please refer to the [14]GCC 7455 project web site or contact the [15]GCC development mailing list. 7456 7457 To obtain GCC please use [16]our mirror sites or [17]our version 7458 control system. 7459 7460 7461 For questions related to the use of GCC, please consult these web 7462 pages and the [18]GCC manuals. If that fails, the 7463 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7464 web pages and the development of GCC are welcome on our developer 7465 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 7466 archives. 7467 7468 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 7469 distribution of this entire article is permitted in any medium, 7470 provided this notice is preserved. 7471 7472 These pages are [23]maintained by the GCC team. Last modified 7473 2024-05-30. 7474 7475 References 7476 7477 1. http://www.gnu.org/ 7478 2. https://gcc.gnu.org/gcc-7/changes.html 7479 3. https://gcc.gnu.org/onlinedocs/7.5.0/ 7480 4. https://gcc.gnu.org/gcc-7/changes.html 7481 5. https://gcc.gnu.org/onlinedocs/7.4.0/ 7482 6. https://gcc.gnu.org/gcc-7/changes.html 7483 7. https://gcc.gnu.org/onlinedocs/7.3.0/ 7484 8. https://gcc.gnu.org/gcc-7/changes.html 7485 9. https://gcc.gnu.org/onlinedocs/7.2.0/ 7486 10. https://gcc.gnu.org/gcc-7/changes.html 7487 11. https://gcc.gnu.org/onlinedocs/7.1.0/ 7488 12. https://gcc.gnu.org/gcc-7/buildstat.html 7489 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7490 14. https://gcc.gnu.org/index.html 7491 15. mailto:gcc (a] gcc.gnu.org 7492 16. https://gcc.gnu.org/mirrors.html 7493 17. https://gcc.gnu.org/git.html 7494 18. https://gcc.gnu.org/onlinedocs/ 7495 19. mailto:gcc-help (a] gcc.gnu.org 7496 20. mailto:gcc (a] gcc.gnu.org 7497 21. https://gcc.gnu.org/lists.html 7498 22. https://www.fsf.org/ 7499 23. https://gcc.gnu.org/about.html 7500 ====================================================================== 7501 http://gcc.gnu.org/gcc-7/changes.html 7502 GCC 7 Release Series 7503 Changes, New Features, and Fixes 7504 7505 This page is a brief summary of some of the huge number of improvements 7506 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 7507 the [2]full GCC documentation. 7508 7509 Caveats 7510 7511 * GCC now uses [3]LRA (a new local register allocator) by default for 7512 new targets. 7513 * The non-standard C++0x type traits has_trivial_default_constructor, 7514 has_trivial_copy_constructor and has_trivial_copy_assign have been 7515 removed. 7516 * The libstdc++ [4]Profile Mode has been deprecated and will be 7517 removed in a future version. 7518 * The Cilk+ extensions to the C and C++ languages have been 7519 deprecated. 7520 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 7521 affects conformance to the procedure call standard (AAPCS) has been 7522 fixed. The bug affects some C++ code where class objects are passed 7523 by value to functions and could result in incorrect or inconsistent 7524 code being generated. This is an ABI change. If the option -Wpsabi 7525 is enabled (on by default) the compiler will emit a diagnostic note 7526 for code that might be affected. 7527 7528 [6]General Optimizer Improvements 7529 7530 * GCC 7 can determine the return value or range of return values of 7531 some calls to the sprintf family of functions and make it available 7532 to other optimization passes. Some calls to the snprintf function 7533 with a zero size argument can be folded into constants. This 7534 optimization is included in -O1 and can be selectively controlled 7535 by the -fprintf-return-value option. 7536 * A new store merging pass has been added. It merges constant stores 7537 to adjacent memory locations into fewer, wider, stores. It is 7538 enabled by the -fstore-merging option and at the -O2 optimization 7539 level or higher (and -Os). 7540 * A new code hoisting optimization has been added to the partial 7541 redundancy elimination pass. It attempts to move evaluation of 7542 expressions executed on all paths to the function exit as early as 7543 possible. This primarily helps improve code size, but can improve 7544 the speed of the generated code as well. It is enabled by the 7545 -fcode-hoisting option and at the -O2 optimization level or higher 7546 (and -Os). 7547 * A new interprocedural bitwise constant propagation optimization has 7548 been added, which propagates knowledge about which bits of 7549 variables are known to be zero (including pointer alignment 7550 information) across the call graph. It is enabled by the 7551 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 7552 at the -O2 optimization level and higher (and -Os). This 7553 optimization supersedes interprocedural alignment propagation of 7554 GCC 6, and therefore the option -fipa-cp-alignment is now 7555 deprecated and ignored. 7556 * A new interprocedural value range propagation optimization has been 7557 added, which propagates integral range information across the call 7558 graph when variable values can be proven to be within those ranges. 7559 It is enabled by the -fipa-vrp option and at the -O2 optimization 7560 level and higher (and -Os). 7561 * A new loop splitting optimization pass has been added. Certain 7562 loops which contain a condition that is always true on one side of 7563 the iteration space and always false on the other are split into 7564 two loops, such that each of the two new loops iterates on just one 7565 side of the iteration space and the condition does not need to be 7566 checked inside of the loop. It is enabled by the -fsplit-loops 7567 option and at the -O3 optimization level or higher. 7568 * The shrink-wrapping optimization can now separate portions of 7569 prologues and epilogues to improve performance if some of the work 7570 done traditionally by prologues and epilogues is not needed on 7571 certain paths. This is controlled by the -fshrink-wrap-separate 7572 option, enabled by default. It requires target support, which is 7573 currently only implemented in the PowerPC and AArch64 ports. 7574 * AddressSanitizer gained a new sanitization option, 7575 -fsanitize-address-use-after-scope, which enables sanitization of 7576 variables whose address is taken and used after a scope where the 7577 variable is defined: 7578 7579 int 7580 main (int argc, char **argv) 7581 { 7582 char *ptr; 7583 { 7584 char my_char; 7585 ptr = &my_char; 7586 } 7587 7588 *ptr = 123; 7589 return *ptr; 7590 } 7591 7592 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 7593 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 7594 WRITE of size 1 at 0x7fffb8dba990 thread T0 7595 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 7596 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 7597 #2 0x400739 in _start (/tmp/a.out+0x400739) 7598 7599 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 7600 #0 0x40067f in main /tmp/use-after-scope-1.c:3 7601 7602 This frame has 1 object(s): 7603 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 7604 7605 The option is enabled by default with -fsanitize=address and 7606 disabled by default with -fsanitize=kernel-address. Compared to the 7607 LLVM compiler, where the option already exists, the implementation 7608 in the GCC compiler has some improvements and advantages: 7609 + Complex uses of gotos and case labels are properly handled and 7610 should not report any false positive or false negatives. 7611 + C++ temporaries are sanitized. 7612 + Sanitization can handle invalid memory stores that are 7613 optimized out by the LLVM compiler when optimization is 7614 enabled. 7615 * The -fsanitize=signed-integer-overflow suboption of the 7616 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 7617 on arithmetic operations with generic vectors. 7618 * Version 5 of the DWARF debugging information standard is supported 7619 through the -gdwarf-5 option. The DWARF version 4 debugging 7620 information remains the default until consumers of debugging 7621 information are adjusted. 7622 7623 [7]New Languages and Language specific improvements 7624 7625 OpenACC support in C, C++, and Fortran continues to be maintained and 7626 improved. See the [8]OpenACC and [9]Offloading wiki pages for further 7627 information. 7628 7629 [10]Ada 7630 7631 * On mainstream native platforms, Ada programs no longer require the 7632 stack to be made executable in order to run properly. 7633 7634 [11]BRIG (HSAIL) 7635 7636 Support for processing BRIG 1.0 files was added in this release. BRIG 7637 is a binary format for HSAIL (Heterogeneous System Architecture 7638 Intermediate Language). The BRIG front end can be used for implementing 7639 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 7640 GCC-supported targets. An implementation of an HSAIL runtime library, 7641 libhsail-rt is also included. 7642 7643 [12]C family 7644 7645 * New command-line options have been added for the C and C++ 7646 compilers: 7647 + -Wimplicit-fallthrough warns when a switch case falls through. 7648 This warning has five different levels. The compiler is able 7649 to parse a wide range of fallthrough comments, depending on 7650 the level. It also handles control-flow statements, such as 7651 ifs. It's possible to suppress the warning by either adding a 7652 fallthrough comment, or by using a null statement: 7653 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 7654 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 7655 is enabled by -Wextra. 7656 + -Wpointer-compare warns when a pointer is compared with a zero 7657 character constant. Such code is now invalid in C++11 and GCC 7658 rejects it. This warning is enabled by default. 7659 + -Wduplicated-branches warns when an if-else has identical 7660 branches. 7661 + -Wrestrict warns when an argument passed to a 7662 restrict-qualified parameter aliases with another argument. 7663 + -Wmemset-elt-size warns for memset calls, when the first 7664 argument references an array, and the third argument is a 7665 number equal to the number of elements of the array, but not 7666 the size of the array. This warning is enabled by -Wall. 7667 + -Wint-in-bool-context warns about suspicious uses of integer 7668 values where boolean values are expected. This warning is 7669 enabled by -Wall. 7670 + -Wswitch-unreachable warns when a switch statement has 7671 statements between the controlling expression and the first 7672 case label which will never be executed. This warning is 7673 enabled by default. 7674 + -Wexpansion-to-defined warns when defined is used outside #if. 7675 This warning is enabled by -Wextra or -Wpedantic. 7676 + -Wregister warns about uses of the register storage specifier. 7677 In C++17 this keyword has been removed and for C++17 this is a 7678 pedantic warning enabled by default. The warning is not 7679 emitted for the GNU Explicit Register Variables extension. 7680 + -Wvla-larger-than=N warns about unbounded uses of 7681 variable-length arrays, and about bounded uses of 7682 variable-length arrays whose bound can be larger than N bytes. 7683 + -Wduplicate-decl-specifier warns when a declaration has 7684 duplicate const, volatile, restrict or _Atomic specifier. This 7685 warning is enabled by -Wall. 7686 * GCC 6's C and C++ front ends were able to offer suggestions for 7687 misspelled field names: 7688 7689 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 7690 you mean 'color'? 7691 return ptr->colour; 7692 ^~~~~~ 7693 7694 GCC 7 greatly expands the scope of these suggestions. Firstly, it 7695 adds fix-it hints to such suggestions: 7696 7697 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 7698 you mean 'color'? 7699 return ptr->colour; 7700 ^~~~~~ 7701 color 7702 7703 The suggestions now cover many other things, such as misspelled 7704 function names: 7705 7706 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 7707 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 7708 ion] 7709 gtk_widget_showall (w); 7710 ^~~~~~~~~~~~~~~~~~ 7711 gtk_widget_show_all 7712 7713 misspelled macro names and enum values: 7714 7715 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 7716 ction); did you mean 'MAX_ITEMS'? 7717 int array[MAX_ITEM]; 7718 ^~~~~~~~ 7719 MAX_ITEMS 7720 7721 misspelled type names: 7722 7723 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 7724 gned'? 7725 void test (singed char e); 7726 ^~~~~~ 7727 signed 7728 7729 and, in the C front end, named initializers: 7730 7731 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 7732 '? 7733 struct s test = { .colour = 3 }; 7734 ^~~~~~ 7735 color 7736 7737 * The preprocessor can now offer suggestions for misspelled 7738 directives, e.g.: 7739 7740 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 7741 #endfi 7742 ^~~~~ 7743 endif 7744 7745 * Warnings about format strings now underline the pertinent part of 7746 the string, and can offer suggested fixes. In some cases, the 7747 pertinent argument is underlined. 7748 7749 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 7750 nt 3 has type 'int' [-Wformat=] 7751 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 7752 ~^ ~~~~~ 7753 %d 7754 7755 * The new -Wdangling-else command-line option has been split out of 7756 -Wparentheses and warns about dangling else. 7757 * The -Wshadow warning has been split into three variants. 7758 -Wshadow=global warns for any shadowing. This is the default when 7759 using -Wshadow without any argument. -Wshadow=local only warns for 7760 a local variable shadowing another local variable or parameter. 7761 -Wshadow=compatible-local only warns for a local variable shadowing 7762 another local variable or parameter whose type is compatible (in 7763 C++ compatible means that the type of the shadowing variable can be 7764 converted to that of the shadowed variable). 7765 The following example shows the different kinds of shadow warnings: 7766 7767 enum operation { add, count }; 7768 struct container { int nr; }; 7769 7770 int 7771 container_count (struct container c, int count) 7772 { 7773 int r = 0; 7774 for (int count = 0; count > 0; count--) 7775 { 7776 struct container count = c; 7777 r += count.nr; 7778 } 7779 return r; 7780 } 7781 7782 -Wshadow=compatible-local will warn for the parameter being 7783 shadowed with the same type: 7784 7785 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 7786 compatible-local] 7787 for (int count = 0; count > 0; count--) 7788 ^~~~~ 7789 warn-test.c:5:42: note: shadowed declaration is here 7790 container_count (struct container c, int count) 7791 ^~~~~ 7792 7793 -Wshadow=local will warn for the above and for the shadowed 7794 declaration with incompatible type: 7795 7796 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 7797 hadow=local] 7798 struct container count = c; 7799 ^~~~~ 7800 warn-test.c:8:12: note: shadowed declaration is here 7801 for (int count = 0; count > 0; count--) 7802 ^~~~~ 7803 7804 -Wshadow=global will warn for all of the above and the shadowing of 7805 the global declaration: 7806 7807 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 7808 -Wshadow] 7809 container_count (struct container c, int count) 7810 ^~~~~ 7811 warn-test.c:1:23: note: shadowed declaration is here 7812 enum operation { add, count }; 7813 ^~~~~ 7814 7815 * GCC 7 contains a number of enhancements that help detect buffer 7816 overflow and other forms of invalid memory accesses. 7817 + The -Walloc-size-larger-than=size option detects calls to 7818 standard and user-defined memory allocation functions 7819 decorated with attribute alloc_size whose argument exceeds the 7820 specified size (PTRDIFF_MAX by default). The option also 7821 detects arithmetic overflow in the computation of the size in 7822 two-argument allocation functions like calloc where the total 7823 size is the product of the two arguments. Since calls with an 7824 excessive size cannot succeed they are typically the result of 7825 programming errors. Such bugs have been known to be the source 7826 of security vulnerabilities and a target of exploits. 7827 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 7828 For example, the following call to malloc incorrectly tries to 7829 avoid passing a negative argument to the function and instead 7830 ends up unconditionally invoking it with an argument less than 7831 or equal to zero. Since after conversion to the type of the 7832 argument of the function (size_t) a negative argument results 7833 in a value in excess of the maximum PTRDIFF_MAX the call is 7834 diagnosed. 7835 7836 void* f (int n) 7837 { 7838 return malloc (n > 0 ? 0 : n); 7839 } 7840 7841 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 7842 147483647 [-Walloc-size-larger-than=] 7843 7844 + The -Walloc-zero option detects calls to standard and 7845 user-defined memory allocation functions decorated with 7846 attribute alloc_size with a zero argument. -Walloc-zero is not 7847 included in either -Wall or -Wextra and must be explicitly 7848 enabled. 7849 + The -Walloca option detects all calls to the alloca function 7850 in the program. -Walloca is not included in either -Wall or 7851 -Wextra and must be explicitly enabled. 7852 + The -Walloca-larger-than=size option detects calls to the 7853 alloca function whose argument either may exceed the specified 7854 size, or that is not known to be sufficiently constrained to 7855 avoid exceeding it. -Walloca-larger-than is not included in 7856 either -Wall or -Wextra and must be explicitly enabled. 7857 For example, compiling the following snippet with 7858 -Walloca-larger-than=1024 results in a warning because even 7859 though the code appears to call alloca only with sizes of 1kb 7860 and less, since n is signed, a negative value would result in 7861 a call to the function well in excess of the limit. 7862 7863 void f (int n) 7864 { 7865 char *d; 7866 if (n < 1025) 7867 d = alloca (n); 7868 else 7869 d = malloc (n); 7870 7871 } 7872 7873 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 7874 ong unsigned int' [-Walloca-larger-than=] 7875 7876 In contrast, a call to alloca that isn't bounded at all such 7877 as in the following function will elicit the warning below 7878 regardless of the size argument to the option. 7879 7880 void f (size_t n) 7881 { 7882 char *d = alloca (n); 7883 7884 } 7885 7886 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 7887 7888 + The -Wformat-overflow=level option detects certain and likely 7889 buffer overflow in calls to the sprintf family of formatted 7890 output functions. Although the option is enabled even without 7891 optimization it works best with -O2 and higher. 7892 For example, in the following snippet the call to sprintf is 7893 diagnosed because even though its output has been constrained 7894 using the modulo operation it could result in as many as three 7895 bytes if mday were negative. The solution is to either 7896 allocate a larger buffer or make sure the argument is not 7897 negative, for example by changing mday's type to unsigned or 7898 by making the type of the second operand of the modulo 7899 expression unsigned: 100U. 7900 7901 void* f (int mday) 7902 { 7903 char *buf = malloc (3); 7904 sprintf (buf, "%02i", mday % 100); 7905 return buf; 7906 } 7907 7908 warning: 'sprintf may write a terminating nul past the end of the destination [- 7909 Wformat-overflow=] 7910 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 7911 7912 + The -Wformat-truncation=level option detects certain and 7913 likely output truncation in calls to the snprintf family of 7914 formatted output functions. -Wformat-truncation=1 is included 7915 in -Wall and enabled without optimization but works best with 7916 -O2 and higher. 7917 For example, the following function attempts to format an 7918 integer between 0 and 255 in hexadecimal, including the 0x 7919 prefix, into a buffer of four characters. But since the 7920 function must always terminate output by the null character 7921 ('\0') such a buffer is only big enough to fit just one digit 7922 plus the prefix. Therefore the snprintf call is diagnosed. To 7923 avoid the warning either use a bigger buffer or handle the 7924 function's return value which indicates whether or not its 7925 output has been truncated. 7926 7927 void f (unsigned x) 7928 { 7929 char d[4]; 7930 snprintf (d, sizeof d, "%#02x", x & 0xff); 7931 7932 } 7933 7934 warning: 'snprintf' output may be truncated before the last format character [-W 7935 format-truncation=] 7936 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 7937 7938 + The -Wnonnull option has been enhanced to detect a broader set 7939 of cases of passing null pointers to functions that expect a 7940 non-null argument (those decorated with attribute nonnull). By 7941 taking advantage of optimizations the option can detect many 7942 more cases of the problem than in prior GCC versions. 7943 + The -Wstringop-overflow=type option detects buffer overflow in 7944 calls to string handling functions like memcpy and strcpy. The 7945 option relies on [13]Object Size Checking and has an effect 7946 similar to defining the _FORTIFY_SOURCE macro. 7947 -Wstringop-overflow=2 is enabled by default. 7948 For example, in the following snippet, because the call to 7949 strncat specifies a maximum that allows the function to write 7950 past the end of the destination, it is diagnosed. To correct 7951 the problem and avoid the overflow the function should be 7952 called with a size of at most sizeof d - strlen(d) - 1. 7953 7954 void f (const char *fname) 7955 { 7956 char d[8]; 7957 strncpy (d, "/tmp/", sizeof d); 7958 strncat (d, fname, sizeof d); 7959 7960 } 7961 7962 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 7963 ow=] 7964 7965 * The <limits.h> header provided by GCC defines macros such as 7966 INT_WIDTH for the width in bits of integer types, if 7967 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 7968 included. The <stdint.h> header defines such macros as SIZE_WIDTH 7969 and INTMAX_WIDTH for the width of some standard typedef names for 7970 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 7971 before the header is included; note that GCC's implementation of 7972 this header is only used for freestanding compilations, not hosted 7973 compilations, on most systems. These macros come from ISO/IEC TS 7974 18661-1:2014. 7975 * The <float.h> header provided by GCC defines the macro 7976 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 7977 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 7978 included. This represents the number of decimal digits for which 7979 conversions between decimal character strings and binary formats, 7980 in both directions, are correctly rounded, and currently has the 7981 value of UINTMAX_MAX on all systems, reflecting that GCC's 7982 compile-time conversions are correctly rounded for any number of 7983 digits. 7984 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 7985 __builtin_mul_overflow_p built-in functions have been added. These 7986 work similarly to their siblings without the _p suffix, but do not 7987 actually store the result of the arithmetics anywhere, just return 7988 whether the operation would overflow. Calls to these built-ins with 7989 integer constant arguments evaluate to integer constants 7990 expressions. 7991 For example, in the following, c is assigned the result of a * b 7992 only if the multiplication does not overflow, otherwise it is 7993 assigned the value zero. The multiplication is performed at 7994 compile-time and without triggering a -Woverflow warning. 7995 7996 enum { 7997 a = 12345678, 7998 b = 87654321, 7999 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 8000 }; 8001 8002 [14]C 8003 8004 * The C front end now supports type names _FloatN for floating-point 8005 types with IEEE interchange formats and _FloatNx for floating-point 8006 types with IEEE extended formats. These type names come from 8007 ISO/IEC TS 18661-3:2015. 8008 The set of types supported depends on the target for which GCC is 8009 configured. Most targets support _Float32, _Float32x and _Float64. 8010 _Float128 is supported on targets where IEEE binary128 encoding was 8011 already supported as long double or __float128. _Float64x is 8012 supported on targets where a type with either binary128 or Intel 8013 extended precision format is available. 8014 Constants with these types are supported using suffixes fN, FN, fNx 8015 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 8016 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 8017 defined before it is included. 8018 These new types are always distinct from each other and from float, 8019 double and long double, even if they have the same encoding. 8020 Complex types such as _Complex _Float128 are also supported. 8021 Type-generic built-in functions such as __builtin_isinf support the 8022 new types, and the following type-specific built-in functions have 8023 versions (suffixed fN or fNx) for the new types: 8024 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 8025 __builtin_inf, __builtin_nan, __builtin_nans. 8026 * Compilation with -fopenmp is now compatible with the C11 _Atomic 8027 keyword. 8028 8029 [15]C++ 8030 8031 * The C++ front end has experimental support for all of the current 8032 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 8033 constexpr, class template argument deduction, auto template 8034 parameters, and structured bindings. For a full list of new 8035 features, see [16]the C++ status page. 8036 * C++17 support for new of over-aligned types can be enabled in other 8037 modes with the -faligned-new flag. 8038 * The C++17 evaluation order requirements can be selected in other 8039 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 8040 with -fno-strong-eval-order. 8041 * The default semantics of inherited constructors has changed in all 8042 modes, following [17]P0136. Essentially, overload resolution 8043 happens as if calling the inherited constructor directly, and the 8044 compiler fills in construction of the other bases and members as 8045 needed. Most uses should not need any changes. The old behavior can 8046 be restored with -fno-new-inheriting-ctors, or -fabi-version less 8047 than 11. 8048 * The resolution of DR 150 on matching of template template 8049 parameters, allowing default template arguments to make a template 8050 match a parameter, is currently enabled by default in C++17 mode 8051 only. The default can be overridden with -f{no-,}new-ttp-matching. 8052 * The C++ front end will now provide fix-it hints for some missing 8053 semicolons, allowing for automatic fixes by IDEs: 8054 8055 test.cc:4:11: error: expected ';' after class definition 8056 class a {} 8057 ^ 8058 ; 8059 8060 * -Waligned-new has been added to the C++ front end. It warns about 8061 new of type with extended alignment without -faligned-new. 8062 8063 [18]Runtime Library (libstdc++) 8064 8065 * The type of exception thrown by iostreams, std::ios_base::failure, 8066 now uses the [19]cxx11 ABI. 8067 * Experimental support for C++17, including the following new 8068 features: 8069 + std::string_view; 8070 + std::any, std::optional, and std::variant; 8071 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 8072 invoke_result; 8073 + std::is_swappable, and std::is_nothrow_swappable; 8074 + std::apply, and std::make_from_tuple; 8075 + std::void_t, std::bool_constant, std::conjunction, 8076 std::disjunction, and std::negation; 8077 + Variable templates for type traits; 8078 + Mathematical Special Functions; 8079 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 8080 std::chrono::abs; 8081 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 8082 + std::scoped_lock, std::shared_mutex, 8083 std::atomic<T>::is_always_lock_free; 8084 + std::sample, std::default_searcher, std::boyer_moore_searcher 8085 and std::boyer_moore_horspool_searcher; 8086 + Extraction and re-insertion of map and set nodes, try_emplace 8087 members for maps, and functions for accessing containers 8088 std::size, std::empty, and std::data; 8089 + std::shared_ptr support for arrays, 8090 std::shared_ptr<T>::weak_type, 8091 std::enable_shared_from_this<T>::weak_from_this(), and 8092 std::owner_less<void>; 8093 + std::byte; 8094 + std::as_const, std::not_fn, 8095 std::has_unique_object_representations, constexpr 8096 std::addressof. 8097 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 8098 Voutilainen for work on the C++17 support. 8099 * A new power-of-two rehashing policy for use with the _Hashtable 8100 internals, thanks to Franois Dumont. 8101 8102 [20]Fortran 8103 8104 * Support for a number of extensions for compatibility with legacy 8105 code with new flags: 8106 + -fdec-structure Support for DEC STRUCTURE and UNION 8107 + -fdec-intrinsic-ints Support for new integer intrinsics with 8108 B/I/J/K prefixes such as BABS, JIAND... 8109 + -fdec-math Support for additional math intrinsics, including 8110 COTAN and degree-valued trigonometric functions such as TAND, 8111 ASIND... 8112 + -fdec Enable the -fdec-* family of extensions. 8113 * New flag -finit-derived to allow default initialization of 8114 derived-type variables. 8115 * Improved DO loops with step equal to 1 or -1, generates faster code 8116 without a loop preheader. A new warning, -Wundefined-do-loop, warns 8117 when a loop iterates either to HUGE(i) (with step equal to 1), or 8118 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 8119 at run time with -fcheck=do enabled: 8120 8121 program test 8122 implicit none 8123 integer(1) :: i 8124 do i = -HUGE(i)+10, -HUGE(i)-1, -1 8125 print *, i 8126 end do 8127 end program test 8128 8129 At line 8 of file do_check_12.f90 8130 Fortran runtime error: Loop iterates infinitely 8131 8132 * Version 4.5 of the [21]OpenMP specification is now partially 8133 supported in the Fortran compiler; the largest missing item is 8134 structure element mapping. 8135 * User-defined derived-type input/output (UDTIO) is added. 8136 * Derived type coarrays with allocatable and pointer components are 8137 partially supported. 8138 * Non-constant stop codes and error stop codes (Fortran 2015 8139 feature). 8140 * Derived types with allocatable components of recursive type. 8141 * Intrinsic assignment to polymorphic variables. 8142 * Improved submodule support. 8143 * Improved diagnostics (polymorphic results in pure functions). 8144 * Coarray: Support for failed images (Fortan 2015 feature). 8145 8146 [22]Go 8147 8148 * GCC 7 provides a complete implementation of the Go 1.8.1 user 8149 packages. 8150 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 8151 conservative and less concurrent. 8152 * Escape analysis is available for experimental use via the 8153 -fgo-optimize-allocs option. The -fgo-debug-escape prints 8154 information useful for debugging escape analysis choices. 8155 8156 [23]Java (GCJ) 8157 8158 The GCC Java front end and associated libjava runtime library have been 8159 removed from GCC. 8160 8161 [24]libgccjit 8162 8163 The libgccjit API gained support for marking calls as requiring 8164 tail-call optimization via a new entry point: 8165 [25]gcc_jit_rvalue_set_bool_require_tail_call. 8166 8167 libgccjit performs numerous checks at the API boundary, but if these 8168 succeed, it previously ignored errors and other diagnostics emitted 8169 within the core of GCC, and treated the compile of a gcc_jit_context as 8170 having succeeded. As of GCC 7 it now ensures that if any diagnostics 8171 are emitted, they are visible from the libgccjit API, and that the the 8172 context is flagged as having failed. 8173 8174 [26]New Targets and Target Specific Improvements 8175 8176 [27]AArch64 8177 8178 * GCC has been updated to the latest revision of the procedure call 8179 standard (AAPCS64) to provide support for parameter passing when 8180 data types have been over-aligned. 8181 * The ARMv8.3-A architecture is now supported. It can be used by 8182 specifying the -march=armv8.3-a option. 8183 * The option -msign-return-address= is supported to enable return 8184 address protection using ARMv8.3-A Pointer Authentication 8185 Extensions. For more information on the arguments accepted by this 8186 option, please refer to [28]AArch64-Options. 8187 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 8188 Extensions are now supported. They can be used by specifying the 8189 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 8190 Floating-Point Extensions introduce new half-precision data 8191 processing floating-point instructions. 8192 * Support has been added for the following processors (GCC 8193 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 8194 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 8195 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 8196 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 8197 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 8198 The GCC identifiers can be used as arguments to the -mcpu or -mtune 8199 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 8200 arguments to the equivalent target attributes and pragmas. 8201 8202 [29]ARC 8203 8204 * Added support for ARC HS and ARC EM processors. 8205 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 8206 * Added support for NPS400 ARC700 based CPUs. 8207 * Thread Local Storage is now supported by ARC CPUs. 8208 * Fixed errors for ARC600 when using 32x16 multiplier option. 8209 * Fixed PIE for ARC CPUs. 8210 * New CPU templates are supported via multilib. 8211 8212 [30]ARM 8213 8214 * Support for the ARMv5 and ARMv5E architectures has been deprecated 8215 (which have no known implementations) and will be removed in a 8216 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 8217 architectures remain supported. The values armv5 and armv5e of 8218 -march are thus deprecated. 8219 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 8220 Extensions are now supported. They can be used by specifying the 8221 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 8222 Floating-Point Extensions introduce new half-precision data 8223 processing floating-point instructions. 8224 * The ARMv8-M architecture is now supported in its two architecture 8225 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 8226 Floating-Point Extensions. They can be used by specifying the 8227 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 8228 * Support has been added for the following processors (GCC 8229 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 8230 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 8231 identifiers can be used as arguments to the -mcpu or -mtune 8232 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 8233 * A new command-line option -mpure-code has been added. It does not 8234 allow constant data to be placed in code sections. This option is 8235 only available when generating non-PIC code for ARMv7-M targets. 8236 * Support for the ACLE Coprocessor Intrinsics has been added. This 8237 enables the generation of coprocessor instructions through the use 8238 of intrinsics such as cdp, ldc, and others. 8239 * The configure option --with-multilib-list now accepts the value 8240 rmprofile to build multilib libraries for a range of embedded 8241 targets. See our [31]installation instructions for details. 8242 8243 [32]AVR 8244 8245 * On the reduced Tiny cores, the progmem [33]variable attribute is 8246 now properly supported. Respective read-only variables are located 8247 in flash memory in section .progmem.data. No special code is needed 8248 to access such variables; the compiler automatically adds an offset 8249 of 0x4000 to all addresses, which is needed to access variables in 8250 flash memory. As opposed to ordinary cores where it is sufficient 8251 to specify the progmem attribute with definitions, on the reduced 8252 Tiny cores the attribute also has to be specified with (external) 8253 declarations: 8254 8255 extern const int array[] __attribute__((__progmem__)); 8256 8257 int get_value2 (void) 8258 { 8259 /* Access via addresses array + 0x4004 and array + 0x4005. */ 8260 return array[2]; 8261 } 8262 8263 const int* get_address (unsigned idx) 8264 { 8265 /* Returns array + 0x4000 + 2 * idx. */ 8266 return &array[idx]; 8267 } 8268 8269 * A new command-line option -Wmisspelled-isr has been added. It turns 8270 off or turns into errors warnings that are reported for 8271 interrupt service routines (ISRs) which don't follow AVR-LibC's 8272 naming convention of prefixing ISR names with __vector. 8273 * __builtin_avr_nops(n) is a new [34]built-in function that inserts n 8274 NOP instructions into the instruction stream. n must be a value 8275 known at compile time. 8276 8277 [35]IA-32/x86-64 8278 8279 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 8280 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 8281 variable precision (4VNNIW), AVX-512 Vector Population Count 8282 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 8283 been added. 8284 8285 [36]NVPTX 8286 8287 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 8288 See the [37]Offloading Wiki on how to configure it. 8289 8290 [38]PowerPC / PowerPC64 / RS6000 8291 8292 * The PowerPC port now uses LRA by default. 8293 * GCC now diagnoses inline assembly that clobbers register r2. This 8294 has always been invalid code, and is no longer quietly tolerated. 8295 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 8296 enhanced to generate more of the new instructions by default, and 8297 to provide more built-in functions to generate code for other new 8298 instructions. 8299 * The configuration option --enable-gnu-indirect-function is now 8300 enabled by default on PowerPC GNU/Linux builds. 8301 * The PowerPC port will now allow 64-bit and 32-bit integer types to 8302 be allocated to the VSX vector registers (ISA 2.06 and above). In 8303 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 8304 allocated in the vector registers. Previously, only 64-bit integer 8305 types were allowed in the traditional floating point registers. 8306 * New options -mstack-protector-guard=global, 8307 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 8308 -mstack-protector-guard-offset= change how the stack protector gets 8309 the value to use as canary. 8310 8311 [39]S/390, System z, IBM z Systems, IBM Z 8312 8313 * Support for the IBM z14 processor has been added. When using the 8314 -march=z14 option, the compiler will generate code making use of 8315 the new instructions introduced with the vector enhancement 8316 facility and the miscellaneous instruction extension facility 2. 8317 The -mtune=z14 option enables z14 specific instruction scheduling 8318 without making use of new instructions. 8319 * Builtins for the new vector instructions have been added and can be 8320 enabled using the -mzvector option. 8321 8322 [40]RISC-V 8323 8324 * Support for the RISC-V instruction set has been added. 8325 8326 [41]RX 8327 8328 Basic support for atomic built-in function has been added. It is 8329 currently implemented by flipping interrupts off and on as needed. 8330 8331 [42]SH 8332 8333 * Support for SH5/SH64 has been removed. 8334 * Improved utilization of delay slots on SH2A. 8335 * Improved utilization of zero-displacement conditional branches. 8336 * The following deprecated options have been removed 8337 + -mcbranchdi 8338 + -mcmpeqdi 8339 + -minvalid-symbols 8340 + -msoft-atomic 8341 + -mspace 8342 + -madjust-unroll 8343 * Support for the following SH2A instructions has been added 8344 + mov.b @-Rm,R0 8345 + mov.w @-Rm,R0 8346 + mov.l @-Rm,R0 8347 + mov.b R0,@Rn+ 8348 + mov.w R0,@Rn+ 8349 + mov.l R0,@Rn+ 8350 8351 [43]SPARC 8352 8353 * The SPARC port now uses LRA by default. 8354 * Support for the new Subtract-Extended-with-Carry instruction 8355 available in SPARC M7 (Niagara 7) has been added. 8356 8357 [44]Operating Systems 8358 8359 [45]AIX 8360 8361 * Visibility support has been enabled for AIX 7.1 and above. 8362 8363 [46]Fuchsia 8364 8365 * Support has been added for the [47]Fuchsia OS. 8366 8367 [48]RTEMS 8368 8369 * The ABI changes on ARM so that no short enums are used by default. 8370 8371 Other significant improvements 8372 8373 * -fverbose-asm previously emitted information on the meanings of 8374 assembly expressions. This has been extended so that it now also 8375 prints comments showing the source lines that correspond to the 8376 assembly, making it easier to read the generated assembly 8377 (especially with larger functions). For example, given this C 8378 source file: 8379 8380 int test (int n) 8381 { 8382 int i; 8383 int total = 0; 8384 8385 for (i = 0; i < n; i++) 8386 total += i * i; 8387 return total; 8388 } 8389 8390 -fverbose-asm now gives output similar to this for the function 8391 body (when compiling for x86_64, with -Os): 8392 8393 .text 8394 .globl test 8395 .type test, @@function 8396 test: 8397 .LFB0: 8398 .cfi_startproc 8399 # example.c:4: int total = 0; 8400 xorl %eax, %eax # <retval> 8401 # example.c:6: for (i = 0; i < n; i++) 8402 xorl %edx, %edx # i 8403 .L2: 8404 # example.c:6: for (i = 0; i < n; i++) 8405 cmpl %edi, %edx # n, i 8406 jge .L5 #, 8407 # example.c:7: total += i * i; 8408 movl %edx, %ecx # i, tmp92 8409 imull %edx, %ecx # i, tmp92 8410 # example.c:6: for (i = 0; i < n; i++) 8411 incl %edx # i 8412 # example.c:7: total += i * i; 8413 addl %ecx, %eax # tmp92, <retval> 8414 jmp .L2 # 8415 .L5: 8416 # example.c:10: } 8417 ret 8418 .cfi_endproc 8419 8420 * Two new options have been added for printing fix-it hints: 8421 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 8422 emitted in a machine-readable form, suitable for consumption 8423 by IDEs. For example, given: 8424 8425 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 8426 you mean 'color'? 8427 return ptr->colour; 8428 ^~~~~~ 8429 color 8430 8431 it will emit: 8432 8433 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 8434 8435 + -fdiagnostics-generate-patch will print a patch in "unified" 8436 format after any diagnostics are printed, showing the result 8437 of applying all fix-it hints. For the above example it would 8438 emit: 8439 8440 --- spellcheck-fields.cc 8441 +++ spellcheck-fields.cc 8442 @@ -49,5 +49,5 @@ 8443 8444 color get_color(struct s *ptr) 8445 { 8446 - return ptr->colour; 8447 + return ptr->color; 8448 } 8449 8450 * The gcc and g++ driver programs will now provide suggestions for 8451 misspelled arguments to command-line options. 8452 8453 $ gcc -c test.c -ftls-model=global-dinamic 8454 gcc: error: unknown TLS model 'global-dinamic' 8455 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 8456 cal-dynamic local-exec; did you mean 'global-dynamic'? 8457 8458 * The compiler will now provide suggestions for misspelled 8459 parameters. 8460 8461 $ gcc -c test.c --param max-early-inliner-iteration=3 8462 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 8463 x-early-inliner-iterations'? 8464 8465 * Profile-guided optimization (PGO) instrumentation, as well as test 8466 coverage (GCOV), can newly instrument constructors (functions marks 8467 with __attribute__((constructor))), destructors and C++ 8468 constructors (and destructors) of classes that are used as the type 8469 of a global variable. 8470 * A new option -fprofile-update=atomic prevents creation of corrupted 8471 profiles created during an instrumentation run (-fprofile=generate) 8472 of an application. The downside of the option is a speed penalty. 8473 Providing -pthread on the command line selects atomic profile 8474 updating (when supported by the target). 8475 * GCC's already extensive testsuite has gained some new capabilities, 8476 to further improve the reliability of the compiler: 8477 + GCC now has an internal unit-testing API and a suite of tests 8478 for programmatic self-testing of subsystems. 8479 + GCC's C front end has been extended so that it can parse dumps 8480 of GCC's internal representations, allowing for DejaGnu tests 8481 that more directly exercise specific optimization passes. This 8482 covers both the [49]GIMPLE representation (for testing 8483 higher-level optimizations) and the [50]RTL representation, 8484 allowing for more direct testing of lower-level details, such 8485 as register allocation and instruction selection. 8486 8487 [51]GCC 7.1 8488 8489 This is the [52]list of problem reports (PRs) from GCC's bug tracking 8490 system that are known to be fixed in the 7.1 release. This list might 8491 not be complete (that is, it is possible that some PRs that have been 8492 fixed are not listed here). 8493 8494 [53]GCC 7.2 8495 8496 This is the [54]list of problem reports (PRs) from GCC's bug tracking 8497 system that are known to be fixed in the 7.2 release. This list might 8498 not be complete (that is, it is possible that some PRs that have been 8499 fixed are not listed here). 8500 8501 Target Specific Changes 8502 8503 SPARC 8504 8505 * Support for the SPARC M8 processor has been added. 8506 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 8507 around an erratum in LEON3FT processors. 8508 * Use of the Floating-point Multiply Single to Double (FsMULd) 8509 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 8510 options. 8511 8512 Operating Systems 8513 8514 RTEMS 8515 8516 * The Ada run-time support uses now thread-local storage (TLS). 8517 * Support for RISC-V has been added. 8518 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 8519 double has been added. 8520 8521 [55]GCC 7.3 8522 8523 This is the [56]list of problem reports (PRs) from GCC's bug tracking 8524 system that are known to be fixed in the 7.3 release. This list might 8525 not be complete (that is, it is possible that some PRs that have been 8526 fixed are not listed here). 8527 8528 Target Specific Changes 8529 8530 SPARC 8531 8532 * Workarounds for the four [57]LEON3FT errata GRLIB-TN-0010..0013 8533 have been added. Relevant errata are activated by the target 8534 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 8535 8536 Operating Systems 8537 8538 RTEMS 8539 8540 * Support has been added for Epiphany target. 8541 8542 [58]GCC 7.4 8543 8544 This is the [59]list of problem reports (PRs) from GCC's bug tracking 8545 system that are known to be fixed in the 7.4 release. This list might 8546 not be complete (that is, it is possible that some PRs that have been 8547 fixed are not listed here). 8548 8549 [60]GCC 7.5 8550 8551 This is the [61]list of problem reports (PRs) from GCC's bug tracking 8552 system that are known to be fixed in the 7.5 release. This list might 8553 not be complete (that is, it is possible that some PRs that have been 8554 fixed are not listed here). 8555 8556 8557 For questions related to the use of GCC, please consult these web 8558 pages and the [62]GCC manuals. If that fails, the 8559 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8560 web pages and the development of GCC are welcome on our developer 8561 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 8562 archives. 8563 8564 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 8565 distribution of this entire article is permitted in any medium, 8566 provided this notice is preserved. 8567 8568 These pages are [67]maintained by the GCC team. Last modified 8569 2026-06-13. 8570 8571 References 8572 8573 1. https://gcc.gnu.org/gcc-7/porting_to.html 8574 2. https://gcc.gnu.org/onlinedocs/index.html#current 8575 3. https://gcc.gnu.org/wiki/LRAIsDefault 8576 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 8577 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 8578 6. https://gcc.gnu.org/gcc-7/changes.html#general 8579 7. https://gcc.gnu.org/gcc-7/changes.html#languages 8580 8. https://gcc.gnu.org/wiki/OpenACC 8581 9. https://gcc.gnu.org/wiki/Offloading 8582 10. https://gcc.gnu.org/gcc-7/changes.html#ada 8583 11. https://gcc.gnu.org/gcc-7/changes.html#brig 8584 12. https://gcc.gnu.org/gcc-7/changes.html#c-family 8585 13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 8586 14. https://gcc.gnu.org/gcc-7/changes.html#c 8587 15. https://gcc.gnu.org/gcc-7/changes.html#cxx 8588 16. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 8589 17. https://wg21.link/p0136 8590 18. https://gcc.gnu.org/gcc-7/changes.html#libstdcxx 8591 19. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 8592 20. https://gcc.gnu.org/gcc-7/changes.html#fortran 8593 21. https://www.openmp.org/specifications/ 8594 22. https://gcc.gnu.org/gcc-7/changes.html#go 8595 23. https://gcc.gnu.org/gcc-7/changes.html#java 8596 24. https://gcc.gnu.org/gcc-7/changes.html#jit 8597 25. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 8598 26. https://gcc.gnu.org/gcc-7/changes.html#targets 8599 27. https://gcc.gnu.org/gcc-7/changes.html#aarch64 8600 28. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 8601 29. https://gcc.gnu.org/gcc-7/changes.html#arc 8602 30. https://gcc.gnu.org/gcc-7/changes.html#arm 8603 31. https://gcc.gnu.org/install/configure.html 8604 32. https://gcc.gnu.org/gcc-7/changes.html#avr 8605 33. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 8606 34. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 8607 35. https://gcc.gnu.org/gcc-7/changes.html#x86 8608 36. https://gcc.gnu.org/gcc-7/changes.html#nvptx 8609 37. https://gcc.gnu.org/wiki/Offloading 8610 38. https://gcc.gnu.org/gcc-7/changes.html#powerpc 8611 39. https://gcc.gnu.org/gcc-7/changes.html#s390 8612 40. https://gcc.gnu.org/gcc-7/changes.html#riscv 8613 41. https://gcc.gnu.org/gcc-7/changes.html#rx 8614 42. https://gcc.gnu.org/gcc-7/changes.html#sh 8615 43. https://gcc.gnu.org/gcc-7/changes.html#sparc 8616 44. https://gcc.gnu.org/gcc-7/changes.html#os 8617 45. https://gcc.gnu.org/gcc-7/changes.html#aix 8618 46. https://gcc.gnu.org/gcc-7/changes.html#fuchsia 8619 47. https://fuchsia.googlesource.com/ 8620 48. https://gcc.gnu.org/gcc-7/changes.html#rtems 8621 49. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 8622 50. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 8623 51. https://gcc.gnu.org/gcc-7/changes.html#GCC7.1 8624 52. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 8625 53. https://gcc.gnu.org/gcc-7/changes.html#GCC7.2 8626 54. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 8627 55. https://gcc.gnu.org/gcc-7/changes.html#GCC7.3 8628 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 8629 57. https://www.gaisler.com/app-notes-tech-notes-and-white-papers 8630 58. https://gcc.gnu.org/gcc-7/changes.html#GCC7.4 8631 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 8632 60. https://gcc.gnu.org/gcc-7/changes.html#GCC7.5 8633 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 8634 62. https://gcc.gnu.org/onlinedocs/ 8635 63. mailto:gcc-help (a] gcc.gnu.org 8636 64. mailto:gcc (a] gcc.gnu.org 8637 65. https://gcc.gnu.org/lists.html 8638 66. https://www.fsf.org/ 8639 67. https://gcc.gnu.org/about.html 8640 ====================================================================== 8641 http://gcc.gnu.org/gcc-6/index.html 8642 GCC 6 Release Series 8643 8644 (This release series is no longer supported.) 8645 8646 October 26, 2018 8647 8648 The [1]GNU project and the GCC developers are pleased to announce the 8649 release of GCC 6.5. 8650 8651 This release is a bug-fix release, containing fixes for regressions in 8652 GCC 6.4 relative to previous releases of GCC. 8653 8654 Release History 8655 8656 GCC 6.5 8657 October 26, 2018 ([2]changes, [3]documentation) 8658 8659 GCC 6.4 8660 July 4, 2017 ([4]changes, [5]documentation) 8661 8662 GCC 6.3 8663 December 21, 2016 ([6]changes, [7]documentation) 8664 8665 GCC 6.2 8666 August 22, 2016 ([8]changes, [9]documentation) 8667 8668 GCC 6.1 8669 April 27, 2016 ([10]changes, [11]documentation) 8670 8671 References and Acknowledgements 8672 8673 GCC used to stand for the GNU C Compiler, but since the compiler 8674 supports several other languages aside from C, it now stands for the 8675 GNU Compiler Collection. 8676 8677 A list of [12]successful builds is updated as new information becomes 8678 available. 8679 8680 The GCC developers would like to thank the numerous people that have 8681 contributed new features, improvements, bug fixes, and other changes as 8682 well as test results to GCC. This [13]amazing group of volunteers is 8683 what makes GCC successful. 8684 8685 For additional information about GCC please refer to the [14]GCC 8686 project web site or contact the [15]GCC development mailing list. 8687 8688 To obtain GCC please use [16]our mirror sites or [17]our version 8689 control system. 8690 8691 8692 For questions related to the use of GCC, please consult these web 8693 pages and the [18]GCC manuals. If that fails, the 8694 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8695 web pages and the development of GCC are welcome on our developer 8696 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 8697 archives. 8698 8699 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8700 distribution of this entire article is permitted in any medium, 8701 provided this notice is preserved. 8702 8703 These pages are [23]maintained by the GCC team. Last modified 8704 2024-05-30. 8705 8706 References 8707 8708 1. http://www.gnu.org/ 8709 2. https://gcc.gnu.org/gcc-6/changes.html 8710 3. https://gcc.gnu.org/onlinedocs/6.5.0/ 8711 4. https://gcc.gnu.org/gcc-6/changes.html 8712 5. https://gcc.gnu.org/onlinedocs/6.4.0/ 8713 6. https://gcc.gnu.org/gcc-6/changes.html 8714 7. https://gcc.gnu.org/onlinedocs/6.3.0/ 8715 8. https://gcc.gnu.org/gcc-6/changes.html 8716 9. https://gcc.gnu.org/onlinedocs/6.2.0/ 8717 10. https://gcc.gnu.org/gcc-6/changes.html 8718 11. https://gcc.gnu.org/onlinedocs/6.1.0/ 8719 12. https://gcc.gnu.org/gcc-6/buildstat.html 8720 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8721 14. https://gcc.gnu.org/index.html 8722 15. mailto:gcc (a] gcc.gnu.org 8723 16. https://gcc.gnu.org/mirrors.html 8724 17. https://gcc.gnu.org/git.html 8725 18. https://gcc.gnu.org/onlinedocs/ 8726 19. mailto:gcc-help (a] gcc.gnu.org 8727 20. mailto:gcc (a] gcc.gnu.org 8728 21. https://gcc.gnu.org/lists.html 8729 22. https://www.fsf.org/ 8730 23. https://gcc.gnu.org/about.html 8731 ====================================================================== 8732 http://gcc.gnu.org/gcc-6/changes.html 8733 GCC 6 Release Series 8734 Changes, New Features, and Fixes 8735 8736 This page is a brief summary of some of the huge number of improvements 8737 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 8738 the [2]full GCC documentation. 8739 8740 Caveats 8741 8742 * The default mode for C++ is now -std=gnu++14 instead of 8743 -std=gnu++98. 8744 * Support for a number of older systems and recently unmaintained or 8745 untested target ports of GCC has been declared obsolete in GCC 6. 8746 Unless there is activity to revive them, the next release of GCC 8747 will have their sources permanently removed. 8748 The following ports for individual systems on particular 8749 architectures have been obsoleted: 8750 + SH5 / SH64 (sh64-*-*) as announced [3]here. 8751 * The AVR port requires binutils version 2.26.1 or later for the fix 8752 for [4]PR71151 to work. 8753 * The GCC 6.5 release has an accidental ABI incompatibility for 8754 nested std::pair objects, for more details see [5]PR 87822. The bug 8755 causes a layout change for pairs where the first member is also a 8756 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 8757 is closed so the bug in GCC 6.5 will not be fixed upstream, but 8758 there is a patch in the bug report to allow it to be fixed by 8759 anybody packaging GCC 6.5 or installing it themselves. 8760 8761 [6]General Optimizer Improvements 8762 8763 * UndefinedBehaviorSanitizer gained a new sanitization option, 8764 -fsanitize=bounds-strict, which enables strict checking of array 8765 bounds. In particular, it enables -fsanitize=bounds as well as 8766 instrumentation of flexible array member-like arrays. 8767 * Type-based alias analysis now disambiguates accesses to different 8768 pointers. This improves precision of the alias oracle by about 8769 20-30% on higher-level C++ programs. Programs doing invalid type 8770 punning of pointer types may now need -fno-strict-aliasing to work 8771 correctly. 8772 * Alias analysis now correctly supports the weakref and alias 8773 attributes. This allows accessing both a variable and its alias in 8774 one translation unit which is common with link-time optimization. 8775 * Value range propagation now assumes that the this pointer in C++ 8776 member functions is non-null. This eliminates common null pointer 8777 checks but also breaks some non-conforming code-bases (such as 8778 Qt-5, Chromium, KDevelop). As a temporary work-around 8779 -fno-delete-null-pointer-checks can be used. Wrong code can be 8780 identified by using -fsanitize=undefined. 8781 * Link-time optimization improvements: 8782 + warning and error attributes are now correctly preserved by 8783 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 8784 supported with -flto. 8785 + Type merging was fixed to handle C and Fortran 8786 interoperability rules as defined by the Fortran 2008 language 8787 standard. 8788 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 8789 with char in all cases because it is an array while char is 8790 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 8791 general, this inter-operability cannot be implemented, for 8792 example on targets where the argument passing convention for 8793 arrays differs from scalars. 8794 + More type information is now preserved at link time, reducing 8795 the loss of accuracy of the type-based alias analysis compared 8796 to builds without link-time optimization. 8797 + Invalid type punning on global variables and declarations is 8798 now reported with -Wodr-type-mismatch. 8799 + The size of LTO object files was reduced by about 11% 8800 (measured by compiling Firefox 46.0). 8801 + Link-time parallelization (enabled using -flto=n) was 8802 significantly improved by decreasing the size of streamed data 8803 when partitioning programs. The size of streamed IL while 8804 compiling Firefox 46.0 was reduced by 66%. 8805 + The linker plugin was extended to pass information about the 8806 type of binary produced to the GCC back end. (That can also be 8807 controlled manually by -flinker-output.) This makes it 8808 possible to properly configure the code generator and support 8809 incremental linking. Incremental linking of LTO objects by gcc 8810 -r is now supported for plugin-enabled setups. 8811 There are two ways to perform incremental linking: 8812 1. Linking by ld -r will result in an object file with all 8813 sections from individual object files mechanically 8814 merged. This delays the actual link-time optimization to 8815 the final linking step and thus permits whole program 8816 optimization. Linking the final binary with such object 8817 files is however slower. 8818 2. Linking by gcc -r will lead to link-time optimization and 8819 emit the final binary into the object file. Linking such 8820 an object file is fast but avoids any benefits from whole 8821 program optimization. 8822 GCC 7 will support incremental link-time optimization with gcc 8823 -r. 8824 * Inter-procedural optimization improvements: 8825 + Basic jump threading is now performed before profile 8826 construction and inline analysis, resulting in more realistic 8827 size and time estimates that drive the heuristics of the 8828 inliner and function cloning passes. 8829 + Function cloning now more aggressively eliminates unused 8830 function parameters. 8831 8832 [7]New Languages and Language specific improvements 8833 8834 Compared to GCC 5, the GCC 6 release series includes a much improved 8835 implementation of the [8]OpenACC 2.0a specification. Highlights are: 8836 * In addition to single-threaded host-fallback execution, offloading 8837 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 8838 little-endian GNU/Linux host systems. For nvptx offloading, with 8839 the OpenACC parallel construct, the execution model allows for an 8840 arbitrary number of gangs, up to 32 workers, and 32 vectors. 8841 * Initial support for parallelized execution of OpenACC kernels 8842 constructs: 8843 + Parallelization of a kernels region is switched on by 8844 -fopenacc combined with -O2 or higher. 8845 + Code is offloaded onto multiple gangs, but executes with just 8846 one worker, and a vector length of 1. 8847 + Directives inside a kernels region are not supported. 8848 + Loops with reductions can be parallelized. 8849 + Only kernels regions with one loop nest are parallelized. 8850 + Only the outer-most loop of a loop nest can be parallelized. 8851 + Loop nests containing sibling loops are not parallelized. 8852 Typically, using the OpenACC parallel construct gives much better 8853 performance, compared to the initial support of the OpenACC kernels 8854 construct. 8855 * The device_type clause is not supported. The bind and nohost 8856 clauses are not supported. The host_data directive is not supported 8857 in Fortran. 8858 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 8859 * Usage of OpenACC constructs inside multithreaded contexts (such as 8860 created by OpenMP, or pthread programming) is not supported. 8861 * If a call to the acc_on_device function has a compile-time constant 8862 argument, the function call evaluates to a compile-time constant 8863 value only for C and C++ but not for Fortran. 8864 8865 See the [9]OpenACC and [10]Offloading wiki pages for further 8866 information. 8867 8868 [11]C family 8869 8870 * Version 4.5 of the [12]OpenMP specification is now supported in the 8871 C and C++ compilers. 8872 * The C and C++ compilers now support attributes on enumerators. For 8873 instance, it is now possible to mark enumerators as deprecated: 8874 8875 enum { 8876 newval, 8877 oldval __attribute__ ((deprecated ("too old"))) 8878 }; 8879 8880 * Source locations for the C and C++ compilers are now tracked as 8881 ranges, rather than just points, making it easier to identify the 8882 subexpression of interest within a complicated expression. For 8883 example: 8884 8885 test.cc: In function 'int test(int, int, foo, int, int)': 8886 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 8887 ) 8888 return p + q * r * s + t; 8889 ~~^~~ 8890 8891 In addition, there is now initial support for precise diagnostic 8892 locations within strings: 8893 8894 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 8895 t' argument [-Wformat=] 8896 printf("%*d"); 8897 ^ 8898 8899 * Diagnostics can now contain "fix-it hints", which are displayed in 8900 context underneath the relevant source code. For example: 8901 8902 fixits.c: In function 'bad_deref': 8903 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 8904 return ptr.x; 8905 ^ 8906 -> 8907 8908 * The C and C++ compilers now offer suggestions for misspelled field 8909 names: 8910 8911 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 8912 you mean 'color'? 8913 return ptr->colour; 8914 ^~~~~~ 8915 8916 * New command-line options have been added for the C and C++ 8917 compilers: 8918 + -Wshift-negative-value warns about left shifting a negative 8919 value. 8920 + -Wshift-overflow warns about left shift overflows. This 8921 warning is enabled by default. -Wshift-overflow=2 also warns 8922 about left-shifting 1 into the sign bit. 8923 + -Wtautological-compare warns if a self-comparison always 8924 evaluates to true or false. This warning is enabled by -Wall. 8925 + -Wnull-dereference warns if the compiler detects paths that 8926 trigger erroneous or undefined behavior due to dereferencing a 8927 null pointer. This option is only active when 8928 -fdelete-null-pointer-checks is active, which is enabled by 8929 optimizations in most targets. The precision of the warnings 8930 depends on the optimization options used. 8931 + -Wduplicated-cond warns about duplicated conditions in an 8932 if-else-if chain. 8933 + -Wmisleading-indentation warns about places where the 8934 indentation of the code gives a misleading idea of the block 8935 structure of the code to a human reader. For example, given 8936 [13]CVE-2014-1266: 8937 8938 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 8939 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 8940 g-indentation] 8941 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 8942 ^~ 8943 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 8944 indented as if it is guarded by the 'if' 8945 goto fail; 8946 ^~~~ 8947 8948 This warning is enabled by -Wall. 8949 * The C and C++ compilers now emit saner error messages if 8950 merge-conflict markers are present in a source file. 8951 8952 test.c:3:1: error: version control conflict marker in file 8953 <<<<<<< HEAD 8954 ^~~~~~~ 8955 8956 [14]C 8957 8958 * It is possible to disable warnings when an initialized field of a 8959 structure or a union with side effects is being overridden when 8960 using designated initializers via a new warning option 8961 -Woverride-init-side-effects. 8962 * A new type attribute scalar_storage_order applying to structures 8963 and unions has been introduced. It specifies the storage order (aka 8964 endianness) in memory of scalar fields in structures or unions. 8965 8966 [15]C++ 8967 8968 * The default mode has been changed to -std=gnu++14. 8969 * [16]C++ Concepts are now supported when compiling with -fconcepts. 8970 * -flifetime-dse is more aggressive in dead-store elimination in 8971 situations where a memory store to a location precedes a 8972 constructor to that memory location. 8973 * G++ now supports [17]C++17 fold expressions, u8 character literals, 8974 extended static_assert, and nested namespace definitions. 8975 * G++ now allows constant evaluation for all non-type template 8976 arguments. 8977 * G++ now supports C++ Transactional Memory when compiling with 8978 -fgnu-tm. 8979 8980 [18]Runtime Library (libstdc++) 8981 8982 * Extensions to the C++ Library to support mathematical special 8983 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 8984 * Experimental support for C++17, including the following new 8985 features: 8986 + std::uncaught_exceptions function (this is also available for 8987 -std=gnu++NN modes); 8988 + new member functions try_emplace and insert_or_assign for 8989 unique_key maps; 8990 + non-member functions std::size, std::empty, and std::data for 8991 accessing containers and arrays; 8992 + std::invoke; 8993 + std::shared_mutex; 8994 + std::void_t and std::bool_constant metaprogramming utilities. 8995 Thanks to Ville Voutilainen for contributing many of the C++17 8996 features. 8997 * An experimental implementation of the File System TS. 8998 * Experimental support for most features of the second version of the 8999 Library Fundamentals TS. This includes polymorphic memory resources 9000 and array support in shared_ptr, thanks to Fan You. 9001 * Some assertions checked by Debug Mode can now also be enabled by 9002 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 9003 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 9004 don't affect the library ABI, so can be enabled per-translation 9005 unit. 9006 * Timed mutex types are supported on more targets, including Darwin. 9007 * Improved std::locale support for DragonFly and FreeBSD, thanks to 9008 John Marino and Andreas Tobler. 9009 9010 [19]Fortran 9011 9012 * Fortran 2008 SUBMODULE support. 9013 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 9014 support. 9015 * Improved support for Fortran 2003 deferred-length character 9016 variables. 9017 * Improved support for OpenMP and OpenACC. 9018 * The MATMUL intrinsic is now inlined for straightforward cases if 9019 front-end optimization is active. The maximum size for inlining can 9020 be set to n with the -finline-matmul-limit=n option and turned off 9021 with -finline-matmul-limit=0. 9022 * The -Wconversion-extra option will warn about REAL constants which 9023 have excess precision for their kind. 9024 * The -Winteger-division option has been added, which warns about 9025 divisions of integer constants which are truncated. This option is 9026 included in -Wall by default. 9027 9028 [20]libgccjit 9029 9030 * The driver code is now run in-process within libgccjit, providing a 9031 small speed-up of the compilation process. 9032 * The API has gained entrypoints for 9033 + [21]timing how long was spent in different parts of code, 9034 + [22]creating switch statements, 9035 + [23]allowing unreachable basic blocks in a function, and 9036 + [24]adding arbitrary command-line options to a compilation. 9037 9038 [25]New Targets and Target Specific Improvements 9039 9040 [26]AArch64 9041 9042 * A number of AArch64-specific options have been added. The most 9043 important ones are summarised in this section; for more detailed 9044 information please refer to the documentation. 9045 * The command-line options -march=native, -mcpu=native and 9046 -mtune=native are now available on native AArch64 GNU/Linux 9047 systems. Specifying these options causes GCC to auto-detect the 9048 host CPU and choose the optimal setting for that system. 9049 * -fpic is now supported when generating code for the small code 9050 model (-mcmodel=small). The size of the global offset table (GOT) 9051 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 9052 ILP32 SysV ABI. 9053 * The AArch64 port now supports target attributes and pragmas. Please 9054 refer to the [27]documentation for details of available attributes 9055 and pragmas as well as usage instructions. 9056 * Link-time optimization across translation units with different 9057 target-specific options is now supported. 9058 * The option -mtls-size= is now supported. It can be used to specify 9059 the bit size of TLS offsets, allowing GCC to generate better TLS 9060 instruction sequences. 9061 * The option -fno-plt is now fully functional. 9062 * The ARMv8.1-A architecture and the Large System Extensions are now 9063 supported. They can be used by specifying the -march=armv8.1-a 9064 option. Additionally, the +lse option extension can be used in a 9065 similar fashion to other option extensions. The Large System 9066 Extensions introduce new instructions that are used in the 9067 implementation of atomic operations. 9068 * The ACLE half-precision floating-point type __fp16 is now supported 9069 in the C and C++ languages. 9070 * The ARM Cortex-A35 processor is now supported via the 9071 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 9072 equivalent target attributes and pragmas. 9073 * The Qualcomm QDF24xx processor is now supported via the 9074 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 9075 target attributes and pragmas. 9076 * Code generation for the ARM Cortex-A57 processor is improved. Among 9077 general code generation improvements, a better algorithm is added 9078 for allocating registers to floating-point multiply-accumulate 9079 instructions offering increased performance when compiling with 9080 -mcpu=cortex-a57 or -mtune=cortex-a57. 9081 * Code generation for the ARM Cortex-A53 processor is improved. A 9082 more accurate instruction scheduling model for the processor is now 9083 used, and a number of compiler tuning parameters have been set to 9084 offer increased performance when compiling with -mcpu=cortex-a53 or 9085 -mtune=cortex-a53. 9086 * Code generation for the Samsung Exynos M1 processor is improved. A 9087 more accurate instruction scheduling model for the processor is now 9088 used, and a number of compiler tuning parameters have been set to 9089 offer increased performance when compiling with -mcpu=exynos-m1 or 9090 -mtune=exynos-m1. 9091 * Improvements in the generation of conditional branches and literal 9092 pools allow the compiler to compile functions of a large size. 9093 Constant pools are now placed into separate rodata sections. The 9094 new option -mpc-relative-literal-loads generates per-function 9095 literal pools, limiting the maximum size of functions to 1MiB. 9096 * Several correctness issues generating Advanced SIMD instructions 9097 for big-endian targets have been fixed resulting in improved code 9098 generation for ACLE intrinsics with -mbig-endian. 9099 9100 [28]ARM 9101 9102 * Support for revisions of the ARM architecture prior to ARMv4t has 9103 been deprecated and will be removed in a future GCC release. The 9104 -mcpu and -mtune values that are deprecated are: arm2, arm250, 9105 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 9106 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 9107 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 9108 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 9109 arm7tdmi is still supported. The values of -march that are 9110 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 9111 * The ARM port now supports target attributes and pragmas. Please 9112 refer to the [29]documentation for details of available attributes 9113 and pragmas as well as usage instructions. 9114 * Support has been added for the following processors (GCC 9115 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 9116 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 9117 identifiers can be used as arguments to the -mcpu or -mtune 9118 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 9119 9120 [30]Heterogeneous Systems Architecture 9121 9122 * GCC can now generate HSAIL (Heterogeneous System Architecture 9123 Intermediate Language) for simple OpenMP device constructs if 9124 configured with --enable-offload-targets=hsa. A new libgomp plugin 9125 then runs the HSA GPU kernels implementing these constructs on HSA 9126 capable GPUs via a standard HSA run time. 9127 If the HSA compilation back end determines it cannot output HSAIL 9128 for a particular input, it gives a warning by default. These 9129 warnings can be suppressed with -Wno-hsa. To give a few examples, 9130 the HSA back end does not implement compilation of code using 9131 function pointers, automatic allocation of variable sized arrays, 9132 functions with variadic arguments as well as a number of other less 9133 common programming constructs. 9134 When compilation for HSA is enabled, the compiler attempts to 9135 compile composite OpenMP constructs 9136 9137 #pragma omp target teams distribute parallel for 9138 9139 into parallel HSA GPU kernels. 9140 9141 [31]IA-32/x86-64 9142 9143 * GCC now supports the Intel CPU named Skylake with AVX-512 9144 extensions through -march=skylake-avx512. The switch enables the 9145 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 9146 AVX-512DQ. 9147 * Support for new AMD instructions monitorx and mwaitx has been 9148 added. This includes new intrinsic and built-in support. It is 9149 enabled through option -mmwaitx. The instructions monitorx and 9150 mwaitx implement the same functionality as the old monitor and 9151 mwait instructions. In addition, mwaitx adds a configurable timer. 9152 The timer value is received as third argument and stored in 9153 register %ebx. 9154 * x86-64 targets now allow stack realignment from a word-aligned 9155 stack pointer using the command-line option -mstackrealign or 9156 __attribute__ ((force_align_arg_pointer)). This allows functions 9157 compiled with a vector-aligned stack to be invoked from objects 9158 that keep only word-alignment. 9159 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 9160 can be used to access data via the %fs and %gs segments without 9161 having to resort to inline assembly. Please refer to the 9162 [32]documentation for usage instructions. 9163 * Support for AMD Zen (family 17h) processors is now available 9164 through the -march=znver1 and -mtune=znver1 options. 9165 9166 [33]MeP 9167 9168 * Support for the MeP (mep-elf) architecture has been deprecated and 9169 will be removed in a future GCC release. 9170 9171 [34]MSP430 9172 9173 * The MSP430 compiler now has the ability to automatically distribute 9174 code and data between low memory (addresses below 64K) and high 9175 memory. This only applies to parts that actually have both memory 9176 regions and only if the linker script for the part has been 9177 specifically set up to support this feature. 9178 A new attribute of either can be applied to both functions and 9179 data, and this tells the compiler to place the object into low 9180 memory if there is room and into high memory otherwise. Two other 9181 new attributes - lower and upper - can be used to explicitly state 9182 that an object should be placed in the specified memory region. If 9183 there is not enough left in that region the compilation will fail. 9184 Two new command-line options - -mcode-region=[lower|upper|either] 9185 and -mdata-region=[lower|upper|either] - can be used to tell the 9186 compiler what to do with objects that do not have one of these new 9187 attributes. 9188 9189 [35]PowerPC / PowerPC64 / RS6000 9190 9191 * PowerPC64 now supports IEEE 128-bit floating-point using the 9192 __float128 data type. In GCC 6, this is not enabled by default, but 9193 you can enable it with -mfloat128. The IEEE 128-bit floating-point 9194 support requires the use of the VSX instruction set. IEEE 128-bit 9195 floating-point values are passed and returned as a single vector 9196 value. The software emulator for IEEE 128-bit floating-point 9197 support is only built on PowerPC GNU/Linux systems where the 9198 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 9199 and later), you will be able to use the -mfloat128-hardware option 9200 to use the ISA 3.0 instructions that support IEEE 128-bit 9201 floating-point. An additional type (__ibm128) has been added to 9202 refer to the IBM extended double type that normally implements long 9203 double. This will allow for a future transition to implementing 9204 long double with IEEE 128-bit floating-point. 9205 * Basic support has been added for POWER9 hardware that will use the 9206 recently published OpenPOWER ISA 3.0 instructions. The following 9207 new switches are available: 9208 + -mcpu=power9: Implement all of the ISA 3.0 instructions 9209 supported by the compiler. 9210 + -mtune=power9: In the future, apply tuning for POWER9 systems. 9211 Currently, POWER8 tunings are used. 9212 + -mmodulo: Generate code using the ISA 3.0 integer instructions 9213 (modulus, count trailing zeros, array index support, integer 9214 multiply/add). 9215 + -mpower9-fusion: Generate code to suitably fuse instruction 9216 sequences for a POWER9 system. 9217 + -mpower9-dform: Generate code to use the new D-form 9218 (register+offset) memory instructions for the vector 9219 registers. 9220 + -mpower9-vector: Generate code using the new ISA 3.0 vector 9221 (VSX or Altivec) instructions. 9222 + -mpower9-minmax: Reserved for future development. 9223 + -mtoc-fusion: Keep TOC entries together to provide more fusion 9224 opportunities. 9225 * New constraints have been added to support IEEE 128-bit 9226 floating-point and ISA 3.0 instructions: 9227 + wb: Altivec register if -mpower9-dform is enabled. 9228 + we: VSX register if -mpower9-vector is enabled for 64-bit code 9229 generation. 9230 + wo: VSX register if -mpower9-vector is enabled. 9231 + wp: Reserved for future use if long double is implemented with 9232 IEEE 128-bit floating-point instead of IBM extended double. 9233 + wq: VSX register if -mfloat128 is enabled. 9234 + wF: Memory operand suitable for POWER9 fusion load/store. 9235 + wG: Memory operand suitable for TOC fusion memory references. 9236 + wL: Integer constant identifying the element number mfvsrld 9237 accesses within a vector. 9238 * Support has been added for __builtin_cpu_is() and 9239 __builtin_cpu_supports(), allowing for very fast access to 9240 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 9241 glibc 2.23 or later. 9242 * All hardware transactional memory builtins now correctly behave as 9243 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 9244 determine whether their "old" compiler treats the builtins as 9245 barriers. 9246 * Split-stack support has been added for gccgo on PowerPC64 for both 9247 big- and little-endian (but not for 32-bit). The gold linker from 9248 at least binutils 2.25.1 must be available in the PATH when 9249 configuring and building gccgo to enable split stack. (The 9250 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 9251 split-stack feature allows a small initial stack size to be 9252 allocated for each goroutine, which increases as needed. 9253 * GCC on PowerPC now supports the standard lround function. 9254 * A new configuration option ---with-advance-toolchain=at was added 9255 for PowerPC 64-bit GNU/Linux systems to use the header files, 9256 library files, and the dynamic linker from a specific Advance 9257 Toolchain release instead of the default versions that are provided 9258 by the GNU/Linux distribution. In general, this option is intended 9259 for the developers of GCC, and it is not intended for general use. 9260 * The "q", "S", "T", and "t" asm-constraints have been removed. 9261 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 9262 9263 [36]S/390, System z, IBM z Systems 9264 9265 * Support for the IBM z13 processor has been added. When using the 9266 -march=z13 option, the compiler will generate code making use of 9267 the new instructions and registers introduced with the vector 9268 extension facility. The -mtune=z13 option enables z13 specific 9269 instruction scheduling without making use of new instructions. 9270 Compiling code with -march=z13 reduces the default alignment of 9271 vector types bigger than 8 bytes to 8. This is an ABI change and 9272 care must be taken when linking modules compiled with different 9273 arch levels which interchange variables containing vector type 9274 values. For newly compiled code the GNU linker will emit a warning. 9275 * The -mzvector option enables a C/C++ language extension. This 9276 extension provides a new keyword vector which can be used to define 9277 vector type variables. (Note: This is not available when enforcing 9278 strict standard compliance e.g. with -std=c99. Either enable GNU 9279 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 9280 Additionally a set of overloaded builtins is provided which is 9281 partially compatible to the PowerPC Altivec builtins. In order to 9282 make use of these builtins the vecintrin.h header file needs to be 9283 included. 9284 * The new command-line options -march=native, and -mtune=native are 9285 now available on native IBM z Systems. Specifying these options 9286 causes GCC to auto-detect the host CPU and choose the optimal 9287 setting for that system. If GCC is unable to detect the host CPU 9288 these options have no effect. 9289 * The IBM z Systems port now supports target attributes and pragmas. 9290 Please refer to the [37]documentation for details of available 9291 attributes and pragmas as well as usage instructions. 9292 * -fsplit-stack is now supported as part of the IBM z Systems port. 9293 This feature requires a recent gold linker to be used. 9294 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 9295 been deprecated and will be removed in a future GCC release. -m31 9296 from now on defaults to -march=z900 if not specified otherwise. 9297 -march=native on a g5/g6 machine will default to -march=z900. 9298 9299 [38]SH 9300 9301 * Support for SH5 / SH64 has been declared obsolete and will be 9302 removed in future releases. 9303 * Support for the FDPIC ABI has been added. It can be enabled using 9304 the new -mfdpic target option and --enable-fdpic configure option. 9305 9306 [39]SPARC 9307 9308 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 9309 change will break binary compatibility with earlier releases for 9310 code it affects, but this should be pretty rare in practice. The 9311 conditions are: a 16-byte structure containing a double or a 8-byte 9312 vector in the second half is passed to a subprogram in slot #15, 9313 for example as 16th parameter if the first 15 ones have at most 8 9314 bytes. The double or vector was wrongly passed in floating-point 9315 register %d32 in lieu of on the stack as per the SPARC calling 9316 conventions. 9317 9318 [40]Operating Systems 9319 9320 [41]AIX 9321 9322 * DWARF debugging support for AIX 7.1 has been enabled as an optional 9323 debugging format. A more recent Technology Level (TL) and GCC built 9324 with that level are required for full exploitation of DWARF 9325 debugging capabilities. 9326 9327 [42]Linux 9328 9329 * Support for the [43]musl C library was added for the AArch64, ARM, 9330 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 9331 x86_64 targets. It can be selected using the new -mmusl option in 9332 case musl is not the default libc. GCC defaults to musl libc if it 9333 is built with a target triplet matching the *-linux-musl* pattern. 9334 9335 [44]RTEMS 9336 9337 * The RTEMS thread model implementation changed. Mutexes now use 9338 self-contained objects defined in newlib <sys/lock.h> instead of 9339 Classic API semaphores. The keys for thread specific data and the 9340 once function are directly defined via <pthread.h>. Self-contained 9341 condition variables are provided via newlib <sys/lock.h>. The RTEMS 9342 thread model also supports C++11 threads. 9343 * OpenMP support now uses self-contained objects provided by newlib 9344 <sys/lock.h> and offers a significantly better performance compared 9345 to the POSIX configuration of libgomp. It is possible to configure 9346 thread pools for each scheduler instance via the environment 9347 variable GOMP_RTEMS_THREAD_POOLS. 9348 9349 [45]Solaris 9350 9351 * Solaris 12 is now fully supported. Minimal support had already been 9352 present in GCC 5.3. 9353 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 9354 crtn.o), which GCC now prefers over its own ones. 9355 * Position independent executables (PIE) are now supported on Solaris 9356 12. 9357 * Constructor priority is now supported on Solaris 12 with the system 9358 linker. 9359 * libvtv has been ported to Solaris 11 and up. 9360 9361 [46]Windows 9362 9363 * The option -mstackrealign is now automatically activated in 32-bit 9364 mode whenever the use of SSE instructions is requested. 9365 9366 Other significant improvements 9367 9368 * The gcc and g++ driver programs will now provide suggestions for 9369 misspelled command-line options. 9370 9371 $ gcc -static-libfortran test.f95 9372 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 9373 '-static-libgfortran'? 9374 9375 * The --enable-default-pie configure option enables generation of PIE 9376 by default. 9377 9378 [47]GCC 6.2 9379 9380 This is the [48]list of problem reports (PRs) from GCC's bug tracking 9381 system that are known to be fixed in the 6.2 release. This list might 9382 not be complete (that is, it is possible that some PRs that have been 9383 fixed are not listed here). 9384 9385 Target Specific Changes 9386 9387 SPARC 9388 9389 * Support for --with-cpu-32 and --with-cpu-64 configure options has 9390 been added on bi-architecture platforms. 9391 * Support for the SPARC M7 (Niagara 7) processor has been added. 9392 * Support for the VIS 4.0 instruction set has been added. 9393 9394 [49]GCC 6.3 9395 9396 This is the [50]list of problem reports (PRs) from GCC's bug tracking 9397 system that are known to be fixed in the 6.3 release. This list might 9398 not be complete (that is, it is possible that some PRs that have been 9399 fixed are not listed here). 9400 9401 Target Specific Changes 9402 9403 IA-32/x86-64 9404 9405 * Support for the deprecated pcommit instruction has been removed. 9406 9407 [51]GCC 6.4 9408 9409 This is the [52]list of problem reports (PRs) from GCC's bug tracking 9410 system that are known to be fixed in the 6.4 release. This list might 9411 not be complete (that is, it is possible that some PRs that have been 9412 fixed are not listed here). 9413 9414 [53]Operating Systems 9415 9416 [54]RTEMS 9417 9418 * The ABI changes on ARM so that no short enums are used by default. 9419 9420 [55]GCC 6.5 9421 9422 This is the [56]list of problem reports (PRs) from GCC's bug tracking 9423 system that are known to be fixed in the 6.5 release. This list might 9424 not be complete (that is, it is possible that some PRs that have been 9425 fixed are not listed here). 9426 9427 9428 For questions related to the use of GCC, please consult these web 9429 pages and the [57]GCC manuals. If that fails, the 9430 [58]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9431 web pages and the development of GCC are welcome on our developer 9432 list at [59]gcc (a] gcc.gnu.org. All of [60]our lists have public 9433 archives. 9434 9435 Copyright (C) [61]Free Software Foundation, Inc. Verbatim copying and 9436 distribution of this entire article is permitted in any medium, 9437 provided this notice is preserved. 9438 9439 These pages are [62]maintained by the GCC team. Last modified 9440 2026-06-20. 9441 9442 References 9443 9444 1. https://gcc.gnu.org/gcc-6/porting_to.html 9445 2. https://gcc.gnu.org/onlinedocs/index.html#current 9446 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 9447 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 9448 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 9449 6. https://gcc.gnu.org/gcc-6/changes.html#general 9450 7. https://gcc.gnu.org/gcc-6/changes.html#languages 9451 8. https://www.openacc.org/ 9452 9. https://gcc.gnu.org/wiki/OpenACC 9453 10. https://gcc.gnu.org/wiki/Offloading 9454 11. https://gcc.gnu.org/gcc-6/changes.html#c-family 9455 12. https://www.openmp.org/specifications/ 9456 13. https://www.cve.org/CVERecord?id=CVE-2014-1266 9457 14. https://gcc.gnu.org/gcc-6/changes.html#c 9458 15. https://gcc.gnu.org/gcc-6/changes.html#cxx 9459 16. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 9460 17. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 9461 18. https://gcc.gnu.org/gcc-6/changes.html#libstdcxx 9462 19. https://gcc.gnu.org/gcc-6/changes.html#fortran 9463 20. https://gcc.gnu.org/gcc-6/changes.html#jit 9464 21. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 9465 22. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 9466 23. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 9467 24. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 9468 25. https://gcc.gnu.org/gcc-6/changes.html#targets 9469 26. https://gcc.gnu.org/gcc-6/changes.html#aarch64 9470 27. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 9471 28. https://gcc.gnu.org/gcc-6/changes.html#arm 9472 29. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 9473 30. https://gcc.gnu.org/gcc-6/changes.html#hsa 9474 31. https://gcc.gnu.org/gcc-6/changes.html#x86 9475 32. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 9476 33. https://gcc.gnu.org/gcc-6/changes.html#mep 9477 34. https://gcc.gnu.org/gcc-6/changes.html#msp430 9478 35. https://gcc.gnu.org/gcc-6/changes.html#powerpc 9479 36. https://gcc.gnu.org/gcc-6/changes.html#s390 9480 37. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 9481 38. https://gcc.gnu.org/gcc-6/changes.html#sh 9482 39. https://gcc.gnu.org/gcc-6/changes.html#sparc 9483 40. https://gcc.gnu.org/gcc-6/changes.html#os 9484 41. https://gcc.gnu.org/gcc-6/changes.html#aix 9485 42. https://gcc.gnu.org/gcc-6/changes.html#linux 9486 43. https://musl.libc.org/ 9487 44. https://gcc.gnu.org/gcc-6/changes.html#rtems 9488 45. https://gcc.gnu.org/gcc-6/changes.html#solaris 9489 46. https://gcc.gnu.org/gcc-6/changes.html#windows 9490 47. https://gcc.gnu.org/gcc-6/changes.html#GCC6.2 9491 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 9492 49. https://gcc.gnu.org/gcc-6/changes.html#GCC6.3 9493 50. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 9494 51. https://gcc.gnu.org/gcc-6/changes.html#GCC6.4 9495 52. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 9496 53. https://gcc.gnu.org/gcc-6/changes.html#os64 9497 54. https://gcc.gnu.org/gcc-6/changes.html#rtems64 9498 55. https://gcc.gnu.org/gcc-6/changes.html#GCC6.5 9499 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 9500 57. https://gcc.gnu.org/onlinedocs/ 9501 58. mailto:gcc-help (a] gcc.gnu.org 9502 59. mailto:gcc (a] gcc.gnu.org 9503 60. https://gcc.gnu.org/lists.html 9504 61. https://www.fsf.org/ 9505 62. https://gcc.gnu.org/about.html 9506 ====================================================================== 9507 http://gcc.gnu.org/gcc-5/index.html 9508 GCC 5 Release Series 9509 9510 (This release series is no longer supported.) 9511 9512 October 10, 2017 9513 9514 The [1]GNU project and the GCC developers are pleased to announce the 9515 release of GCC 5.5. 9516 9517 This release is a bug-fix release, containing fixes for regressions in 9518 GCC 5.4 relative to previous releases of GCC. 9519 9520 Release History 9521 9522 GCC 5.5 9523 October 10, 2017 ([2]changes, [3]documentation) 9524 9525 GCC 5.4 9526 June 3, 2016 ([4]changes, [5]documentation) 9527 9528 GCC 5.3 9529 December 4, 2015 ([6]changes, [7]documentation) 9530 9531 GCC 5.2 9532 July 16, 2015 ([8]changes, [9]documentation) 9533 9534 GCC 5.1 9535 April 22, 2015 ([10]changes, [11]documentation) 9536 9537 References and Acknowledgements 9538 9539 GCC used to stand for the GNU C Compiler, but since the compiler 9540 supports several other languages aside from C, it now stands for the 9541 GNU Compiler Collection. 9542 9543 A list of [12]successful builds is updated as new information becomes 9544 available. 9545 9546 The GCC developers would like to thank the numerous people that have 9547 contributed new features, improvements, bug fixes, and other changes as 9548 well as test results to GCC. This [13]amazing group of volunteers is 9549 what makes GCC successful. 9550 9551 For additional information about GCC please refer to the [14]GCC 9552 project web site or contact the [15]GCC development mailing list. 9553 9554 To obtain GCC please use [16]our mirror sites or [17]our version 9555 control system. 9556 9557 9558 For questions related to the use of GCC, please consult these web 9559 pages and the [18]GCC manuals. If that fails, the 9560 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9561 web pages and the development of GCC are welcome on our developer 9562 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 9563 archives. 9564 9565 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9566 distribution of this entire article is permitted in any medium, 9567 provided this notice is preserved. 9568 9569 These pages are [23]maintained by the GCC team. Last modified 9570 2024-05-30. 9571 9572 References 9573 9574 1. http://www.gnu.org/ 9575 2. https://gcc.gnu.org/gcc-5/changes.html 9576 3. https://gcc.gnu.org/onlinedocs/5.5.0/ 9577 4. https://gcc.gnu.org/gcc-5/changes.html 9578 5. https://gcc.gnu.org/onlinedocs/5.4.0/ 9579 6. https://gcc.gnu.org/gcc-5/changes.html 9580 7. https://gcc.gnu.org/onlinedocs/5.3.0/ 9581 8. https://gcc.gnu.org/gcc-5/changes.html 9582 9. https://gcc.gnu.org/onlinedocs/5.2.0/ 9583 10. https://gcc.gnu.org/gcc-5/changes.html 9584 11. https://gcc.gnu.org/onlinedocs/5.1.0/ 9585 12. https://gcc.gnu.org/gcc-5/buildstat.html 9586 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9587 14. https://gcc.gnu.org/index.html 9588 15. mailto:gcc (a] gcc.gnu.org 9589 16. https://gcc.gnu.org/mirrors.html 9590 17. https://gcc.gnu.org/git.html 9591 18. https://gcc.gnu.org/onlinedocs/ 9592 19. mailto:gcc-help (a] gcc.gnu.org 9593 20. mailto:gcc (a] gcc.gnu.org 9594 21. https://gcc.gnu.org/lists.html 9595 22. https://www.fsf.org/ 9596 23. https://gcc.gnu.org/about.html 9597 ====================================================================== 9598 http://gcc.gnu.org/gcc-5/changes.html 9599 GCC 5 Release Series 9600 Changes, New Features, and Fixes 9601 9602 Caveats 9603 9604 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 9605 * The C++ runtime library (libstdc++) uses a new ABI by default (see 9606 [1]below). 9607 * The Graphite framework for loop optimizations no longer requires 9608 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 9609 The installation manual contains more information about 9610 requirements to build GCC. 9611 * The non-standard C++0x type traits has_trivial_default_constructor, 9612 has_trivial_copy_constructor and has_trivial_copy_assign have been 9613 deprecated and will be removed in a future version. The standard 9614 C++11 traits is_trivially_default_constructible, 9615 is_trivially_copy_constructible and is_trivially_copy_assignable 9616 should be used instead. 9617 9618 [2]General Optimizer Improvements 9619 9620 * Inter-procedural optimization improvements: 9621 + An Identical Code Folding (ICF) pass (controlled via 9622 -fipa-icf) has been added. Compared to the identical code 9623 folding performed by the Gold linker this pass does not 9624 require function sections. It also performs merging before 9625 inlining, so inter-procedural optimizations are aware of the 9626 code re-use. On the other hand not all unifications performed 9627 by a linker are doable by GCC which must honor aliasing 9628 information. During link-time optimization of Firefox, this 9629 pass unifies about 31000 functions, that is 14% overall. 9630 + The devirtualization pass was significantly improved by adding 9631 better support for speculative devirtualization and dynamic 9632 type detection. About 50% of virtual calls in Firefox are now 9633 speculatively devirtualized during link-time optimization. 9634 + A new comdat localization pass allows the linker to eliminate 9635 more dead code in presence of C++ inline functions. 9636 + Virtual tables are now optimized. Local aliases are used to 9637 reduce dynamic linking time of C++ virtual tables on ELF 9638 targets and data alignment has been reduced to limit data 9639 segment bloat. 9640 + A new -fno-semantic-interposition option can be used to 9641 improve code quality of shared libraries where interposition 9642 of exported symbols is not allowed. 9643 + Write-only variables are now detected and optimized out. 9644 + With profile feedback the function inliner can now bypass 9645 --param inline-insns-auto and --param inline-insns-single 9646 limits for hot calls. 9647 + The IPA reference pass was significantly sped up making it 9648 feasible to enable -fipa-reference with -fprofile-generate. 9649 This also solves a bottleneck seen when building Chromium with 9650 link-time optimization. 9651 + The symbol table and call-graph API was reworked to C++ and 9652 simplified. 9653 + The interprocedural propagation of constants now also 9654 propagates alignments of pointer parameters. This for example 9655 means that the vectorizer often does not need to generate loop 9656 prologues and epilogues to make up for potential 9657 misalignments. 9658 * Link-time optimization improvements: 9659 + One Definition Rule based merging of C++ types has been 9660 implemented. Type merging enables better devirtualization and 9661 alias analysis. Streaming extra information needed to merge 9662 types adds about 2-6% of memory size and object size increase. 9663 This can be controlled by -flto-odr-type-merging. 9664 + Command-line optimization and target options are now streamed 9665 on a per-function basis and honored by the link-time 9666 optimizer. This change makes link-time optimization a more 9667 transparent replacement of per-file optimizations. It is now 9668 possible to build projects that require different optimization 9669 settings for different translation units (such as -ffast-math, 9670 -mavx, or -finline). Contrary to earlier GCC releases, the 9671 optimization and target options passed on the link command 9672 line are ignored. 9673 Note that this applies only to those command-line options that 9674 can be passed to optimize and target attributes. Command-line 9675 options affecting global code generation (such as -fpic), 9676 warnings (such as -Wodr), optimizations affecting the way 9677 static variables are optimized (such as -fcommon), debug 9678 output (such as -g), and --param parameters can be applied 9679 only to the whole link-time optimization unit. In these cases, 9680 it is recommended to consistently use the same options at both 9681 compile time and link time. 9682 + GCC bootstrap now uses slim LTO object files. 9683 + Memory usage and link times were improved. Tree merging was 9684 sped up, memory usage of GIMPLE declarations and types was 9685 reduced, and, support for on-demand streaming of variable 9686 constructors was added. 9687 * Feedback directed optimization improvements: 9688 + A new auto-FDO mode uses profiles collected by low overhead 9689 profiling tools (perf) instead of more expensive program 9690 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 9691 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 9692 traditional feedback directed optimization. 9693 + Profile precision was improved in presence of C++ inline and 9694 extern inline functions. 9695 + The new gcov-tool utility allows manipulating profiles. 9696 + Profiles are now more tolerant to source file changes (this 9697 can be controlled by --param profile-func-internal-id). 9698 * Register allocation improvements: 9699 + A new local register allocator (LRA) sub-pass, controlled by 9700 -flra-remat, implements control-flow sensitive global register 9701 rematerialization. Instead of spilling and restoring a 9702 register value, it is recalculated if it is profitable. The 9703 sub-pass improved SPEC2000 generated code by 1% and 0.5% 9704 correspondingly on ARM and x86-64. 9705 + Reuse of the PIC hard register, instead of using a fixed 9706 register, was implemented on x86/x86-64 targets. This improves 9707 generated PIC code performance as more hard registers can be 9708 used. Shared libraries can significantly benefit from this 9709 optimization. Currently it is switched on only for x86/x86-64 9710 targets. As RA infrastructure is already implemented for PIC 9711 register reuse, other targets might follow this in the future. 9712 + A simple form of inter-procedural RA was implemented. When it 9713 is known that a called function does not use caller-saved 9714 registers, save/restore code is not generated around the call 9715 for such registers. This optimization can be controlled by 9716 -fipa-ra 9717 + LRA is now much more effective at generating spills of general 9718 registers into vector registers instead of memory on 9719 architectures (e.g., modern Intel processors) where this is 9720 profitable. 9721 * UndefinedBehaviorSanitizer gained a few new sanitization options: 9722 + -fsanitize=float-divide-by-zero: detect floating-point 9723 division by zero; 9724 + -fsanitize=float-cast-overflow: check that the result of 9725 floating-point type to integer conversions do not overflow; 9726 + -fsanitize=bounds: enable instrumentation of array bounds and 9727 detect out-of-bounds accesses; 9728 + -fsanitize=alignment: enable alignment checking, detect 9729 various misaligned objects; 9730 + -fsanitize=object-size: enable object size checking, detect 9731 various out-of-bounds accesses. 9732 + -fsanitize=vptr: enable checking of C++ member function calls, 9733 member accesses and some conversions between pointers to base 9734 and derived classes, detect if the referenced object does not 9735 have the correct dynamic type. 9736 * Pointer Bounds Checker, a bounds violation detector, has been added 9737 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 9738 instrumented with run-time checks of used pointers against their 9739 bounds to detect pointer bounds violations (overflows). The Pointer 9740 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 9741 new ISA extension Intel MPX support. See the Pointer Bounds Checker 9742 [3]Wiki page for more details. 9743 9744 [4]New Languages and Language specific improvements 9745 9746 * [5]OpenMP 4.0 specification offloading features are now supported 9747 by the C, C++, and Fortran compilers. Generic changes: 9748 + Infrastructure (suitable for any vendor). 9749 + Testsuite which covers offloading from the [6]OpenMP 4.0 9750 Examples document. 9751 Specific for upcoming Intel Xeon Phi products: 9752 + Run-time library. 9753 + Card emulator. 9754 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 9755 specification. OpenACC is intended for programming accelerator 9756 devices such as GPUs. See [7]the OpenACC wiki page for more 9757 information. 9758 9759 [8]C family 9760 9761 * The default setting of the -fdiagnostics-color= command-line option 9762 is now [9]configurable when building GCC using configuration option 9763 --with-diagnostics-color=. The possible values are: never, always, 9764 auto and auto-if-env. The new default auto uses color only when the 9765 standard error is a terminal. The default in GCC 4.9 was 9766 auto-if-env, which is equivalent to auto if there is a non-empty 9767 GCC_COLORS environment variable, and never otherwise. As in GCC 9768 4.9, an empty GCC_COLORS variable in the environment will always 9769 disable colors, no matter what the default is or what command-line 9770 options are used. 9771 * A new command-line option -Wswitch-bool has been added for the C 9772 and C++ compilers, which warns whenever a switch statement has an 9773 index of boolean type. 9774 * A new command-line option -Wlogical-not-parentheses has been added 9775 for the C and C++ compilers, which warns about "logical not" used 9776 on the left hand side operand of a comparison. 9777 * A new command-line option -Wsizeof-array-argument has been added 9778 for the C and C++ compilers, which warns when the sizeof operator 9779 is applied to a parameter that has been declared as an array in a 9780 function definition. 9781 * A new command-line option -Wbool-compare has been added for the C 9782 and C++ compilers, which warns about boolean expressions compared 9783 with an integer value different from true/false. 9784 * Full support for Cilk Plus has been added to the GCC compiler. Cilk 9785 Plus is an extension to the C and C++ languages to support data and 9786 task parallelism. 9787 * A new attribute no_reorder prevents reordering of selected symbols 9788 against other such symbols or inline assembler. This enables to 9789 link-time optimize the Linux kernel without having to resort to 9790 -fno-toplevel-reorder that disables several optimizations. 9791 * New preprocessor constructs, __has_include and __has_include_next, 9792 to test the availability of headers have been added. 9793 This demonstrates a way to include the header <optional> only if it 9794 is available: 9795 9796 #ifdef __has_include 9797 # if __has_include(<optional>) 9798 # include <optional> 9799 # define have_optional 1 9800 # elif __has_include(<experimental/optional>) 9801 # include <experimental/optional> 9802 # define have_optional 1 9803 # define experimental_optional 9804 # else 9805 # define have_optional 0 9806 # endif 9807 #endif 9808 9809 The header search paths for __has_include and __has_include_next 9810 are equivalent to those of the standard directive #include and the 9811 extension #include_next respectively. 9812 * A new built-in function-like macro to determine the existence of an 9813 attribute, __has_attribute, has been added. The equivalent built-in 9814 macro __has_cpp_attribute was added to C++ to support 9815 [10]Feature-testing recommendations for C++. The macro 9816 __has_attribute is added to all C-like languages as an extension: 9817 9818 int 9819 #ifdef __has_attribute 9820 # if __has_attribute(__noinline__) 9821 __attribute__((__noinline__)) 9822 # endif 9823 #endif 9824 foo(int x); 9825 9826 If an attribute exists, a nonzero constant integer is returned. For 9827 standardized C++ attributes a date is returned, otherwise the 9828 constant returned is 1. Both __has_attribute and 9829 __has_cpp_attribute will add underscores to an attribute name if 9830 necessary to resolve the name. For C++11 and onwards the attribute 9831 may be scoped. 9832 * A new set of built-in functions for arithmetics with overflow 9833 checking has been added: __builtin_add_overflow, 9834 __builtin_sub_overflow and __builtin_mul_overflow and for 9835 compatibility with clang also other variants. These builtins have 9836 two integral arguments (which don't need to have the same type), 9837 the arguments are extended to infinite precision signed type, +, - 9838 or * is performed on those, and the result is stored in an integer 9839 variable pointed to by the last argument. If the stored value is 9840 equal to the infinite precision result, the built-in functions 9841 return false, otherwise true. The type of the integer variable that 9842 will hold the result can be different from the types of the first 9843 two arguments. The following snippet demonstrates how this can be 9844 used in computing the size for the calloc function: 9845 9846 void * 9847 calloc (size_t x, size_t y) 9848 { 9849 size_t sz; 9850 if (__builtin_mul_overflow (x, y, &sz)) 9851 return NULL; 9852 void *ret = malloc (sz); 9853 if (ret) memset (res, 0, sz); 9854 return ret; 9855 } 9856 9857 On e.g. i?86 or x86-64 the above will result in a mul instruction 9858 followed by a jump on overflow. 9859 * The option -fextended-identifiers is now enabled by default for 9860 C++, and for C99 and later C versions. Various bugs in the 9861 implementation of extended identifiers have been fixed. 9862 9863 [11]C 9864 9865 * The default mode has been changed to -std=gnu11. 9866 * A new command-line option -Wc90-c99-compat has been added to warn 9867 about features not present in ISO C90, but present in ISO C99. 9868 * A new command-line option -Wc99-c11-compat has been added to warn 9869 about features not present in ISO C99, but present in ISO C11. 9870 * It is possible to disable warnings about conversions between 9871 pointers that have incompatible types via a new warning option 9872 -Wno-incompatible-pointer-types; warnings about implicit 9873 incompatible integer to pointer and pointer to integer conversions 9874 via a new warning option -Wno-int-conversion; and warnings about 9875 qualifiers on pointers being discarded via a new warning option 9876 -Wno-discarded-qualifiers. 9877 * To allow proper use of const qualifiers with multidimensional 9878 arrays, GCC will not warn about incompatible pointer types anymore 9879 for conversions between pointers to arrays with and without const 9880 qualifier (except when using -pedantic). Instead, a new warning is 9881 emitted only if the const qualifier is lost. This can be controlled 9882 with a new warning option -Wno-discarded-array-qualifiers. 9883 * The C front end now generates more precise caret diagnostics. 9884 * The -pg command-line option now only affects the current file in an 9885 LTO build. 9886 9887 [12]C++ 9888 9889 * G++ now supports [13]C++14 variable templates. 9890 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 9891 * Excessive template instantiation depth is now a fatal error. This 9892 prevents excessive diagnostics that usually do not help to identify 9893 the problem. 9894 * G++ and libstdc++ now implement the feature-testing macros from 9895 [14]Feature-testing recommendations for C++. 9896 * G++ now allows typename in a template template parameter. 9897 9898 template<template<typename> typename X> struct D; // OK 9899 9900 * G++ now supports [15]C++14 aggregates with non-static data member 9901 initializers. 9902 9903 struct A { int i, j = i; }; 9904 A a = { 42 }; // a.j is also 42 9905 9906 * G++ now supports [16]C++14 extended constexpr. 9907 9908 constexpr int f (int i) 9909 { 9910 int j = 0; 9911 for (; i > 0; --i) 9912 ++j; 9913 return j; 9914 } 9915 9916 constexpr int i = f(42); // i is 42 9917 9918 * G++ now supports the [17]C++14 sized deallocation functions. 9919 9920 void operator delete (void *, std::size_t) noexcept; 9921 void operator delete[] (void *, std::size_t) noexcept; 9922 9923 * A new One Definition Rule violation warning (controlled by -Wodr) 9924 detects mismatches in type definitions and virtual table contents 9925 during link-time optimization. 9926 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 9927 developers to annotate programs with final specifiers (or anonymous 9928 namespaces) to improve code generation. These warnings can be used 9929 at compile time, but they are more useful in combination with 9930 link-time optimization. 9931 * G++ no longer supports [18]N3639 variable length arrays, as they 9932 were removed from the C++14 working paper prior to ratification. 9933 GNU VLAs are still supported, so VLA support is now the same in 9934 C++14 mode as in C++98 and C++11 modes. 9935 * G++ now allows passing a non-trivially-copyable class via C 9936 varargs, which is conditionally-supported with 9937 implementation-defined semantics in the standard. This uses the 9938 same calling convention as a normal value parameter. 9939 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 9940 various mangling bugs are fixed, but G++ will still emit aliases 9941 with the old, wrong mangling where feasible. -Wabi=2 will warn 9942 about differences between ABI version 2 and the current setting. 9943 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 9944 to be unaffected, but -Wabi=8 will warn about a non-static data 9945 member with type std::nullptr_t which changes position due to this 9946 change. 9947 9948 [19]Runtime Library (libstdc++) 9949 9950 * A [20]Dual ABI is provided by the library. A new ABI is enabled by 9951 default. The old ABI is still supported and can be used by defining 9952 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 9953 standard library headers. 9954 * A new implementation of std::string is enabled by default, using 9955 the small string optimization instead of copy-on-write reference 9956 counting. 9957 * A new implementation of std::list is enabled by default, with an 9958 O(1) size() function; 9959 * [21]Full support for C++11, including the following new features: 9960 + std::deque and std::vector<bool> meet the allocator-aware 9961 container requirements; 9962 + movable and swappable iostream classes; 9963 + support for std::align and std::aligned_union; 9964 + type traits std::is_trivially_copyable, 9965 std::is_trivially_constructible, std::is_trivially_assignable 9966 etc.; 9967 + I/O manipulators std::put_time, std::get_time, std::hexfloat 9968 and std::defaultfloat; 9969 + generic locale-aware std::isblank; 9970 + locale facets for Unicode conversion; 9971 + atomic operations for std::shared_ptr; 9972 + std::notify_all_at_thread_exit() and functions for making 9973 futures ready at thread exit. 9974 * Support for the C++11 hexfloat manipulator changes how the num_put 9975 facet formats floating point types when 9976 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 9977 This change affects all language modes, even though the C++98 9978 standard gave no special meaning to that combination of flags. To 9979 prevent the use of hexadecimal notation for floating point types 9980 use str.unsetf(std::ios_base::floatfield) to clear the relevant 9981 bits in str.flags(). 9982 * [22]Full experimental support for C++14, including the following 9983 new features: 9984 + std::is_final type trait; 9985 + heterogeneous comparison lookup in associative containers. 9986 + global functions cbegin, cend, rbegin, rend, crbegin, and 9987 crend for range access to containers, arrays and initializer 9988 lists. 9989 * [23]Improved experimental support for the Library Fundamentals TS, 9990 including: 9991 + class std::experimental::any; 9992 + function template std::experimental::apply; 9993 + function template std::experimental::sample; 9994 + function template std::experimental::search and related 9995 searcher types; 9996 + variable templates for type traits; 9997 + function template std::experimental::not_fn. 9998 * New random number distributions logistic_distribution and 9999 uniform_on_sphere_distribution as extensions. 10000 * [24]GDB Xmethods for containers and std::unique_ptr. 10001 10002 [25]Fortran 10003 10004 * Compatibility notice: 10005 + The version of the module files (.mod) has been incremented. 10006 + For free-form source files [26]-Werror=line-truncation is now 10007 enabled by default. Note that comments exceeding the line 10008 length are not diagnosed. (For fixed-form source code, the 10009 same warning is available but turned off by default, such that 10010 excess characters are ignored. -ffree-line-length-n and 10011 -ffixed-line-length-n can be used to modify the default line 10012 lengths of 132 and 72 columns, respectively.) 10013 + The -Wtabs option is now more sensible: with -Wtabs the 10014 compiler warns if it encounters tabs and with -Wno-tabs this 10015 warning is turned off. Before, -Wno-tabs warned and -Wtabs 10016 disabled the warning. As before, this warning is also enabled 10017 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 10018 options of -std=. 10019 * Incomplete support for colorizing diagnostics emitted by gfortran 10020 has been added. The option [27]-fdiagnostics-color controls when 10021 color is used in diagnostics. The default value of this option can 10022 be [28]configured when building GCC. The GCC_COLORS environment 10023 variable can be used to customize the colors or disable coloring 10024 completely. Sample diagnostics output: 10025 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 10026 test.f90:6:1: 10027 10028 0 continue 10029 1 10030 Error: Zero is not a valid statement label at (1) 10031 test.f90:9:6: 10032 10033 USE foo 10034 1 10035 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 10036 10037 * The -Wuse-without-only option has been added to warn when a USE 10038 statement has no ONLY qualifier and thus implicitly imports all 10039 public entities of the used module. 10040 * Formatted READ and WRITE statements now work correctly in 10041 locale-aware programs. For more information and potential caveats, 10042 see [29]Section 5.3 Thread-safety of the runtime library in the 10043 manual. 10044 * [30]Fortran 2003: 10045 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 10046 IEEE_ARITHMETIC) are now supported. 10047 * [31]Fortran 2008: 10048 + [32]Coarrays: Full experimental support of Fortran 2008's 10049 coarrays with -fcoarray=lib except for allocatable/pointer 10050 components of derived-type coarrays. GCC currently only ships 10051 with a single-image library (libcaf_single), but multi-image 10052 support based on MPI and GASNet is provided by the libraries 10053 of the [33]OpenCoarrays project. 10054 * TS18508 Additional Parallel Features in Fortran: 10055 + Support for the collective intrinsic subroutines CO_MAX, 10056 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 10057 including -fcoarray=lib support. 10058 + Support for the new atomic intrinsics has been added, 10059 including -fcoarray=lib support. 10060 * Fortran 2015: 10061 + Support for IMPLICIT NONE (external, type). 10062 + ERROR STOP is now permitted in pure procedures. 10063 10064 [34]Go 10065 10066 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 10067 * Building GCC 5 with Go enabled will install two new programs: 10068 [35]go and [36]gofmt. 10069 10070 [37]libgccjit 10071 10072 New in GCC 5 is the ability to build GCC as a shared library for 10073 embedding in other processes (such as interpreters), suitable for 10074 Just-In-Time compilation to machine code. 10075 10076 The shared library has a [38]C API and a [39]C++ wrapper API providing 10077 some "syntactic sugar". There are also bindings available from 3rd 10078 parties for [40]Python and for [41]D. 10079 10080 For example, this library can be used by interpreters for [42]compiling 10081 functions from bytecode to machine code. 10082 10083 The library can also be used for ahead-of-time compilation, enabling 10084 GCC to be plugged into a pre-existing front end. An example of using 10085 this to build a compiler for an esoteric language we'll refer to as 10086 "brainf" can be seen [43]here. 10087 10088 libgccjit is licensed under the GPLv3 (or at your option, any later 10089 version) 10090 10091 It should be regarded as experimental at this time. 10092 10093 [44]New Targets and Target Specific Improvements 10094 10095 [45]Reporting stack usage 10096 10097 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 10098 -fstack-usage option. 10099 10100 [46]AArch64 10101 10102 * Code generation for the ARM Cortex-A57 processor has been improved. 10103 A more accurate instruction scheduling model for the processor is 10104 now used, and a number of compiler tuning parameters have been set 10105 to offer increased performance when compiling with -mcpu=cortex-a57 10106 or -mtune=cortex-a57. 10107 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 10108 and can be enabled by giving the -mfix-cortex-a53-835769 option. 10109 Alternatively it can be enabled by default by configuring GCC with 10110 the --enable-fix-cortex-a53-835769 option. 10111 * The optional cryptographic extensions to the ARMv8-A architecture 10112 are no longer enabled by default when specifying the 10113 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 10114 options. To enable these extensions add +crypto to the value of 10115 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 10116 * Support has been added for the following processors (GCC 10117 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 10118 initial support for its big.LITTLE combination with the ARM 10119 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 10120 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 10121 The GCC identifiers can be used as arguments to the -mcpu or -mtune 10122 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 10123 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 10124 support for the Cortex-A72. 10125 * The transitional options -mlra and -mno-lra have been removed. The 10126 AArch64 back end now uses the local register allocator (LRA) only. 10127 10128 [47]ARM 10129 10130 * Thumb-1 assembly code is now generated in unified syntax. The new 10131 option -masm-syntax-unified specifies whether inline assembly code 10132 is using unified syntax. By default the option is off which means 10133 non-unified syntax is used. However this is subject to change in 10134 future releases. Eventually the non-unified syntax will be 10135 deprecated. 10136 * It is now a configure-time error to use the --with-cpu configure 10137 option with either of --with-tune or --with-arch. 10138 * Code generation for the ARM Cortex-A57 processor has been improved. 10139 A more accurate instruction scheduling model for the processor is 10140 now used, and a number of compiler tuning parameters have been set 10141 to offer increased performance when compiling with -mcpu=cortex-a57 10142 or -mtune=cortex-a57. 10143 * Support has been added for the following processors (GCC 10144 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 10145 initial support for its big.LITTLE combination with the ARM 10146 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 10147 initial support for its big.LITTLE combination with the ARM 10148 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 10149 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 10150 The GCC identifiers can be used as arguments to the -mcpu or -mtune 10151 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 10152 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 10153 support for the Cortex-A72. 10154 * The deprecated option -mwords-little-endian has been removed. 10155 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 10156 -mtpcs-leaf-frame which are only applicable to the old ABI have 10157 been deprecated. 10158 * The transitional options -mlra and -mno-lra have been removed. The 10159 ARM back end now uses the local register allocator (LRA) only. 10160 10161 [48]AVR 10162 10163 * Support has been added for the devices ATtiny4/5/9/10/20/40. This 10164 requires Binutils 2.25 or newer. 10165 * The port uses a new scheme to describe supported devices: For each 10166 supported device, the compiler provides a device-specific [49]spec 10167 file. If the compiler is used together with AVR-LibC, this requires 10168 at least GCC 5.2 and a version of AVR-LibC which implements 10169 [50]feature #44574. 10170 As a consequence, the compiler no more supports individual devices 10171 like ATmega8. Specifying, say, -mmcu=atmega8 triggers the usage of 10172 the device-specific [51]spec file specs-atmega8 which is part of 10173 the installation and describes options for the sub-processes like 10174 compiler proper, assembler and linker. You can add support for a 10175 new device -mmcu=mydevice as follows: 10176 1. In an empty directory /someplace, create a new directory 10177 device-specs. 10178 2. Copy a device spec file from the installed device-specs 10179 folder, follow the comments in that file and then save it as 10180 /someplace/device-specs/specs-mydevice. 10181 3. Add -B /someplace -mmcu=mydevice to the compiler's 10182 command-line options. Notice that /someplace must specify an 10183 absolute path and that mydevice must not start with "avr". 10184 4. Provided you have a device-specific library libmydevice.a 10185 available, you can put it at /someplace, dito for a 10186 device-specific startup file crtmydevice.o. 10187 The contents of the device spec files depend on the compiler's 10188 configuration, in particular on --with-avrlibc=no and whether or 10189 not it is configured for RTEMS. 10190 * A new command-line option -nodevicelib has been added. It prevents 10191 the compiler from linking against AVR-LibC's device-specific 10192 library libdevice.a. 10193 * The following three command-line options have been added: 10194 10195 -mrmw 10196 Set if the device supports the read-modify-write 10197 instructions LAC, LAS, LAT and XCH. 10198 10199 -mn-flash=size 10200 Specify the flash size of the device in units of 64 KiB, 10201 rounded up to the next integer as needed. This option 10202 affects the availability of the [52]AVR address-spaces. 10203 10204 -mskip-bug 10205 Set if the device is affected by the respective silicon 10206 bug. 10207 10208 These options are used internally in order to communicate between 10209 the compiler and the device-spces file. They are set in the 10210 device-specs file as needed. Don't set them by hand. 10211 10212 [53]IA-32/x86-64 10213 10214 * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's 10215 CPU codenamed Skylake Server was added to GCC. That includes inline 10216 assembly support, new intrinsics, and basic autovectorization. 10217 These new AVX-512 extensions are available via the following GCC 10218 switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512 10219 Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword 10220 instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma 10221 and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi. 10222 * New ISA extensions support Intel MPX was added to GCC. This new 10223 extension is available via the -mmpx compiler switch. Intel MPX is 10224 a set of processor features which, with compiler, run-time library 10225 and OS support, brings increased robustness to software by run-time 10226 checking pointer references against their bounds. In GCC Intel MPX 10227 is supported by Pointer Bounds Checker and libmpx run-time 10228 libraries. 10229 * The new -mrecord-mcount option for -pg generates a Linux kernel 10230 style table of pointers to mcount or __fentry__ calls at the 10231 beginning of functions. The new -mnop-mcount option in addition 10232 also generates nops in place of the __fentry__ or mcount call, so 10233 that a call per function can be later patched in. This can be used 10234 for low overhead tracing or hot code patching. 10235 * The new -malign-data option controls how GCC aligns variables. 10236 -malign-data=compat uses increased alignment compatible with GCC 10237 4.8 and earlier, -malign-data=abi uses alignment as specified by 10238 the psABI, and -malign-data=cacheline uses increased alignment to 10239 match the cache line size. -malign-data=compat is the default. 10240 * The new -mskip-rax-setup option skips setting up the RAX register 10241 when SSE is disabled and there are no variable arguments passed in 10242 vector registers. This can be used to optimize the Linux kernel. 10243 10244 [54]MIPS 10245 10246 * MIPS Releases 3 and 5 are now directly supported. Use the 10247 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 10248 to enable code-generation for these processors. 10249 * The Imagination P5600 processor is now supported using the 10250 -march=p5600 command-line option. 10251 * The Cavium Octeon3 processor is now supported using the 10252 -march=octeon3 command-line option. 10253 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 10254 command-line options. 10255 * The o32 ABI has been modified and extended. The o32 64-bit 10256 floating-point register support is now obsolete and has been 10257 removed. It has been replaced by three ABI extensions FPXX, FP64A, 10258 and FP64. The meaning of the -mfp64 command-line option has 10259 changed. It is now used to enable the FP64A and FP64 ABI 10260 extensions. 10261 + The FPXX extension requires that code generated to access 10262 double-precision values use even-numbered registers. Code that 10263 adheres to this extension is link-compatible with all other 10264 o32 double-precision ABI variants and will execute correctly 10265 in all hardware FPU modes. The command-line options -mabi=32 10266 -mfpxx can be used to enable this extension. MIPS II is the 10267 minimum processor required. 10268 + The o32 FP64A extension requires that floating-point registers 10269 be 64-bit and odd-numbered single-precision registers are not 10270 allowed. Code that adheres to the o32 FP64A variant is 10271 link-compatible with all other o32 double-precision ABI 10272 variants. The command-line options -mabi=32 -mfp64 10273 -mno-odd-spreg can be used to enable this extension. MIPS32R2 10274 is the minimum processor required. 10275 + The o32 FP64 extension also requires that floating-point 10276 registers be 64-bit, but permits the use of single-precision 10277 registers. Code that adheres to the o32 FP64 variant is 10278 link-compatible with o32 FPXX and o32 FP64A variants only, 10279 i.e. it is not compatible with the original o32 10280 double-precision ABI. The command-line options -mabi=32 -mfp64 10281 -modd-spreg can be used to enable this extension. MIPS32R2 is 10282 the minimum processor required. 10283 The new ABI variants can be enabled by default using the configure 10284 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 10285 It is strongly recommended that all vendors begin to set o32 FPXX 10286 as the default ABI. This will be required to run the generated code 10287 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 10288 MIPSR6 cores. 10289 * GCC will now pass all floating-point options to the assembler if 10290 GNU binutils 2.25 is used. As a result, any inline assembly code 10291 that uses hard-float instructions should be amended to include a 10292 .set directive to override the global assembler options when 10293 compiling for soft-float targets. 10294 10295 [55]NDS32 10296 10297 * The variadic function ABI implementation is now compatible with 10298 past Andes toolchains where the caller uses registers to pass 10299 arguments and the callee is in charge of pushing them on stack. 10300 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 10301 been removed since they are not yet available in the nds32 port of 10302 GNU binutils. 10303 * A new option -mcmodel=[small|medium|large] supports varied code 10304 models on code generation. The -mgp-direct option became 10305 meaningless and can be discarded. 10306 10307 [56]RX 10308 10309 * A new command-line option -mno-allow-string-insns can be used to 10310 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 10311 SWHILE and RMPA instructions. An erratum released by Renesas shows 10312 that it is unsafe to use these instructions on addresses within the 10313 I/O space of the processor. The new option can be used when the 10314 programmer is concerned that the I/O space might be accessed. The 10315 default is still to enable these instructions. 10316 10317 [57]SH 10318 10319 * The compiler will now pass the appropriate --isa= option to the 10320 assembler. 10321 * The default handling for the GBR has been changed from call 10322 clobbered to call preserved. The old behavior can be reinstated by 10323 specifying the option -fcall-used-gbr. 10324 * Support for the SH4A fpchg instruction has been added which will be 10325 utilized when switching between single and double precision FPU 10326 modes. 10327 * The compiler no longer uses the __fpscr_values array for switching 10328 between single and double FPU precision modes on non-SH4A targets. 10329 Instead mode switching will now be performed by storing, modifying 10330 and reloading the FPSCR, so that other FPSCR bits are preserved 10331 across mode switches. The __fpscr_values array that is defined in 10332 libgcc is still present for backwards compatibility, but it will 10333 not be referenced by compiler generated code anymore. 10334 * New builtin functions __builtin_sh_get_fpscr and 10335 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 10336 function will mask the specified bits in such a way that the SZ, PR 10337 and FR mode bits will be preserved, while changing the other bits. 10338 These new functions do not reference the __fpscr_values array. The 10339 old functions __set_fpscr and __get_fpscr in libgcc which access 10340 the __fpscr_values array are still present for backwards 10341 compatibility, but their usage is highly discouraged. 10342 * Some improvements to code generated for __atomic built-in 10343 functions. 10344 * When compiling for SH2E the compiler will no longer force the usage 10345 of delay slots for conditional branch instructions bt and bf. The 10346 old behavior can be reinstated (e.g. to work around a hardware bug 10347 in the original SH7055) by specifying the new option 10348 -mcbranch-force-delay-slot. 10349 10350 [58]Operating Systems 10351 10352 [59]AIX 10353 10354 * GCC now supports stabs debugging continuation lines to allow long 10355 stabs debug information without overflow that generates AIX linker 10356 errors. 10357 10358 [60]DragonFly BSD 10359 10360 * GCC now supports the DragonFly BSD operating system. 10361 10362 [61]FreeBSD 10363 10364 * GCC now supports the FreeBSD operating system for the arm port 10365 through the arm*-*-freebsd* target triplets. 10366 10367 [62]VxWorks MILS 10368 10369 * GCC now supports the MILS (Multiple Independent Levels of Security) 10370 variant of WindRiver's VxWorks operating system for PowerPC 10371 targets. 10372 10373 Other significant improvements 10374 10375 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 10376 to set the compiler to use. 10377 10378 * When the new command-line option -freport-bug is used, GCC 10379 automatically generates a developer-friendly reproducer whenever an 10380 internal compiler error is encountered. 10381 10382 [63]GCC 5.2 10383 10384 This is the [64]list of problem reports (PRs) from GCC's bug tracking 10385 system that are known to be fixed in the 5.2 release. This list might 10386 not be complete (that is, it is possible that some PRs that have been 10387 fixed are not listed here). 10388 10389 Target Specific Changes 10390 10391 IA-32/x86-64 10392 10393 * Support for new AMD instructions monitorx and mwaitx has been 10394 added. This includes new intrinsic and built-in support. It is 10395 enabled through option -mmwaitx. The instructions monitorx and 10396 mwaitx implement the same functionality as the old monitor and 10397 mwait instructions. In addition, mwaitx adds a configurable timer. 10398 The timer value is received as third argument and stored in 10399 register %ebx. 10400 10401 S/390, System z, IBM z Systems 10402 10403 * Support for the IBM z13 processor has been added. When using the 10404 -march=z13 option, the compiler will generate code making use of 10405 the new instructions and registers introduced with the vector 10406 extension facility. The -mtune=z13 option enables z13 specific 10407 instruction scheduling without making use of new instructions. 10408 Compiling code with -march=z13 reduces the default alignment of 10409 vector types bigger than 8 bytes to 8. This is an ABI change and 10410 care must be taken when linking modules compiled with different 10411 arch levels which interchange variables containing vector type 10412 values. For newly compiled code the GNU linker will emit a warning. 10413 * The -mzvector option enables a C/C++ language extension. This 10414 extension provides a new keyword vector which can be used to define 10415 vector type variables. (Note: This is not available when enforcing 10416 strict standard compliance e.g. with -std=c99. Either enable GNU 10417 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 10418 Additionally a set of overloaded builtins is provided which is 10419 partially compatible to the PowerPC Altivec builtins. In order to 10420 make use of these builtins the vecintrin.h header file needs to be 10421 included. 10422 10423 [65]GCC 5.3 10424 10425 This is the [66]list of problem reports (PRs) from GCC's bug tracking 10426 system that are known to be fixed in the 5.3 release. This list might 10427 not be complete (that is, it is possible that some PRs that have been 10428 fixed are not listed here). 10429 10430 Target Specific Changes 10431 10432 IA-32/x86-64 10433 10434 * GCC now supports the Intel CPU named Skylake with AVX-512 10435 extensions through -march=skylake-avx512. The switch enables the 10436 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 10437 AVX-512DQ. 10438 10439 S/390, System z, IBM z Systems 10440 10441 * With this version of GCC IBM z Systems support has been added to 10442 the GO runtime environment. GCC 5.3 has proven to be able to 10443 compile larger GO applications on IBM z Systems. 10444 10445 [67]GCC 5.4 10446 10447 This is the [68]list of problem reports (PRs) from GCC's bug tracking 10448 system that are known to be fixed in the 5.4 release. This list might 10449 not be complete (that is, it is possible that some PRs that have been 10450 fixed are not listed here). 10451 10452 [69]GCC 5.5 10453 10454 This is the [70]list of problem reports (PRs) from GCC's bug tracking 10455 system that are known to be fixed in the 5.5 release. This list might 10456 not be complete (that is, it is possible that some PRs that have been 10457 fixed are not listed here). 10458 10459 Target Specific Changes 10460 10461 IA-32/x86-64 10462 10463 * Support for the pcommit instruction has been removed. 10464 10465 10466 For questions related to the use of GCC, please consult these web 10467 pages and the [71]GCC manuals. If that fails, the 10468 [72]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10469 web pages and the development of GCC are welcome on our developer 10470 list at [73]gcc (a] gcc.gnu.org. All of [74]our lists have public 10471 archives. 10472 10473 Copyright (C) [75]Free Software Foundation, Inc. Verbatim copying and 10474 distribution of this entire article is permitted in any medium, 10475 provided this notice is preserved. 10476 10477 These pages are [76]maintained by the GCC team. Last modified 10478 2026-06-03. 10479 10480 References 10481 10482 1. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 10483 2. https://gcc.gnu.org/gcc-5/changes.html#general 10484 3. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler 10485 4. https://gcc.gnu.org/gcc-5/changes.html#languages 10486 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 10487 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 10488 7. https://gcc.gnu.org/wiki/OpenACC 10489 8. https://gcc.gnu.org/gcc-5/changes.html#c-family 10490 9. https://gcc.gnu.org/install/configure.html 10491 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 10492 11. https://gcc.gnu.org/gcc-5/changes.html#c 10493 12. https://gcc.gnu.org/gcc-5/changes.html#cxx 10494 13. https://gcc.gnu.org/projects/cxx1y.html 10495 14. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 10496 15. https://gcc.gnu.org/projects/cxx1y.html 10497 16. https://gcc.gnu.org/projects/cxx1y.html 10498 17. https://gcc.gnu.org/projects/cxx1y.html 10499 18. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 10500 19. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 10501 20. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 10502 21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 10503 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 10504 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 10505 24. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python 10506 25. https://gcc.gnu.org/gcc-5/changes.html#fortran 10507 26. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 10508 27. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 10509 28. https://gcc.gnu.org/install/configure.html 10510 29. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 10511 30. https://gcc.gnu.org/wiki/Fortran2003Status 10512 31. https://gcc.gnu.org/wiki/Fortran2008Status 10513 32. https://gcc.gnu.org/wiki/Coarray 10514 33. http://www.opencoarrays.org/ 10515 34. https://gcc.gnu.org/gcc-5/changes.html#go 10516 35. https://pkg.go.dev/cmd/go 10517 36. https://pkg.go.dev/cmd/gofmt 10518 37. https://gcc.gnu.org/gcc-5/changes.html#jit 10519 38. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 10520 39. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 10521 40. https://github.com/davidmalcolm/pygccjit 10522 41. https://github.com/ibuclaw/gccjitd 10523 42. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 10524 43. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 10525 44. https://gcc.gnu.org/gcc-5/changes.html#targets 10526 45. https://gcc.gnu.org/gcc-5/changes.html#stack-usage 10527 46. https://gcc.gnu.org/gcc-5/changes.html#aarch64 10528 47. https://gcc.gnu.org/gcc-5/changes.html#arm 10529 48. https://gcc.gnu.org/gcc-5/changes.html#avr 10530 49. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 10531 50. https://savannah.nongnu.org/bugs/?44574 10532 51. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 10533 52. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 10534 53. https://gcc.gnu.org/gcc-5/changes.html#x86 10535 54. https://gcc.gnu.org/gcc-5/changes.html#mips 10536 55. https://gcc.gnu.org/gcc-5/changes.html#nds32 10537 56. https://gcc.gnu.org/gcc-5/changes.html#rx 10538 57. https://gcc.gnu.org/gcc-5/changes.html#sh 10539 58. https://gcc.gnu.org/gcc-5/changes.html#os 10540 59. https://gcc.gnu.org/gcc-5/changes.html#aix 10541 60. https://gcc.gnu.org/gcc-5/changes.html#dragonfly 10542 61. https://gcc.gnu.org/gcc-5/changes.html#freebsd 10543 62. https://gcc.gnu.org/gcc-5/changes.html#vxmils 10544 63. https://gcc.gnu.org/gcc-5/changes.html#GCC5.2 10545 64. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 10546 65. https://gcc.gnu.org/gcc-5/changes.html#GCC5.3 10547 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 10548 67. https://gcc.gnu.org/gcc-5/changes.html#GCC5.4 10549 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 10550 69. https://gcc.gnu.org/gcc-5/changes.html#GCC5.5 10551 70. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 10552 71. https://gcc.gnu.org/onlinedocs/ 10553 72. mailto:gcc-help (a] gcc.gnu.org 10554 73. mailto:gcc (a] gcc.gnu.org 10555 74. https://gcc.gnu.org/lists.html 10556 75. https://www.fsf.org/ 10557 76. https://gcc.gnu.org/about.html 10558 ====================================================================== 10559 http://gcc.gnu.org/gcc-4.9/index.html 10560 GCC 4.9 Release Series 10561 10562 (This release series is no longer supported.) 10563 10564 Aug 3, 2016 10565 10566 The [1]GNU project and the GCC developers are pleased to announce the 10567 release of GCC 4.9.4. 10568 10569 This release is a bug-fix release, containing fixes for regressions in 10570 GCC 4.9.3 relative to previous releases of GCC. 10571 10572 Release History 10573 10574 GCC 4.9.4 10575 Aug 3, 2016 ([2]changes, [3]documentation) 10576 10577 GCC 4.9.3 10578 June 26, 2015 ([4]changes, [5]documentation) 10579 10580 GCC 4.9.2 10581 October 30, 2014 ([6]changes, [7]documentation) 10582 10583 GCC 4.9.1 10584 July 16, 2014 ([8]changes, [9]documentation) 10585 10586 GCC 4.9.0 10587 April 22, 2014 ([10]changes, [11]documentation) 10588 10589 References and Acknowledgements 10590 10591 GCC used to stand for the GNU C Compiler, but since the compiler 10592 supports several other languages aside from C, it now stands for the 10593 GNU Compiler Collection. 10594 10595 A list of [12]successful builds is updated as new information becomes 10596 available. 10597 10598 The GCC developers would like to thank the numerous people that have 10599 contributed new features, improvements, bug fixes, and other changes as 10600 well as test results to GCC. This [13]amazing group of volunteers is 10601 what makes GCC successful. 10602 10603 For additional information about GCC please refer to the [14]GCC 10604 project web site or contact the [15]GCC development mailing list. 10605 10606 To obtain GCC please use [16]our mirror sites or [17]our version 10607 control system. 10608 10609 10610 For questions related to the use of GCC, please consult these web 10611 pages and the [18]GCC manuals. If that fails, the 10612 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10613 web pages and the development of GCC are welcome on our developer 10614 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 10615 archives. 10616 10617 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 10618 distribution of this entire article is permitted in any medium, 10619 provided this notice is preserved. 10620 10621 These pages are [23]maintained by the GCC team. Last modified 10622 2024-05-30. 10623 10624 References 10625 10626 1. http://www.gnu.org/ 10627 2. https://gcc.gnu.org/gcc-4.9/changes.html 10628 3. https://gcc.gnu.org/onlinedocs/4.9.4/ 10629 4. https://gcc.gnu.org/gcc-4.9/changes.html 10630 5. https://gcc.gnu.org/onlinedocs/4.9.3/ 10631 6. https://gcc.gnu.org/gcc-4.9/changes.html 10632 7. https://gcc.gnu.org/onlinedocs/4.9.2/ 10633 8. https://gcc.gnu.org/gcc-4.9/changes.html 10634 9. https://gcc.gnu.org/onlinedocs/4.9.1/ 10635 10. https://gcc.gnu.org/gcc-4.9/changes.html 10636 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 10637 12. https://gcc.gnu.org/gcc-4.9/buildstat.html 10638 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10639 14. https://gcc.gnu.org/index.html 10640 15. mailto:gcc (a] gcc.gnu.org 10641 16. https://gcc.gnu.org/mirrors.html 10642 17. https://gcc.gnu.org/git.html 10643 18. https://gcc.gnu.org/onlinedocs/ 10644 19. mailto:gcc-help (a] gcc.gnu.org 10645 20. mailto:gcc (a] gcc.gnu.org 10646 21. https://gcc.gnu.org/lists.html 10647 22. https://www.fsf.org/ 10648 23. https://gcc.gnu.org/about.html 10649 ====================================================================== 10650 http://gcc.gnu.org/gcc-4.9/changes.html 10651 GCC 4.9 Release Series 10652 Changes, New Features, and Fixes 10653 10654 Caveats 10655 10656 * The mudflap run time checker has been removed. The mudflap options 10657 remain, but do nothing. 10658 * Support for a number of older systems and recently unmaintained or 10659 untested target ports of GCC has been declared obsolete in GCC 4.9. 10660 Unless there is activity to revive them, the next release of GCC 10661 will have their sources permanently removed. 10662 The following ports for individual systems on particular 10663 architectures have been obsoleted: 10664 + Solaris 9 (*-*-solaris2.9). Details can be found in the 10665 [1]announcement. 10666 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 10667 float64x1_t exported by arm_neon.h are defined to be the same as 10668 their base types. This results in incorrect application of 10669 parameter passing rules to arguments of types int64x1_t and 10670 uint64x1_t, with respect to the AAPCS64 ABI specification. In 10671 addition, names of C++ functions with parameters of these types 10672 (including float64x1_t) are not mangled correctly. The current 10673 typedef declarations also unintentionally allow implicit casting 10674 between singleton vector types and their base types. These issues 10675 will be resolved in a near future release. See [2]PR60825 for more 10676 information. 10677 10678 More information on porting to GCC 4.9 from previous versions of GCC 10679 can be found in the [3]porting guide for this release. 10680 10681 General Optimizer Improvements 10682 10683 * AddressSanitizer, a fast memory error detector, is now available on 10684 ARM. 10685 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 10686 detector, has been added and can be enabled via 10687 -fsanitize=undefined. Various computations will be instrumented to 10688 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 10689 currently available for the C and C++ languages. 10690 * Link-time optimization (LTO) improvements: 10691 + Type merging was rewritten. The new implementation is 10692 significantly faster and uses less memory. 10693 + Better partitioning algorithm resulting in less streaming 10694 during link time. 10695 + Early removal of virtual methods reduces the size of object 10696 files and improves link-time memory usage and compile time. 10697 + Function bodies are now loaded on-demand and released early 10698 improving overall memory usage at link time. 10699 + C++ hidden keyed methods can now be optimized out. 10700 + When using a linker plugin, compiling with the -flto option 10701 now generates slim object files (.o) which only contain 10702 intermediate language representation for LTO. Use 10703 -ffat-lto-objects to create files which contain additionally 10704 the object code. To generate static libraries suitable for LTO 10705 processing, use gcc-ar and gcc-ranlib; to list symbols from a 10706 slim object file use gcc-nm. (This requires that ar, ranlib 10707 and nm have been compiled with plugin support.) 10708 Memory usage building Firefox with debug enabled was reduced from 10709 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 10710 * Inter-procedural optimization improvements: 10711 + New type inheritance analysis module improving 10712 devirtualization. Devirtualization now takes into account 10713 anonymous name-spaces and the C++11 final keyword. 10714 + New speculative devirtualization pass (controlled by 10715 -fdevirtualize-speculatively. 10716 + Calls that were speculatively made direct are turned back to 10717 indirect where direct call is not cheaper. 10718 + Local aliases are introduced for symbols that are known to be 10719 semantically equivalent across shared libraries improving 10720 dynamic linking times. 10721 * Feedback directed optimization improvements: 10722 + Profiling of programs using C++ inline functions is now more 10723 reliable. 10724 + New time profiling determines typical order in which functions 10725 are executed. 10726 + A new function reordering pass (controlled by 10727 -freorder-functions) significantly reduces startup time of 10728 large applications. Until binutils support is completed, it is 10729 effective only with link-time optimization. 10730 + Feedback driven indirect call removal and devirtualization now 10731 handle cross-module calls when link-time optimization is 10732 enabled. 10733 10734 [4]New Languages and Language specific improvements 10735 10736 * Version 4.0 of the [5]OpenMP specification is now supported in the 10737 C and C++ compilers and starting with the 4.9.1 release also in the 10738 Fortran compiler. The new -fopenmp-simd option can be used to 10739 enable OpenMP's SIMD directives while ignoring other OpenMP 10740 directives. The new [6]-fsimd-cost-model= option permits to tune 10741 the vectorization cost model for loops annotated with OpenMP and 10742 Cilk Plus simd directives. -Wopenmp-simd warns when the current 10743 cost model overrides simd directives set by the user. 10744 * The -Wdate-time option has been added for the C, C++ and Fortran 10745 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 10746 macros are used. Those macros might prevent bit-wise-identical 10747 reproducible compilations. 10748 10749 [7]Ada 10750 10751 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 10752 10753 [8]C family 10754 10755 * Support for colorizing diagnostics emitted by GCC has been added. 10756 The [9]-fdiagnostics-color=auto will enable it when outputting to 10757 terminals, -fdiagnostics-color=always unconditionally. The 10758 GCC_COLORS environment variable can be used to customize the colors 10759 or disable coloring. If GCC_COLORS variable is present in the 10760 environment, the default is -fdiagnostics-color=auto, otherwise 10761 -fdiagnostics-color=never. 10762 Sample diagnostics output: 10763 $ g++ -fdiagnostics-color=always -S -Wall test.C 10764 test.C: In function int foo(): 10765 test.C:1:14: warning: no return statement in function returning non-void [-W 10766 return-type] 10767 int foo () { } 10768 ^ 10769 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 10770 -ftemplate-depth= to increase the maximum) instantiating struct X<100> 10771 template <int N> struct X { static const int value = X<N-1>::value; }; temp 10772 late struct X<1000>; 10773 ^ 10774 test.C:2:46: recursively required from const int X<999>::value 10775 test.C:2:46: required from const int X<1000>::value 10776 test.C:2:88: required from here 10777 10778 test.C:2:46: error: incomplete type X<100> used in nested name specifier 10779 10780 * With the new [10]#pragma GCC ivdep, the user can assert that there 10781 are no loop-carried dependencies which would prevent concurrent 10782 execution of consecutive iterations using SIMD (single instruction 10783 multiple data) instructions. 10784 * Support for Cilk Plus has been added and can be enabled with the 10785 -fcilkplus option. Cilk Plus is an extension to the C and C++ 10786 languages to support data and task parallelism. The present 10787 implementation follows ABI version 1.2; all features but _Cilk_for 10788 have been implemented. 10789 10790 [11]C 10791 10792 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 10793 <stdatomic.h> header) are now supported. 10794 * ISO C11 generic selections (_Generic keyword) are now supported. 10795 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 10796 __thread) is now supported. 10797 * ISO C11 support is now at a similar level of completeness to ISO 10798 C99 support: substantially complete modulo bugs, extended 10799 identifiers (supported except for corner cases when 10800 -fextended-identifiers is used), floating-point issues (mainly but 10801 not entirely relating to optional C99 features from Annexes F and 10802 G) and the optional Annexes K (Bounds-checking interfaces) and L 10803 (Analyzability). 10804 * A new C extension __auto_type provides a subset of the 10805 functionality of C++11 auto in GNU C. 10806 10807 [12]C++ 10808 10809 * The G++ implementation of [13]C++1y return type deduction for 10810 normal functions has been updated to conform to [14]N3638, the 10811 proposal accepted into the working paper. Most notably, it adds 10812 decltype(auto) for getting decltype semantics rather than the 10813 template argument deduction semantics of plain auto: 10814 10815 int& f(); 10816 auto i1 = f(); // int 10817 decltype(auto) i2 = f(); // int& 10818 10819 * G++ supports [15]C++1y lambda capture initializers: 10820 10821 [x = 42]{ ... }; 10822 10823 Actually, they have been accepted since GCC 4.5, but now the 10824 compiler doesn't warn about them with -std=c++1y, and supports 10825 parenthesized and brace-enclosed initializers as well. 10826 * G++ supports [16]C++1y variable length arrays. G++ has supported 10827 GNU/C99-style VLAs for a long time, but now additionally supports 10828 initializers and lambda capture by reference. In C++1y mode G++ 10829 will complain about VLA uses that are not permitted by the draft 10830 standard, such as forming a pointer to VLA type or applying sizeof 10831 to a VLA variable. Note that it now appears that VLAs will not be 10832 part of C++14, but will be part of a separate document and then 10833 perhaps C++17. 10834 10835 void f(int n) { 10836 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 10837 [&a]{ for (int i : a) { cout << i << endl; } }(); 10838 &a; // error, taking address of VLA 10839 } 10840 10841 * G++ supports the [17]C++1y [[deprecated]] attribute modulo bugs in 10842 the underlying [[gnu::deprecated]] attribute. Classes and functions 10843 can be marked deprecated and a diagnostic message added: 10844 10845 class A; 10846 int bar(int n); 10847 #if __cplusplus > 201103 10848 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 10849 [[deprecated("bar is unsafe; use foo() instead")]] 10850 int bar(int n); 10851 10852 int foo(int n); 10853 class B; 10854 #endif 10855 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 10856 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 10857 o() instead 10858 10859 * G++ supports [18]C++1y digit separators. Long numeric literals can 10860 be subdivided with a single quote ' to enhance readability: 10861 10862 int i = 1048576; 10863 int j = 1'048'576; 10864 int k = 0x10'0000; 10865 int m = 0'004'000'000; 10866 int n = 0b0001'0000'0000'0000'0000'0000; 10867 10868 double x = 1.602'176'565e-19; 10869 double y = 1.602'176'565e-1'9; 10870 10871 * G++ supports [19]C++1y generic (polymorphic) lambdas. 10872 10873 // a functional object that will increment any type 10874 auto incr = [](auto x) { return x++; }; 10875 10876 * As a GNU extension, G++ supports explicit template parameter syntax 10877 for generic lambdas. This can be combined in the expected way with 10878 the standard auto syntax. 10879 10880 // a functional object that will add two like-type objects 10881 auto add = [] <typename T> (T a, T b) { return a + b; }; 10882 10883 * G++ supports unconstrained generic functions as specified by 4.1.2 10884 and 5.1.1 of [20]N3889: Concepts Lite Specification. Briefly, auto 10885 may be used as a type-specifier in a parameter declaration of any 10886 function declarator in order to introduce an implicit function 10887 template parameter, akin to generic lambdas. 10888 10889 // the following two function declarations are equivalent 10890 auto incr(auto x) { return x++; } 10891 template <typename T> 10892 auto incr(T x) { return x++; } 10893 10894 Runtime Library (libstdc++) 10895 10896 * [21]Improved support for C++11, including: 10897 + support for <regex>; 10898 + The associative containers in <map> and <set> and the 10899 unordered associative containers in <unordered_map> and 10900 <unordered_set> meet the allocator-aware container 10901 requirements; 10902 * [22]Improved experimental support for the upcoming ISO C++ 10903 standard, C++14, including: 10904 + fixing constexpr member functions without const; 10905 + implementation of the std::exchange() utility function; 10906 + addressing tuples by type; 10907 + implemention of std::make_unique; 10908 + implemention of std::shared_lock; 10909 + making std::result_of SFINAE-friendly; 10910 + adding operator() to std::integral_constant; 10911 + adding user-defined literals for standard library types 10912 std::basic_string, std::chrono::duration, and std::complex; 10913 + adding two range overloads to non-modifying sequence oprations 10914 std::equal and std::mismatch; 10915 + adding IO manipulators for quoted strings; 10916 + adding constexpr members to <utility>, <complex>, <chrono>, 10917 and some containers; 10918 + adding compile-time std::integer_sequence; 10919 + adding cleaner transformation traits; 10920 + making <functional>s operator functors easier to use and more 10921 generic; 10922 * An implementation of std::experimental::optional. 10923 * An implementation of std::experimental::string_view. 10924 * The non-standard function std::copy_exception has been deprecated 10925 and will be removed in a future version. std::make_exception_ptr 10926 should be used instead. 10927 10928 [23]Fortran 10929 10930 * Compatibility notice: 10931 + Module files: The version of the module files (.mod) has been 10932 incremented; additionally, module files are now compressed. 10933 Fortran MODULEs compiled by earlier GCC versions have to be 10934 recompiled, when they are USEd by files compiled with GCC 4.9. 10935 GCC 4.9 is not able to read .mod files of earlier GCC 10936 versions; attempting to do so gives an error message. Note: 10937 The ABI of the produced assembler data itself has not changed: 10938 object files and libraries are fully compatible with older 10939 versions (except as stated below). 10940 + ABI changes: 10941 o The [24]argument passing ABI has changed for scalar dummy 10942 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 10943 which have both the VALUE and the OPTIONAL attributes. 10944 o To support finalization the virtual table associated with 10945 polymorphic variables has changed. Code containing CLASS 10946 should be recompiled, including all files which define 10947 derived types involved in the type definition used by 10948 polymorphic variables. (Note: Due to the incremented 10949 module version, trying to mix old code with new code will 10950 usually give an error message.) 10951 + GNU Fortran no longer deallocates allocatable variables or 10952 allocatable components of variables declared in the main 10953 program. Since Fortran 2008, the standard explicitly states 10954 that variables declared in the Fortran main program 10955 automatically have the SAVE attribute. 10956 + When opening files, the close-on-exec flag is set if the 10957 system supports such a feature. This is generally considered 10958 good practice these days, but if there is a need to pass file 10959 descriptors to child processes the parent process must now 10960 remember to clear the close-on-exec flag by calling fcntl(), 10961 e.g. via ISO_C_BINDING, before executing the child process. 10962 * The deprecated command-line option -fno-whole-file has been 10963 removed. (-fwhole-file is the default since GCC 4.6.) 10964 -fwhole-file/-fno-whole-file continue to be accepted but do not 10965 influence the code generation. 10966 * The compiler no longer unconditionally warns about DO loops with 10967 zero iterations. This warning is now controlled by the -Wzerotrip 10968 option, which is implied by -Wall. 10969 * The new NO_ARG_CHECK attribute of the [25]!GCC$ directive can be 10970 used to disable the type-kind-rank (TKR) argument check for a dummy 10971 argument. The feature is similar to ISO/IEC TS 29133:2012's 10972 TYPE(*), except that it additionally also disables the rank check. 10973 Variables with NO_ARG_CHECK have to be dummy arguments and may only 10974 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 10975 to another NO_ARG_CHECK dummy argument; also the other constraints 10976 of TYPE(*) apply. The dummy arguments should be declared as scalar 10977 or assumed-size variable of type type(*) (recommended) or of type 10978 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 10979 the data without further type or shape information is passed, 10980 similar to C's void*. Note that also TS 29113's 10981 type(*),dimension(..) accepts arguments of any type and rank; 10982 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 10983 descriptor which contains the array shape and stride of the 10984 argument. 10985 * [26]Fortran 2003: 10986 + Finalization is now supported. It is currently only done for a 10987 subset of those situations in which it should occur. 10988 + Experimental support for scalar character components with 10989 deferred length (i.e. allocatable string length) in derived 10990 types has been added. (Deferred-length character variables are 10991 supported since GCC 4.6.) 10992 * [27]Fortran 2008: 10993 + When STOP or ERROR STOP are used to terminate the execution 10994 and any exception (but inexact) is signaling, a warning is 10995 printed to ERROR_UNIT, indicating which exceptions are 10996 signaling. The [28]-ffpe-summary= command-line option can be 10997 used to fine-tune for which exceptions the warning should be 10998 shown. 10999 + Rounding on input (READ) is now handled on systems where 11000 strtod honours the rounding mode. (For output, rounding is 11001 supported since GCC 4.5.) Note that for input, the compatible 11002 rounding mode is handled as nearest (i.e., rounding to an even 11003 least significant [cf. IEC 60559:1989] for a tie, while 11004 compatible rounds away from zero in that case). 11005 11006 [29]Go 11007 11008 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 11009 11010 [30]New Targets and Target Specific Improvements 11011 11012 [31]AArch64 11013 11014 * The ARMv8-A crypto and CRC instructions are now supported through 11015 intrinsics. These are enabled when the architecture supports these 11016 and are available through the -march=armv8-a+crc and 11017 -march=armv8-a+crypto options. 11018 * Initial support for ILP32 has now been added to the compiler. This 11019 is now available through the command-line option -mabi=ilp32. 11020 Support for ILP32 is considered experimental as the ABI 11021 specification is still beta. 11022 * Coverage of more of the ISA including the SIMD extensions has been 11023 added. The Advanced SIMD intrinsics have also been improved. 11024 * The new local register allocator (LRA) is now on by default for the 11025 AArch64 back end. 11026 * The REE (Redundant extension elimination) pass has now been enabled 11027 by default for the AArch64 back end. 11028 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 11029 * Initial big.LITTLE tuning support for the combination of Cortex-A57 11030 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 11031 option. 11032 * A number of structural changes have been made to both the ARM and 11033 AArch64 back ends to facilitate improved code-generation. 11034 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 11035 has been added and can be enabled by giving the 11036 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 11037 default by configuring GCC with the --enable-fix-cortex-a53-835769 11038 option. 11039 11040 [32]ARC 11041 11042 * A port for Synopsys Designware ARC has been contributed by Embecosm 11043 and Synopsys Inc. 11044 11045 [33]ARM 11046 11047 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 11048 disabled by default. This was found to generate better code in only 11049 a small number of cases. It can be turned back on with the 11050 -mneon-for-64bits option. 11051 * Further support for the ARMv8-A architecture, notably implementing 11052 the restriction around IT blocks in the Thumb32 instruction set has 11053 been added. The -mrestrict-it option can be used with 11054 -march=armv7-a or the -march=armv7ve options to make code 11055 generation fully compatible with the deprecated instructions in 11056 ARMv8-A. 11057 * Support has now been added for the ARMv7ve variant of the 11058 architecture. This can be used by the -march=armv7ve option. 11059 * The ARMv8-A crypto and CRC instructions are now supported through 11060 intrinsics and are available through the -march=armv8-a+crc and 11061 mfpu=crypto-neon-fp-armv8 options. 11062 * LRA is now on by default for the ARM target. This can be turned off 11063 using the -mno-lra option. This option is a purely transitionary 11064 command-line option and will be removed in a future release. We are 11065 interested in any bug reports regarding functional and performance 11066 regressions with LRA. 11067 * A new option -mslow-flash-data to improve performance of programs 11068 fetching data on slow flash memory has now been introduced for the 11069 ARMv7-M profile cores. 11070 * A new option -mpic-data-is-text-relative for targets that allows 11071 data segments to be relative to text segments has been added. This 11072 is on by default for all targets except VxWorks RTP. 11073 * A number of infrastructural changes have been made to both the ARM 11074 and AArch64 back ends to facilitate improved code-generation. 11075 * GCC now supports Cortex-A12 and the Cortex-R7 through the 11076 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 11077 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 11078 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 11079 * Initial big.LITTLE tuning support for the combination of Cortex-A57 11080 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 11081 option. Similar support was added for the combination of Cortex-A15 11082 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 11083 * Further performance optimizations for the Cortex-A15 and the 11084 Cortex-M4 have been added. 11085 * A number of code generation improvements for Thumb2 to reduce code 11086 size when compiling for the M-profile processors. 11087 11088 [34]AVR 11089 11090 * A new command-line option -mfract-convert-truncate has been added. 11091 It allows compiler to use truncation instead of rounding towards 11092 zero for fractional fixed-point types. 11093 11094 [35]IA-32/x86-64 11095 11096 * -mfpmath=sse is now implied by -ffast-math on all targets where 11097 SSE2 is supported. 11098 * Intel AVX-512 support was added to GCC. That includes inline 11099 assembly support, new registers and extending existing ones, new 11100 intrinsics (covered by corresponding testsuite), and basic 11101 autovectorization. AVX-512 instructions are available via the 11102 following GCC switches: AVX-512 foundation instructions: -mavx512f, 11103 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 11104 reciprocal instructions: -mavx512er, AVX-512 conflict detection 11105 instructions: -mavx512cd. 11106 * It is now possible to call x86 intrinsics from select functions in 11107 a file that are tagged with the corresponding target attribute 11108 without having to compile the entire file with the -mxxx option. 11109 This improves the usability of x86 intrinsics and is particularly 11110 useful when doing [36]Function Multiversioning. 11111 * GCC now supports the new Intel microarchitecture named Silvermont 11112 through -march=silvermont. 11113 * GCC now supports the new Intel microarchitecture named Broadwell 11114 through -march=broadwell. 11115 * Optimizing for other Intel microarchitectures have been renamed to 11116 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 11117 * -march=generic has been retuned for better support of Intel core 11118 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 11119 Pentium-M, and Pentium4 based CPUs is no longer considered 11120 important for generic. 11121 * -mtune=intel can now be used to generate code running well on the 11122 most current Intel processors, which are Haswell and Silvermont for 11123 GCC 4.9. 11124 * Support to encode 32-bit assembly instructions in 16-bit format is 11125 now available through the -m16 command-line option. 11126 * Better inlining of memcpy and memset that is aware of value ranges 11127 and produces shorter alignment prologues. 11128 * -mno-accumulate-outgoing-args is now honored when unwind 11129 information is output. Argument accumulation is also now turned off 11130 for portions of programs optimized for size. 11131 * Support for new AMD family 15h processors (Excavator core) is now 11132 available through the -march=bdver4 and -mtune=bdver4 options. 11133 11134 [37]MSP430 11135 11136 * A new command-line option -mcpu= has been added to the MSP430 back 11137 end. This option is used to specify the ISA to be used. Accepted 11138 values are msp430 (the default), msp430x and msp430xv2. The ISA is 11139 no longer deduced from the -mmcu= option as there are far too many 11140 different MCU names. The -mmcu= option is still supported, and this 11141 is still used to select linker scripts and generate a C 11142 preprocessor symbol that will be recognised by the msp430.h header 11143 file. 11144 11145 [38]NDS32 11146 11147 * A new nds32 port supports the 32-bit architecture from Andes 11148 Technology Corporation. 11149 * The port provides initial support for the V2, V3, V3m instruction 11150 set architectures. 11151 11152 [39]Nios II 11153 11154 * A port for the Altera Nios II has been contributed by Mentor 11155 Graphics. 11156 11157 [40]PowerPC / PowerPC64 / RS6000 11158 11159 * GCC now supports Power ISA 2.07, which includes support for 11160 Hardware Transactional Memory (HTM), Quadword atomics and several 11161 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 11162 integer and decimal integer operations. 11163 * Support for the POWER8 processor is now available through the 11164 -mcpu=power8 and -mtune=power8 options. 11165 * The libitm library has been modified to add a HTM fastpath that 11166 automatically uses POWER's HTM hardware instructions when it is 11167 executing on a HTM enabled processor. 11168 * Support for the new powerpc64le-linux platform has been added. It 11169 defaults to generating code that conforms to the ELFV2 ABI. 11170 11171 S/390, System z 11172 11173 * Support for the Transactional Execution Facility included with the 11174 IBM zEnterprise zEC12 processor has been added. A set of GCC style 11175 builtins as well as XLC style builtins are provided. The builtins 11176 are enabled by default when using the -march=zEC12 option but can 11177 explicitly be disabled with -mno-htm. Using the GCC builtins also 11178 libitm supports hardware transactions on S/390. 11179 * The hotpatch features allows to prepare functions for hotpatching. 11180 A certain amount of bytes is reserved before the function entry 11181 label plus a NOP is inserted at its very beginning to implement a 11182 backward jump when applying a patch. The feature can either be 11183 enabled per compilation unit via the command-line option -mhotpatch 11184 or per function using the hotpatch attribute. 11185 * The shrink wrap optimization is now supported on S/390 and enabled 11186 by default. 11187 * A major rework of the routines to determine which registers need to 11188 be saved and restored in function prologue/epilogue now allow to 11189 use floating point registers as save slots. This will happen for 11190 certain leaf function with -march=z10 or higher. 11191 * The LRA rtl pass replaces reload by default on S/390. 11192 11193 [41]RX 11194 11195 * The port now allows to specify the RX100, RX200, and RX600 11196 processors with the command-line options -mcpu=rx100, -mcpu=rx200 11197 and -mcpu=rx600. 11198 11199 [42]SH 11200 11201 * Minor improvements to code generated for integer arithmetic and 11202 code that involves the T bit. 11203 * Added support for the SH2A clips and clipu instructions. The 11204 compiler will now try to utilize them for min/max expressions such 11205 as max (-128, min (127, x)). 11206 * Added support for the cmp/str instruction through built-in 11207 functions such as __builtin_strlen. When not optimizing for size, 11208 the compiler will now expand calls to e.g. strlen as an inlined 11209 sequences which utilize the cmp/str instruction. 11210 * Improved code generated around volatile memory loads and stores. 11211 * The option -mcbranchdi has been deprecated. Specifying it will 11212 result in a warning and will not influence code generation. 11213 * The option -mcmpeqdi has been deprecated. Specifying it will result 11214 in a warning and will not influence code generation. 11215 11216 [43]GCC 4.9.1 11217 11218 This is the [44]list of problem reports (PRs) from GCC's bug tracking 11219 system that are known to be fixed in the 4.9.1 release. This list might 11220 not be complete (that is, it is possible that some PRs that have been 11221 fixed are not listed here). 11222 11223 Version 4.0 of the OpenMP specification is supported even in Fortran, 11224 not just C and C++. 11225 11226 [45]GCC 4.9.2 11227 11228 This is the [46]list of problem reports (PRs) from GCC's bug tracking 11229 system that are known to be fixed in the 4.9.2 release. This list might 11230 not be complete (that is, it is possible that some PRs that have been 11231 fixed are not listed here). 11232 11233 [47]GCC 4.9.3 11234 11235 This is the [48]list of problem reports (PRs) from GCC's bug tracking 11236 system that are known to be fixed in the 4.9.3 release. This list might 11237 not be complete (that is, it is possible that some PRs that have been 11238 fixed are not listed here). 11239 11240 [49]GCC 4.9.4 11241 11242 This is the [50]list of problem reports (PRs) from GCC's bug tracking 11243 system that are known to be fixed in the 4.9.4 release. This list might 11244 not be complete (that is, it is possible that some PRs that have been 11245 fixed are not listed here). 11246 11247 11248 For questions related to the use of GCC, please consult these web 11249 pages and the [51]GCC manuals. If that fails, the 11250 [52]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11251 web pages and the development of GCC are welcome on our developer 11252 list at [53]gcc (a] gcc.gnu.org. All of [54]our lists have public 11253 archives. 11254 11255 Copyright (C) [55]Free Software Foundation, Inc. Verbatim copying and 11256 distribution of this entire article is permitted in any medium, 11257 provided this notice is preserved. 11258 11259 These pages are [56]maintained by the GCC team. Last modified 11260 2026-06-01. 11261 11262 References 11263 11264 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 11265 2. https://gcc.gnu.org/PR60825 11266 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 11267 4. https://gcc.gnu.org/gcc-4.9/changes.html#languages 11268 5. https://www.openmp.org/specifications/ 11269 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 11270 7. https://gcc.gnu.org/gcc-4.9/changes.html#ada 11271 8. https://gcc.gnu.org/gcc-4.9/changes.html#c-family 11272 9. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 11273 10. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 11274 11. https://gcc.gnu.org/gcc-4.9/changes.html#c 11275 12. https://gcc.gnu.org/gcc-4.9/changes.html#cxx 11276 13. https://gcc.gnu.org/projects/cxx1y.html 11277 14. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 11278 15. https://gcc.gnu.org/projects/cxx1y.html 11279 16. https://gcc.gnu.org/projects/cxx1y.html 11280 17. https://gcc.gnu.org/projects/cxx1y.html 11281 18. https://gcc.gnu.org/projects/cxx1y.html 11282 19. https://gcc.gnu.org/projects/cxx1y.html 11283 20. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 11284 21. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 11285 22. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 11286 23. https://gcc.gnu.org/gcc-4.9/changes.html#fortran 11287 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 11288 25. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 11289 26. https://gcc.gnu.org/wiki/Fortran2003Status 11290 27. https://gcc.gnu.org/wiki/Fortran2008Status 11291 28. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 11292 29. https://gcc.gnu.org/gcc-4.9/changes.html#go 11293 30. https://gcc.gnu.org/gcc-4.9/changes.html#targets 11294 31. https://gcc.gnu.org/gcc-4.9/changes.html#aarch64 11295 32. https://gcc.gnu.org/gcc-4.9/changes.html#arc 11296 33. https://gcc.gnu.org/gcc-4.9/changes.html#arm 11297 34. https://gcc.gnu.org/gcc-4.9/changes.html#avr 11298 35. https://gcc.gnu.org/gcc-4.9/changes.html#x86 11299 36. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 11300 37. https://gcc.gnu.org/gcc-4.9/changes.html#msp430 11301 38. https://gcc.gnu.org/gcc-4.9/changes.html#nds32 11302 39. https://gcc.gnu.org/gcc-4.9/changes.html#nios2 11303 40. https://gcc.gnu.org/gcc-4.9/changes.html#powerpc 11304 41. https://gcc.gnu.org/gcc-4.9/changes.html#rx 11305 42. https://gcc.gnu.org/gcc-4.9/changes.html#sh 11306 43. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.1 11307 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 11308 45. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.2 11309 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 11310 47. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.3 11311 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 11312 49. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.4 11313 50. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 11314 51. https://gcc.gnu.org/onlinedocs/ 11315 52. mailto:gcc-help (a] gcc.gnu.org 11316 53. mailto:gcc (a] gcc.gnu.org 11317 54. https://gcc.gnu.org/lists.html 11318 55. https://www.fsf.org/ 11319 56. https://gcc.gnu.org/about.html 11320 ====================================================================== 11321 http://gcc.gnu.org/gcc-4.8/index.html 11322 GCC 4.8 Release Series 11323 11324 (This release series is no longer supported.) 11325 11326 June 23, 2015 11327 11328 The [1]GNU project and the GCC developers are pleased to announce the 11329 release of GCC 4.8.5. 11330 11331 This release is a bug-fix release, containing fixes for regressions in 11332 GCC 4.8.4 relative to previous releases of GCC. 11333 11334 Release History 11335 11336 GCC 4.8.5 11337 June 23, 2015 ([2]changes, [3]documentation) 11338 11339 GCC 4.8.4 11340 December 19, 2014 ([4]changes, [5]documentation) 11341 11342 GCC 4.8.3 11343 May 22, 2014 ([6]changes, [7]documentation) 11344 11345 GCC 4.8.2 11346 October 16, 2013 ([8]changes, [9]documentation) 11347 11348 GCC 4.8.1 11349 May 31, 2013 ([10]changes, [11]documentation) 11350 11351 GCC 4.8.0 11352 March 22, 2013 ([12]changes, [13]documentation) 11353 11354 References and Acknowledgements 11355 11356 GCC used to stand for the GNU C Compiler, but since the compiler 11357 supports several other languages aside from C, it now stands for the 11358 GNU Compiler Collection. 11359 11360 A list of [14]successful builds is updated as new information becomes 11361 available. 11362 11363 The GCC developers would like to thank the numerous people that have 11364 contributed new features, improvements, bug fixes, and other changes as 11365 well as test results to GCC. This [15]amazing group of volunteers is 11366 what makes GCC successful. 11367 11368 For additional information about GCC please refer to the [16]GCC 11369 project web site or contact the [17]GCC development mailing list. 11370 11371 To obtain GCC please use [18]our mirror sites or [19]our version 11372 control system. 11373 11374 11375 For questions related to the use of GCC, please consult these web 11376 pages and the [20]GCC manuals. If that fails, the 11377 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11378 web pages and the development of GCC are welcome on our developer 11379 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 11380 archives. 11381 11382 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 11383 distribution of this entire article is permitted in any medium, 11384 provided this notice is preserved. 11385 11386 These pages are [25]maintained by the GCC team. Last modified 11387 2022-10-26. 11388 11389 References 11390 11391 1. http://www.gnu.org/ 11392 2. https://gcc.gnu.org/gcc-4.8/changes.html 11393 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 11394 4. https://gcc.gnu.org/gcc-4.8/changes.html 11395 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 11396 6. https://gcc.gnu.org/gcc-4.8/changes.html 11397 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 11398 8. https://gcc.gnu.org/gcc-4.8/changes.html 11399 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 11400 10. https://gcc.gnu.org/gcc-4.8/changes.html 11401 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 11402 12. https://gcc.gnu.org/gcc-4.8/changes.html 11403 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 11404 14. https://gcc.gnu.org/gcc-4.8/buildstat.html 11405 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11406 16. https://gcc.gnu.org/index.html 11407 17. mailto:gcc (a] gcc.gnu.org 11408 18. https://gcc.gnu.org/mirrors.html 11409 19. https://gcc.gnu.org/git.html 11410 20. https://gcc.gnu.org/onlinedocs/ 11411 21. mailto:gcc-help (a] gcc.gnu.org 11412 22. mailto:gcc (a] gcc.gnu.org 11413 23. https://gcc.gnu.org/lists.html 11414 24. https://www.fsf.org/ 11415 25. https://gcc.gnu.org/about.html 11416 ====================================================================== 11417 http://gcc.gnu.org/gcc-4.8/changes.html 11418 GCC 4.8 Release Series 11419 Changes, New Features, and Fixes 11420 11421 Caveats 11422 11423 GCC now uses C++ as its implementation language. This means that to 11424 build GCC from sources, you will need a C++ compiler that understands 11425 C++ 2003. For more details on the rationale and specific changes, 11426 please refer to the [1]C++ conversion page. 11427 11428 To enable the Graphite framework for loop optimizations you now need 11429 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 11430 the [2]GCC infrastructure directory. The installation manual contains 11431 more information about requirements to build GCC. 11432 11433 GCC now uses a more aggressive analysis to derive an upper bound for 11434 the number of iterations of loops using constraints imposed by language 11435 standards. This may cause non-conforming programs to no longer work as 11436 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 11437 option, -fno-aggressive-loop-optimizations, was added to disable this 11438 aggressive analysis. In some loops that have known constant number of 11439 iterations, but undefined behavior is known to occur in the loop before 11440 reaching or during the last iteration, GCC will warn about the 11441 undefined behavior in the loop instead of deriving lower upper bound of 11442 the number of iterations for the loop. The warning can be disabled with 11443 -Wno-aggressive-loop-optimizations. 11444 11445 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 11446 for the layout of vectors that could lead to wrong code being 11447 generated. Vectors larger than 8 bytes in size are now by default 11448 aligned to an 8-byte boundary. This is an ABI change: code that makes 11449 explicit use of vector types may be incompatible with binary objects 11450 built with older versions of GCC. Auto-vectorized code is not affected 11451 by this change. 11452 11453 More information on porting to GCC 4.8 from previous versions of GCC 11454 can be found in the [3]porting guide for this release. 11455 11456 General Optimizer Improvements (and Changes) 11457 11458 * DWARF4 is now the default when generating DWARF debug information. 11459 When -g is used on a platform that uses DWARF debugging 11460 information, GCC will now default to -gdwarf-4 11461 -fno-debug-types-section. 11462 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 11463 consumers support DWARF4 by default. Before GCC 4.8 the default 11464 version used was DWARF2. To make GCC 4.8 generate an older DWARF 11465 version use -g together with -gdwarf-2 or -gdwarf-3. The default 11466 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 11467 * A new general optimization level, -Og, has been introduced. It 11468 addresses the need for fast compilation and a superior debugging 11469 experience while providing a reasonable level of run-time 11470 performance. Overall experience for development should be better 11471 than the default optimization level -O0. 11472 * A new option -ftree-partial-pre was added to control the partial 11473 redundancy elimination (PRE) optimization. This option is enabled 11474 by default at the -O3 optimization level, and it makes PRE more 11475 aggressive. 11476 * The option -fconserve-space has been removed; it was no longer 11477 useful on most targets since GCC supports putting variables into 11478 BSS without making them common. 11479 * The struct reorg and matrix reorg optimizations (command-line 11480 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 11481 removed. They did not always work correctly, nor did they work with 11482 link-time optimization (LTO), hence were only applicable to 11483 programs consisting of a single translation unit. 11484 * Several scalability bottle-necks have been removed from GCC's 11485 optimization passes. Compilation of extremely large functions, e.g. 11486 due to the use of the flatten attribute in the "Eigen" C++ linear 11487 algebra templates library, is significantly faster than previous 11488 releases of GCC. 11489 * Link-time optimization (LTO) improvements: 11490 + LTO partitioning has been rewritten for better reliability and 11491 maintanibility. Several important bugs leading to link 11492 failures have been fixed. 11493 * Interprocedural optimization improvements: 11494 + A new symbol table has been implemented. It builds on existing 11495 callgraph and varpool modules and provide a new API. Unusual 11496 symbol visibilities and aliases are handled more consistently 11497 leading to, for example, more aggressive unreachable code 11498 removal with LTO. 11499 + The inline heuristic can now bypass limits on the size of of 11500 inlined functions when the inlining is particularly 11501 profitable. This happens, for example, when loop bounds or 11502 array strides get propagated. 11503 + Values passed through aggregates (either by value or 11504 reference) are now propagated at the inter-procedural level 11505 leading to better inlining decisions (for example in the case 11506 of Fortran array descriptors) and devirtualization. 11507 * [4]AddressSanitizer , a fast memory error detector, has been added 11508 and can be enabled via -fsanitize=address. Memory access 11509 instructions will be instrumented to detect heap-, stack-, and 11510 global-buffer overflow as well as use-after-free bugs. To get nicer 11511 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 11512 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 11513 x86-64 Darwin. 11514 * [5]ThreadSanitizer has been added and can be enabled via 11515 -fsanitize=thread. Instructions will be instrumented to detect data 11516 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 11517 * A new local register allocator (LRA) has been implemented, which 11518 replaces the 26 year old reload pass and improves generated code 11519 quality. For now it is active on the IA-32 and x86-64 targets. 11520 * Support for transactional memory has been implemented on the 11521 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 11522 Alpha. 11523 11524 New Languages and Language specific improvements 11525 11526 C family 11527 11528 * Each diagnostic emitted now includes the original source line and a 11529 caret '^' indicating the column. The option 11530 -fno-diagnostics-show-caret suppresses this information. 11531 * The option -ftrack-macro-expansion=2 is now enabled by default. 11532 This allows the compiler to display the macro expansion stack in 11533 diagnostics. Combined with the caret information, an example 11534 diagnostic showing these two features is: 11535 11536 t.c:1:94: error: invalid operands to binary < (have struct mystruct and float 11537 ) 11538 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 11539 _b = (B); __a < __b ? __b : __a; }) 11540 11541 ^ 11542 t.c:7:7: note: in expansion of macro 'MYMAX' 11543 X = MYMAX(P, F); 11544 ^ 11545 11546 * A new -Wsizeof-pointer-memaccess warning has been added (also 11547 enabled by -Wall) to warn about suspicious length parameters to 11548 certain string and memory built-in functions if the argument uses 11549 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 11550 (ptr)); if ptr is not an array, but a pointer, and suggests a 11551 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 11552 * The new option -Wpedantic is an alias for -pedantic, which is now 11553 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 11554 -Wno-error=pedantic work in the same way as for any other -W 11555 option. One caveat is that -Werror=pedantic is not equivalent to 11556 -pedantic-errors, since the latter makes into errors some warnings 11557 that are not controlled by -Wpedantic, and the former only affects 11558 diagnostics that are disabled when using -Wno-pedantic. 11559 * The option -Wshadow no longer warns if a declaration shadows a 11560 function declaration, unless the former declares a function or 11561 pointer to function, because this is [6]a common and valid case in 11562 real-world code. 11563 11564 [7]C++ 11565 11566 * G++ now implements the [8]C++11 thread_local keyword; this differs 11567 from the GNU __thread keyword primarily in that it allows dynamic 11568 initialization and destruction semantics. Unfortunately, this 11569 support requires a run-time penalty for references to 11570 non-function-local thread_local variables defined in a different 11571 translation unit even if they don't need dynamic initialization, so 11572 users may want to continue to use __thread for TLS variables with 11573 static initialization semantics. 11574 If the programmer can be sure that no use of the variable in a 11575 non-defining TU needs to trigger dynamic initialization (either 11576 because the variable is statically initialized, or a use of the 11577 variable in the defining TU will be executed before any uses in 11578 another TU), they can avoid this overhead with the 11579 -fno-extern-tls-init option. 11580 OpenMP threadprivate variables now also support dynamic 11581 initialization and destruction by the same mechanism. 11582 * G++ now implements the [9]C++11 attribute syntax, e.g. 11583 11584 [[noreturn]] void f(); 11585 11586 and also the alignment specifier, e.g. 11587 11588 alignas(double) int i; 11589 11590 * G++ now implements [10]C++11 inheriting constructors, e.g. 11591 11592 struct A { A(int); }; 11593 struct B: A { using A::A; }; // defines B::B(int) 11594 B b(42); // OK 11595 11596 * As of GCC 4.8.1, G++ implements the change to decltype semantics 11597 from [11]N3276. 11598 11599 struct A f(); 11600 decltype(f()) g(); // OK, return type of f() is not required to be complete. 11601 11602 * As of GCC 4.8.1, G++ implements [12]C++11 ref-qualifiers, e.g. 11603 11604 struct A { int f() &; }; 11605 int i = A().f(); // error, f() requires an lvalue object 11606 11607 * G++ now supports a -std=c++1y option for experimentation with 11608 features proposed for the next revision of the standard, expected 11609 around 2014. Currently the only difference from -std=c++11 is 11610 support for return type deduction in normal functions, as proposed 11611 in [13]N3386. Status of C++1y features in GCC 4.8 can be found 11612 [14]here. 11613 * The G++ namespace association extension, __attribute ((strong)), 11614 has been deprecated. Inline namespaces should be used instead. 11615 * G++ now supports a -fext-numeric-literal option to control whether 11616 GNU numeric literal suffixes are accepted as extensions or 11617 processed as C++11 user-defined numeric literal suffixes. The flag 11618 is on (use suffixes for GNU literals) by default for -std=gnu++*, 11619 and -std=c++98. The flag is off (use suffixes for user-defined 11620 literals) by default for -std=c++11 and later. 11621 11622 Runtime Library (libstdc++) 11623 11624 * [15]Improved experimental support for the new ISO C++ standard, 11625 C++11, including: 11626 + forward_list meets the allocator-aware container requirements; 11627 + this_thread::sleep_for(), this_thread::sleep_until() and 11628 this_thread::yield() are defined without requiring the 11629 configure option --enable-libstdcxx-time; 11630 * Improvements to <random>: 11631 + SSE optimized normal_distribution. 11632 + Use of hardware RNG instruction for random_device on new x86 11633 processors (requires the assembler to support the 11634 instruction.) 11635 and <ext/random>: 11636 + New random number engine simd_fast_mersenne_twister_engine 11637 with an optimized SSE implementation. 11638 + New random number distributions beta_distribution, 11639 normal_mv_distribution, rice_distribution, 11640 nakagami_distribution, pareto_distribution, k_distribution, 11641 arcsine_distribution, hoyt_distribution. 11642 * Added --disable-libstdcxx-verbose configure option to disable 11643 diagnostic messages issued when a process terminates abnormally. 11644 This may be useful for embedded systems to reduce the size of 11645 executables that link statically to the library. 11646 11647 [16]Fortran 11648 11649 * Compatibility notice: 11650 + Module files: The version of module files (.mod) has been 11651 incremented. Fortran MODULEs compiled by earlier GCC versions 11652 have to be recompiled, when they are USEd by files compiled 11653 with GCC 4.8. GCC 4.8 is not able to read .mod files created 11654 by earlier versions; attempting to do so gives an error 11655 message. 11656 Note: The ABI of the produced assembler data itself has not 11657 changed; object files and libraries are fully compatible with 11658 older versions except as noted below. 11659 + ABI: Some internal names (used in the assembler/object file) 11660 have changed for symbols declared in the specification part of 11661 a module. If an affected module or a file using it via use 11662 association is recompiled, the module and all files which 11663 directly use such symbols have to be recompiled as well. This 11664 change only affects the following kind of module symbols: 11665 o Procedure pointers. Note: C-interoperable function 11666 pointers (type(c_funptr)) are not affected nor are 11667 procedure-pointer components. 11668 o Deferred-length character strings. 11669 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 11670 backtrace at an arbitrary place in user code; program execution 11671 continues normally afterwards. 11672 * The [18]-Wc-binding-type warning option has been added (disabled by 11673 default). It warns if the a variable might not be C interoperable; 11674 in particular, if the variable has been declared using an intrinsic 11675 type with default kind instead of using a kind parameter defined 11676 for C interoperability in the intrinsic ISO_C_Binding module. 11677 Before, this warning was always printed. The -Wc-binding-type 11678 option is enabled by -Wall. 11679 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 11680 options have been added, which diagnose when code is inserted for 11681 automatic (re)allocation of a variable during assignment. This 11682 option can be used to decide whether it is safe to use 11683 [20]-fno-realloc-lhs. Additionally, it can be used to find 11684 automatic (re)allocation in hot loops. (For arrays, replacing 11685 "var=" by "var(:)=" disables the automatic reallocation.) 11686 * The [21]-Wcompare-reals command-line option has been added. When 11687 this is set, warnings are issued when comparing REAL or COMPLEX 11688 types for equality and inequality; consider replacing a == b by 11689 abs(ab) < eps with a suitable eps. -Wcompare-reals is enabled by 11690 -Wextra. 11691 * The [22]-Wtarget-lifetime command-line option has been added 11692 (enabled with -Wall), which warns if the pointer in a pointer 11693 assignment might outlive its target. 11694 * Reading floating point numbers which use "q" for the exponential 11695 (such as 4.0q0) is now supported as vendor extension for better 11696 compatibility with old data files. It is strongly recommended to 11697 use for I/O the equivalent but standard conforming "e" (such as 11698 4.0e0). 11699 (For Fortran source code, consider replacing the "q" in 11700 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 11701 suitable qp). Note that in Fortran source code replacing "q" by 11702 a simple "e" is not equivalent.) 11703 * The GFORTRAN_TMPDIR environment variable for specifying a 11704 non-default directory for files opened with STATUS="SCRATCH", is 11705 not used anymore. Instead gfortran checks the POSIX/GNU standard 11706 TMPDIR environment variable. If TMPDIR is not defined, gfortran 11707 falls back to other methods to determine the directory for 11708 temporary files as documented in the [23]user manual. 11709 * [24]Fortran 2003: 11710 + Support for unlimited polymorphic variables (CLASS(*)) has 11711 been added. Nonconstant character lengths are not yet 11712 supported. 11713 * [25]TS 29113: 11714 + Assumed types (TYPE(*)) are now supported. 11715 + Experimental support for assumed-rank arrays (dimension(..)) 11716 has been added. Note that currently gfortran's own array 11717 descriptor is used, which is different from the one defined in 11718 TS29113, see [26]gfortran's header file or use the [27]Chasm 11719 Language Interoperability Tools. 11720 11721 [28]Go 11722 11723 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 11724 release. 11725 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 11726 release. The library support is not quite complete. 11727 * Go has been tested on GNU/Linux and Solaris platforms for various 11728 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 11729 work on other platforms as well. 11730 11731 [29]New Targets and Target Specific Improvements 11732 11733 [30]AArch64 11734 11735 * A new port has been added to support AArch64, the new 64-bit 11736 architecture from ARM. Note that this is a separate port from the 11737 existing 32-bit ARM port. 11738 * The port provides initial support for the Cortex-A53 and the 11739 Cortex-A57 processors with the command-line options 11740 -mcpu=cortex-a53 and -mcpu=cortex-a57. 11741 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 11742 has been added and can be enabled by giving the 11743 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 11744 default by configuring GCC with the --enable-fix-cortex-a53-835769 11745 option. 11746 11747 [31]ARM 11748 11749 * Initial support has been added for the AArch32 extensions defined 11750 in the ARMv8 architecture. 11751 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 11752 * A new option, -mcpu=marvell-pj4, has been added to generate code 11753 for the Marvell PJ4 processor. 11754 * The compiler can now automatically generate the VFMA, VFMS, REVSH 11755 and REV16 instructions. 11756 * A new vectorizer cost model for Advanced SIMD configurations to 11757 improve the auto-vectorization strategies used. 11758 * The scheduler now takes into account the number of live registers 11759 to reduce the amount of spilling that can occur. This should 11760 improve code performance in large functions. The limit can be 11761 removed by using the option -fno-sched-pressure. 11762 * Improvements have been made to the Marvell iWMMX code generation 11763 and support for the iWMMX2 SIMD unit has been added. The option 11764 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 11765 * A number of code generation improvements for Thumb2 to reduce code 11766 size when compiling for the M-profile processors. 11767 * The RTEMS (arm-rtems) port has been updated to use the EABI. 11768 * Code generation support for the old FPA and Maverick floating-point 11769 architectures has been removed. Ports that previously relied on 11770 these features have also been removed. This includes the targets: 11771 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 11772 + arm*-*-elf (use arm*-*-eabi) 11773 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 11774 + arm*-*-ecos-elf (no alternative) 11775 + arm*-*-freebsd (no alternative) 11776 + arm*-wince-pe* (no alternative). 11777 11778 [32]AVR 11779 11780 * Support for the command-line option -mshort-calls has been removed. 11781 It was deprecated in GCC 4.7. 11782 * The configure option --with-avrlibc supported since GCC 4.7.2 is 11783 turned on per default for all non-RTEMS configurations. This option 11784 arranges for a better integration of [33]AVR-LibC with avr-gcc. For 11785 technical details, see [34]PR54461. To turn off the option in 11786 non-AVR-LibC configurations, use --with-avrlibc=no. If the compiler 11787 is configured for RTEMS, the option is always turned off. 11788 * Support for the "Embedded C" fixed-point has been added. For 11789 details, see the [35]GCC wiki and the [36]user manual. The support 11790 is not complete. 11791 * A new print modifier %r for register operands in inline assembler 11792 is supported. It will print the raw register number without the 11793 register prefix 'r': 11794 /* Return the most significant byte of 'val', a 64-bit value. */ 11795 11796 unsigned char msb (long long val) 11797 { 11798 unsigned char c; 11799 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 11800 return c; 11801 } 11802 The inline assembler in this example will generate code like 11803 mov r24, 8+7 11804 provided c is allocated to R24 and val is allocated to R8R15. This 11805 works because the GNU assembler accepts plain register numbers 11806 without register prefix. 11807 * Static initializers with 3-byte symbols are supported now: 11808 extern const __memx char foo; 11809 const __memx void *pfoo = &foo; 11810 This requires at least Binutils 2.23. 11811 11812 IA-32/x86-64 11813 11814 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 11815 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 11816 stack alignment, this is ABI incompatible and intended to be used 11817 in controlled environments where stack space is an important 11818 limitation. This option will lead to wrong code when functions 11819 compiled with 16 byte stack alignment (such as functions from a 11820 standard library) are called with misaligned stack. In this case, 11821 SSE instructions may lead to misaligned memory access traps. In 11822 addition, variable arguments will be handled incorrectly for 16 11823 byte aligned objects (including x87 long double and __int128), 11824 leading to wrong results. You must build all modules with 11825 -mpreferred-stack-boundary=3, including any libraries. This 11826 includes the system libraries and startup modules. 11827 * Support for the new Intel processor codename Broadwell with RDSEED, 11828 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 11829 -mrdseed command-line options. 11830 * Support for the Intel RTM and HLE intrinsics, built-in functions 11831 and code generation is available via -mrtm and -mhle. 11832 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 11833 Intrinsics and built-in functions are available via -mfxsr, -mxsave 11834 and -mxsaveopt respectively. 11835 * New -maddress-mode=[short|long] options for x32. 11836 -maddress-mode=short overrides default 64-bit addresses to 32-bit 11837 by emitting the 0x67 address-size override prefix. This is the 11838 default address mode for x32. 11839 * New built-in functions to detect run-time CPU type and ISA: 11840 + A built-in function __builtin_cpu_is has been added to detect 11841 if the run-time CPU is of a particular type. It returns a 11842 positive integer on a match and zero otherwise. It accepts one 11843 string literal argument, the CPU name. For example, 11844 __builtin_cpu_is("westmere") returns a positive integer if the 11845 run-time CPU is an Intel Core i7 Westmere processor. Please 11846 refer to the [37]user manual for the list of valid CPU names 11847 recognized. 11848 + A built-in function __builtin_cpu_supports has been added to 11849 detect if the run-time CPU supports a particular ISA feature. 11850 It returns a positive integer on a match and zero otherwise. 11851 It accepts one string literal argument, the ISA feature. For 11852 example, __builtin_cpu_supports("ssse3") returns a positive 11853 integer if the run-time CPU supports SSSE3 instructions. 11854 Please refer to the [38]user manual for the list of valid ISA 11855 names recognized. 11856 Caveat: If these built-in functions are called before any static 11857 constructors are invoked, like during IFUNC initialization, then 11858 the CPU detection initialization must be explicitly run using this 11859 newly provided built-in function, __builtin_cpu_init. The 11860 initialization needs to be done only once. For example, this is how 11861 the invocation would look like inside an IFUNC initializer: 11862 static void (*some_ifunc_resolver(void))(void) 11863 { 11864 __builtin_cpu_init(); 11865 if (__builtin_cpu_is("amdfam10h") ... 11866 if (__builtin_cpu_supports("popcnt") ... 11867 } 11868 11869 * Function Multiversioning Support with G++: 11870 It is now possible to create multiple function versions each 11871 targeting a specific processor and/or ISA. Function versions have 11872 the same signature but different target attributes. For example, 11873 here is a program with function versions: 11874 __attribute__ ((target ("default"))) 11875 int foo(void) 11876 { 11877 return 1; 11878 } 11879 11880 __attribute__ ((target ("sse4.2"))) 11881 int foo(void) 11882 { 11883 return 2; 11884 } 11885 11886 int main (void) 11887 { 11888 int (*p) = &foo; 11889 assert ((*p)() == foo()); 11890 return 0; 11891 } 11892 11893 Please refer to this [39]wiki for more information. 11894 * The x86 back end has been improved to allow option -fschedule-insns 11895 to work reliably. This option can be used to schedule instructions 11896 better and leads to improved performace in certain cases. 11897 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 11898 from the Mingw-w64 trunk. 11899 * Support for new AMD family 15h processors (Steamroller core) is now 11900 available through the -march=bdver3 and -mtune=bdver3 options. 11901 * Support for new AMD family 16h processors (Jaguar core) is now 11902 available through the -march=btver2 and -mtune=btver2 options. 11903 11904 [40]FRV 11905 11906 * This target now supports the -fstack-usage command-line option. 11907 11908 [41]MIPS 11909 11910 * GCC can now generate code specifically for the R4700, Broadcom XLP 11911 and MIPS 34kn processors. The associated -march options are 11912 -march=r4700, -march=xlp and -march=34kn respectively. 11913 * GCC now generates better DSP code for MIPS 74k cores thanks to 11914 further scheduling optimizations. 11915 * The MIPS port now supports the -fstack-check option. 11916 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 11917 * Previous versions of GCC would silently accept -fpic and -fPIC for 11918 -mno-abicalls targets like mips*-elf. This combination was not 11919 intended or supported, and did not generate position-independent 11920 code. GCC 4.8 now reports an error when this combination is used. 11921 11922 [42]PowerPC / PowerPC64 / RS6000 11923 11924 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 11925 restore or update the VRSAVE register by default. The respective 11926 operating systems manage the VRSAVE register directly. 11927 * Large TOC support has been added for AIX through the command line 11928 option -mcmodel=large. 11929 * Native Thread-Local Storage support has been added for AIX. 11930 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 11931 when targetting processors that support those hardware features on 11932 AIX 6.1 and above. 11933 11934 [43]RX 11935 11936 * This target will now issue a warning message whenever multiple fast 11937 interrupt handlers are found in the same compilation unit. This 11938 feature can be turned off by the new 11939 -mno-warn-multiple-fast-interrupts command-line option. 11940 11941 S/390, System z 11942 11943 * Support for the IBM zEnterprise zEC12 processor has been added. 11944 When using the -march=zEC12 option, the compiler will generate code 11945 making use of the following new instructions: 11946 + load and trap instructions 11947 + 2 new compare and trap instructions 11948 + rotate and insert selected bits - without CC clobber 11949 The -mtune=zEC12 option enables zEC12 specific instruction 11950 scheduling without making use of new instructions. 11951 * Register pressure sensitive instruction scheduling is enabled by 11952 default. 11953 * The ifunc function attribute is enabled by default. 11954 * memcpy and memcmp invokations on big memory chunks or with run time 11955 lengths are not generated inline anymore when tuning for z10 or 11956 higher. The purpose is to make use of the IFUNC optimized versions 11957 in Glibc. 11958 11959 [44]SH 11960 11961 * The default alignment settings have been reduced to be less 11962 aggressive. This results in more compact code for optimization 11963 levels other than -Os. 11964 * Improved support for the __atomic built-in functions: 11965 + A new option -matomic-model=model selects the model for the 11966 generated atomic sequences. The following models are 11967 supported: 11968 11969 soft-gusa 11970 Software gUSA sequences (SH3* and SH4* only). On 11971 SH4A targets this will now also partially utilize 11972 the movco.l and movli.l instructions. This is the 11973 default when the target is sh3*-*-linux* or 11974 sh4*-*-linux*. 11975 11976 hard-llcs 11977 Hardware movco.l / movli.l sequences (SH4A only). 11978 11979 soft-tcb 11980 Software thread control block sequences. 11981 11982 soft-imask 11983 Software interrupt flipping sequences (privileged 11984 mode only). This is the default when the target is 11985 sh1*-*-linux* or sh2*-*-linux*. 11986 11987 none 11988 Generates function calls to the respective __atomic 11989 built-in functions. This is the default for SH64 11990 targets or when the target is not sh*-*-linux*. 11991 11992 + The option -msoft-atomic has been deprecated. It is now an 11993 alias for -matomic-model=soft-gusa. 11994 + A new option -mtas makes the compiler generate the tas.b 11995 instruction for the __atomic_test_and_set built-in function 11996 regardless of the selected atomic model. 11997 + The __sync functions in libgcc now reflect the selected atomic 11998 model when building the toolchain. 11999 * Added support for the mov.b and mov.w instructions with 12000 displacement addressing. 12001 * Added support for the SH2A instructions movu.b and movu.w. 12002 * Various improvements to code generated for integer arithmetic. 12003 * Improvements to conditional branches and code that involves the T 12004 bit. A new option -mzdcbranch tells the compiler to favor 12005 zero-displacement branches. This is enabled by default for SH4* 12006 targets. 12007 * The pref instruction will now be emitted by the __builtin_prefetch 12008 built-in function for SH3* targets. 12009 * The fmac instruction will now be emitted by the fmaf standard 12010 function and the __builtin_fmaf built-in function. 12011 * The -mfused-madd option has been deprecated in favor of the 12012 machine-independent -ffp-contract option. Notice that the fmac 12013 instruction will now be generated by default for expressions like a 12014 * b + c. This is due to the compiler default setting 12015 -ffp-contract=fast. 12016 * Added new options -mfsrra and -mfsca to allow the compiler using 12017 the fsrra and fsca instructions on targets other than SH4A (where 12018 they are already enabled by default). 12019 * Added support for the __builtin_bswap32 built-in function. It is 12020 now expanded as a sequence of swap.b and swap.w instructions 12021 instead of a library function call. 12022 * The behavior of the -mieee option has been fixed and the negative 12023 form -mno-ieee has been added to control the IEEE conformance of 12024 floating point comparisons. By default -mieee is now enabled and 12025 the option -ffinite-math-only implicitly sets -mno-ieee. 12026 * Added support for the built-in functions __builtin_thread_pointer 12027 and __builtin_set_thread_pointer. This assumes that GBR is used to 12028 hold the thread pointer of the current thread. Memory loads and 12029 stores relative to the address returned by __builtin_thread_pointer 12030 will now also utilize GBR based displacement address modes. 12031 * The -mdiv= option for targets other than SHmedia has been fixed and 12032 documented. 12033 12034 [45]SPARC 12035 12036 * Added optimized instruction scheduling for Niagara4. 12037 12038 [46]TILE-Gx 12039 12040 * Added support for the -mcmodel=MODEL command-line option. The 12041 models supported are small and large. 12042 12043 [47]V850 12044 12045 * This target now supports the E3V5 architecture via the use of the 12046 new -mv850e3v5 command-line option. It also has experimental 12047 support for the e3v5 LOOP instruction which can be enabled via the 12048 new -mloop command-line option. 12049 12050 [48]XStormy16 12051 12052 * This target now supports the -fstack-usage command-line option. 12053 12054 [49]Operating Systems 12055 12056 [50]OpenBSD 12057 12058 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 12059 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 12060 12061 [51]Windows (Cygwin) 12062 12063 * Executables are now linked against shared libgcc by default. The 12064 previous default was to link statically, which can still be done by 12065 explicitly specifying -static or static-libgcc on the command line. 12066 However it is strongly advised against, as it will cause problems 12067 for any application that makes use of DLLs compiled by GCC. It 12068 should be alright for a monolithic stand-alone application that 12069 only links against the Windows DLLs, but offers little or no 12070 benefit. 12071 12072 [52]GCC 4.8.1 12073 12074 This is the [53]list of problem reports (PRs) from GCC's bug tracking 12075 system that are known to be fixed in the 4.8.1 release. This list might 12076 not be complete (that is, it is possible that some PRs that have been 12077 fixed are not listed here). 12078 12079 The C++11 <chrono> std::chrono::system_clock and 12080 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 12081 both are now separate (never typedefs of each other), both use 12082 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 12083 std::chrono::steady_clock is now finally monotonic, and both classes 12084 are mangled differently than in the previous GCC releases. 12085 std::chrono::system_clock::now() with std::chrono::microseconds resp. 12086 std::chrono::seconds resolution is still exported for backwards 12087 compatibility with default configured libstdc++. Note that libstdc++ 12088 configured with --enable-libstdcxx-time= used to be ABI incompatible 12089 with default configured libstdc++ for those two classes and no ABI 12090 compatibility can be offered for those configurations, so any C++11 12091 code that uses those classes and has been compiled and linked against 12092 libstdc++ configured with the non-default --enable-libstdcxx-time= 12093 configuration option needs to be recompiled. 12094 12095 [54]GCC 4.8.2 12096 12097 This is the [55]list of problem reports (PRs) from GCC's bug tracking 12098 system that are known to be fixed in the 4.8.2 release. This list might 12099 not be complete (that is, it is possible that some PRs that have been 12100 fixed are not listed here). 12101 12102 [56]GCC 4.8.3 12103 12104 This is the [57]list of problem reports (PRs) from GCC's bug tracking 12105 system that are known to be fixed in the 4.8.3 release. This list might 12106 not be complete (that is, it is possible that some PRs that have been 12107 fixed are not listed here). 12108 12109 Support for the new powerpc64le-linux platform has been added. It 12110 defaults to generating code that conforms to the ELFV2 ABI. 12111 12112 [58]GCC 4.8.4 12113 12114 This is the [59]list of problem reports (PRs) from GCC's bug tracking 12115 system that are known to be fixed in the 4.8.4 release. This list might 12116 not be complete (that is, it is possible that some PRs that have been 12117 fixed are not listed here). 12118 12119 [60]GCC 4.8.5 12120 12121 This is the [61]list of problem reports (PRs) from GCC's bug tracking 12122 system that are known to be fixed in the 4.8.5 release. This list might 12123 not be complete (that is, it is possible that some PRs that have been 12124 fixed are not listed here). 12125 12126 12127 For questions related to the use of GCC, please consult these web 12128 pages and the [62]GCC manuals. If that fails, the 12129 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12130 web pages and the development of GCC are welcome on our developer 12131 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 12132 archives. 12133 12134 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 12135 distribution of this entire article is permitted in any medium, 12136 provided this notice is preserved. 12137 12138 These pages are [67]maintained by the GCC team. Last modified 12139 2026-06-08. 12140 12141 References 12142 12143 1. https://gcc.gnu.org/wiki/cxx-conversion 12144 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 12145 3. https://gcc.gnu.org/gcc-4.8/porting_to.html 12146 4. https://github.com/google/sanitizers 12147 5. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 12148 6. https://lkml.org/lkml/2006/11/28/239 12149 7. https://gcc.gnu.org/gcc-4.8/changes.html#cxx 12150 8. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12151 9. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12152 10. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12153 11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 12154 12. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12155 13. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 12156 14. https://gcc.gnu.org/projects/cxx1y.html 12157 15. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 12158 16. https://gcc.gnu.org/gcc-4.8/changes.html#fortran 12159 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 12160 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12161 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12162 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 12163 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12164 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12165 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 12166 24. https://gcc.gnu.org/wiki/Fortran2003Status 12167 25. https://gcc.gnu.org/wiki/TS29113Status 12168 26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h 12169 27. https://chasm-interop.sourceforge.net/ 12170 28. https://gcc.gnu.org/gcc-4.8/changes.html#go 12171 29. https://gcc.gnu.org/gcc-4.8/changes.html#targets 12172 30. https://gcc.gnu.org/gcc-4.8/changes.html#aarch64 12173 31. https://gcc.gnu.org/gcc-4.8/changes.html#arm 12174 32. https://gcc.gnu.org/gcc-4.8/changes.html#avr 12175 33. http://www.nongnu.org/avr-libc/ 12176 34. https://gcc.gnu.org/PR54461 12177 35. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 12178 36. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 12179 37. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 12180 38. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 12181 39. https://gcc.gnu.org/wiki/FunctionMultiVersioning 12182 40. https://gcc.gnu.org/gcc-4.8/changes.html#frv 12183 41. https://gcc.gnu.org/gcc-4.8/changes.html#mips 12184 42. https://gcc.gnu.org/gcc-4.8/changes.html#powerpc 12185 43. https://gcc.gnu.org/gcc-4.8/changes.html#rx 12186 44. https://gcc.gnu.org/gcc-4.8/changes.html#sh 12187 45. https://gcc.gnu.org/gcc-4.8/changes.html#sparc 12188 46. https://gcc.gnu.org/gcc-4.8/changes.html#tilegx 12189 47. https://gcc.gnu.org/gcc-4.8/changes.html#v850 12190 48. https://gcc.gnu.org/gcc-4.8/changes.html#xstormy16 12191 49. https://gcc.gnu.org/gcc-4.8/changes.html#os 12192 50. https://gcc.gnu.org/gcc-4.8/changes.html#openbsd 12193 51. https://gcc.gnu.org/gcc-4.8/changes.html#windows 12194 52. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.1 12195 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 12196 54. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.2 12197 55. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 12198 56. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.3 12199 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 12200 58. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.4 12201 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 12202 60. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.5 12203 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 12204 62. https://gcc.gnu.org/onlinedocs/ 12205 63. mailto:gcc-help (a] gcc.gnu.org 12206 64. mailto:gcc (a] gcc.gnu.org 12207 65. https://gcc.gnu.org/lists.html 12208 66. https://www.fsf.org/ 12209 67. https://gcc.gnu.org/about.html 12210 ====================================================================== 12211 http://gcc.gnu.org/gcc-4.7/index.html 12212 GCC 4.7 Release Series 12213 12214 (This release series is no longer supported.) 12215 12216 June 12, 2014 12217 12218 The [1]GNU project and the GCC developers are pleased to announce the 12219 release of GCC 4.7.4. 12220 12221 This release is a bug-fix release, containing fixes for regressions in 12222 GCC 4.7.3 relative to previous releases of GCC. 12223 12224 Release History 12225 12226 GCC 4.7.4 12227 June 12, 2014 ([2]changes, [3]documentation) 12228 12229 GCC 4.7.3 12230 April 11, 2013 ([4]changes, [5]documentation) 12231 12232 GCC 4.7.2 12233 September 20, 2012 ([6]changes, [7]documentation) 12234 12235 GCC 4.7.1 12236 June 14, 2012 ([8]changes, [9]documentation) 12237 12238 GCC 4.7.0 12239 March 22, 2012 ([10]changes, [11]documentation) 12240 12241 References and Acknowledgements 12242 12243 GCC used to stand for the GNU C Compiler, but since the compiler 12244 supports several other languages aside from C, it now stands for the 12245 GNU Compiler Collection. 12246 12247 A list of [12]successful builds is updated as new information becomes 12248 available. 12249 12250 The GCC developers would like to thank the numerous people that have 12251 contributed new features, improvements, bug fixes, and other changes as 12252 well as test results to GCC. This [13]amazing group of volunteers is 12253 what makes GCC successful. 12254 12255 For additional information about GCC please refer to the [14]GCC 12256 project web site or contact the [15]GCC development mailing list. 12257 12258 To obtain GCC please use [16]our mirror sites or [17]our version 12259 control system. 12260 12261 12262 For questions related to the use of GCC, please consult these web 12263 pages and the [18]GCC manuals. If that fails, the 12264 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12265 web pages and the development of GCC are welcome on our developer 12266 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 12267 archives. 12268 12269 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 12270 distribution of this entire article is permitted in any medium, 12271 provided this notice is preserved. 12272 12273 These pages are [23]maintained by the GCC team. Last modified 12274 2022-10-26. 12275 12276 References 12277 12278 1. http://www.gnu.org/ 12279 2. https://gcc.gnu.org/gcc-4.7/changes.html 12280 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 12281 4. https://gcc.gnu.org/gcc-4.7/changes.html 12282 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 12283 6. https://gcc.gnu.org/gcc-4.7/changes.html 12284 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 12285 8. https://gcc.gnu.org/gcc-4.7/changes.html 12286 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 12287 10. https://gcc.gnu.org/gcc-4.7/changes.html 12288 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 12289 12. https://gcc.gnu.org/gcc-4.7/buildstat.html 12290 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12291 14. https://gcc.gnu.org/index.html 12292 15. mailto:gcc (a] gcc.gnu.org 12293 16. https://gcc.gnu.org/mirrors.html 12294 17. https://gcc.gnu.org/git.html 12295 18. https://gcc.gnu.org/onlinedocs/ 12296 19. mailto:gcc-help (a] gcc.gnu.org 12297 20. mailto:gcc (a] gcc.gnu.org 12298 21. https://gcc.gnu.org/lists.html 12299 22. https://www.fsf.org/ 12300 23. https://gcc.gnu.org/about.html 12301 ====================================================================== 12302 http://gcc.gnu.org/gcc-4.7/changes.html 12303 GCC 4.7 Release Series 12304 Changes, New Features, and Fixes 12305 12306 Caveats 12307 12308 * The -fconserve-space flag has been deprecated. The flag had no 12309 effect for most targets: only targets without a global .bss section 12310 and without support for switchable sections. Furthermore, the flag 12311 only had an effect for G++, where it could result in wrong 12312 semantics (please refer to the GCC manual for further details). The 12313 flag will be removed in GCC 4.8 12314 * Support for a number of older systems and recently unmaintained or 12315 untested target ports of GCC has been declared obsolete in GCC 4.7. 12316 Unless there is activity to revive them, the next release of GCC 12317 will have their sources permanently removed. 12318 All GCC ports for the following processor architectures have been 12319 declared obsolete: 12320 + picoChip (picochip-*) 12321 The following ports for individual systems on particular 12322 architectures have been obsoleted: 12323 + IRIX 6.5 (mips-sgi-irix6.5) 12324 + MIPS OpenBSD (mips*-*-openbsd*) 12325 + Solaris 8 (*-*-solaris2.8). Details can be found in the 12326 [1]announcement. 12327 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 12328 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 12329 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 12330 default, which for some sources generates code that accesses memory 12331 on unaligned addresses. This requires the kernel of those systems 12332 to enable such accesses (controlled by CP15 register c1, refer to 12333 ARM documentation). Alternatively, or for compatibility with 12334 kernels where unaligned accesses are not supported, all code has to 12335 be compiled with -mno-unaligned-access. Upstream Linux kernel 12336 releases have automatically and unconditionally supported unaligned 12337 accesses as emitted by GCC due to this option being active since 12338 version 2.6.28. 12339 * Support on ARM for the legacy floating-point accelerator (FPA) and 12340 the mixed-endian floating-point format that it used has been 12341 obsoleted. The ports that still use this format have been obsoleted 12342 as well. Many legacy ARM ports already provide an alternative that 12343 uses the VFP floating-point format. The obsolete ports will be 12344 deleted in the next release. 12345 The obsolete ports with alternatives are: 12346 + arm*-*-rtems (use arm*-*-rtemseabi) 12347 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 12348 + arm*-*-elf (use arm*-*-eabi) 12349 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 12350 Note, however, that these alternatives are not binary compatible 12351 with their legacy counterparts (although some can support running 12352 legacy applications). 12353 The obsolete ports that currently lack a modern alternative are: 12354 + arm*-*-ecos-elf 12355 + arm*-*-freebsd 12356 + arm*-wince-pe* 12357 New ports that support more recent versions of the architecture are 12358 welcome. 12359 * Support for the Maverick co-processor on ARM has been obsoleted. 12360 Code to support it will be deleted in the next release. 12361 * Support has been removed for Unix International threads on Solaris 12362 2, so the --enable-threads=solaris configure option and the 12363 -threads compiler option don't work any longer. 12364 * Support has been removed for the Solaris BSD Compatibility Package, 12365 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 12366 from Solaris 11, and was only intended as a migration aid from 12367 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 12368 recognized any longer. 12369 * The ARM port's -mwords-little-endian option has been deprecated. It 12370 will be removed in a future release. 12371 * Support has been removed for the NetWare x86 configuration 12372 obsoleted in GCC 4.6. 12373 * It is no longer possible to use the "l" constraint in MIPS16 asm 12374 statements. 12375 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 12376 library which affected the ABI in C++11 mode: a data member was 12377 added to std::list changing its size and altering the definitions 12378 of some member functions, and std::pair's move constructor was 12379 non-trivial which altered the calling convention for functions with 12380 std::pair arguments or return types. The ABI incompatibilities have 12381 been fixed for GCC version 4.7.2 but as a result C++11 code 12382 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 12383 code compiled with different GCC versions and with C++98/C++03 code 12384 compiled with any version. 12385 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 12386 rules for the layout of vectors that could lead to wrong code being 12387 generated. Vectors larger than 8 bytes in size are now by default 12388 aligned to an 8-byte boundary. This is an ABI change: code that 12389 makes explicit use of vector types may be incompatible with binary 12390 objects built with older versions of GCC. Auto-vectorized code is 12391 not affected by this change. (This change affects GCC versions 12392 4.7.2 and later.) 12393 * More information on porting to GCC 4.7 from previous versions of 12394 GCC can be found in the [2]porting guide for this release. 12395 12396 General Optimizer Improvements 12397 12398 * Support for a new parameter --param case-values-threshold=n was 12399 added to allow users to control the cutoff between doing switch 12400 statements as a series of if statements and using a jump table. 12401 * Link-time optimization (LTO) improvements: 12402 + Improved scalability and reduced memory usage. Link time 12403 optimization of Firefox now requires 3GB of RAM on a 64-bit 12404 system, while over 8GB was needed previously. Linking time has 12405 been improved, too. The serial stage of linking Firefox has 12406 been sped up by about a factor of 10. 12407 + Reduced size of object files and temporary storage used during 12408 linking. 12409 + Streaming performance (both outbound and inbound) has been 12410 improved. 12411 + ld -r is now supported with LTO. 12412 + Several bug fixes, especially in symbol table handling and 12413 merging. 12414 * Interprocedural optimization improvements: 12415 + Heuristics now take into account that after inlining code will 12416 be optimized out because of known values (or properties) of 12417 function parameters. For example: 12418 void foo(int a) 12419 { 12420 if (a > 10) 12421 ... huge code ... 12422 } 12423 void bar (void) 12424 { 12425 foo (0); 12426 } 12427 12428 The call of foo will be inlined into bar even when optimizing 12429 for code size. Constructs based on __builtin_constant_p are 12430 now understood by the inliner and code size estimates are 12431 evaluated a lot more realistically. 12432 + The representation of C++ virtual thunks and aliases (both 12433 implicit and defined via the alias attribute) has been 12434 re-engineered. Aliases no longer pose optimization barriers 12435 and calls to an alias can be inlined and otherwise optimized. 12436 + The inter-procedural constant propagation pass has been 12437 rewritten. It now performs generic function specialization. 12438 For example when compiling the following: 12439 void foo(bool flag) 12440 { 12441 if (flag) 12442 ... do something ... 12443 else 12444 ... do something else ... 12445 } 12446 void bar (void) 12447 { 12448 foo (false); 12449 foo (true); 12450 foo (false); 12451 foo (true); 12452 foo (false); 12453 foo (true); 12454 } 12455 12456 GCC will now produce two copies of foo. One with flag being 12457 true, while other with flag being false. This leads to 12458 performance improvements previously possible only by inlining 12459 all calls. Cloning causes a lot less code size growth. 12460 * A string length optimization pass has been added. It attempts to 12461 track string lengths and optimize various standard C string 12462 functions like strlen, strchr, strcpy, strcat, stpcpy and their 12463 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 12464 enabled by default at -O2 or above, unless optimizing for size, and 12465 can be disabled by the -fno-optimize-strlen option. The pass can 12466 e.g. optimize 12467 char *bar (const char *a) 12468 { 12469 size_t l = strlen (a) + 2; 12470 char *p = malloc (l); if (p == NULL) return p; 12471 strcpy (p, a); strcat (p, "/"); return p; 12472 } 12473 12474 into: 12475 char *bar (const char *a) 12476 { 12477 size_t tmp = strlen (a); 12478 char *p = malloc (tmp + 2); if (p == NULL) return p; 12479 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 12480 } 12481 12482 or for hosted compilations where stpcpy is available in the runtime 12483 and headers provide its prototype, e.g. 12484 void foo (char *a, const char *b, const char *c, const char *d) 12485 { 12486 strcpy (a, b); strcat (a, c); strcat (a, d); 12487 } 12488 12489 can be optimized into: 12490 void foo (char *a, const char *b, const char *c, const char *d) 12491 { 12492 strcpy (stpcpy (stpcpy (a, b), c), d); 12493 } 12494 12495 New Languages and Language specific improvements 12496 12497 * Version 3.1 of the OpenMP specification is now supported for the C, 12498 C++, and Fortran compilers. 12499 12500 Ada 12501 12502 * The command-line option -feliminate-unused-debug-types has been 12503 re-enabled by default, as it is for the other languages, leading to 12504 a reduction in debug info size of 12.5% and more for relevant 12505 cases, as well as to a small compilation speedup. 12506 12507 C family 12508 12509 * A new built-in, __builtin_assume_aligned, has been added, through 12510 which the compiler can be hinted about pointer alignment and can 12511 use it to improve generated code. 12512 * A new warning option -Wunused-local-typedefs was added for C, C++, 12513 Objective-C and Objective-C++. This warning diagnoses typedefs 12514 locally defined in a function, and otherwise not used. 12515 * A new experimental command-line option -ftrack-macro-expansion was 12516 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 12517 the compiler to emit diagnostic about the current macro expansion 12518 stack when a compilation error occurs in a macro expansion. 12519 * Experimental support for transactional memory has been added. It 12520 includes support in the compiler, as well as a supporting runtime 12521 library called libitm. To compile code with transactional memory 12522 constructs, use the -fgnu-tm option. 12523 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 12524 and 32-bit/64-bit x86 platforms. 12525 For more details on transactional memory see [3]the GCC WiKi. 12526 * Support for atomic operations specifying the C++11/C11 memory model 12527 has been added. These new __atomic routines replace the existing 12528 __sync built-in routines. 12529 Atomic support is also available for memory blocks. Lock-free 12530 instructions will be used if a memory block is the same size and 12531 alignment as a supported integer type. Atomic operations which do 12532 not have lock-free support are left as function calls. A set of 12533 library functions is available on the GCC atomic wiki in the 12534 "External Atomics Library" section. 12535 For more details on the memory models and features, see the 12536 [4]atomic wiki. 12537 * When a binary operation is performed on vector types and one of the 12538 operands is a uniform vector, it is possible to replace the vector 12539 with the generating element. For example: 12540 typedef int v4si __attribute__ ((vector_size (16))); 12541 v4si res, a = {1,2,3,4}; 12542 int x; 12543 12544 res = 2 + a; /* means {2,2,2,2} + a */ 12545 res = a - x; /* means a - {x,x,x,x} */ 12546 12547 [5]C 12548 12549 * There is support for some more features from the C11 revision of 12550 the ISO C standard. GCC now accepts the options -std=c11 and 12551 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 12552 + Unicode strings (previously supported only with options such 12553 as -std=gnu11, now supported with -std=c11), and the 12554 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 12555 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 12556 + Alignment support (_Alignas, _Alignof, max_align_t, 12557 <stdalign.h>). 12558 + A built-in function __builtin_complex is provided to support C 12559 library implementation of the CMPLX family of macros. 12560 12561 [6]C++ 12562 12563 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 12564 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 12565 -Wc++0x-compat, respectively. 12566 * G++ now implements [7]C++11 extended friend syntax: 12567 12568 template<class W> 12569 class Q 12570 { 12571 static const int I = 2; 12572 public: 12573 friend W; 12574 }; 12575 12576 struct B 12577 { 12578 int ar[Q<B>::I]; 12579 }; 12580 12581 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 12582 override control. 12583 12584 struct B { 12585 virtual void f() const final; 12586 virtual void f(int); 12587 }; 12588 12589 struct D : B { 12590 void f() const; // error: D::f attempts to override final B::f 12591 void f(long) override; // error: doesn't override anything 12592 void f(int) override; // ok 12593 }; 12594 12595 struct E final { }; 12596 struct F: E { }; // error: deriving from final class 12597 12598 * G++ now implements [9]C++11 non-static data member initializers. 12599 12600 struct A { 12601 int i = 42; 12602 } a; // initializes a.i to 42 12603 12604 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 12605 user-defined literals. 12606 12607 // Not actually a good approximation. :) 12608 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 12609 long double pi = 180.0_degrees; 12610 12611 * G++ now implements [11]C++11 alias-declarations. 12612 12613 template <class T> using Ptr = T*; 12614 Ptr<int> ip; // decltype(ip) is int* 12615 12616 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 12617 [12]C++11 delegating constructors. 12618 12619 struct A { 12620 A(int); 12621 A(): A(42) { } // delegate to the A(int) constructor 12622 }; 12623 12624 * G++ now fully implements C++11 atomic classes rather than just 12625 integer derived classes. 12626 12627 class POD { 12628 int a; 12629 int b; 12630 }; 12631 std::atomic<POD> my_atomic_POD; 12632 12633 * G++ now sets the predefined macro __cplusplus to the correct value, 12634 199711L for C++98/03, and 201103L for C++11. 12635 * G++ now correctly implements the two-phase lookup rules such that 12636 an unqualified name used in a template must have an appropriate 12637 declaration found either in scope at the point of definition of the 12638 template or by argument-dependent lookup at the point of 12639 instantiation. As a result, code that relies on a second 12640 unqualified lookup at the point of instantiation to find functions 12641 declared after the template or in dependent bases will be rejected. 12642 The compiler will suggest ways to fix affected code, and using the 12643 -fpermissive compiler flag will allow the code to compile with a 12644 warning. 12645 12646 template <class T> 12647 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 12648 void g(int) { } // fix by moving this declaration before the declaration of f 12649 12650 template <class T> 12651 struct A: T { 12652 // error, B::g(B) not found by argument-dependent lookup 12653 void f() { g(T()); } // fix by using this->g or A::g 12654 }; 12655 12656 struct B { void g(B); }; 12657 12658 int main() 12659 { 12660 f<int>(); 12661 A<B>().f(); 12662 } 12663 12664 * G++ now properly re-uses stack space allocated for temporary 12665 objects when their lifetime ends, which can significantly lower 12666 stack consumption for some C++ functions. As a result of this, some 12667 code with undefined behavior will now break: 12668 12669 const int &f(const int &i) { return i; } 12670 .... 12671 const int &x = f(1); 12672 const int &y = f(2); 12673 12674 Here, x refers to the temporary allocated to hold the 1 argument, 12675 which only lives until the end of the initialization; it 12676 immediately becomes a dangling reference. So the next statement 12677 re-uses the stack slot to hold the 2 argument, and users of x get 12678 that value instead. 12679 Note that this should not cause any change of behavior for 12680 temporaries of types with non-trivial destructors, as they are 12681 already destroyed at end of full-expression; the change is that now 12682 the storage is released as well. 12683 * A new command-line option -Wdelete-non-virtual-dtor has been added 12684 to warn when delete is used to destroy an instance of a class which 12685 has virtual functions and non-virtual destructor. It is unsafe to 12686 delete an instance of a derived class through a pointer to a base 12687 class if the base class does not have a virtual destructor. This 12688 warning is enabled by -Wall. 12689 * A new command-line option -Wzero-as-null-pointer-constant has been 12690 added to warn when a literal '0' is used as null pointer constant. 12691 It can be useful to facilitate the conversion to nullptr in C++11. 12692 * As per C++98, access-declarations are now deprecated by G++. 12693 Using-declarations are to be used instead. Furthermore, some 12694 efforts have been made to improve the support of class scope 12695 using-declarations. In particular, using-declarations referring to 12696 a dependent type now work as expected ([13]bug c++/14258). 12697 * The ELF symbol visibility of a template instantiation is now 12698 properly constrained by the visibility of its template arguments 12699 ([14]bug c++/35688). 12700 12701 Runtime Library (libstdc++) 12702 12703 * [15]Improved experimental support for the new ISO C++ standard, 12704 C++11, including: 12705 + using noexcept in most of the library; 12706 + implementations of pointer_traits, allocator_traits and 12707 scoped_allocator_adaptor; 12708 + uses-allocator construction for tuple; 12709 + vector meets the allocator-aware container requirements; 12710 + replacing monotonic_clock with steady_clock; 12711 + enabling the thread support library on most POSIX targets; 12712 + many small improvements to conform to the FDIS. 12713 * Added --enable-clocale=newlib configure option. 12714 * Debug Mode iterators for unordered associative containers. 12715 * Avoid polluting the global namespace and do not include <unistd.h>. 12716 12717 [16]Fortran 12718 12719 * The compile flag [17]-fstack-arrays has been added, which causes 12720 all local arrays to be put on stack memory. For some programs this 12721 will improve the performance significantly. If your program uses 12722 very large local arrays, it is possible that you will have to 12723 extend your runtime limits for stack memory. 12724 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 12725 [20]-fstack-arrays. 12726 * Front-end optimizations can now be selected by the 12727 [21]-ffrontend-optimize option and deselected by the 12728 -fno-frontend-optimize option. 12729 * When front-end optimization removes a function call, 12730 [22]-Wfunction-elimination warns about that. 12731 * When performing front-end-optimization, the 12732 [23]-faggressive-function-elimination option allows the removal of 12733 duplicate function calls even for impure functions. 12734 * The flag [24]-Wreal-q-constant has been added, which warns if 12735 floating-point literals have been specified using q (such as 12736 1.0q0); the q marker is now supported as a vendor extension to 12737 denote quad precision (REAL(16) or, if not available, REAL(10)). 12738 Consider using a kind parameter (such as in 1.0_qp) instead, which 12739 can be obtained via [25]SELECTED_REAL_KIND. 12740 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 12741 Fortran now always prints error messages to standard error. If you 12742 wish to redirect standard error, please consult the manual for your 12743 OS, shell, batch environment etc. as appropriate. 12744 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 12745 variable have been removed. When encountering a serious error, 12746 gfortran will now always abort the program. Whether a core dump is 12747 generated depends on the user environment settings; see the ulimit 12748 -c setting for POSIX shells, limit coredumpsize for C shells, and 12749 the [26]WER user-mode dumps settings on Windows. 12750 * The [27]-fbacktrace option is now enabled by default. When 12751 encountering a fatal error, gfortran will attempt to print a 12752 backtrace to standard error before aborting. It can be disabled 12753 with -fno-backtrace. Note: On POSIX targets with the addr2line 12754 utility from GNU binutils, GNU Fortran can print a backtrace with 12755 function name, file name, line number information in addition to 12756 the addresses; otherwise only the addresses are printed. 12757 * [28]Fortran 2003: 12758 + Generic interface names which have the same name as derived 12759 types are now supported, which allows to write constructor 12760 functions. Note that Fortran does not support static 12761 constructor functions; only default initialization or an 12762 explicit structure-constructor initialization are available. 12763 + [29]Polymorphic (class) arrays are now supported. 12764 * [30]Fortran 2008: 12765 + Support for the DO CONCURRENT construct has been added, which 12766 allows the user to specify that individual loop iterations 12767 have no interdependencies. 12768 + [31]Coarrays: Full single-image support except for polymorphic 12769 coarrays. Additionally, preliminary support for multiple 12770 images via an MPI-based [32]coarray communication library has 12771 been added. Note: The library version is not yet usable as 12772 remote coarray access is not yet possible. 12773 * [33]TS 29113: 12774 + New flag [34]-std=f2008ts permits programs that are expected 12775 to conform to the Fortran 2008 standard and the draft 12776 Technical Specification (TS) 29113 on Further Interoperability 12777 of Fortran with C. 12778 + The OPTIONAL attribute is now allowed for dummy arguments of 12779 BIND(C) procedures. 12780 + The RANK intrinsic has been added. 12781 + The implementation of the ASYNCHRONOUS attribute in GCC is 12782 compatible with the candidate draft of TS 29113 (since GCC 12783 4.6). 12784 12785 [35]Go 12786 12787 * GCC 4.7 implements the [36]Go 1 language standard. The library 12788 support in 4.7.0 is not quite complete, due to release timing. 12789 Release 4.7.1 includes complete support for Go 1. The Go library is 12790 from the Go 1.0.1 release. 12791 * Go has been tested on GNU/Linux and Solaris platforms. It may work 12792 on other platforms as well. 12793 12794 [37]New Targets and Target Specific Improvements 12795 12796 [38]ARM 12797 12798 * GCC now supports the Cortex-A7 processor implementing the v7-a 12799 version of the architecture using the option -mcpu=cortex-a7. 12800 * The default vector size in auto-vectorization for NEON is now 128 12801 bits. If vectorization fails thusly, the vectorizer tries again 12802 with 64-bit vectors. 12803 * A new option -mvectorize-with-neon-double was added to allow users 12804 to change the vector size to 64 bits. 12805 12806 [39]AVR 12807 12808 * The AVR port's libgcc has been improved and its multilib structure 12809 has been enhanced. As a result, all objects contributing to an 12810 application must either be compiled with GCC versions up to 4.6.x 12811 or with GCC versions 4.7.1 or later. If the compiler is used with 12812 AVR-LibC, you need a version that supports the new layout, i.e. 12813 implements [40]#35407. 12814 * The -mshort-calls command-line option has been deprecated. It will 12815 be removed in the GCC 4.8 release. See -mrelax for a replacement. 12816 * The compiled code only references startup code that clears .bss and 12817 the common section resp. initializes the .data and .rodata section 12818 provided respective sections (or subsections thereof) are not 12819 empty, see [41]PR18145. Applications that put all static storage 12820 objects into non-standard sections and / or define all static 12821 storage objects in assembler modules, must reference __do_clear_bss 12822 resp. __do_copy_data by hand or undefine the symbol(s) by means of 12823 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 12824 * GCC now supports the XMEGA architecture. This requires GNU Binutils 12825 2.22 or later. 12826 * Support for the [42]named address spaces __flash, __flash1, , 12827 __flash5 and __memx has been added. These address spaces locate 12828 read-only data in flash memory and allow reading from flash memory 12829 by means of ordinary C code, i.e. without the need of (inline) 12830 assembler code: 12831 12832 const __flash int values[] = { 42, 31 }; 12833 12834 int add_values (const __flash int *p, int i) 12835 { 12836 return values[i] + *p; 12837 } 12838 12839 * Support has been added for the AVR-specific configure option 12840 --with-avrlibc=yes in order to arrange for better integration of 12841 [43]AVR-Libc. This configure option is supported in avr-gcc v4.7.2 12842 and newer and will only take effect in non-RTEMS configurations. If 12843 avr-gcc is configured for RTEMS, the option will be ignored which 12844 is the same as specifying --with-avrlibc=no. See [44]PR54461 for 12845 more technical details. 12846 * Support for AVR-specific [45]built-in functions has been added. 12847 * Support has been added for the signed and unsigned 24-bit scalar 12848 integer types __int24 and __uint24. 12849 * New command-line options -maccumulate-args, -mbranch-cost=cost and 12850 -mstrict-X were added to allow better fine-tuning of code 12851 optimization. 12852 * The command-line option -fdata-sections now also takes affect on 12853 the section names of variables with the progmem attribute. 12854 * A new inline assembler print modifier %i to print a RAM address as 12855 I/O address has been added: 12856 12857 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 12858 12859 void set_portb (uint8_t value) 12860 { 12861 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 12862 } 12863 12864 The offset between an I/O address and the RAM address for that I/O 12865 location is device-specific. This offset is taken into account when 12866 printing a RAM address with the %i modifier so that the address is 12867 suitable to be used as operand in an I/O command. The address must 12868 be a constant integer known at compile time. 12869 * The inline assembler constraint "R" to represent integers in the 12870 range 6 5 has been removed without replacement. 12871 * Many optimizations to: 12872 + 64-bit integer arithmetic 12873 + Widening multiplication 12874 + Integer division by a constant 12875 + Avoid constant reloading in multi-byte instructions. 12876 + Micro-optimizations for special instruction sequences. 12877 + Generic built-in functions like __builtin_ffs*, 12878 __builtin_clz*, etc. 12879 + If-else decision trees generated by switch instructions 12880 + Merging of data located in flash memory 12881 + New libgcc variants for devices with 8-bit wide stack pointer 12882 + 12883 * Better documentation: 12884 + Handling of EIND and indirect jumps on devices with more than 12885 128 KiB of program memory. 12886 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 12887 registers. 12888 + Function attributes OS_main and OS_task. 12889 + AVR-specific built-in macros. 12890 12891 C6X 12892 12893 * Support has been added for the Texas Instruments C6X family of 12894 processors. 12895 12896 CR16 12897 12898 * Support has been added for National Semiconductor's CR16 12899 architecture. 12900 12901 Epiphany 12902 12903 * Support has been added for Adapteva's Epiphany architecture. 12904 12905 IA-32/x86-64 12906 12907 * Support for Intel AVX2 intrinsics, built-in functions and code 12908 generation is available via -mavx2. 12909 * Support for Intel BMI2 intrinsics, built-in functions and code 12910 generation is available via -mbmi2. 12911 * Implementation and automatic generation of __builtin_clz* using the 12912 lzcnt instruction is available via -mlzcnt. 12913 * Support for Intel FMA3 intrinsics and code generation is available 12914 via -mfma. 12915 * A new -mfsgsbase command-line option is available that makes GCC 12916 generate new segment register read/write instructions through 12917 dedicated built-ins. 12918 * Support for the new Intel rdrnd instruction is available via 12919 -mrdrnd. 12920 * Two additional AVX vector conversion instructions are available via 12921 -mf16c. 12922 * Support for new Intel processor codename IvyBridge with RDRND, 12923 FSGSBASE and F16C is available through -march=core-avx-i. 12924 * Support for the new Intel processor codename Haswell with AVX2, 12925 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 12926 * Support for new AMD family 15h processors (Piledriver core) is now 12927 available through -march=bdver2 and -mtune=bdver2 options. 12928 * Support for [46]the x32 psABI is now available through the -mx32 12929 option. 12930 * Windows mingw targets are using the -mms-bitfields option by 12931 default. 12932 * Windows x86 targets are using the __thiscall calling convention for 12933 C++ class-member functions. 12934 * Support for the configure option --with-threads=posix for Windows 12935 mingw targets. 12936 12937 [47]MIPS 12938 12939 * GCC now supports thread-local storage (TLS) for MIPS16. This 12940 requires GNU binutils 2.22 or later. 12941 * GCC can now generate code specifically for the Cavium Octeon+ and 12942 Octeon2 processors. The associated command-line options are 12943 -march=octeon+ and -march=octeon2 respectively. Both options 12944 require GNU binutils 2.22 or later. 12945 * GCC can now work around certain 24k errata, under the control of 12946 the command-line option -mfix-24k. These workarounds require GNU 12947 binutils 2.20 or later. 12948 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 12949 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 12950 toolchain that generates 32-bit code by default. Use the 12951 configure-time option --enable-targets=all to select these extra 12952 multilibs. 12953 * Passing -fno-delayed-branch now also stops the assembler from 12954 automatically filling delay slots. 12955 12956 PowerPC/PowerPC64 12957 12958 * Vectors of type vector long long or vector long are passed and 12959 returned using the same method as other vectors with the VSX 12960 instruction set. Previously GCC did not adhere to the ABI for 12961 128-bit vectors with 64-bit integer base types (PR 48857). This 12962 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 12963 * A new option -mno-pointers-to-nested-functions was added to allow 12964 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 12965 that the compiler should not load up the chain register (r11) 12966 before calling a function through a pointer. If you use this 12967 option, you cannot call nested functions through a pointer, or call 12968 other languages that might use the static chain. 12969 * A new option msave-toc-indirect was added to allow AIX 12970 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 12971 save the TOC in the prologue for indirect calls or generate the 12972 save inline. This can speed up some programs that call through a 12973 function pointer a lot, but it can slow down other functions that 12974 only call through a function pointer in exceptional cases. 12975 * The PowerPC port will now enable machine-specific built-in 12976 functions when the user switches the target machine using the 12977 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 12978 sequences. In addition, the target macros are updated. However, due 12979 to the way the -save-temps switch is implemented, you won't see the 12980 effect of these additional macros being defined in preprocessor 12981 output. 12982 12983 SH 12984 12985 * A new option -msoft-atomic has been added. When it is specified, 12986 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 12987 the new __atomic routines. 12988 * Since it is neither supported by GAS nor officially documented, 12989 code generation for little endian SH2A has been disabled. 12990 Specifying -ml with -m2a* will now result in a compiler error. 12991 * The defunct -mbranch-cost option has been fixed. 12992 * Some improvements to the generated code of: 12993 + Utilization of the tst #imm,R0 instruction. 12994 + Dynamic shift instructions on SH2A. 12995 + Integer absolute value calculations. 12996 * The -mdiv= option for targets other than SHmedia has been fixed and 12997 documented. 12998 12999 SPARC 13000 13001 * The option -mflat has been reinstated. When it is specified, the 13002 compiler will generate code for a single register window model. 13003 This is essentially a new implementation and the corresponding 13004 debugger support has been added to GDB 7.4. 13005 * Support for the options -mtune=native and -mcpu=native has been 13006 added on selected native platforms (GNU/Linux and Solaris). 13007 * Support for the SPARC T3 (Niagara 3) processor has been added. 13008 * VIS: 13009 + An intrinsics header visintrin.h has been added. 13010 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 13011 compare instructions have been added. 13012 + The little-endian version of alignaddr is now supported. 13013 + When possible, VIS builtins are marked const, which should 13014 increase the compiler's ability to optimize VIS operations. 13015 + The compiler now properly tracks the %gsr register and how it 13016 behaves as an input for various VIS instructions. 13017 + Akin to fzero, the compiler can now generate fone instructions 13018 in order to set all of the bits of a floating-point register 13019 to 1. 13020 + The documentation for the VIS intrinsics in the GCC manual has 13021 been brought up to date and many inaccuracies were fixed. 13022 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 13023 non-condition-code setting edge instructions have been added. 13024 Their availability is controlled by the new -mvis2 and 13025 -mno-vis2 options. They are enabled by default on 13026 UltraSPARC-III and later CPUs. 13027 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 13028 has been added. These instructions are enabled by default on SPARC 13029 T3 (Niagara 3) and later CPUs. 13030 13031 TILE-Gx/TILEPro 13032 13033 * Support has been added for the Tilera TILE-Gx and TILEPro families 13034 of processors. 13035 13036 Other significant improvements 13037 13038 * A new option (-grecord-gcc-switches) was added that appends 13039 compiler command-line options that might affect code generation to 13040 the DW_AT_producer attribute string in the DWARF debugging 13041 information. 13042 * GCC now supports various new GNU extensions to the DWARF debugging 13043 information format, like [48]entry value and [49]call site 13044 information, [50]typed DWARF stack or [51]a more compact macro 13045 representation. Support for these extensions has been added to GDB 13046 7.4. They can be disabled through the -gstrict-dwarf command-line 13047 option. 13048 13049 [52]GCC 4.7.1 13050 13051 This is the [53]list of problem reports (PRs) from GCC's bug tracking 13052 system that are known to be fixed in the 4.7.1 release. This list might 13053 not be complete (that is, it is possible that some PRs that have been 13054 fixed are not listed here). 13055 13056 The Go front end in the 4.7.1 release fully supports the [54]Go 1 13057 language standard. 13058 13059 [55]GCC 4.7.2 13060 13061 This is the [56]list of problem reports (PRs) from GCC's bug tracking 13062 system that are known to be fixed in the 4.7.2 release. This list might 13063 not be complete (that is, it is possible that some PRs that have been 13064 fixed are not listed here). 13065 13066 [57]GCC 4.7.3 13067 13068 This is the [58]list of problem reports (PRs) from GCC's bug tracking 13069 system that are known to be fixed in the 4.7.3 release. This list might 13070 not be complete (that is, it is possible that some PRs that have been 13071 fixed are not listed here). 13072 13073 [59]GCC 4.7.4 13074 13075 This is the [60]list of problem reports (PRs) from GCC's bug tracking 13076 system that are known to be fixed in the 4.7.4 release. This list might 13077 not be complete (that is, it is possible that some PRs that have been 13078 fixed are not listed here). 13079 13080 13081 For questions related to the use of GCC, please consult these web 13082 pages and the [61]GCC manuals. If that fails, the 13083 [62]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13084 web pages and the development of GCC are welcome on our developer 13085 list at [63]gcc (a] gcc.gnu.org. All of [64]our lists have public 13086 archives. 13087 13088 Copyright (C) [65]Free Software Foundation, Inc. Verbatim copying and 13089 distribution of this entire article is permitted in any medium, 13090 provided this notice is preserved. 13091 13092 These pages are [66]maintained by the GCC team. Last modified 13093 2026-04-30. 13094 13095 References 13096 13097 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 13098 2. https://gcc.gnu.org/gcc-4.7/porting_to.html 13099 3. https://gcc.gnu.org/wiki/TransactionalMemory 13100 4. https://gcc.gnu.org/wiki/Atomic/GCCMM 13101 5. https://gcc.gnu.org/gcc-4.7/changes.html#c 13102 6. https://gcc.gnu.org/gcc-4.7/changes.html#cxx 13103 7. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13104 8. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13105 9. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13106 10. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13107 11. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13108 12. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13109 13. https://gcc.gnu.org/PR14258 13110 14. https://gcc.gnu.org/PR35688 13111 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 13112 16. https://gcc.gnu.org/gcc-4.7/changes.html#fortran 13113 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 13114 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 13115 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 13116 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 13117 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 13118 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 13119 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 13120 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 13121 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 13122 26. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 13123 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 13124 28. https://gcc.gnu.org/wiki/Fortran2003Status 13125 29. https://gcc.gnu.org/wiki/OOP 13126 30. https://gcc.gnu.org/wiki/Fortran2008Status 13127 31. https://gcc.gnu.org/wiki/Coarray 13128 32. https://gcc.gnu.org/wiki/CoarrayLib 13129 33. https://gcc.gnu.org/wiki/TS29113Status 13130 34. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 13131 35. https://gcc.gnu.org/gcc-4.7/changes.html#go 13132 36. https://go.dev/doc/go1 13133 37. https://gcc.gnu.org/gcc-4.7/changes.html#targets 13134 38. https://gcc.gnu.org/gcc-4.7/changes.html#arm 13135 39. https://gcc.gnu.org/gcc-4.7/changes.html#avr 13136 40. http://savannah.nongnu.org/bugs/?35407 13137 41. https://gcc.gnu.org/PR18145 13138 42. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 13139 43. http://www.nongnu.org/avr-libc/ 13140 44. https://gcc.gnu.org/PR54461 13141 45. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html 13142 46. https://sites.google.com/site/x32abi/ 13143 47. https://gcc.gnu.org/gcc-4.7/changes.html#mips 13144 48. https://dwarfstd.org/issues/100909.1.html 13145 49. https://dwarfstd.org/issues/100909.2.html 13146 50. https://dwarfstd.org/issues/140425.1.html 13147 51. https://dwarfstd.org/issues/110722.1.html 13148 52. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.1 13149 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 13150 54. https://go.dev/doc/go1 13151 55. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.2 13152 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 13153 57. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.3 13154 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 13155 59. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.4 13156 60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 13157 61. https://gcc.gnu.org/onlinedocs/ 13158 62. mailto:gcc-help (a] gcc.gnu.org 13159 63. mailto:gcc (a] gcc.gnu.org 13160 64. https://gcc.gnu.org/lists.html 13161 65. https://www.fsf.org/ 13162 66. https://gcc.gnu.org/about.html 13163 ====================================================================== 13164 http://gcc.gnu.org/gcc-4.6/index.html 13165 GCC 4.6 Release Series 13166 13167 (This release series is no longer supported.) 13168 13169 April 12, 2013 13170 13171 The [1]GNU project and the GCC developers are pleased to announce the 13172 release of GCC 4.6.4. 13173 13174 This release is a bug-fix release, containing fixes for regressions in 13175 GCC 4.6.3 relative to previous releases of GCC. 13176 13177 Release History 13178 13179 GCC 4.6.4 13180 April 12, 2013 ([2]changes, [3]documentation) 13181 13182 GCC 4.6.3 13183 March 1, 2012 ([4]changes, [5]documentation) 13184 13185 GCC 4.6.2 13186 October 26, 2011 ([6]changes, [7]documentation) 13187 13188 GCC 4.6.1 13189 June 27, 2011 ([8]changes, [9]documentation) 13190 13191 GCC 4.6.0 13192 March 25, 2011 ([10]changes, [11]documentation) 13193 13194 References and Acknowledgements 13195 13196 GCC used to stand for the GNU C Compiler, but since the compiler 13197 supports several other languages aside from C, it now stands for the 13198 GNU Compiler Collection. 13199 13200 A list of [12]successful builds is updated as new information becomes 13201 available. 13202 13203 The GCC developers would like to thank the numerous people that have 13204 contributed new features, improvements, bug fixes, and other changes as 13205 well as test results to GCC. This [13]amazing group of volunteers is 13206 what makes GCC successful. 13207 13208 For additional information about GCC please refer to the [14]GCC 13209 project web site or contact the [15]GCC development mailing list. 13210 13211 To obtain GCC please use [16]our mirror sites or [17]our version 13212 control system. 13213 13214 13215 For questions related to the use of GCC, please consult these web 13216 pages and the [18]GCC manuals. If that fails, the 13217 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13218 web pages and the development of GCC are welcome on our developer 13219 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 13220 archives. 13221 13222 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 13223 distribution of this entire article is permitted in any medium, 13224 provided this notice is preserved. 13225 13226 These pages are [23]maintained by the GCC team. Last modified 13227 2022-10-26. 13228 13229 References 13230 13231 1. http://www.gnu.org/ 13232 2. https://gcc.gnu.org/gcc-4.6/changes.html 13233 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 13234 4. https://gcc.gnu.org/gcc-4.6/changes.html 13235 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 13236 6. https://gcc.gnu.org/gcc-4.6/changes.html 13237 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 13238 8. https://gcc.gnu.org/gcc-4.6/changes.html 13239 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 13240 10. https://gcc.gnu.org/gcc-4.6/changes.html 13241 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 13242 12. https://gcc.gnu.org/gcc-4.6/buildstat.html 13243 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13244 14. https://gcc.gnu.org/index.html 13245 15. mailto:gcc (a] gcc.gnu.org 13246 16. https://gcc.gnu.org/mirrors.html 13247 17. https://gcc.gnu.org/git.html 13248 18. https://gcc.gnu.org/onlinedocs/ 13249 19. mailto:gcc-help (a] gcc.gnu.org 13250 20. mailto:gcc (a] gcc.gnu.org 13251 21. https://gcc.gnu.org/lists.html 13252 22. https://www.fsf.org/ 13253 23. https://gcc.gnu.org/about.html 13254 ====================================================================== 13255 http://gcc.gnu.org/gcc-4.6/changes.html 13256 GCC 4.6 Release Series 13257 Changes, New Features, and Fixes 13258 13259 Caveats 13260 13261 * The options -b <machine> and -V <version> have been removed because 13262 they were unreliable. Instead, users should directly run 13263 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 13264 run a different version of gcc. 13265 * GCC now has stricter checks for invalid command-line options. In 13266 particular, when gcc was called to link object files rather than 13267 compile source code, it would previously accept and ignore all 13268 options starting with --, including linker options such as 13269 --as-needed and --export-dynamic, although such options would 13270 result in errors if any source code was compiled. Such options, if 13271 unknown to the compiler, are now rejected in all cases; if the 13272 intent was to pass them to the linker, options such as 13273 -Wl,--as-needed should be used. 13274 * Versions of the GNU C library up to and including 2.11.1 included 13275 an [1]incorrect implementation of the cproj function. GCC optimizes 13276 its builtin cproj according to the behavior specified and allowed 13277 by the ISO C99 standard. If you want to avoid discrepancies between 13278 the C library and GCC's builtin transformations when using cproj in 13279 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 13280 and actually rely on the incorrect behavior of cproj, then you can 13281 disable GCC's transformations using -fno-builtin-cproj. 13282 * The C-only intermodule optimization framework (IMA, enabled by 13283 -combine) has been removed in favor of the new generic link-time 13284 optimization framework (LTO) introduced in [2]GCC 4.5.0. 13285 * GCC now ships with the LGPL-licensed libquadmath library, which 13286 provides quad-precision mathematical functions for targets with a 13287 __float128 datatype. __float128 is available for targets on 32-bit 13288 x86, x86-64 and Itanium architectures. The libquadmath library is 13289 automatically built on such targets when building the Fortran 13290 compiler. 13291 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 13292 warnings were added for C, C++, Objective-C and Objective-C++. 13293 These warnings diagnose variables respective parameters which are 13294 only set in the code and never otherwise used. Usually such 13295 variables are useless and often even the value assigned to them is 13296 computed needlessly, sometimes expensively. The 13297 -Wunused-but-set-variable warning is enabled by default by -Wall 13298 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 13299 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 13300 rules for the layout of vectors that could lead to wrong code being 13301 generated. Vectors larger than 8 bytes in size are now by default 13302 aligned to an 8-byte boundary. This is an ABI change: code that 13303 makes explicit use of vector types may be incompatible with binary 13304 objects built with older versions of GCC. Auto-vectorized code is 13305 not affected by this change. (This change affects GCC versions 13306 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 13307 * On AVR, variables with the progmem attribute to locate data in 13308 flash memory must be qualified as const. 13309 * Support for a number of older systems and recently unmaintained or 13310 untested target ports of GCC has been declared obsolete in GCC 4.6. 13311 Unless there is activity to revive them, the next release of GCC 13312 will have their sources permanently removed. 13313 All GCC ports for the following processor architectures have been 13314 declared obsolete: 13315 + Argonaut ARC (arc-*) 13316 + National Semiconductor CRX (crx-*) 13317 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 13318 m68hc12-*-*, m6812-*-*) 13319 + Sunplus S+core (score-*) 13320 The following ports for individual systems on particular 13321 architectures have been obsoleted: 13322 + Interix (i[34567]86-*-interix3*) 13323 + NetWare x86 (i[3456x]86-*-netware*) 13324 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 13325 + MCore PE (mcore-*-pe*) 13326 + SH SymbianOS (sh*-*-symbianelf*) 13327 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 13328 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 13329 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 13330 vax-*-netbsd*, but not *-*-netbsdelf*) 13331 The i[34567]86-*-pe alias for Cygwin targets has also been 13332 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 13333 Certain configure options to control the set of libraries built 13334 with GCC on some targets have been obsoleted. On ARM targets, the 13335 options --disable-fpu, --disable-26bit, --disable-underscore, 13336 --disable-interwork, --disable-biendian and --disable-nofmult have 13337 been obsoleted. On MIPS targets, the options 13338 --disable-single-float, --disable-biendian and --disable-softfloat 13339 have been obsoleted. 13340 * Support has been removed for all the [3]configurations obsoleted in 13341 GCC 4.5. 13342 * More information on porting to GCC 4.6 from previous versions of 13343 GCC can be found in the [4]porting guide for this release. 13344 13345 General Optimizer Improvements 13346 13347 * A new general optimization level, -Ofast, has been introduced. It 13348 combines the existing optimization level -O3 with options that can 13349 affect standards compliance but result in better optimized code. 13350 For example, -Ofast enables -ffast-math. 13351 * Link-time optimization improvements: 13352 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 13353 stabilized to the point of being usable. It has become the 13354 default mode when using the LTO optimization model. Link time 13355 optimization can now split itself into multiple parallel 13356 compilations. Parallelism is controlled with -flto=n (where n 13357 specifies the number of compilations to execute in parallel). 13358 GCC can also cooperate with a GNU make job server by 13359 specifying the -flto=jobserver option and adding + to the 13360 beginning of the Makefile rule executing the linker. 13361 Classical LTO mode can be enforced by -flto-partition=none. 13362 This may result in small code quality improvements. 13363 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 13364 and other large applications can be built with LTO enabled. 13365 + The linker plugin support improvements 13366 o Linker plugin is now enabled by default when the linker 13367 is detected to have plugin support. This is the case for 13368 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 13369 the Gold linker on ELF targets. Plugin support of the 13370 Apple linker on Darwin is not compatible with GCC. The 13371 linker plugin can also be controlled by the 13372 -fuse-linker-plugin command-line option. 13373 o Resolution information from the linker plugin is used to 13374 drive whole program assumptions. Use of the linker plugin 13375 results in more aggressive optimization on binaries and 13376 on shared libraries that use the hidden visibility 13377 attribute. Consequently the use of -fwhole-program is not 13378 necessary in addition to LTO. 13379 + Hidden symbols used from non-LTO objects now have to be 13380 explicitly annotated with externally_visible when the linker 13381 plugin is not used. 13382 + C++ inline functions and virtual tables are now privatized 13383 more aggressively, leading to better inter-procedural 13384 optimization and faster dynamic linking. 13385 + Memory usage and intermediate language streaming performance 13386 have been improved. 13387 + Static constructors and destructors from individual units are 13388 inlined into a single function. This can significantly improve 13389 startup times of large C++ applications where static 13390 constructors are very common. For example, static constructors 13391 are used when including the iostream header. 13392 + Support for the Ada language has been added. 13393 * Interprocedural optimization improvements 13394 + The interprocedural framework was re-tuned for link time 13395 optimization. Several scalability issues were resolved. 13396 + Improved auto-detection of const and pure functions. Newly, 13397 noreturn functions are auto-detected. 13398 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 13399 available that informs users when adding attributes to headers 13400 might improve code generation. 13401 + A number of inlining heuristic improvements. In particular: 13402 o Partial inlining is now supported and enabled by default 13403 at -O2 and greater. The feature can be controlled via 13404 -fpartial-inlining. 13405 Partial inlining splits functions with short hot path to 13406 return. This allows more aggressive inlining of the hot 13407 path leading to better performance and often to code size 13408 reductions (because cold parts of functions are not 13409 duplicated). 13410 o Scalability for large compilation units was improved 13411 significantly. 13412 o Inlining of callbacks is now more aggressive. 13413 o Virtual methods are considered for inlining when the 13414 caller is inlined and devirtualization is then possible. 13415 o Inlining when optimizing for size (either in cold regions 13416 of a program or when compiling with -Os) was improved to 13417 better handle C++ programs with larger abstraction 13418 penalty, leading to smaller and faster code. 13419 + The IPA reference optimization pass detecting global variables 13420 used or modified by functions was strengthened and sped up. 13421 + Functions whose address was taken are now optimized out when 13422 all references to them are dead. 13423 + A new inter-procedural static profile estimation pass detects 13424 functions that are executed once or unlikely to be executed. 13425 Unlikely executed functions are optimized for size. Functions 13426 executed once are optimized for size except for the inner 13427 loops. 13428 + On most targets with named section support, functions used 13429 only at startup (static constructors and main), functions used 13430 only at exit and functions detected to be cold are placed into 13431 separate text segment subsections. This extends the 13432 -freorder-functions feature and is controlled by the same 13433 switch. The goal is to improve the startup time of large C++ 13434 programs. 13435 Proper function placement requires linker support. GNU ld 13436 2.21.51 on ELF targets was updated to place those functions 13437 together within the text section leading to better code 13438 locality and faster startup times of large C++ programs. The 13439 feature is also supported in the Apple linker. Support in the 13440 gold linker is planned. 13441 * A new switch -fstack-usage has been added. It makes the compiler 13442 output stack usage information for the program, on a per-function 13443 basis, in an auxiliary file. 13444 * A new switch -fcombine-stack-adjustments has been added. It can be 13445 used to enable or disable the compiler's stack-slot combining pass 13446 which before was enabled automatically at -O1 and above, but could 13447 not be controlled on its own. 13448 * A new switch -fstrict-volatile-bitfields has been added. Using it 13449 indicates that accesses to volatile bitfields should use a single 13450 access of the width of the field's type. This option can be useful 13451 for precisely defining and accessing memory-mapped peripheral 13452 registers from C or C++. 13453 13454 Compile time and memory usage improvements 13455 13456 * Datastructures used by the dataflow framework in GCC were 13457 reorganized for better memory usage and more cache locality. 13458 Compile time is improved especially on units with large functions 13459 (possibly resulting from a lot of inlining) not fitting into the 13460 processor cache. The compile time of the GCC C compiler binary with 13461 link-time optimization went down by over 10% (benchmarked on x86-64 13462 target). 13463 13464 New Languages and Language specific improvements 13465 13466 Ada 13467 13468 * Stack checking has been improved on selected architectures (Alpha, 13469 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 13470 overflows in all cases on these architectures. 13471 * Initial support for Ada 2012 has been added. 13472 13473 C family 13474 13475 * A new warning, enabled by -Wdouble-promotion, has been added that 13476 warns about cases where a value of type float is implicitly 13477 promoted to double. This is especially helpful for CPUs that handle 13478 the former in hardware, but emulate the latter in software. 13479 * A new function attribute leaf was introduced. This attribute allows 13480 better inter-procedural optimization across calls to functions that 13481 return to the current unit only via returning or exception 13482 handling. This is the case for most library functions that have no 13483 callbacks. 13484 * Support for a new data type __int128 for targets having wide enough 13485 machine-mode support. 13486 * The new function attribute callee_pop_aggregate allows to specify 13487 if the caller or callee is responsible for popping the aggregate 13488 return pointer value from the stack. 13489 * Support for selectively enabling and disabling warnings via #pragma 13490 GCC diagnostic has been added. For instance: 13491 #pragma GCC diagnostic error "-Wuninitialized" 13492 foo(a); /* error is given for this one */ 13493 #pragma GCC diagnostic push 13494 #pragma GCC diagnostic ignored "-Wuninitialized" 13495 foo(b); /* no diagnostic for this one */ 13496 #pragma GCC diagnostic pop 13497 foo(c); /* error is given for this one */ 13498 #pragma GCC diagnostic pop 13499 foo(d); /* depends on command-line options */ 13500 13501 * The -fmax-errors=N option is now supported. Using this option 13502 causes the compiler to exit after N errors have been issued. 13503 13504 [7]C 13505 13506 * There is now experimental support for some features from the 13507 upcoming C1X revision of the ISO C standard. This support may be 13508 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 13509 Note that this support is experimental and may change incompatibly 13510 in future releases for consistency with changes to the C1X standard 13511 draft. The following features are newly supported as described in 13512 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 13513 meeting); some other features were already supported with no 13514 compiler changes being needed, or have some support but not in full 13515 accord with N1539 (as amended). 13516 + Static assertions (_Static_assert keyword) 13517 + Typedef redefinition 13518 + New macros in <float.h> 13519 + Anonymous structures and unions 13520 * The new -fplan9-extensions option directs the compiler to support 13521 some extensions for anonymous struct fields which are implemented 13522 by the Plan 9 compiler. A pointer to a struct may be automatically 13523 converted to a pointer to an anonymous field when calling a 13524 function, in order to make the types match. An anonymous struct 13525 field whose type is a typedef name may be referred to using the 13526 typedef name. 13527 13528 [8]C++ 13529 13530 * Improved [9]experimental support for the upcoming C++0x ISO C++ 13531 standard, including support for constexpr (thanks to Gabriel Dos 13532 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 13533 noexcept, unrestricted unions, range-based for loops (thanks to 13534 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 13535 Rodrigo), implicitly deleted functions and implicit move 13536 constructors. 13537 * When an extern declaration within a function does not match a 13538 declaration in the enclosing context, G++ now properly declares the 13539 name within the namespace of the function rather than the namespace 13540 which was open just before the function definition ([10]c++/43145). 13541 * GCC now warns by default when casting integers to larger pointer 13542 types. These warnings can be disabled with the option 13543 -Wno-int-to-pointer-cast, which is now also available in C++. 13544 * G++ no longer optimizes using the assumption that a value of 13545 enumeration type will fall within the range specified by the 13546 standard, since that assumption is easily violated with a 13547 conversion from integer type ([11]c++/43680). The old behavior can 13548 be restored with -fstrict-enums. 13549 * The new -fnothrow-opt flag changes the semantics of a throw() 13550 exception specification to match the proposed semantics of the 13551 noexcept specification: just call terminate if an exception tries 13552 to propagate out of a function with such an exception 13553 specification. This dramatically reduces or eliminates the code 13554 size overhead from adding the exception specification. 13555 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 13556 a function that the compiler can tell doesn't throw if it would 13557 change the value of a noexcept expression. 13558 * The -Wshadow option now warns if a local variable or type 13559 declaration shadows another type in C++. Note that the compiler 13560 will not warn if a local variable shadows a struct/class/enum, but 13561 will warn if it shadows an explicit typedef. 13562 * When an identifier is not found in the current scope, G++ now 13563 offers suggestions about which identifier might have been intended. 13564 * G++ now issues clearer diagnostics for missing semicolons after 13565 class, struct, and union definitions. 13566 * G++ now issues clearer diagnostics for missing semicolons after 13567 class member declarations. 13568 * G++ now issues clearer diagnostics when a colon is used in a place 13569 where a double-colon was intended. 13570 * G++ no longer accepts mutable on reference members ([12]c++/33558). 13571 Use -fpermissive to allow the old, non-conforming behavior. 13572 * A few mangling fixes have been made, to attribute const/volatile on 13573 function pointer types, decltype of a plain decl, and use of a 13574 function parameter in the declaration of another parameter. By 13575 default the compiler still uses the old mangling, but emits aliases 13576 with the new mangling on targets that support strong aliases. Users 13577 can switch over entirely to the new mangling with -fabi-version=5 13578 or -fabi-version=0. -Wabi will now warn about code that uses the 13579 old mangling. 13580 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 13581 type to be default initialized unless the type has a user-declared 13582 default constructor. In 4.6.2 G++ implements the proposed 13583 resolution of [13]DR 253, so default initialization is allowed if 13584 it initializes all subobjects. Code that fails to compile can be 13585 fixed by providing an initializer e.g. 13586 struct A { A(); }; 13587 struct B : A { int i; }; 13588 const B b = B(); 13589 Use -fpermissive to allow the old, non-conforming behavior. 13590 13591 Runtime Library (libstdc++) 13592 13593 * [14]Improved experimental support for the upcoming ISO C++ 13594 standard, C++0x, including using constexpr and nullptr. 13595 * Performance improvements to the [15]Debug Mode, thanks to Franois 13596 Dumont. 13597 * Atomic operations used for reference-counting are annotated so that 13598 they can be understood by race detectors such as Helgrind, see 13599 [16]Data Race Hunting. 13600 * Most libstdc++ standard headers have been changed to no longer 13601 include the cstddef header as an implementation detail. Code that 13602 relied on that header being included as side-effect of including 13603 other standard headers will need to include cstddef explicitly. 13604 13605 [17]Fortran 13606 13607 * On systems supporting the libquadmath library, GNU Fortran now also 13608 supports a quad-precision, kind=16 floating-point data type 13609 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 13610 hardware, calculations might be one to two orders of magnitude 13611 slower than with the 4, 8 or 10 bytes floating-point data types. 13612 This change does not affect systems which support REAL(16) in 13613 hardware nor those which do not support libquadmath. 13614 * Much improved compile time for large array constructors. 13615 * In order to reduce execution time and memory consumption, use of 13616 temporary arrays in assignment expressions is avoided for many 13617 cases. The compiler now reverses loops in order to avoid generating 13618 a temporary array where possible. 13619 * Improved diagnostics, especially with -fwhole-file. 13620 * The -fwhole-file flag is now enabled by default. This improves code 13621 generation and diagnostics. It can be disabled using the deprecated 13622 -fno-whole-file flag. 13623 * Support the generation of Makefile dependencies via the [18]-M... 13624 flags of GCC; you may need to specify the -cpp option in addition. 13625 The dependencies take modules, Fortran's include, and CPP's 13626 #include into account. Note: Using -M for the module path is no 13627 longer supported, use -J instead. 13628 * The flag -Wconversion has been modified to only issue warnings 13629 where a conversion leads to information loss. This drastically 13630 reduces the number of warnings; -Wconversion is thus now enabled 13631 with -Wall. The flag -Wconversion-extra has been added and also 13632 warns about other conversions; -Wconversion-extra typically issues 13633 a huge number of warnings, most of which can be ignored. 13634 * A new command-line option -Wunused-dummy-argument warns about 13635 unused dummy arguments and is included in -Wall. Before, 13636 -Wunused-variable also warned about unused dummy arguments. 13637 * Fortran 2003 support has been extended: 13638 + Improved support for polymorphism between libraries and 13639 programs and for complicated inheritance patterns (cf. 13640 [19]object-oriented programming). 13641 + Experimental support of the ASSOCIATE construct. 13642 + In pointer assignments it is now possible to specify the lower 13643 bounds of the pointer and, for a rank-1 or a simply contiguous 13644 data-target, to remap the bounds. 13645 + Automatic (re)allocation: In intrinsic assignments to 13646 allocatable variables the left-hand side will be automatically 13647 allocated (if unallocated) or reallocated (if the shape or 13648 type parameter is different). To avoid the small performance 13649 penalty, you can use a(:) = ... instead of a = ... for arrays 13650 and character strings or disable the feature using -std=f95 13651 or -fno-realloc-lhs. 13652 + Deferred type parameter: For scalar allocatable and pointer 13653 variables the character length can be deferred. 13654 + Namelist variables with allocatable and pointer attribute and 13655 nonconstant length type parameter are supported. 13656 * Fortran 2008 support has been extended: 13657 + Experimental [20]coarray support (for one image only, i.e. 13658 num_images() == 1); use the [21]-fcoarray=single flag to 13659 enable it. 13660 + The STOP and the new ERROR STOP statements now support all 13661 constant expressions. 13662 + Support for the CONTIGUOUS attribute. 13663 + Support for ALLOCATE with MOLD. 13664 + Support for the STORAGE_SIZE intrinsic inquiry function. 13665 + Support of the NORM2 and PARITY intrinsic functions. 13666 + The following bit intrinsics were added: POPCNT and POPPAR for 13667 counting the number of 1 bits and returning the parity; BGE, 13668 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 13669 for combined left and right shifts, MASKL and MASKR for simple 13670 left and right justified masks, MERGE_BITS for a bitwise merge 13671 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 13672 and the transformational bit intrinsics IALL, IANY and 13673 IPARITY. 13674 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 13675 + Support for the IMPURE attribute for procedures, which allows 13676 for ELEMENTAL procedures without the restrictions of PURE. 13677 + Null pointers (including NULL()) and not allocated variables 13678 can be used as actual argument to optional non-pointer, 13679 non-allocatable dummy arguments, denoting an absent argument. 13680 + Non-pointer variables with TARGET attribute can be used as 13681 actual argument to POINTER dummies with INTENT(IN) 13682 + Pointers including procedure pointers and those in a derived 13683 type (pointer components) can now be initialized by a target 13684 instead of only by NULL. 13685 + The EXIT statement (with construct-name) can now be used to 13686 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 13687 SELECT CASE and SELECT TYPE constructs. 13688 + Internal procedures can now be used as actual argument. 13689 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 13690 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 13691 have been added; these arrays contain the supported kind 13692 values for the respective types. 13693 + The module procedures C_SIZEOF of the intrinsic module 13694 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 13695 ISO_FORTRAN_ENV have been implemented. 13696 + Minor changes: obsolescence diagnostics for ENTRY was added 13697 for -std=f2008; a line may start with a semicolon; for 13698 internal and module procedures END can be used instead of END 13699 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 13700 a RADIX argument; intrinsic types are supported for 13701 TYPE(intrinsic-type-spec); multiple type-bound procedures can 13702 be declared in a single PROCEDURE statement; implied-shape 13703 arrays are supported for named constants (PARAMETER). The 13704 transformational, three argument versions of BESSEL_JN and 13705 BESSEL_YN were added the elemental, two-argument version had 13706 been added in GCC 4.4; note that the transformational 13707 functions use a recurrence algorithm. 13708 13709 [22]Go 13710 13711 Support for the Go programming language has been added to GCC. It is 13712 not enabled by default when you build GCC; use the --enable-languages 13713 configure option to build it. The driver program for compiling Go code 13714 is gccgo. 13715 13716 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 13717 is in progress. It may or may not work on other platforms. 13718 13719 [23]Objective-C and Objective-C++ 13720 13721 * The -fobjc-exceptions flag is now required to enable Objective-C 13722 exception and synchronization syntax (introduced by the keywords 13723 @try, @catch, @finally and @synchronized). 13724 * A number of Objective-C 2.0 features and extensions are now 13725 supported by GCC. These features are enabled by default; you can 13726 disable them by using the new -fobjc-std=objc1 command-line option. 13727 * The Objective-C 2.0 dot-syntax is now supported. It is an 13728 alternative syntax for using getters and setters; object.count is 13729 automatically converted into [object count] or [object setCount: 13730 ...] depending on context; for example if (object.count > 0) is 13731 automatically compiled into the equivalent of if ([object count] > 13732 0) while object.count = 0; is automatically compiled into the 13733 equivalent ot [object setCount: 0];. The dot-syntax can be used 13734 with instance and class objects and with any setters or getters, no 13735 matter if they are part of a declared property or not. 13736 * Objective-C 2.0 declared properties are now supported. They are 13737 declared using the new @property keyword, and are most commonly 13738 used in conjunction with the new Objective-C 2.0 dot-syntax. The 13739 nonatomic, readonly, readwrite, assign, retain, copy, setter and 13740 getter attributes are all supported. Marking declared properties 13741 with __attribute__ ((deprecated)) is supported too. 13742 * The Objective-C 2.0 @synthesize and @dynamic keywords are 13743 supported. @synthesize causes the compiler to automatically 13744 synthesize a declared property, while @dynamic is used to disable 13745 all warnings for a declared property for which no implementation is 13746 provided at compile time. Synthesizing declared properties requires 13747 runtime support in most useful cases; to be able to use it with the 13748 GNU runtime, appropriate helper functions have been added to the 13749 GNU Objective-C runtime ABI, and are implemented by the GNU 13750 Objective-C runtime library shipped with GCC. 13751 * The Objective-C 2.0 fast enumeration syntax is supported in 13752 Objective-C. This is currently not yet available in Objective-C++. 13753 Fast enumeration requires support in the runtime, and such support 13754 has been added to the GNU Objective-C runtime library (shipped with 13755 GCC). 13756 * The Objective-C 2.0 @optional keyword is supported. It allows you 13757 to mark methods or properties in a protocol as optional as opposed 13758 to required. 13759 * The Objective-C 2.0 @package keyword is supported. It has currently 13760 the same effect as the @public keyword. 13761 * Objective-C 2.0 method attributes are supported. Currently the 13762 supported attributes are deprecated, sentinel, noreturn and format. 13763 * Objective-C 2.0 method argument attributes are supported. The most 13764 widely used attribute is unused, to mark an argument as unused in 13765 the implementation. 13766 * Objective-C 2.0 class and protocol attributes are supported. 13767 Currently the only supported attribute is deprecated. 13768 * Objective-C 2.0 class extensions are supported. A class extension 13769 has the same syntax as a category declaration with no category 13770 name, and the methods and properties declared in it are added 13771 directly to the main class. It is mostly used as an alternative to 13772 a category to add methods to a class without advertising them in 13773 the public headers, with the advantage that for class extensions 13774 the compiler checks that all the privately declared methods are 13775 actually implemented. 13776 * As a result of these enhancements, GCC can now be used to build 13777 Objective-C and Objective-C++ software that uses Foundation and 13778 other important system frameworks with the NeXT runtime on Darwin 9 13779 and Darwin 10 (OSX 10.5 and 10.6). 13780 * Many bugs in the compiler have been fixed in this release; in 13781 particular, LTO can now be used when compiling Objective-C and 13782 Objective-C++ and the parser is much more robust in dealing with 13783 invalid code. 13784 13785 Runtime Library (libobjc) 13786 13787 * The GNU Objective-C runtime library now defines the macro 13788 __GNU_LIBOBJC__ (with a value that is increased at every release 13789 where there is any change to the API) in objc/objc.h, making it 13790 easy to determine if the GNU Objective-C runtime library is being 13791 used, and if so, which version. Previous versions of the GNU 13792 Objective-C runtime library (and other Objective-C runtime 13793 libraries such as the Apple one) do not define this macro. 13794 * A new Objective-C 2.0 API, almost identical to the one implemented 13795 by the Apple Objective-C runtime, has been implemented in the GNU 13796 Objective-C runtime library. The new API hides the internals of 13797 most runtime structures but provides a more extensive set of 13798 functions to operate on them. It is much easier, for example, to 13799 create or modify classes at runtime. The new API also makes it 13800 easier to port software from Apple to GNU as almost no changes 13801 should be required. The old API is still supported for backwards 13802 compatibility; including the old objc/objc-api.h header file 13803 automatically selects the old API, while including the new 13804 objc/runtime.h header file automatically selects the new API. 13805 Support for the old API is being phased out and upgrading the 13806 software to use the new API is strongly recommended. To check for 13807 the availability of the new API, the __GNU_LIBOBJC__ macro can be 13808 used as older versions of the GNU Objective-C runtime library, 13809 which do not support the new API, do not define such a macro. 13810 * Runtime support for @synchronized has been added. 13811 * Runtime support for Objective-C 2.0 synthesized property accessors 13812 has been added. 13813 * Runtime support for Objective-C 2.0 fast enumeration has been 13814 added. 13815 13816 [24]New Targets and Target Specific Improvements 13817 13818 [25]ARM 13819 13820 * GCC now supports the Cortex-M4 processor implementing the v7-em 13821 version of the architecture using the option -mcpu=cortex-m4. 13822 * Scheduling descriptions for the Cortex-M4, the Neon and the 13823 floating point units of the Cortex-A9 and a pipeline description 13824 for the Cortex-A5 have been added. 13825 * Synchronization primitives such as __sync_fetch_and_add and friends 13826 are now inlined for supported architectures rather than calling 13827 into a kernel helper function. 13828 * SSA loop prefetching is enabled by default for the Cortex-A9 at 13829 -O3. 13830 * Several improvements were committed to improve code generation for 13831 the ARM architecture including a rewritten implementation for load 13832 and store multiples. 13833 * Several enhancements were committed to improve SIMD code generation 13834 for NEON by adding support for widening instructions, misaligned 13835 loads and stores, vector conditionals and support for 64-bit 13836 arithmetic. 13837 * Support was added for the Faraday cores fa526, fa606te, fa626te, 13838 fmp626te, fmp626 and fa726te and can be used with the respective 13839 names as parameters to the -mcpu= option. 13840 * Basic support was added for Cortex-A15 and is available through 13841 -mcpu=cortex-a15. 13842 * GCC for AAPCS configurations now more closely adheres to the AAPCS 13843 specification by enabling -fstrict-volatile-bitfields by default. 13844 13845 IA-32/x86-64 13846 13847 * The new -fsplit-stack option permits programs to use a 13848 discontiguous stack. This is useful for threaded programs, in that 13849 it is no longer necessary to specify the maximum stack size when 13850 creating a thread. This feature is currently only implemented for 13851 32-bit and 64-bit x86 GNU/Linux targets. 13852 * Support for emitting profiler counter calls before function 13853 prologues. This is enabled via a new command-line option -mfentry. 13854 * Optimization for the Intel Core 2 processors is now available 13855 through the -march=core2 and -mtune=core2 options. 13856 * Support for Intel Core i3/i5/i7 processors is now available through 13857 the -march=corei7 and -mtune=corei7 options. 13858 * Support for Intel Core i3/i5/i7 processors with AVX is now 13859 available through the -march=corei7-avx and -mtune=corei7-avx 13860 options. 13861 * Support for AMD Bobcat (family 14) processors is now available 13862 through the -march=btver1 and -mtune=btver1 options. 13863 * Support for AMD Bulldozer (family 15) processors is now available 13864 through the -march=bdver1 and -mtune=bdver1 options. 13865 * The default setting (when not optimizing for size) for 32-bit 13866 GNU/Linux and Darwin x86 targets has been changed to 13867 -fomit-frame-pointer. The default can be reverted to 13868 -fno-omit-frame-pointer by configuring GCC with the 13869 --enable-frame-pointer configure option. 13870 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 13871 __float128 on 32-bit and 64-bit x86 targets. 13872 * AVX floating-point arithmetic can now be enabled by default at 13873 configure time with the new --with-fpmath=avx option. 13874 * The SSA loop prefetching pass is enabled when using -O3 when 13875 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 13876 than K6). 13877 * Support for TBM (Trailing Bit Manipulation) built-in functions and 13878 code generation is available via -mtbm. 13879 * Support for AMD's BMI (Bit Manipulation) built-in functions and 13880 code generation is available via -mbmi. 13881 13882 [26]MicroBlaze 13883 13884 * Support has been added for the Xilinx MicroBlaze softcore processor 13885 (microblaze-elf) embedded target. This configurable processor is 13886 supported on several Xilinx Spartan and Virtex FPGAs. 13887 13888 MIPS 13889 13890 * GCC now supports the Loongson 3A processor. Its canonical -march= 13891 and -mtune= name is loongson3a. 13892 13893 [27]MN10300 / AM33 13894 13895 * The inline assembly register constraint "A" has been renamed "c". 13896 This constraint is used to select a floating-point register that 13897 can be used as the destination of a multiply-accumulate 13898 instruction. 13899 * New inline assembly register constraints "A" and "D" have been 13900 added. These constraint letters resolve to all general registers 13901 when compiling for AM33, and resolve to address registers only or 13902 data registers only when compiling for MN10300. 13903 * The MDR register is represented in the compiler. One can access the 13904 register via the "z" constraint in inline assembly. It can be 13905 marked as clobbered or used as a local register variable via the 13906 "mdr" name. The compiler uses the RETF instruction if the function 13907 does not modify the MDR register, so it is important that inline 13908 assembly properly annotate any usage of the register. 13909 13910 PowerPC/PowerPC64 13911 13912 * GCC now supports the Applied Micro Titan processor with 13913 -mcpu=titan. 13914 * The -mrecip option has been added, which indicates whether the 13915 reciprocal and reciprocal square root instructions should be used. 13916 * The -mveclibabi=mass option can be used to enable the compiler to 13917 autovectorize mathematical functions using the Mathematical 13918 Acceleration Subsystem library. 13919 * The -msingle-pic-base option has been added, which instructs the 13920 compiler to avoid loading the PIC base register in function 13921 prologues. The PIC base register must be initialized by the runtime 13922 system. 13923 * The -mblock-move-inline-limit option has been added, which enables 13924 the user to control the maximum size of inlined memcpy calls and 13925 similar. 13926 * PowerPC64 GNU/Linux support for applications requiring a large TOC 13927 section has been improved. A new command-line option, 13928 -mcmodel=MODEL, controls this feature; valid values for MODEL are 13929 small, medium, or large. 13930 * The Altivec builtin functions vec_ld and vec_st have been modified 13931 to generate the Altivec memory instructions LVX and STVX, even if 13932 the -mvsx option is used. In the initial GCC 4.5 release, these 13933 builtin functions were changed to generate VSX memory reference 13934 instructions instead of Altivec memory instructions, but there are 13935 differences between the two instructions. If the VSX instruction 13936 set is available, you can now use the new builtin functions 13937 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 13938 instructions. 13939 * The GCC compiler on AIX now defaults to a process layout with a 13940 larger data space allowing larger programs to be compiled. 13941 * The GCC long double type on AIX 6.1 and above has reverted to 13942 64-bit double precision, matching the AIX XL compiler default, 13943 because of missing C99 symbols required by the GCC runtime. 13944 * The default processor scheduling model and tuning for PowerPC64 13945 GNU/Linux and for AIX 6.1 and above now is POWER7. 13946 * Starting with GCC 4.6.1, vectors of type vector long long or vector 13947 long are passed and returned in the same method as other vectors 13948 with the VSX instruction set. Previously the GCC compiler did not 13949 adhere to the ABI for 128-bit vectors with 64-bit integer base 13950 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 13951 13952 S/390, zSeries and System z9/z10, IBM zEnterprise z196 13953 13954 * Support for the zEnterprise z196 processor has been added. When 13955 using the -march=z196 option, the compiler will generate code 13956 making use of the following instruction facilities: 13957 + Conditional load/store 13958 + Distinct-operands 13959 + Floating-point-extension 13960 + Interlocked-access 13961 + Population-count 13962 The -mtune=z196 option avoids the compare and branch instructions 13963 as well as the load address instruction with an index register as 13964 much as possible and performs instruction scheduling appropriate 13965 for the new out-of-order pipeline architecture. 13966 * When using the -m31 -mzarch options the generated code still 13967 conforms to the 32-bit ABI but uses the general purpose registers 13968 as 64-bit registers internally. This requires a Linux kernel saving 13969 the whole 64-bit registers when doing a context switch. Kernels 13970 providing that feature indicate that by the 'highgprs' string in 13971 /proc/cpuinfo. 13972 * The SSA loop prefetching pass is enabled when using -O3. 13973 13974 SPARC 13975 13976 * GCC now supports the LEON series of SPARC V8 processors. The code 13977 generated by the compiler can either be tuned to it by means of the 13978 --with-tune=leon configure option and -mtune=leon compilation 13979 option, or the compiler can be built for the sparc-leon-{elf,linux} 13980 and sparc-leon3-{elf,linux} targets directly. 13981 * GCC has stopped sign/zero-extending parameter registers in the 13982 callee for functions taking parameters with sub-word size in 32-bit 13983 mode, since this is redundant with the specification of the ABI. 13984 GCC has never done so in 64-bit mode since this is also redundant. 13985 * The command-line option -mfix-at697f has been added to enable the 13986 documented workaround for the single erratum of the Atmel AT697F 13987 processor. 13988 13989 [28]Operating Systems 13990 13991 [29]Android 13992 13993 * GCC now supports the Bionic C library and provides a convenient way 13994 of building native libraries and applications for the Android 13995 platform. Refer to the documentation of the -mandroid and -mbionic 13996 options for details on building native code. At the moment, Android 13997 support is enabled only for ARM. 13998 13999 [30]Darwin/Mac OS X 14000 14001 * General 14002 + Initial support for CFString types has been added. 14003 This allows GCC to build projects including the system Core 14004 Foundation frameworks. The GCC Objective-C family supports 14005 CFString "toll-free bridged" as per the Mac OS X system tools. 14006 CFString is also recognized in the context of format 14007 attributes and arguments (see the documentation for format 14008 attributes for limitations). At present, 8-bit character types 14009 are supported. 14010 + Object file size reduction. 14011 The Darwin zeroed memory allocators have been re-written to 14012 make more use of .zerofill sections. For non-debug code, this 14013 can reduce object file size significantly. 14014 + Objective-C family 64-bit support (NeXT ABI 2). 14015 Initial support has been added to support 64-bit Objective-C 14016 code using the Darwin/OS X native (NeXT) runtime. ABI version 14017 2 will be selected automatically when 64-bit code is built. 14018 + Objective-C family 32-bit ABI 1. 14019 For 32-bit code ABI 1 is also now also allowed. At present it 14020 must be selected manually using -fobjc-abi-version=1 where 14021 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 14022 * x86 Architecture 14023 + The -mdynamic-no-pic option has been enabled. 14024 Code supporting -mdynamic-no-pic optimization has been added 14025 and is applicable to -m32 builds. The compiler bootstrap uses 14026 the option where appropriate. 14027 + The default value for -mtune= has been changed. 14028 Since Darwin systems are primarily Xeon, Core-2 or similar the 14029 default tuning has been changed to -mtune=core2. 14030 + Enable 128-bit long double (__float128) support on Darwin. 14031 * PPC Architecture 14032 + Darwin64 ABI. 14033 Several significant bugs have been fixed, such that GCC now 14034 produces code compatible with the Darwin64 PowerPC ABI. 14035 + libffi and boehm-gc. 14036 The Darwin ports of the libffi and boehm-gc libraries have 14037 been upgraded to include a Darwin64 implementation. This means 14038 that powerpc*-*-darwin9 platforms may now, for example, build 14039 Java applications with -m64 enabled. 14040 + Plug-in support has been enabled. 14041 + The -fsection-anchors option is now available although, 14042 presently, not heavily tested. 14043 14044 [31]Solaris 2 14045 14046 New Features 14047 14048 * Support symbol versioning with the Sun linker. 14049 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 14050 * Support thread-local storage (TLS) with the Sun assembler on 14051 Solaris 2/x86. 14052 * Support TLS on Solaris 8/9 if prerequisites are met. 14053 * Support COMDAT group with the GNU assembler and recent Sun linker. 14054 * Support the Sun assembler visibility syntax. 14055 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 14056 -march=pentiumpro (Solaris 8/9). 14057 * Don't use SSE on Solaris 8/9 x86 by default. 14058 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 14059 14060 ABI Change 14061 14062 * Change the ABI for returning 8-byte vectors like __m64 in MMX 14063 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 14064 compilers. This is an incompatible change. If you use such types, 14065 you must either recompile all your code with the new compiler or 14066 use the new -mvect8-ret-in-mem option to remain compatible with 14067 previous versions of GCC and Sun Studio. 14068 14069 [32]Windows x86/x86_64 14070 14071 * Initial support for decimal floating point. 14072 * Support for the __thiscall calling-convention. 14073 * Support for hot-patchable function prologues via the 14074 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 14075 * Improvements of stack-probing and stack-allocation mechanisms. 14076 * Support of push/pop-macro pragma as preprocessor command. 14077 With #pragma push_macro("macro-name") the current definition of 14078 macro-name is saved and can be restored with #pragma 14079 pop_macro("macro-name") to its saved definition. 14080 * Enable 128-bit long double (__float128) support on MinGW and 14081 Cygwin. 14082 14083 Other significant improvements 14084 14085 Installation changes 14086 14087 * An install-strip make target is provided that installs stripped 14088 executables, and may install libraries with unneeded or debugging 14089 sections stripped. 14090 * On Power7 systems, there is a potential problem if you build the 14091 GCC compiler with a host compiler using options that enable the VSX 14092 instruction set generation. If the host compiler has been patched 14093 so that the vec_ld and vec_st builtin functions generate Altivec 14094 memory instructions instead of VSX memory instructions, then you 14095 should be able to build the compiler with VSX instruction 14096 generation. 14097 14098 Changes for GCC Developers 14099 14100 Note: these changes concern developers that develop GCC itself or 14101 software that integrates with GCC, such as plugins, and not the general 14102 GCC users. 14103 * The gengtype utility, which previously was internal to the GCC 14104 build process, has been enchanced to provide GC root information 14105 for plugins as necessary. 14106 * The old GC allocation interface of ggc_alloc and friends was 14107 replaced with a type-safe alternative. 14108 14109 [33]GCC 4.6.1 14110 14111 This is the [34]list of problem reports (PRs) from GCC's bug tracking 14112 system that are known to be fixed in the 4.6.1 release. This list might 14113 not be complete (that is, it is possible that some PRs that have been 14114 fixed are not listed here). 14115 14116 [35]GCC 4.6.2 14117 14118 This is the [36]list of problem reports (PRs) from GCC's bug tracking 14119 system that are known to be fixed in the 4.6.2 release. This list might 14120 not be complete (that is, it is possible that some PRs that have been 14121 fixed are not listed here). 14122 14123 [37]GCC 4.6.3 14124 14125 This is the [38]list of problem reports (PRs) from GCC's bug tracking 14126 system that are known to be fixed in the 4.6.3 release. This list might 14127 not be complete (that is, it is possible that some PRs that have been 14128 fixed are not listed here). 14129 14130 [39]GCC 4.6.4 14131 14132 This is the [40]list of problem reports (PRs) from GCC's bug tracking 14133 system that are known to be fixed in the 4.6.4 release. This list might 14134 not be complete (that is, it is possible that some PRs that have been 14135 fixed are not listed here). 14136 14137 14138 For questions related to the use of GCC, please consult these web 14139 pages and the [41]GCC manuals. If that fails, the 14140 [42]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14141 web pages and the development of GCC are welcome on our developer 14142 list at [43]gcc (a] gcc.gnu.org. All of [44]our lists have public 14143 archives. 14144 14145 Copyright (C) [45]Free Software Foundation, Inc. Verbatim copying and 14146 distribution of this entire article is permitted in any medium, 14147 provided this notice is preserved. 14148 14149 These pages are [46]maintained by the GCC team. Last modified 14150 2026-06-07. 14151 14152 References 14153 14154 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 14155 2. https://gcc.gnu.org/gcc-4.5/changes.html 14156 3. https://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 14157 4. https://gcc.gnu.org/gcc-4.6/porting_to.html 14158 5. https://gcc.gnu.org/projects/lto/whopr.pdf 14159 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 14160 7. https://gcc.gnu.org/gcc-4.6/changes.html#c 14161 8. https://gcc.gnu.org/gcc-4.6/changes.html#cplusplus 14162 9. https://gcc.gnu.org/gcc-4.6/cxx0x_status.html 14163 10. https://gcc.gnu.org/PR43145 14164 11. https://gcc.gnu.org/PR43680 14165 12. https://gcc.gnu.org/PR33558 14166 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 14167 14. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 14168 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 14169 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 14170 17. https://gcc.gnu.org/gcc-4.6/changes.html#fortran 14171 18. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 14172 19. https://gcc.gnu.org/wiki/OOP 14173 20. https://gcc.gnu.org/wiki/Coarray 14174 21. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 14175 22. https://gcc.gnu.org/gcc-4.6/changes.html#go 14176 23. https://gcc.gnu.org/gcc-4.6/changes.html#objective-c 14177 24. https://gcc.gnu.org/gcc-4.6/changes.html#targets 14178 25. https://gcc.gnu.org/gcc-4.6/changes.html#arm 14179 26. https://gcc.gnu.org/gcc-4.6/changes.html#microblaze 14180 27. https://gcc.gnu.org/gcc-4.6/changes.html#mn10300 14181 28. https://gcc.gnu.org/gcc-4.6/changes.html#os 14182 29. https://gcc.gnu.org/gcc-4.6/changes.html#android 14183 30. https://gcc.gnu.org/gcc-4.6/changes.html#darwin 14184 31. https://gcc.gnu.org/gcc-4.6/changes.html#solaris 14185 32. https://gcc.gnu.org/gcc-4.6/changes.html#windows 14186 33. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.1 14187 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 14188 35. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.2 14189 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 14190 37. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.3 14191 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 14192 39. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.4 14193 40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 14194 41. https://gcc.gnu.org/onlinedocs/ 14195 42. mailto:gcc-help (a] gcc.gnu.org 14196 43. mailto:gcc (a] gcc.gnu.org 14197 44. https://gcc.gnu.org/lists.html 14198 45. https://www.fsf.org/ 14199 46. https://gcc.gnu.org/about.html 14200 ====================================================================== 14201 http://gcc.gnu.org/gcc-4.5/index.html 14202 GCC 4.5 Release Series 14203 14204 (This release series is no longer supported.) 14205 14206 Jul 2, 2012 14207 14208 The [1]GNU project and the GCC developers are pleased to announce the 14209 release of GCC 4.5.4. 14210 14211 This release is a bug-fix release, containing fixes for regressions in 14212 GCC 4.5.3 relative to previous releases of GCC. 14213 14214 Release History 14215 14216 GCC 4.5.4 14217 Jul 2, 2012 ([2]changes) 14218 14219 GCC 4.5.3 14220 Apr 28, 2011 ([3]changes) 14221 14222 GCC 4.5.2 14223 Dec 16, 2010 ([4]changes) 14224 14225 GCC 4.5.1 14226 Jul 31, 2010 ([5]changes) 14227 14228 GCC 4.5.0 14229 April 14, 2010 ([6]changes) 14230 14231 References and Acknowledgements 14232 14233 GCC used to stand for the GNU C Compiler, but since the compiler 14234 supports several other languages aside from C, it now stands for the 14235 GNU Compiler Collection. 14236 14237 A list of [7]successful builds is updated as new information becomes 14238 available. 14239 14240 The GCC developers would like to thank the numerous people that have 14241 contributed new features, improvements, bug fixes, and other changes as 14242 well as test results to GCC. This [8]amazing group of volunteers is 14243 what makes GCC successful. 14244 14245 For additional information about GCC please refer to the [9]GCC project 14246 web site or contact the [10]GCC development mailing list. 14247 14248 To obtain GCC please use [11]our mirror sites or [12]our version 14249 control system. 14250 14251 14252 For questions related to the use of GCC, please consult these web 14253 pages and the [13]GCC manuals. If that fails, the 14254 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14255 web pages and the development of GCC are welcome on our developer 14256 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 14257 archives. 14258 14259 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14260 distribution of this entire article is permitted in any medium, 14261 provided this notice is preserved. 14262 14263 These pages are [18]maintained by the GCC team. Last modified 14264 2022-10-26. 14265 14266 References 14267 14268 1. http://www.gnu.org/ 14269 2. https://gcc.gnu.org/gcc-4.5/changes.html 14270 3. https://gcc.gnu.org/gcc-4.5/changes.html 14271 4. https://gcc.gnu.org/gcc-4.5/changes.html 14272 5. https://gcc.gnu.org/gcc-4.5/changes.html 14273 6. https://gcc.gnu.org/gcc-4.5/changes.html 14274 7. https://gcc.gnu.org/gcc-4.5/buildstat.html 14275 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14276 9. https://gcc.gnu.org/index.html 14277 10. mailto:gcc (a] gcc.gnu.org 14278 11. https://gcc.gnu.org/mirrors.html 14279 12. https://gcc.gnu.org/git.html 14280 13. https://gcc.gnu.org/onlinedocs/ 14281 14. mailto:gcc-help (a] gcc.gnu.org 14282 15. mailto:gcc (a] gcc.gnu.org 14283 16. https://gcc.gnu.org/lists.html 14284 17. https://www.fsf.org/ 14285 18. https://gcc.gnu.org/about.html 14286 ====================================================================== 14287 http://gcc.gnu.org/gcc-4.5/changes.html 14288 GCC 4.5 Release Series 14289 Changes, New Features, and Fixes 14290 14291 Caveats 14292 14293 * GCC now requires the [1]MPC library in order to build. See the 14294 [2]prerequisites page for version requirements. 14295 * Support for a number of older systems and recently unmaintained or 14296 untested target ports of GCC has been declared obsolete in GCC 4.5. 14297 Unless there is activity to revive them, the next release of GCC 14298 will have their sources permanently removed. 14299 The following ports for individual systems on particular 14300 architectures have been obsoleted: 14301 + IRIX releases before 6.5 (mips-sgi-irix5*, 14302 mips-sgi-irix6.[0-4]) 14303 + Solaris 7 (*-*-solaris2.7) 14304 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 14305 alpha-dec-osf5.0*) 14306 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 14307 can be found in the [3]announcement. 14308 Support for the classic POWER architecture implemented in the 14309 original RIOS and RIOS2 processors of the old IBM RS/6000 product 14310 line has been obsoleted in the rs6000 port. This does not affect 14311 the new generation Power and PowerPC architectures. 14312 * Support has been removed for all the [4]configurations obsoleted in 14313 GCC 4.4. 14314 * Support has been removed for the protoize and unprotoize utilities, 14315 obsoleted in GCC 4.4. 14316 * Support has been removed for tuning for Itanium1 (Merced) variants. 14317 Note that code tuned for Itanium2 should also run correctly on 14318 Itanium1. 14319 * GCC now generates unwind info also for epilogues. DWARF debuginfo 14320 generated by GCC now uses more features of DWARF3 than before, and 14321 also some DWARF4 features. GDB older than 7.0 is not able to handle 14322 either of these, so to debug GCC 4.5 generated binaries or 14323 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 14324 features with the -gdwarf-3 -gstrict-dwarf options, or use 14325 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 14326 epilogue unwind info is emitted unconditionally whenever unwind 14327 info is emitted. 14328 * On x86 targets, code containing floating-point calculations may run 14329 significantly more slowly when compiled with GCC 4.5 in strict C99 14330 conformance mode than they did with earlier GCC versions. This is 14331 due to stricter standard conformance of the compiler and can be 14332 avoided by using the option -fexcess-precision=fast; also see 14333 [5]below. 14334 * The function attribute noinline no longer prevents GCC from cloning 14335 the function. A new attribute noclone has been introduced for this 14336 purpose. Cloning a function means that it is duplicated and the new 14337 copy is specialized for certain contexts (for example when a 14338 parameter is a known constant). 14339 14340 General Optimizer Improvements 14341 14342 * The -save-temps now takes an optional argument. The -save-temps and 14343 -save-temps=cwd switches write the temporary files in the current 14344 working directory based on the original source file. The 14345 -save-temps=obj switch will write files into the directory 14346 specified with the -o option, and the intermediate filenames are 14347 based on the output file. This will allow the user to get the 14348 compiler intermediate files when doing parallel builds without two 14349 builds of the same filename located in different directories from 14350 interfering with each other. 14351 * Debugging dumps are now created in the same directory as the object 14352 file rather than in the current working directory. This allows the 14353 user to get debugging dumps when doing parallel builds without two 14354 builds of the same filename interfering with each other. 14355 * GCC has been integrated with the MPC library. This allows GCC to 14356 evaluate complex arithmetic at compile time [6]more accurately. It 14357 also allows GCC to evaluate calls to complex built-in math 14358 functions having constant arguments and replace them at compile 14359 time with their mathematically equivalent results. In doing so, GCC 14360 can generate correct results regardless of the math library 14361 implementation or floating point precision of the host platform. 14362 This also allows GCC to generate identical results regardless of 14363 whether one compiles in native or cross-compile configurations to a 14364 particular target. The following built-in functions take advantage 14365 of this new capability: cacos, cacosh, casin, casinh, catan, 14366 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 14367 and ctanh. The float and long double variants of these functions 14368 (e.g. csinf and csinl) are also handled. 14369 * A new link-time optimizer has been added ([7]-flto). When this 14370 option is used, GCC generates a bytecode representation of each 14371 input file and writes it to specially-named sections in each object 14372 file. When the object files are linked together, all the function 14373 bodies are read from these named sections and instantiated as if 14374 they had been part of the same translation unit. This enables 14375 interprocedural optimizations to work across different files (and 14376 even different languages), potentially improving the performance of 14377 the generated code. To use the link-timer optimizer, -flto needs to 14378 be specified at compile time and during the final link. If the 14379 program does not require any symbols to be exported, it is possible 14380 to combine -flto and the experimental [8]-fwhopr with 14381 [9]-fwhole-program to allow the interprocedural optimizers to use 14382 more aggressive assumptions. 14383 * The automatic parallelization pass was enhanced to support 14384 parallelization of outer loops. 14385 * Automatic parallelization can be enabled as part of Graphite. In 14386 addition to -ftree-parallelize-loops=, specify 14387 -floop-parallelize-all to enable the Graphite-based optimization. 14388 * The infrastructure for optimizing based on [10]restrict qualified 14389 pointers has been rewritten and should result in code generation 14390 improvements. Optimizations based on restrict qualified pointers 14391 are now also available when using -fno-strict-aliasing. 14392 * There is a new optimization pass that attempts to change prototype 14393 of functions to avoid unused parameters, pass only relevant parts 14394 of structures and turn arguments passed by reference to arguments 14395 passed by value when possible. It is enabled by -O2 and above as 14396 well as -Os and can be manually invoked using the new command-line 14397 switch -fipa-sra. 14398 * GCC now optimize exception handling code. In particular cleanup 14399 regions that are proved to not have any effect are optimized out. 14400 14401 New Languages and Language specific improvements 14402 14403 All languages 14404 14405 * The -fshow-column option is now on by default. This means error 14406 messages now have a column associated with them. 14407 14408 Ada 14409 14410 * Compilation of programs heavily using discriminated record types 14411 with variant parts has been sped up and generates more compact 14412 code. 14413 * Stack checking now works reasonably well on most plaforms. In some 14414 specific cases, stack overflows may still fail to be detected, but 14415 a compile-time warning will be issued for these cases. 14416 14417 C family 14418 14419 * If a header named in a #include directive is not found, the 14420 compiler exits immediately. This avoids a cascade of errors arising 14421 from declarations expected to be found in that header being 14422 missing. 14423 * A new built-in function __builtin_unreachable() has been added that 14424 tells the compiler that control will never reach that point. It may 14425 be used after asm statements that terminate by transferring control 14426 elsewhere, and in other places that are known to be unreachable. 14427 * The -Wlogical-op option now warns for logical expressions such as 14428 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 14429 mistakes. This option is disabled by default. 14430 * An asm goto feature has been added to allow asm statements that 14431 jump to C labels. 14432 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 14433 * The deprecated attribute now takes an optional string argument, for 14434 example, __attribute__((deprecated("text string"))), that will be 14435 printed together with the deprecation warning. 14436 14437 C 14438 14439 * The -Wenum-compare option, which warns when comparing values of 14440 different enum types, now works for C. It formerly only worked for 14441 C++. This warning is enabled by -Wall. It may be avoided by using a 14442 type cast. 14443 * The -Wcast-qual option now warns about casts which are unsafe in 14444 that they permit const-correctness to be violated without further 14445 warnings. Specifically, it warns about cases where a qualifier is 14446 added when all the lower types are not const. For example, it warns 14447 about a cast from char ** to const char **. 14448 * The -Wc++-compat option is significantly improved. It issues new 14449 warnings for: 14450 + Using C++ reserved operator names as identifiers. 14451 + Conversions to enum types without explicit casts. 14452 + Using va_arg with an enum type. 14453 + Using different enum types in the two branches of ?:. 14454 + Using ++ or -- on a variable of enum type. 14455 + Using the same name as both a struct, union or enum tag and a 14456 typedef, unless the typedef refers to the tagged type itself. 14457 + Using a struct, union, or enum which is defined within another 14458 struct or union. 14459 + A struct field defined using a typedef if there is a field in 14460 the struct, or an enclosing struct, whose name is the typedef 14461 name. 14462 + Duplicate definitions at file scope. 14463 + Uninitialized const variables. 14464 + A global variable with an anonymous struct, union, or enum 14465 type. 14466 + Using a string constant to initialize a char array whose size 14467 is the length of the string. 14468 * The new -Wjump-misses-init option warns about cases where a goto or 14469 switch skips the initialization of a variable. This sort of branch 14470 is an error in C++ but not in C. This warning is enabled by 14471 -Wc++-compat. 14472 * GCC now ensures that a C99-conforming <stdint.h> is present on most 14473 targets, and uses information about the types in this header to 14474 implement the Fortran bindings to those types. GCC does not ensure 14475 the presence of such a header, and does not implement the Fortran 14476 bindings, on the following targets: NetBSD, VxWorks, VMS, 14477 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 14478 * GCC now implements C90- and C99-conforming rules for constant 14479 expressions. This may cause warnings or errors for some code using 14480 expressions that can be folded to a constant but are not constant 14481 expressions as defined by ISO C. 14482 * All known target-independent C90 and C90 Amendment 1 conformance 14483 bugs, and all known target-independent C99 conformance bugs not 14484 related to floating point or extended identifiers, have been fixed. 14485 * The C decimal floating point support now includes support for the 14486 FLOAT_CONST_DECIMAL64 pragma. 14487 * The named address space feature from ISO/IEC TR 18037 is now 14488 supported. This is currently only implemented for the SPU 14489 processor. 14490 14491 [11]C++ 14492 14493 * Improved [12]experimental support for the upcoming C++0x ISO C++ 14494 standard, including support for raw strings, lambda expressions and 14495 explicit type conversion operators. 14496 * When printing the name of a class template specialization, G++ will 14497 now omit any template arguments which come from default template 14498 arguments. This behavior (and the pretty-printing of function 14499 template specializations as template signature and arguments) can 14500 be disabled with the -fno-pretty-templates option. 14501 * Access control is now applied to typedef names used in a template, 14502 which may cause G++ to reject some ill-formed code that was 14503 accepted by earlier releases. The -fno-access-control option can be 14504 used as a temporary workaround until the code is corrected. 14505 * Compilation time for code that uses templates should now scale 14506 linearly with the number of instantiations rather than 14507 quadratically, as template instantiations are now looked up using 14508 hash tables. 14509 * Declarations of functions that look like builtin declarations of 14510 library functions are only considered to be redeclarations if they 14511 are declared with extern "C". This may cause problems with code 14512 that omits extern "C" on hand-written declarations of C library 14513 functions such as abort or memcpy. Such code is ill-formed, but was 14514 accepted by earlier releases. 14515 * Diagnostics that used to complain about passing non-POD types to 14516 ... or jumping past the declaration of a non-POD variable now check 14517 for triviality rather than PODness, as per C++0x. 14518 * In C++0x mode local and anonymous classes are now allowed as 14519 template arguments, and in declarations of variables and functions 14520 with linkage, so long as any such declaration that is used is also 14521 defined ([13]DR 757). 14522 * Labels may now have attributes, as has been permitted for a while 14523 in C. This is only permitted when the label definition and the 14524 attribute specifier is followed by a semicoloni.e., the label 14525 applies to an empty statement. The only useful attribute for a 14526 label is unused. 14527 * G++ now implements [14]DR 176. Previously G++ did not support using 14528 the injected-class-name of a template base class as a type name, 14529 and lookup of the name found the declaration of the template in the 14530 enclosing scope. Now lookup of the name finds the 14531 injected-class-name, which can be used either as a type or as a 14532 template, depending on whether or not the name is followed by a 14533 template argument list. As a result of this change, some code that 14534 was previously accepted may be ill-formed because 14535 1. The injected-class-name is not accessible because it's from a 14536 private base, or 14537 2. The injected-class-name cannot be used as an argument for a 14538 template template parameter. 14539 In either of these cases, the code can be fixed by adding a 14540 nested-name-specifier to explicitly name the template. The first 14541 can be worked around with -fno-access-control; the second is only 14542 rejected with -pedantic. 14543 * A new standard mangling for SIMD vector types has been added, to 14544 avoid name clashes on systems with vectors of varying length. By 14545 default the compiler still uses the old mangling, but emits aliases 14546 with the new mangling on targets that support strong aliases. Users 14547 can switch over entirely to the new mangling with -fabi-version=4 14548 or -fabi-version=0. -Wabi will now warn about code that uses the 14549 old mangling. 14550 * The command-line option -ftemplate-depth-N is now written as 14551 -ftemplate-depth=N and the old form is deprecated. 14552 * Conversions between NULL and non-pointer types are now warned by 14553 default. The new option -Wno-conversion-null disables these 14554 warnings. Previously these warnings were only available when using 14555 -Wconversion explicitly. 14556 14557 Runtime Library (libstdc++) 14558 14559 * Improved experimental support for the upcoming ISO C++ standard, 14560 C++0x, including: 14561 + Support for <future>, <functional>, and <random>. 14562 + Existing facilities now exploit explicit operators and the 14563 newly implemented core C++0x features. 14564 + The header <cstdatomic> has been renamed to <atomic>. 14565 * An experimental profile mode has been added. This is an 14566 implementation of many C++ standard library constructs with an 14567 additional analysis layer that gives performance improvement advice 14568 based on recognition of suboptimal usage patterns. For example, 14569 #include <vector> 14570 int main() 14571 { 14572 std::vector<int> v; 14573 for (int k = 0; k < 1024; ++k) 14574 v.insert(v.begin(), k); 14575 } 14576 14577 When instrumented via the profile mode, can return suggestions 14578 about the initial size and choice of the container used as follows: 14579 vector-to-list: improvement = 5: call stack = 0x804842c ... 14580 : advice = change std::vector to std::list 14581 vector-size: improvement = 3: call stack = 0x804842c ... 14582 : advice = change initial container size from 0 to 1024 14583 14584 These constructs can be substituted for the normal libstdc++ 14585 constructs on a piecemeal basis, or all existing components can be 14586 transformed via the -D_GLIBCXX_PROFILE macro. 14587 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 14588 24733) has been added. This support is in header file 14589 <decimal/decimal>, uses namespace std::decimal, and includes 14590 classes decimal32, decimal64, and decimal128. 14591 * Sources have been audited for application of function attributes 14592 nothrow, const, pure, and noreturn. 14593 * Python pretty-printers have been added for many standard library 14594 components that simplify the internal representation and present a 14595 more intuitive view of components when used with 14596 appropriately-advanced versions of GDB. For more information, 14597 please consult the more [16]detailed description. 14598 * The default behavior for comparing typeinfo names has changed, so 14599 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 14600 * The new -static-libstdc++ option directs g++ to link the C++ 14601 library statically, even if the default would normally be to link 14602 it dynamically. 14603 14604 [17]Fortran 14605 14606 * The COMMON default padding has been changed instead of adding the 14607 padding before a variable it is now added afterwards, which 14608 increases the compatibility with other vendors and helps to obtain 14609 the correct output in some cases. Cf. also the -falign-commons 14610 option ([18]added in 4.4). 14611 * The -finit-real= option now also supports the value snan for 14612 signaling not-a-number; to be effective, one additionally needs to 14613 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 14614 optimizations can turn a signaling NaN into a quiet one. 14615 * The new option -fcheck= has been added with the options bounds, 14616 array-temps, do, pointer, and recursive. The bounds and array-temps 14617 options are equivalent to -fbounds-check and 14618 -fcheck-array-temporaries. The do option checks for invalid 14619 modification of loop iteration variables, and the recursive option 14620 tests for recursive calls to subroutines/functions which are not 14621 marked as recursive. With pointer pointer association checks in 14622 calls are performed; however, neither undefined pointers nor 14623 pointers in expressions are handled. Using -fcheck=all enables all 14624 these run-time checks. 14625 * The run-time checking -fcheck=bounds now warns about invalid string 14626 lengths of character dummy arguments. Additionally, more 14627 compile-time checks have been added. 14628 * The new option [19]-fno-protect-parens has been added; if set, the 14629 compiler may reorder REAL and COMPLEX expressions without regard to 14630 parentheses. 14631 * GNU Fortran no longer links against libgfortranbegin. As before, 14632 MAIN__ (assembler symbol name) is the actual Fortran main program, 14633 which is invoked by the main function. However, main is now 14634 generated and put in the same object file as MAIN__. For the time 14635 being, libgfortranbegin still exists for backward compatibility. 14636 For details see the new [20]Mixed-Language Programming chapter in 14637 the manual. 14638 * The I/O library was restructured for performance and cleaner code. 14639 * Array assignments and WHERE are now run in parallel when OpenMP's 14640 WORKSHARE is used. 14641 * The experimental option -fwhole-file was added. The option allows 14642 whole-file checking of procedure arguments and allows for better 14643 optimizations. It can also be used with -fwhole-program, which is 14644 now also supported in gfortran. 14645 * More Fortran 2003 and Fortran 2008 mathematical functions can now 14646 be used as initialization expressions. 14647 * Some extended attributes such as STDCALL are now supported via the 14648 [21]GCC$ compiler directive. 14649 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 14650 intrinsic behaves now as if zero were always positive. 14651 * For legacy compatibiliy: On Cygwin and MinGW, the special files 14652 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 14653 supported. 14654 * Fortran 2003 support has been extended: 14655 + Procedure-pointer function results and procedure-pointer 14656 components (including PASS), 14657 + allocatable scalars (experimental), 14658 + DEFERRED type-bound procedures, 14659 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 14660 have been implemented. 14661 + The ALLOCATE statement supports type-specs and the SOURCE= 14662 argument. 14663 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 14664 type-bound procedure (i.e. as type-bound operators). 14665 + Rounding (ROUND=, RZ, ...) for output is now supported. 14666 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 14667 intrinsic module ISO_C_BINDING are now supported, except for 14668 the targets listed above as ones where GCC does not have 14669 <stdint.h> type information. 14670 + Extensible derived types with type-bound procedure or 14671 procedure pointer with PASS attribute now have to use CLASS in 14672 line with the Fortran 2003 standard; the workaround to use 14673 TYPE is no longer supported. 14674 + [22]Experimental, incomplete support for polymorphism, 14675 including CLASS, SELECT TYPE and dynamic dispatch of 14676 type-bound procedure calls. Some features do not work yet such 14677 as unlimited polymorphism (CLASS(*)). 14678 * Fortran 2008 support has been extended: 14679 + The OPEN statement now supports the NEWUNIT= option, which 14680 returns a unique file unit, thus preventing inadvertent use of 14681 the same unit in different parts of the program. 14682 + Support for unlimited format items has been added. 14683 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 14684 the intrinsic module ISO_FORTRAN_ENV are now supported. 14685 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 14686 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 14687 and ATANH have been added (for real and complex arguments) and 14688 ATAN(Y,X) is now an alias for ATAN2(Y,X). 14689 + The BLOCK construct has been implemented. 14690 14691 [23]New Targets and Target Specific Improvements 14692 14693 AIX 14694 14695 * Full cross-toolchain support now available with GNU Binutils 14696 14697 ARM 14698 14699 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 14700 * GCC now supports the ARM v7E-M architecture. 14701 * GCC now supports VFPv4-based FPUs and FPUs with 14702 single-precision-only VFP. 14703 * GCC has many improvements to optimization for other ARM processors, 14704 including scheduling support for the integer pipeline on Cortex-A9. 14705 * GCC now supports the IEEE 754-2008 half-precision floating-point 14706 type, and a variant ARM-specific half-precision type. This type is 14707 specified using __fp16, with the layout determined by 14708 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 14709 VFPv4 half-precision instructions will be used. 14710 * GCC now supports the variant of AAPCS that uses VFP registers for 14711 parameter passing and return values. 14712 14713 AVR 14714 14715 * The -mno-tablejump option has been removed because it has the same 14716 effect as the -fno-jump-tables option. 14717 * Added support for these new AVR devices: 14718 + ATmega8U2 14719 + ATmega16U2 14720 + ATmega32U2 14721 14722 [24]IA-32/x86-64 14723 14724 * GCC now will set the default for -march= based on the configure 14725 target. 14726 * GCC now supports handling floating-point excess precision arising 14727 from use of the x87 floating-point unit in a way that conforms to 14728 ISO C99. This is enabled with -fexcess-precision=standard and with 14729 standards conformance options such as -std=c99, and may be disabled 14730 using -fexcess-precision=fast. 14731 * Support for the Intel Atom processor is now available through the 14732 -march=atom and -mtune=atom options. 14733 * A new -mcrc32 option is now available to enable crc32 intrinsics. 14734 * A new -mmovbe option is now available to enable GCC to use the 14735 movbe instruction to implement __builtin_bswap32 and 14736 __builtin_bswap64. 14737 * SSE math now can be enabled by default at configure time with the 14738 new --with-fpmath=sse option. 14739 * There is a new intrinsic header file, <x86intrin.h>. It should be 14740 included before using any IA-32/x86-64 intrinsics. 14741 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 14742 Orochi processors are now available with the -mxop, -mfma4, and 14743 -mlwp options. 14744 * The -mabm option enables GCC to use the popcnt and lzcnt 14745 instructions on AMD processors. 14746 * The -mpopcnt option enables GCC to use the popcnt instructions on 14747 both AMD and Intel processors. 14748 14749 M68K/ColdFire 14750 14751 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 14752 and 5441x devices. 14753 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 14754 processors. 14755 14756 [25]MeP 14757 14758 Support has been added for the Toshiba Media embedded Processor (MeP, 14759 or mep-elf) embedded target. 14760 14761 MIPS 14762 14763 * GCC now supports MIPS 1004K processors. 14764 * GCC can now be configured with options --with-arch-32, 14765 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 14766 default optimization separately for 32-bit and 64-bit modes. 14767 * MIPS targets now support an alternative _mcount interface, in which 14768 register $12 points to the function's save slot for register $31. 14769 This interface is selected by the -mcount-ra-address option; see 14770 the documentation for more details. 14771 * GNU/Linux targets can now generate read-only .eh_frame sections. 14772 This optimization requires GNU binutils 2.20 or above, and is only 14773 available if GCC is configured with a suitable version of binutils. 14774 * GNU/Linux targets can now attach special relocations to indirect 14775 calls, so that the linker can turn them into direct jumps or 14776 branches. This optimization requires GNU binutils 2.20 or later, 14777 and is automatically selected if GCC is configured with an 14778 appropriate version of binutils. It can be explicitly enabled or 14779 disabled using the -mrelax-pic-calls command-line option. 14780 * GCC now generates more heavily-optimized atomic operations on 14781 Octeon processors. 14782 * MIPS targets now support the -fstack-protector option. 14783 * GCC now supports an -msynci option, which specifies that synci is 14784 enough to flush the instruction cache, without help from the 14785 operating system. GCC uses this information to optimize 14786 automatically-generated cache flush operations, such as those used 14787 for nested functions in C. There is also a --with-synci 14788 configure-time option, which makes -msynci the default. 14789 * GCC supports four new function attributes for interrupt handlers: 14790 interrupt, use_shadow_register_set, keep_interrupts_masked and 14791 use_debug_exception_return. See the documentation for more details 14792 about these attributes. 14793 14794 [26]RS/6000 (POWER/PowerPC) 14795 14796 * GCC now supports the Power ISA 2.06, which includes the VSX 14797 instructions that add vector 64-bit floating point support, new 14798 population count instructions, and conversions between floating 14799 point and unsigned types. 14800 * Support for the power7 processor is now available through the 14801 -mcpu=power7 and -mtune=power7. 14802 * GCC will now vectorize loops that contain simple math functions 14803 like copysign when generating code for altivec or VSX targets. 14804 * Support for the A2 processor is now available through the -mcpu=a2 14805 and -mtune=a2 options. 14806 * Support for the 476 processor is now available through the 14807 -mcpu={476,476fp} and -mtune={476,476fp} options. 14808 * Support for the e500mc64 processor is now available through the 14809 -mcpu=e500mc64 and -mtune=e500mc64 options. 14810 * GCC can now be configured with options --with-cpu-32, 14811 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 14812 default optimization separately for 32-bit and 64-bit modes. 14813 * Starting with GCC 4.5.4, vectors of type vector long long or vector 14814 long are passed and returned in the same method as other vectors 14815 with the VSX instruction set. Previously the GCC compiler did not 14816 adhere to the ABI for 128-bit vectors with 64-bit integer base 14817 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 14818 14819 [27]RX 14820 14821 Support has been added for the Renesas RX Processor (rx-elf) target. 14822 14823 [28]Operating Systems 14824 14825 [29]Windows (Cygwin and MinGW) 14826 14827 * GCC now installs all the major language runtime libraries as DLLs 14828 when configured with the --enable-shared option. 14829 * GCC now makes use of the new support for aligned common variables 14830 in versions of binutils >= 2.20 to fix bugs in the support for SSE 14831 data types. 14832 * Improvements to the libffi support library increase the reliability 14833 of code generated by GCJ on all Windows platforms. Libgcj is 14834 enabled by default for the first time. 14835 * Libtool improvements simplify installation by placing the generated 14836 DLLs in the correct binaries directory. 14837 * Numerous other minor bugfixes and improvements, and substantial 14838 enhancements to the Fortran language support library. 14839 14840 > 14841 14842 Other significant improvements 14843 14844 Plugins 14845 14846 * It is now possible to extend the compiler without having to modify 14847 its source code. A new option -fplugin=file.so tells GCC to load 14848 the shared object file.so and execute it as part of the compiler. 14849 The internal documentation describes the details on how plugins can 14850 interact with the compiler. 14851 14852 Installation changes 14853 14854 * The move to newer autotools changed default installation 14855 directories and switches to control them: The --with-datarootdir, 14856 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 14857 used any more. Instead, you can now use --datarootdir, --docdir, 14858 --htmldir, and --pdfdir. The default installation directories have 14859 changed as follows according to the GNU Coding Standards: 14860 14861 datarootdir read-only architecture-independent data root [PREFIX/share] 14862 localedir locale-specific message catalogs [DATAROOTDIR/locale] 14863 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 14864 htmldir html documentation [DOCDIR] 14865 dvidir dvi documentation [DOCDIR] 14866 pdfdir pdf documentation [DOCDIR] 14867 psdir ps documentation [DOCDIR] 14868 The following variables have new default values: 14869 14870 datadir read-only architecture-independent data [DATAROOTDIR] 14871 infodir info documentation [DATAROOTDIR/info] 14872 mandir man documentation [DATAROOTDIR/man] 14873 14874 [30]GCC 4.5.1 14875 14876 This is the [31]list of problem reports (PRs) from GCC's bug tracking 14877 system that are known to be fixed in the 4.5.1 release. This list might 14878 not be complete (that is, it is possible that some PRs that have been 14879 fixed are not listed here). 14880 14881 All languages 14882 14883 * GCC's new link-time optimizer ([32]-flto) now also works on a few 14884 non-ELF targets: 14885 + Cygwin (*-cygwin*) 14886 + MinGW (*-mingw*) 14887 + Darwin on x86-64 (x86_64-apple-darwin*) 14888 LTO is not enabled by default for these targets. To enable LTO, you 14889 should configure with the --enable-lto option. 14890 14891 [33]GCC 4.5.2 14892 14893 This is the [34]list of problem reports (PRs) from GCC's bug tracking 14894 system that are known to be fixed in the 4.5.2 release. This list might 14895 not be complete (that is, it is possible that some PRs that have been 14896 fixed are not listed here). 14897 14898 [35]GCC 4.5.3 14899 14900 This is the [36]list of problem reports (PRs) from GCC's bug tracking 14901 system that are known to be fixed in the 4.5.3 release. This list might 14902 not be complete (that is, it is possible that some PRs that have been 14903 fixed are not listed here). 14904 14905 On the PowerPC compiler, the Altivec builtin functions vec_ld and 14906 vec_st have been modified to generate the Altivec memory instructions 14907 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 14908 release, these builtin functions were changed to generate VSX memory 14909 reference instructions instead of Altivec memory instructions, but 14910 there are differences between the two instructions. If the VSX 14911 instruction set is available, you can now use the new builtin functions 14912 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 14913 instructions. 14914 14915 [37]GCC 4.5.4 14916 14917 This is the [38]list of problem reports (PRs) from GCC's bug tracking 14918 system that are known to be fixed in the 4.5.4 release. This list might 14919 not be complete (that is, it is possible that some PRs that have been 14920 fixed are not listed here). 14921 14922 14923 For questions related to the use of GCC, please consult these web 14924 pages and the [39]GCC manuals. If that fails, the 14925 [40]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14926 web pages and the development of GCC are welcome on our developer 14927 list at [41]gcc (a] gcc.gnu.org. All of [42]our lists have public 14928 archives. 14929 14930 Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and 14931 distribution of this entire article is permitted in any medium, 14932 provided this notice is preserved. 14933 14934 These pages are [44]maintained by the GCC team. Last modified 14935 2025-09-29. 14936 14937 References 14938 14939 1. https://www.multiprecision.org/ 14940 2. https://gcc.gnu.org/install/prerequisites.html 14941 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 14942 4. https://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 14943 5. https://gcc.gnu.org/gcc-4.5/changes.html#x86 14944 6. https://gcc.gnu.org/PR30789 14945 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 14946 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 14947 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 14948 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 14949 11. https://gcc.gnu.org/gcc-4.5/changes.html#cplusplus 14950 12. https://gcc.gnu.org/gcc-4.5/cxx0x_status.html 14951 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 14952 14. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 14953 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 14954 16. https://sourceware.org/gdb/wiki/STLSupport 14955 17. https://gcc.gnu.org/gcc-4.5/changes.html#Fortran 14956 18. https://gcc.gnu.org/gcc-4.4/changes.html 14957 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 14958 20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 14959 21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 14960 22. https://gcc.gnu.org/wiki/OOP 14961 23. https://gcc.gnu.org/gcc-4.5/changes.html#targets 14962 24. https://gcc.gnu.org/gcc-4.5/changes.html#x86 14963 25. https://gcc.gnu.org/gcc-4.5/changes.html#mep 14964 26. https://gcc.gnu.org/gcc-4.5/changes.html#rs6000 14965 27. https://gcc.gnu.org/gcc-4.5/changes.html#rx 14966 28. https://gcc.gnu.org/gcc-4.5/changes.html#os 14967 29. https://gcc.gnu.org/gcc-4.5/changes.html#windows 14968 30. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.1 14969 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 14970 32. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 14971 33. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.2 14972 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 14973 35. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.3 14974 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 14975 37. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.4 14976 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 14977 39. https://gcc.gnu.org/onlinedocs/ 14978 40. mailto:gcc-help (a] gcc.gnu.org 14979 41. mailto:gcc (a] gcc.gnu.org 14980 42. https://gcc.gnu.org/lists.html 14981 43. https://www.fsf.org/ 14982 44. https://gcc.gnu.org/about.html 14983 ====================================================================== 14984 http://gcc.gnu.org/gcc-4.4/index.html 14985 GCC 4.4 Release Series 14986 14987 This release series is no longer maintained. 14988 14989 March 13, 2012 14990 14991 The [1]GNU project and the GCC developers are pleased to announce the 14992 release of GCC 4.4.7. 14993 14994 This release is a bug-fix release, containing fixes for regressions in 14995 GCC 4.4.6 relative to previous releases of GCC. 14996 14997 Release History 14998 14999 GCC 4.4.7 15000 March 13, 2012 ([2]changes) 15001 15002 GCC 4.4.6 15003 April 16, 2011 ([3]changes) 15004 15005 GCC 4.4.5 15006 October 1, 2010 ([4]changes) 15007 15008 GCC 4.4.4 15009 April 29, 2010 ([5]changes) 15010 15011 GCC 4.4.3 15012 January 21, 2010 ([6]changes) 15013 15014 GCC 4.4.2 15015 October 15, 2009 ([7]changes) 15016 15017 GCC 4.4.1 15018 July 22, 2009 ([8]changes) 15019 15020 GCC 4.4.0 15021 April 21, 2009 ([9]changes) 15022 15023 References and Acknowledgements 15024 15025 GCC used to stand for the GNU C Compiler, but since the compiler 15026 supports several other languages aside from C, it now stands for the 15027 GNU Compiler Collection. 15028 15029 A list of [10]successful builds is updated as new information becomes 15030 available. 15031 15032 The GCC developers would like to thank the numerous people that have 15033 contributed new features, improvements, bug fixes, and other changes as 15034 well as test results to GCC. This [11]amazing group of volunteers is 15035 what makes GCC successful. 15036 15037 For additional information about GCC please refer to the [12]GCC 15038 project web site or contact the [13]GCC development mailing list. 15039 15040 To obtain GCC please use [14]our mirror sites or [15]our version 15041 control system. 15042 15043 15044 For questions related to the use of GCC, please consult these web 15045 pages and the [16]GCC manuals. If that fails, the 15046 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15047 web pages and the development of GCC are welcome on our developer 15048 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 15049 archives. 15050 15051 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 15052 distribution of this entire article is permitted in any medium, 15053 provided this notice is preserved. 15054 15055 These pages are [21]maintained by the GCC team. Last modified 15056 2022-10-26. 15057 15058 References 15059 15060 1. http://www.gnu.org/ 15061 2. https://gcc.gnu.org/gcc-4.4/changes.html 15062 3. https://gcc.gnu.org/gcc-4.4/changes.html 15063 4. https://gcc.gnu.org/gcc-4.4/changes.html 15064 5. https://gcc.gnu.org/gcc-4.4/changes.html 15065 6. https://gcc.gnu.org/gcc-4.4/changes.html 15066 7. https://gcc.gnu.org/gcc-4.4/changes.html 15067 8. https://gcc.gnu.org/gcc-4.4/changes.html 15068 9. https://gcc.gnu.org/gcc-4.4/changes.html 15069 10. https://gcc.gnu.org/gcc-4.4/buildstat.html 15070 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15071 12. https://gcc.gnu.org/index.html 15072 13. mailto:gcc (a] gcc.gnu.org 15073 14. https://gcc.gnu.org/mirrors.html 15074 15. https://gcc.gnu.org/git.html 15075 16. https://gcc.gnu.org/onlinedocs/ 15076 17. mailto:gcc-help (a] gcc.gnu.org 15077 18. mailto:gcc (a] gcc.gnu.org 15078 19. https://gcc.gnu.org/lists.html 15079 20. https://www.fsf.org/ 15080 21. https://gcc.gnu.org/about.html 15081 ====================================================================== 15082 http://gcc.gnu.org/gcc-4.4/changes.html 15083 GCC 4.4 Release Series 15084 Changes, New Features, and Fixes 15085 15086 The latest release in the 4.4 release series is [1]GCC 4.4.7. 15087 15088 Caveats 15089 15090 * __builtin_stdarg_start has been completely removed from GCC. 15091 Support for <varargs.h> had been deprecated since GCC 4.0. Use 15092 __builtin_va_start as a replacement. 15093 * Some of the errors issued by the C++ front end that could be 15094 downgraded to warnings in previous releases by using -fpermissive 15095 are now warnings by default. They can be converted into errors by 15096 using -pedantic-errors. 15097 * Use of the cpp assertion extension will now emit a warning when 15098 -Wdeprecated or -pedantic is used. This extension has been 15099 deprecated for many years, but never warned about. 15100 * Packed bit-fields of type char were not properly bit-packed on many 15101 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 15102 causes an ABI change. For example there is no longer a 4-bit 15103 padding between field a and b in this structure: 15104 struct foo 15105 { 15106 char a:4; 15107 char b:8; 15108 } __attribute__ ((packed)); 15109 There is a new warning to help identify fields that are affected: 15110 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 15111 The warning can be disabled with -Wno-packed-bitfield-compat. 15112 * On ARM EABI targets, the C++ mangling of the va_list type has been 15113 changed to conform to the current revision of the EABI. This does 15114 not affect the libstdc++ library included with GCC. 15115 * The SCOUNT and POS bits of the MIPS DSP control register are now 15116 treated as global. Previous versions of GCC treated these fields as 15117 call-clobbered instead. 15118 * The MIPS port no longer recognizes the h asm constraint. It was 15119 necessary to remove this constraint in order to avoid generating 15120 unpredictable code sequences. 15121 One of the main uses of the h constraint was to extract the high 15122 part of a multiplication on 64-bit targets. For example: 15123 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 15124 You can now achieve the same effect using 128-bit types: 15125 typedef unsigned int uint128_t __attribute__((mode(TI))); 15126 result = ((uint128_t) x * y) >> 64; 15127 The second sequence is better in many ways. For example, if x and y 15128 are constants, the compiler can perform the multiplication at 15129 compile time. If x and y are not constants, the compiler can 15130 schedule the runtime multiplication better than it can schedule an 15131 asm statement. 15132 * Support for a number of older systems and recently unmaintained or 15133 untested target ports of GCC has been declared obsolete in GCC 4.4. 15134 Unless there is activity to revive them, the next release of GCC 15135 will have their sources permanently removed. 15136 The following ports for individual systems on particular 15137 architectures have been obsoleted: 15138 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 15139 m68k-*-aout*) 15140 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 15141 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 15142 sh-*-*). This does not affect other more specific targets 15143 using the COFF object format on those architectures, or the 15144 more specific H8300 and SH targets (h8300-*-rtems*, 15145 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 15146 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 15147 + 2BSD on PDP-11 (pdp11-*-bsd) 15148 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 15149 powerpc-ibm-aix4.[12]*) 15150 + Tuning support for Itanium1 (Merced) variants. Note that code 15151 tuned for Itanium2 should also run correctly on Itanium1. 15152 * The protoize and unprotoize utilities have been obsoleted and will 15153 be removed in GCC 4.5. These utilities have not been installed by 15154 default since GCC 3.0. 15155 * Support has been removed for all the [2]configurations obsoleted in 15156 GCC 4.3. 15157 * Unknown -Wno-* options are now silently ignored by GCC if no other 15158 diagnostics are issued. If other diagnostics are issued, then GCC 15159 warns about the unknown options. 15160 * More information on porting to GCC 4.4 from previous versions of 15161 GCC can be found in the [3]porting guide for this release. 15162 15163 General Optimizer Improvements 15164 15165 * A new command-line switch -findirect-inlining has been added. When 15166 turned on it allows the inliner to also inline indirect calls that 15167 are discovered to have known targets at compile time thanks to 15168 previous inlining. 15169 * A new command-line switch -ftree-switch-conversion has been added. 15170 This new pass turns simple initializations of scalar variables in 15171 switch statements into initializations from a static array, given 15172 that all the values are known at compile time and the ratio between 15173 the new array size and the original switch branches does not exceed 15174 the parameter --param switch-conversion-max-branch-ratio (default 15175 is eight). 15176 * A new command-line switch -ftree-builtin-call-dce has been added. 15177 This optimization eliminates unnecessary calls to certain builtin 15178 functions when the return value is not used, in cases where the 15179 calls can not be eliminated entirely because the function may set 15180 errno. This optimization is on by default at -O2 and above. 15181 * A new command-line switch -fconserve-stack directs the compiler to 15182 minimize stack usage even if it makes the generated code slower. 15183 This affects inlining decisions. 15184 * When the assembler supports it, the compiler will now emit unwind 15185 information using assembler .cfi directives. This makes it possible 15186 to use such directives in inline assembler code. The new option 15187 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 15188 directives. 15189 * The [4]Graphite branch has been merged. This merge has brought in a 15190 new framework for loop optimizations based on a polyhedral 15191 intermediate representation. These optimizations apply to all the 15192 languages supported by GCC. The following new code transformations 15193 are available in GCC 4.4: 15194 + -floop-interchange performs loop interchange transformations 15195 on loops. Interchanging two nested loops switches the inner 15196 and outer loops. For example, given a loop like: 15197 DO J = 1, M 15198 DO I = 1, N 15199 A(J, I) = A(J, I) * C 15200 ENDDO 15201 ENDDO 15202 15203 loop interchange will transform the loop as if the user had 15204 written: 15205 DO I = 1, N 15206 DO J = 1, M 15207 A(J, I) = A(J, I) * C 15208 ENDDO 15209 ENDDO 15210 15211 which can be beneficial when N is larger than the caches, 15212 because in Fortran, the elements of an array are stored in 15213 memory contiguously by column, and the original loop iterates 15214 over rows, potentially creating at each access a cache miss. 15215 + -floop-strip-mine performs loop strip mining transformations 15216 on loops. Strip mining splits a loop into two nested loops. 15217 The outer loop has strides equal to the strip size and the 15218 inner loop has strides of the original loop within a strip. 15219 For example, given a loop like: 15220 DO I = 1, N 15221 A(I) = A(I) + C 15222 ENDDO 15223 15224 loop strip mining will transform the loop as if the user had 15225 written: 15226 DO II = 1, N, 4 15227 DO I = II, min (II + 3, N) 15228 A(I) = A(I) + C 15229 ENDDO 15230 ENDDO 15231 15232 + -floop-block performs loop blocking transformations on loops. 15233 Blocking strip mines each loop in the loop nest such that the 15234 memory accesses of the element loops fit inside caches. For 15235 example, given a loop like: 15236 DO I = 1, N 15237 DO J = 1, M 15238 A(J, I) = B(I) + C(J) 15239 ENDDO 15240 ENDDO 15241 15242 loop blocking will transform the loop as if the user had 15243 written: 15244 DO II = 1, N, 64 15245 DO JJ = 1, M, 64 15246 DO I = II, min (II + 63, N) 15247 DO J = JJ, min (JJ + 63, M) 15248 A(J, I) = B(I) + C(J) 15249 ENDDO 15250 ENDDO 15251 ENDDO 15252 ENDDO 15253 15254 which can be beneficial when M is larger than the caches, 15255 because the innermost loop will iterate over a smaller amount 15256 of data that can be kept in the caches. 15257 * A new register allocator has replaced the old one. It is called 15258 integrated register allocator (IRA) because coalescing, register 15259 live range splitting, and hard register preferencing are done 15260 on-the-fly during coloring. It also has better integration with the 15261 reload pass. IRA is a regional register allocator which uses modern 15262 Chaitin-Briggs coloring instead of Chow's priority coloring used in 15263 the old register allocator. More info about IRA internals and 15264 options can be found in the GCC manuals. 15265 * A new instruction scheduler and software pipeliner, based on the 15266 selective scheduling approach, has been added. The new pass 15267 performs instruction unification, register renaming, substitution 15268 through register copies, and speculation during scheduling. The 15269 software pipeliner is able to pipeline non-countable loops. The new 15270 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 15271 it is available for the Intel Itanium platform working by default 15272 as the second scheduling pass (after register allocation) at the 15273 -O3 optimization level. 15274 * When using -fprofile-generate with a multi-threaded program, the 15275 profile counts may be slightly wrong due to race conditions. The 15276 new -fprofile-correction option directs the compiler to apply 15277 heuristics to smooth out the inconsistencies. By default the 15278 compiler will give an error message when it finds an inconsistent 15279 profile. 15280 * The new -fprofile-dir=PATH option permits setting the directory 15281 where profile data files are stored when using -fprofile-generate 15282 and friends, and the directory used when reading profile data files 15283 using -fprofile-use and friends. 15284 15285 New warning options 15286 15287 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 15288 warning if any stack frame is larger than NUMBER bytes. This may be 15289 used to help ensure that code fits within a limited amount of stack 15290 space. 15291 * The command-line option -Wlarger-than-N is now written as 15292 -Wlarger-than=N and the old form is deprecated. 15293 * The new -Wno-mudflap option disables warnings about constructs 15294 which can not be instrumented when using -fmudflap. 15295 15296 New Languages and Language specific improvements 15297 15298 * Version 3.0 of the OpenMP specification is now supported for the C, 15299 C++, and Fortran compilers. 15300 * New character data types, per [5]TR 19769: New character types in 15301 C, are now supported for the C compiler in -std=gnu99 mode, as 15302 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 15303 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 15304 15305 C family 15306 15307 * A new optimize attribute was added to allow programmers to change 15308 the optimization level and particular optimization options for an 15309 individual function. You can also change the optimization options 15310 via the GCC optimize pragma for functions defined after the pragma. 15311 The GCC push_options pragma and the GCC pop_options pragma allow 15312 you temporarily save and restore the options used. The GCC 15313 reset_options pragma restores the options to what was specified on 15314 the command line. 15315 * Uninitialized warnings do not require enabling optimization 15316 anymore, that is, -Wuninitialized can be used together with -O0. 15317 Nonetheless, the warnings given by -Wuninitialized will probably be 15318 more accurate if optimization is enabled. 15319 * -Wparentheses now warns about expressions such as (!x | y) and (!x 15320 & y). Using explicit parentheses, such as in ((!x) | y), silences 15321 this warning. 15322 * -Wsequence-point now warns within if, while,do while and for 15323 conditions, and within for begin/end expressions. 15324 * A new option -dU is available to dump definitions of preprocessor 15325 macros that are tested or expanded. 15326 15327 C++ 15328 15329 * [6]Improved experimental support for the upcoming ISO C++ standard, 15330 C++0x. Including support for auto, inline namespaces, generalized 15331 initializer lists, defaulted and deleted functions, new character 15332 types, and scoped enums. 15333 * Those errors that may be downgraded to warnings to build legacy 15334 code now mention -fpermissive when -fdiagnostics-show-option is 15335 enabled. 15336 * -Wconversion now warns if the result of a static_cast to enumeral 15337 type is unspecified because the value is outside the range of the 15338 enumeral type. 15339 * -Wuninitialized now warns if a non-static reference or non-static 15340 const member appears in a class without constructors. 15341 * G++ now properly implements value-initialization, so objects with 15342 an initializer of () and an implicitly defined default constructor 15343 will be zero-initialized before the default constructor is called. 15344 15345 Runtime Library (libstdc++) 15346 15347 * Improved experimental support for the upcoming ISO C++ standard, 15348 C++0x, including: 15349 + Support for <chrono>, <condition_variable>, <cstdatomic>, 15350 <forward_list>, <initializer_list>, <mutex>, <ratio>, 15351 <system_error>, and <thread>. 15352 + unique_ptr, <algorithm> additions, exception propagation, and 15353 support for the new character types in <string> and <limits>. 15354 + Existing facilities now exploit initializer lists, defaulted 15355 and deleted functions, and the newly implemented core C++0x 15356 features. 15357 + Some standard containers are more efficient together with 15358 stateful allocators, i.e., no allocator is constructed on the 15359 fly at element construction time. 15360 * Experimental support for non-standard pointer types in containers. 15361 * The long standing libstdc++/30928 has been fixed for targets 15362 running glibc 2.10 or later. 15363 * As usual, many small and larger bug fixes, in particular quite a 15364 few corner cases in <locale>. 15365 15366 Fortran 15367 15368 * GNU Fortran now employs libcpp directly instead of using cc1 as an 15369 external preprocessor. The [7]-cpp option was added to allow manual 15370 invocation of the preprocessor without relying on filename 15371 extensions. 15372 * The [8]-Warray-temporaries option warns about array temporaries 15373 generated by the compiler, as an aid to optimization. 15374 * The [9]-fcheck-array-temporaries option has been added, printing a 15375 notification at run time, when an array temporary had to be created 15376 for an function argument. Contrary to -Warray-temporaries the 15377 warning is only printed if the array is noncontiguous. 15378 * Improved generation of DWARF debugging symbols 15379 * If using an intrinsic not part of the selected standard (via -std= 15380 and -fall-intrinsics) gfortran will now treat it as if this 15381 procedure were declared EXTERNAL and try to link to a user-supplied 15382 procedure. -Wintrinsics-std will warn whenever this happens. The 15383 now-useless option -Wnonstd-intrinsic was removed. 15384 * The flag -falign-commons has been added to control the alignment of 15385 variables in COMMON blocks, which is enabled by default in line 15386 with previous GCC version. Using -fno-align-commons one can force 15387 commons to be contiguous in memory as required by the Fortran 15388 standard, however, this slows down the memory access. The option 15389 -Walign-commons, which is enabled by default, warns when padding 15390 bytes were added for alignment. The proper solution is to sort the 15391 common objects by decreasing storage size, which avoids the 15392 alignment problems. 15393 * Fortran 2003 support has been extended: 15394 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 15395 now supported (except internal reads from/writes to wide 15396 strings). [10]-fbackslash now supports also \unnnn and 15397 \Unnnnnnnn to enter Unicode characters. 15398 + Asynchronous I/O (implemented as synchronous I/O) and the 15399 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 15400 are now supported in I/O statements. 15401 + Support for Fortran 2003 structure constructors and for array 15402 constructor with typespec has been added. 15403 + Procedure Pointers (but not yet as component in derived types 15404 and as function results) are now supported. 15405 + Abstract types, type extension, and type-bound procedures 15406 (both PROCEDURE and GENERIC but not as operators). Note: As 15407 CLASS/polymorphyic types are not implemented, type-bound 15408 procedures with PASS accept as non-standard extension TYPE 15409 arguments. 15410 * Fortran 2008 support has been added: 15411 + The -std=f2008 option and support for the file extensions 15412 .f2008 and .F2008 has been added. 15413 + The g0 format descriptor is now supported. 15414 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 15415 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 15416 are now available (some of them existed as GNU extension 15417 before). Note: The hyperbolic functions are not yet supporting 15418 complex arguments and the three- argument version of BESSEL_*N 15419 is not available. 15420 + The bit intrinsics LEADZ and TRAILZ have been added. 15421 15422 Java (GCJ) 15423 15424 Ada 15425 15426 * The Ada runtime now supports multilibs on many platforms including 15427 x86_64, SPARC and PowerPC. Their build is enabled by default. 15428 15429 [11]New Targets and Target Specific Improvements 15430 15431 ARM 15432 15433 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 15434 Cortex-R4F processors and has many other improvements to 15435 optimization for ARM processors. 15436 * GCC now supports the VFPv3 variant with 16 double-precision 15437 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 15438 renamed to -mfpu=vfpv3. 15439 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 15440 erratum on Cortex-M3 processors. 15441 * GCC now supports the __sync_* atomic operations for ARM EABI 15442 GNU/Linux. 15443 * The section anchors optimization is now enabled by default when 15444 optimizing for ARM. 15445 * GCC now uses a new EABI-compatible profiling interface for EABI 15446 targets. This requires a function __gnu_mcount_nc, which is 15447 provided by GNU libc versions 2.8 and later. 15448 15449 AVR 15450 15451 * The -mno-tablejump option has been deprecated because it has the 15452 same effect as the -fno-jump-tables option. 15453 * Added support for these new AVR devices: 15454 + ATA6289 15455 + ATtiny13A 15456 + ATtiny87 15457 + ATtiny167 15458 + ATtiny327 15459 + ATmega8C1 15460 + ATmega16C1 15461 + ATmega32C1 15462 + ATmega8M1 15463 + ATmega16M1 15464 + ATmega32M1 15465 + ATmega32U4 15466 + ATmega16HVB 15467 + ATmega4HVD 15468 + ATmega8HVD 15469 + ATmega64C1 15470 + ATmega64M1 15471 + ATmega16U4 15472 + ATmega32U6 15473 + ATmega128RFA1 15474 + AT90PWM81 15475 + AT90SCR100 15476 + M3000F 15477 + M3000S 15478 + M3001B 15479 15480 IA-32/x86-64 15481 15482 * Support for Intel AES built-in functions and code generation is 15483 available via -maes. 15484 * Support for Intel PCLMUL built-in function and code generation is 15485 available via -mpclmul. 15486 * Support for Intel AVX built-in functions and code generation is 15487 available via -mavx. 15488 * Automatically align the stack for local variables with alignment 15489 requirement. 15490 * GCC can now utilize the SVML library for vectorizing calls to a set 15491 of C99 functions if -mveclibabi=svml is specified and you link to 15492 an SVML ABI compatible library. 15493 * On x86-64, the ABI has been changed in the following cases to 15494 conform to the x86-64 ABI: 15495 + Passing/returning structures with flexible array member: 15496 struct foo 15497 { 15498 int i; 15499 int flex[]; 15500 }; 15501 + Passing/returning structures with complex float member: 15502 struct foo 15503 { 15504 int i; 15505 __complex__ float f; 15506 }; 15507 + Passing/returning unions with long double member: 15508 union foo 15509 { 15510 int x; 15511 long double ld; 15512 }; 15513 Code built with previous versions of GCC that uses any of these is 15514 not compatible with code built with GCC 4.4.0 or later. 15515 * A new target attribute was added to allow programmers to change the 15516 target options like -msse2 or -march=k8 for an individual function. 15517 You can also change the target options via the GCC target pragma 15518 for functions defined after the pragma. 15519 * GCC can now be configured with options --with-arch-32, 15520 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 15521 --with-tune-64 to control the default optimization separately for 15522 32-bit and 64-bit modes. 15523 15524 IA-32/IA64 15525 15526 * Support for __float128 (TFmode) IEEE quad type and corresponding 15527 TCmode IEEE complex quad type is available via the soft-fp library 15528 on IA-32/IA64 targets. This includes basic arithmetic operations 15529 (addition, subtraction, negation, multiplication and division) on 15530 __float128 real and TCmode complex values, the full set of IEEE 15531 comparisons between __float128 values, conversions to and from 15532 float, double and long double floating point types, as well as 15533 conversions to and from signed or unsigned integer, signed or 15534 unsigned long integer and signed or unsigned quad (TImode, IA64 15535 only) integer types. Additionally, all operations generate the full 15536 set of IEEE exceptions and support the full set of IEEE rounding 15537 modes. 15538 15539 M68K/ColdFire 15540 15541 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 15542 processors. (Scheduling support for ColdFire V2 processors was 15543 added in GCC 4.3.) 15544 * GCC now supports the -mxgot option to support programs requiring 15545 many GOT entries on ColdFire. 15546 * The m68k-*-linux-gnu target now builds multilibs by default. 15547 15548 MIPS 15549 15550 * MIPS Technologies have extended the original MIPS SVR4 ABI to 15551 include support for procedure linkage tables (PLTs) and copy 15552 relocations. These extensions allow GNU/Linux executables to use a 15553 significantly more efficient code model than the one defined by the 15554 original ABI. 15555 GCC support for this code model is available via a new command-line 15556 option, -mplt. There is also a new configure-time option, 15557 --with-mips-plt, to make -mplt the default. 15558 The new code model requires support from the assembler, the linker, 15559 and the runtime C library. This support is available in binutils 15560 2.19 and GLIBC 2.9. 15561 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 15562 and 32-bit GNU/Linux shared libraries. This feature requires GNU 15563 binutils 2.19 or above. 15564 * Support for RMI's XLR processor is now available through the 15565 -march=xlr and -mtune=xlr options. 15566 * 64-bit targets can now perform 128-bit multiplications inline, 15567 instead of relying on a libgcc function. 15568 * Native GNU/Linux toolchains now support -march=native and 15569 -mtune=native, which select the host processor. 15570 * GCC now supports the R10K, R12K, R14K and R16K processors. The 15571 canonical -march= and -mtune= names for these processors are 15572 r10000, r12000, r14000 and r16000 respectively. 15573 * GCC can now work around the side effects of speculative execution 15574 on R10K processors. Please see the documentation of the 15575 -mr10k-cache-barrier option for details. 15576 * Support for the MIPS64 Release 2 instruction set has been added. 15577 The option -march=mips64r2 enables generation of these 15578 instructions. 15579 * GCC now supports Cavium Networks' Octeon processor. This support is 15580 available through the -march=octeon and -mtune=octeon options. 15581 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 15582 canonical -march= and -mtune= names for these processors are 15583 loongson2e and loongson2f. 15584 15585 [12]picochip 15586 15587 Picochip is a 16-bit processor. A typical picoChip contains over 250 15588 small cores, each with small amounts of memory. There are three 15589 processor variants (STAN, MEM and CTRL) with different instruction sets 15590 and memory configurations and they can be chosen using the -mae option. 15591 15592 This port is intended to be a "C" only port. 15593 15594 Power Architecture and PowerPC 15595 15596 * GCC now supports the e300c2, e300c3 and e500mc processors. 15597 * GCC now supports Xilinx processors with a single-precision FPU. 15598 * Decimal floating point is now supported for e500 processors. 15599 15600 S/390, zSeries and System z9/z10 15601 15602 * Support for the IBM System z10 EC/BC processor has been added. When 15603 using the -march=z10 option, the compiler will generate code making 15604 use of instructions provided by the General-Instruction-Extension 15605 Facility and the Execute-Extension Facility. 15606 15607 VxWorks 15608 15609 * GCC now supports the thread-local storage mechanism used on 15610 VxWorks. 15611 15612 Xtensa 15613 15614 * GCC now supports thread-local storage (TLS) for Xtensa processor 15615 configurations that include the Thread Pointer option. TLS also 15616 requires support from the assembler and linker; this support is 15617 provided in the GNU binutils beginning with version 2.19. 15618 15619 Documentation improvements 15620 15621 Other significant improvements 15622 15623 [13]GCC 4.4.1 15624 15625 This is the [14]list of problem reports (PRs) from GCC's bug tracking 15626 system that are known to be fixed in the 4.4.1 release. This list might 15627 not be complete (that is, it is possible that some PRs that have been 15628 fixed are not listed here). 15629 15630 [15]GCC 4.4.2 15631 15632 This is the [16]list of problem reports (PRs) from GCC's bug tracking 15633 system that are known to be fixed in the 4.4.2 release. This list might 15634 not be complete (that is, it is possible that some PRs that have been 15635 fixed are not listed here). 15636 15637 [17]GCC 4.4.3 15638 15639 This is the [18]list of problem reports (PRs) from GCC's bug tracking 15640 system that are known to be fixed in the 4.4.3 release. This list might 15641 not be complete (that is, it is possible that some PRs that have been 15642 fixed are not listed here). 15643 15644 [19]GCC 4.4.4 15645 15646 This is the [20]list of problem reports (PRs) from GCC's bug tracking 15647 system that are known to be fixed in the 4.4.4 release. This list might 15648 not be complete (that is, it is possible that some PRs that have been 15649 fixed are not listed here). 15650 15651 [21]GCC 4.4.5 15652 15653 This is the [22]list of problem reports (PRs) from GCC's bug tracking 15654 system that are known to be fixed in the 4.4.5 release. This list might 15655 not be complete (that is, it is possible that some PRs that have been 15656 fixed are not listed here). 15657 15658 [23]GCC 4.4.6 15659 15660 This is the [24]list of problem reports (PRs) from GCC's bug tracking 15661 system that are known to be fixed in the 4.4.6 release. This list might 15662 not be complete (that is, it is possible that some PRs that have been 15663 fixed are not listed here). 15664 15665 [25]GCC 4.4.7 15666 15667 This is the [26]list of problem reports (PRs) from GCC's bug tracking 15668 system that are known to be fixed in the 4.4.7 release. This list might 15669 not be complete (that is, it is possible that some PRs that have been 15670 fixed are not listed here). 15671 15672 15673 For questions related to the use of GCC, please consult these web 15674 pages and the [27]GCC manuals. If that fails, the 15675 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15676 web pages and the development of GCC are welcome on our developer 15677 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 15678 archives. 15679 15680 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 15681 distribution of this entire article is permitted in any medium, 15682 provided this notice is preserved. 15683 15684 These pages are [32]maintained by the GCC team. Last modified 15685 2025-01-31. 15686 15687 References 15688 15689 1. https://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 15690 2. https://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 15691 3. https://gcc.gnu.org/gcc-4.4/porting_to.html 15692 4. https://gcc.gnu.org/wiki/Graphite 15693 5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 15694 6. https://gcc.gnu.org/gcc-4.4/cxx0x_status.html 15695 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 15696 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 15697 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 15698 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 15699 11. https://gcc.gnu.org/gcc-4.4/changes.html#targets 15700 12. https://gcc.gnu.org/gcc-4.4/changes.html#picochip 15701 13. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.1 15702 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 15703 15. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.2 15704 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 15705 17. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.3 15706 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 15707 19. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.4 15708 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 15709 21. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.5 15710 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 15711 23. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.6 15712 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 15713 25. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.7 15714 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 15715 27. https://gcc.gnu.org/onlinedocs/ 15716 28. mailto:gcc-help (a] gcc.gnu.org 15717 29. mailto:gcc (a] gcc.gnu.org 15718 30. https://gcc.gnu.org/lists.html 15719 31. https://www.fsf.org/ 15720 32. https://gcc.gnu.org/about.html 15721 ====================================================================== 15722 http://gcc.gnu.org/gcc-4.3/index.html 15723 GCC 4.3 Release Series 15724 15725 (This release series is no longer supported.) 15726 15727 Jun 27, 2011 15728 15729 The [1]GNU project and the GCC developers are pleased to announce the 15730 release of GCC 4.3.6. 15731 15732 This release is a bug-fix release, containing fixes for regressions in 15733 GCC 4.3.5 relative to previous releases of GCC. 15734 15735 Release History 15736 15737 GCC 4.3.6 15738 Jun 27, 2011 ([2]changes) 15739 15740 GCC 4.3.5 15741 May 22, 2010 ([3]changes) 15742 15743 GCC 4.3.4 15744 August 4, 2009 ([4]changes) 15745 15746 GCC 4.3.3 15747 January 24, 2009 ([5]changes) 15748 15749 GCC 4.3.2 15750 August 27, 2008 ([6]changes) 15751 15752 GCC 4.3.1 15753 June 6, 2008 ([7]changes) 15754 15755 GCC 4.3.0 15756 March 5, 2008 ([8]changes) 15757 15758 References and Acknowledgements 15759 15760 GCC used to stand for the GNU C Compiler, but since the compiler 15761 supports several other languages aside from C, it now stands for the 15762 GNU Compiler Collection. 15763 15764 A list of [9]successful builds is updated as new information becomes 15765 available. 15766 15767 The GCC developers would like to thank the numerous people that have 15768 contributed new features, improvements, bug fixes, and other changes as 15769 well as test results to GCC. This [10]amazing group of volunteers is 15770 what makes GCC successful. 15771 15772 For additional information about GCC please refer to the [11]GCC 15773 project web site or contact the [12]GCC development mailing list. 15774 15775 To obtain GCC please use [13]our mirror sites or [14]our version 15776 control system. 15777 15778 15779 For questions related to the use of GCC, please consult these web 15780 pages and the [15]GCC manuals. If that fails, the 15781 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15782 web pages and the development of GCC are welcome on our developer 15783 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 15784 archives. 15785 15786 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 15787 distribution of this entire article is permitted in any medium, 15788 provided this notice is preserved. 15789 15790 These pages are [20]maintained by the GCC team. Last modified 15791 2022-10-26. 15792 15793 References 15794 15795 1. http://www.gnu.org/ 15796 2. https://gcc.gnu.org/gcc-4.3/changes.html 15797 3. https://gcc.gnu.org/gcc-4.3/changes.html 15798 4. https://gcc.gnu.org/gcc-4.3/changes.html 15799 5. https://gcc.gnu.org/gcc-4.3/changes.html 15800 6. https://gcc.gnu.org/gcc-4.3/changes.html 15801 7. https://gcc.gnu.org/gcc-4.3/changes.html 15802 8. https://gcc.gnu.org/gcc-4.3/changes.html 15803 9. https://gcc.gnu.org/gcc-4.3/buildstat.html 15804 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15805 11. https://gcc.gnu.org/index.html 15806 12. mailto:gcc (a] gcc.gnu.org 15807 13. https://gcc.gnu.org/mirrors.html 15808 14. https://gcc.gnu.org/git.html 15809 15. https://gcc.gnu.org/onlinedocs/ 15810 16. mailto:gcc-help (a] gcc.gnu.org 15811 17. mailto:gcc (a] gcc.gnu.org 15812 18. https://gcc.gnu.org/lists.html 15813 19. https://www.fsf.org/ 15814 20. https://gcc.gnu.org/about.html 15815 ====================================================================== 15816 http://gcc.gnu.org/gcc-4.3/changes.html 15817 GCC 4.3 Release Series 15818 Changes, New Features, and Fixes 15819 15820 The latest release in the 4.3 release series is [1]GCC 4.3.5. 15821 15822 Caveats 15823 15824 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 15825 various front-end languages it supports. See the [4]prerequisites 15826 page for version requirements. 15827 * ColdFire targets now treat long double as having the same format as 15828 double. In earlier versions of GCC, they used the 68881 long double 15829 format instead. 15830 * The m68k-uclinux target now uses the same calling conventions as 15831 m68k-linux-gnu. You can select the original calling conventions by 15832 configuring for m68k-uclinuxoldabi instead. Note that 15833 m68k-uclinuxoldabi also retains the original 80-bit long double on 15834 ColdFire targets. 15835 * The -fforce-mem option has been removed because it has had no 15836 effect in the last few GCC releases. 15837 * The i386 -msvr3-shlib option has been removed since it is no longer 15838 used. 15839 * Fastcall for i386 has been changed not to pass aggregate arguments 15840 in registers, following Microsoft compilers. 15841 * Support for the AOF assembler has been removed from the ARM back 15842 end; this affects only the targets arm-semi-aof and armel-semi-aof, 15843 which are no longer recognized. We removed these targets without a 15844 deprecation period because we discovered that they have been 15845 unusable since GCC 4.0.0. 15846 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 15847 has been removed. This support had been deprecated since GCC 4.0.0. 15848 * Support for a number of older systems and recently unmaintained or 15849 untested target ports of GCC has been declared obsolete in GCC 4.3. 15850 Unless there is activity to revive them, the next release of GCC 15851 will have their sources permanently removed. 15852 All GCC ports for the following processor architectures have been 15853 declared obsolete: 15854 + Morpho MT (mt-*) 15855 The following aliases for processor architectures have been 15856 declared obsolete. Users should use the indicated generic target 15857 names instead, with compile-time options such as -mcpu or 15858 configure-time options such as --with-cpu to control the 15859 configuration more precisely. 15860 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 15861 instead). 15862 + parisc*-*-* (use hppa*-*-* instead). 15863 + m680[012]0-*-* (use m68k-*-* instead). 15864 All GCC ports for the following operating systems have been 15865 declared obsolete: 15866 + BeOS (*-*-beos*) 15867 + kaOS (*-*-kaos*) 15868 + GNU/Linux using the a.out object format (*-*-linux*aout*) 15869 + GNU/Linux using version 1 of the GNU C Library 15870 (*-*-linux*libc1*) 15871 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 15872 *-*-solaris2.[0-6].*) 15873 + Miscellaneous System V (*-*-sysv*) 15874 + WindISS (*-*-windiss*) 15875 Also, those for some individual systems on particular architectures 15876 have been obsoleted: 15877 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 15878 + CRIS with a.out object format (cris-*-aout) 15879 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 15880 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 15881 + PRO on PA-RISC (hppa1.1-*-pro*) 15882 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 15883 i[34567]86-sequent-sysv4*) 15884 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 15885 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 15886 was previously [5]removed in 2001, leaving only the support 15887 for UWIN as a target now being deprecated) 15888 + ChorusOS on PowerPC (powerpc-*-chorusos*) 15889 + All VAX configurations apart from NetBSD and OpenBSD 15890 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 15891 * The [6]-Wconversion option has been modified. Its purpose now is to 15892 warn for implicit conversions that may alter a value. This new 15893 behavior is available for both C and C++. Warnings about 15894 conversions between signed and unsigned integers can be disabled by 15895 using -Wno-sign-conversion. In C++, they are disabled by default 15896 unless -Wsign-conversion is explicitly requested. The old behavior 15897 of -Wconversion, that is, warn for prototypes causing a type 15898 conversion that is different from what would happen to the same 15899 argument in the absence of a prototype, has been moved to a new 15900 option -Wtraditional-conversion, which is only available for C. 15901 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 15902 been removed because they were deprecated for more than 3 GCC major 15903 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 15904 -mtune=pentiumpro as a replacement. 15905 * The -funsafe-math-optimizations option now automatically turns on 15906 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 15907 reassociation and thus may introduce or remove traps. 15908 * The -ftree-vectorize option is now on by default under -O3. In 15909 order to generate code for a SIMD extension, it has to be enabled 15910 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 15911 i?86 and x86_64. 15912 * More information on porting to GCC 4.3 from previous versions of 15913 GCC can be found in the [7]porting guide for this release. 15914 15915 General Optimizer Improvements 15916 15917 * The GCC middle-end has been integrated with the MPFR library. This 15918 allows GCC to evaluate and replace at compile-time calls to 15919 built-in math functions having constant arguments with their 15920 mathematically equivalent results. In making use of MPFR, GCC can 15921 generate correct results regardless of the math library 15922 implementation or floating point precision of the host platform. 15923 This also allows GCC to generate identical results regardless of 15924 whether one compiles in native or cross-compile configurations to a 15925 particular target. The following built-in functions take advantage 15926 of this new capability: acos, acosh, asin, asinh, atan2, atan, 15927 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 15928 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 15929 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 15930 tan, tanh, tgamma, y0, y1 and yn. The float and long double 15931 variants of these functions (e.g. sinf and sinl) are also handled. 15932 The sqrt and cabs functions with constant arguments were already 15933 optimized in prior GCC releases. Now they also use MPFR. 15934 * A new forward propagation pass on RTL was added. The new pass 15935 replaces several slower transformations, resulting in compile-time 15936 improvements as well as better code generation in some cases. 15937 * A new command-line switch -frecord-gcc-switches has been added to 15938 GCC, although it is only enabled for some targets. The switch 15939 causes the command line that was used to invoke the compiler to be 15940 recorded into the object file that is being created. The exact 15941 format of this recording is target and binary file format 15942 dependent, but it usually takes the form of a note section 15943 containing ASCII text. The switch is related to the -fverbose-asm 15944 switch, but that one only records the information in the assembler 15945 output file as comments, so the information never reaches the 15946 object file. 15947 * The inliner heuristic is now aware of stack frame consumption. New 15948 command-line parameters --param large-stack-frame and --param 15949 large-stack-frame-growth can be used to limit stack frame size 15950 growth caused by inlining. 15951 * During feedback directed optimizations, the expected block size the 15952 memcpy, memset and bzero functions operate on is discovered and for 15953 cases of commonly used small sizes, specialized inline code is 15954 generated. 15955 * __builtin_expect no longer requires its argument to be a compile 15956 time constant. 15957 * Interprocedural optimization was reorganized to work on functions 15958 in SSA form. This enables more precise and cheaper dataflow 15959 analysis and makes writing interprocedural optimizations easier. 15960 The following improvements have been implemented on top of this 15961 framework: 15962 + Pre-inline optimization: Selected local optimization passes 15963 are run before the inliner (and other interprocedural passes) 15964 are executed. This significantly improves the accuracy of code 15965 growth estimates used by the inliner and reduces the overall 15966 memory footprint for large compilation units. 15967 + Early inlining (a simple bottom-up inliner pass inlining only 15968 functions whose body is smaller than the expected call 15969 overhead) is now executed with the early optimization passes, 15970 thus inlining already optimized function bodies into an 15971 unoptimized function that is subsequently optimized by early 15972 optimizers. This enables the compiler to quickly eliminate 15973 abstraction penalty in C++ programs. 15974 + Interprocedural constant propagation now operate on SSA form 15975 increasing accuracy of the analysis. 15976 * A new internal representation for GIMPLE statements has been 15977 contributed, resulting in compile-time memory savings. 15978 * The vectorizer was enhanced to support vectorization of outer 15979 loops, intra-iteration parallelism (loop-aware SLP), vectorization 15980 of strided accesses and loops with multiple data-types. Run-time 15981 dependency testing using loop versioning was added. The cost model, 15982 turned on by -fvect-cost-model, was developed. 15983 15984 New Languages and Language specific improvements 15985 15986 * We have added new command-line options 15987 -finstrument-functions-exclude-function-list and 15988 -finstrument-functions-exclude-file-list. They provide more control 15989 over which functions are annotated by the -finstrument-functions 15990 option. 15991 15992 C family 15993 15994 * Implicit conversions between generic vector types are now only 15995 permitted when the two vectors in question have the same number of 15996 elements and compatible element types. (Note that the restriction 15997 involves compatible element types, not implicitly-convertible 15998 element types: thus, a vector type with element type int may not be 15999 implicitly converted to a vector type with element type unsigned 16000 int.) This restriction, which is in line with specifications for 16001 SIMD architectures such as AltiVec, may be relaxed using the flag 16002 -flax-vector-conversions. This flag is intended only as a 16003 compatibility measure and should not be used for new code. 16004 * -Warray-bounds has been added and is now enabled by default for 16005 -Wall . It produces warnings for array subscripts that can be 16006 determined at compile time to be always out of bounds. 16007 -Wno-array-bounds will disable the warning. 16008 * The constructor and destructor function attributes now accept 16009 optional priority arguments which control the order in which the 16010 constructor and destructor functions are run. 16011 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 16012 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 16013 -Wignored-qualifiers have been added for finer control of the 16014 diverse warnings enabled by -Wextra. 16015 * A new function attribute alloc_size has been added to mark up 16016 malloc style functions. For constant sized allocations this can be 16017 used to find out the size of the returned pointer using the 16018 __builtin_object_size() function for buffer overflow checking and 16019 similar. This supplements the already built-in malloc and calloc 16020 constant size handling. 16021 * Integer constants written in binary are now supported as a GCC 16022 extension. They consist of a prefix 0b or 0B, followed by a 16023 sequence of 0 and 1 digits. 16024 * A new predefined macro __COUNTER__ has been added. It expands to 16025 sequential integral values starting from 0. In conjunction with the 16026 ## operator, this provides a convenient means to generate unique 16027 identifiers. 16028 * A new command-line option -fdirectives-only has been added. It 16029 enables a special preprocessing mode which improves the performance 16030 of applications like distcc and ccache. 16031 * Fixed-point data types and operators have been added. They are 16032 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 16033 Currently, only MIPS targets are supported. 16034 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 16035 N1241, is now supported as a GCC extension to C for targets 16036 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 16037 and x86_64-*-linux-gnu. The feature introduces new data types 16038 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 16039 DD, and DL. 16040 16041 C++ 16042 16043 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 16044 * -Wc++0x-compat has been added and is now enabled by default for 16045 -Wall. It produces warnings for constructs whose meaning differs 16046 between ISO C++ 1998 and C++0x. 16047 * The -Wparentheses option now works for C++ as it does for C. It 16048 warns if parentheses are omitted when operators with confusing 16049 precedence are nested. It also warns about ambiguous else 16050 statements. Since -Wparentheses is enabled by -Wall, this may cause 16051 additional warnings with existing C++ code which uses -Wall. These 16052 new warnings may be disabled by using -Wall -Wno-parentheses. 16053 * The -Wmissing-declarations now works for C++ as it does for C. 16054 * The -fvisibility-ms-compat flag was added, to make it easier to 16055 port larger projects using shared libraries from Microsoft's Visual 16056 Studio to ELF and Mach-O systems. 16057 * C++ attribute handling has been overhauled for template arguments 16058 (ie dependent types). In particular, __attribute__((aligned(T))); 16059 works for C++ types. 16060 16061 Runtime Library (libstdc++) 16062 16063 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 16064 * Support for TR1 mathematical special functions and regular 16065 expressions. 16066 * Default what implementations give more elaborate exception strings 16067 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 16068 * Header dependencies have been streamlined, reducing unnecessary 16069 includes and pre-processed bloat. 16070 * Variadic template implementations of items in <tuple> and 16071 <functional>. 16072 * An experimental [11]parallel mode has been added. This is a 16073 parallel implementation of many C++ Standard library algorithms, 16074 like std::accumulate, std::for_each, std::transform, or std::sort, 16075 to give but four examples. These algorithms can be substituted for 16076 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 16077 or all existing algorithms can be transformed via the 16078 -D_GLIBCXX_PARALLEL macro. 16079 * Debug mode versions of classes in <unordered_set> and 16080 <unordered_map>. 16081 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 16082 now <backward/hash_set> and <backward/hash_map>. This code: 16083 #include <ext/hash_set> 16084 __gnu_cxx::hash_set<int> s; 16085 16086 Can be transformed (in order of preference) to: 16087 #include <tr1/unordered_set> 16088 std::tr1::unordered_set<int> s; 16089 16090 or 16091 #include <backward/hash_set> 16092 __gnu_cxx::hash_set<int> s; 16093 16094 Similar transformations apply to __gnu_cxx::hash_map, 16095 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 16096 __gnu_cxx::hash_multiset. 16097 16098 Fortran 16099 16100 * Due to the fact that the GMP and MPFR libraries are required for 16101 all languages, Fortran is no longer special in this regard and is 16102 available by default. 16103 * The [12]-fexternal-blas option has been added, which generates 16104 calls to BLAS routines for intrinsic matrix operations such as 16105 matmul rather than using the built-in algorithms. 16106 * Support to give a backtrace (compiler flag -fbacktrace or 16107 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 16108 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 16109 run-time error occured. 16110 * GNU Fortran now defines __GFORTRAN__ when it runs the C 16111 preprocessor (CPP). 16112 * The [13]-finit-local-zero, -finit-real, -finit-integer, 16113 -finit-character, and -finit-logical options have been added, which 16114 can be used to initialize local variables. 16115 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 16116 which calculate the Gamma function and its logarithm. Use EXTERNAL 16117 gamma if you want to use your own gamma function. 16118 * GNU Fortran now regards the backslash character as literal (as 16119 required by the Fortran 2003 standard); using [16]-fbackslash GNU 16120 Fortran interprets backslashes as C-style escape characters. 16121 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 16122 literal constants has been changed. Before they were always 16123 interpreted as integer; now they are bit-wise transferred as 16124 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 16125 2003 standard, and for real and complex variables in DATA 16126 statements or when directly assigned to real and complex variables. 16127 Everywhere else and especially in expressions they are still 16128 regarded as integer constants. 16129 * Fortran 2003 support has been extended: 16130 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 16131 + Pointer intent 16132 + Intrinsic module ISO_ENV_FORTRAN 16133 + Interoperability with C (ISO C Bindings) 16134 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 16135 attribute) 16136 + Fortran 2003 BOZ 16137 16138 Java (GCJ) 16139 16140 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 16141 This enables the use of all 1.5 language features, and fixes most 16142 existing front end bugs. 16143 * libgcj now supports all 1.5 language features which require runtime 16144 support: foreach, enum, annotations, generics, and auto-boxing. 16145 * We've made many changes to the tools shipped with gcj. 16146 + The old jv-scan tool has been removed. This tool never really 16147 worked properly. There is no replacement. 16148 + gcjh has been rewritten. Some of its more obscure options no 16149 longer work, but are still recognized in an attempt at 16150 compatibility. gjavah is a new program with similar 16151 functionality but different command-line options. 16152 + grmic and grmiregistry have been rewritten. grmid has been 16153 added. 16154 + gjar replaces the old fastjar. 16155 + gjarsigner (used for signing jars), gkeytool (used for key 16156 management), gorbd (for CORBA), gserialver (computes 16157 serialization UIDs), and gtnameserv (also for CORBA) are now 16158 installed. 16159 * The ability to dump the contents of the java run time heap to a 16160 file for off-line analysis has been added. The heap dumps may be 16161 analyzed with the new gc-analyze tool. They may be generated on 16162 out-of-memory conditions or on demand and are controlled by the new 16163 run time class gnu.gcj.util.GCInfo. 16164 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 16165 provide correct, updated, timezone information. This means that 16166 packagers no longer have to update libgcj when a time zone change 16167 is published. 16168 16169 [18]New Targets and Target Specific Improvements 16170 16171 IA-32/x86-64 16172 16173 * Tuning for Intel Core 2 processors is available via -mtune=core2 16174 and -march=core2. 16175 * Tuning for AMD Geode processors is available via -mtune=geode and 16176 -march=geode. 16177 * Code generation of block move (memcpy) and block set (memset) was 16178 rewritten. GCC can now pick the best algorithm (loop, unrolled 16179 loop, instruction with rep prefix or a library call) based on the 16180 size of the block being copied and the CPU being optimized for. A 16181 new option -minline-stringops-dynamically has been added. With this 16182 option string operations of unknown size are expanded such that 16183 small blocks are copied by in-line code, while for large blocks a 16184 library call is used. This results in faster code than 16185 -minline-all-stringops when the library implementation is capable 16186 of using cache hierarchy hints. The heuristic choosing the 16187 particular algorithm can be overwritten via -mstringop-strategy. 16188 Newly also memset of values different from 0 is inlined. 16189 * GCC no longer places the cld instruction before string operations. 16190 Both i386 and x86-64 ABI documents mandate the direction flag to be 16191 clear at the entry of a function. It is now invalid to set the flag 16192 in asm statement without reseting it afterward. 16193 * Support for SSSE3 built-in functions and code generation are 16194 available via -mssse3. 16195 * Support for SSE4.1 built-in functions and code generation are 16196 available via -msse4.1. 16197 * Support for SSE4.2 built-in functions and code generation are 16198 available via -msse4.2. 16199 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 16200 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 16201 allow explicit control of x87 floating point precision. 16202 * Support for __float128 (TFmode) IEEE quad type and corresponding 16203 TCmode IEEE complex quad type is available via the soft-fp library 16204 on x86_64 targets. This includes basic arithmetic operations 16205 (addition, subtraction, negation, multiplication and division) on 16206 __float128 real and TCmode complex values, the full set of IEEE 16207 comparisons between __float128 values, conversions to and from 16208 float, double and long double floating point types, as well as 16209 conversions to and from signed or unsigned integer, signed or 16210 unsigned long integer and signed or unsigned quad (TImode) integer 16211 types. Additionally, all operations generate the full set of IEEE 16212 exceptions and support the full set of IEEE rounding modes. 16213 * GCC can now utilize the ACML library for vectorizing calls to a set 16214 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 16215 link to an ACML ABI compatible library. 16216 16217 ARM 16218 16219 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 16220 has been added. 16221 16222 [19]CRIS 16223 16224 New features 16225 16226 * Compiler and Library support for the CRIS v32 architecture, as 16227 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 16228 added. 16229 16230 Configuration changes 16231 16232 * The cris-*-elf target now includes support for CRIS v32, including 16233 libraries, through the -march=v32 option. 16234 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 16235 * A new crisv32-*-linux* target defaults to generate code for CRIS 16236 v32. 16237 * The cris-*-aout target has been obsoleted. 16238 16239 Improved support for built-in functions 16240 16241 * GCC can now use the lz and swapwbr instructions to implement the 16242 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 16243 * __builtin_bswap32 is now implemented using the swapwb instruction, 16244 when available. 16245 16246 [20]m68k and ColdFire 16247 16248 New features 16249 16250 * Support for several new ColdFire processors has been added. You can 16251 generate code for them using the new -mcpu option. 16252 * All targets now support ColdFire processors. 16253 * m68k-uclinux targets have improved support for C++ constructors and 16254 destructors, and for shared libraries. 16255 * It is now possible to set breakpoints on the first or last line of 16256 a function, even if there are no statements on that line. 16257 16258 Optimizations 16259 16260 * Support for sibling calls has been added. 16261 * More use is now made of the ColdFire mov3q instruction. 16262 * __builtin_clz is now implemented using the ff1 ColdFire 16263 instruction, when available. 16264 * GCC now honors the -m68010 option. 68010 code now uses clr rather 16265 than move to zero volatile memory. 16266 * 68020 targets and above can now use symbol(index.size*scale) 16267 addresses for indexed array accesses. Earlier compilers would 16268 always load the symbol into a base register first. 16269 16270 Configuration changes 16271 16272 * All m68k and ColdFire targets now allow the default processor to be 16273 set at configure time using --with-cpu. 16274 * A --with-arch configuration option has been added. This option 16275 allows you to restrict a target to ColdFire or non-ColdFire 16276 processors. 16277 16278 Preprocessor macros 16279 16280 * An __mcfv*__ macro is now defined for all ColdFire targets. 16281 (Earlier versions of GCC only defined __mcfv4e__.) 16282 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 16283 * All targets now define __mc68010 and __mc68010__ when generating 16284 68010 code. 16285 16286 Command-line changes 16287 16288 * New command-line options -march, -mcpu, -mtune and -mhard-float 16289 have been added. These options apply to both m68k and ColdFire 16290 targets. 16291 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 16292 versions of -mshort, etc. 16293 * -fforce-addr has been removed. It is now ignored by the compiler. 16294 16295 Other improvements 16296 16297 * ColdFire targets now try to maintain a 4-byte-aligned stack where 16298 possible. 16299 * m68k-uclinux targets now try to avoid situations that lead to the 16300 load-time error: BINFMT_FLAT: reloc outside program. 16301 16302 MIPS 16303 16304 Changes to existing configurations 16305 16306 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 16307 and n64. Every GNU/Linux configuration now builds these libraries 16308 by default. 16309 * GNU/Linux configurations now generate -mno-shared code unless 16310 overridden by -fpic, -fPIC, -fpie or -fPIE. 16311 * mipsisa32*-linux-gnu configurations now generate hard-float code by 16312 default, just like other mipsisa32* and mips*-linux-gnu 16313 configurations. You can build a soft-float version of any 16314 mips*-linux-gnu configuration by passing --with-float=soft to 16315 configure. 16316 * mips-wrs-vxworks now supports run-time processes (RTPs). 16317 16318 Changes to existing command-line options 16319 16320 * The -march and -mtune options no longer accept 24k as a processor 16321 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 16322 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 16323 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 16324 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 16325 24kx, 24kex and 34kx. 16326 16327 New configurations 16328 16329 GCC now supports the following configurations: 16330 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 16331 default. Earlier releases also recognized this configuration, but 16332 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 16333 you can customize any mips*-linux-gnu* configuration to a 16334 particular ISA or processor by passing an appropriate --with-arch 16335 option to configure. 16336 * mipsisa*-sde-elf*, which provides compatibility with MIPS 16337 Technologies' SDE toolchains. The configuration uses the SDE 16338 libraries by default, but you can use it like other newlib-based 16339 ELF configurations by passing --with-newlib to configure. It is the 16340 only configuration besides mips64vr*-elf* to build MIPS16 as well 16341 as non-MIPS16 libraries. 16342 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 16343 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 16344 and 64-bit forms of the EABI. 16345 16346 New processors and application-specific extensions 16347 16348 * Support for the SmartMIPS ASE is available through the new 16349 -msmartmips option. 16350 * Support for revision 2 of the DSP ASE is available through the new 16351 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 16352 indicates the revision of the ASE in use. 16353 * Support for the 4KS and 74K families of processors is available 16354 through the -march and -mtune options. 16355 16356 Improved support for built-in functions 16357 16358 * GCC can now use load-linked, store-conditional and sync 16359 instructions to implement atomic built-in functions such as 16360 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 16361 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 16362 * GCC can now use the clz and dclz instructions to implement the 16363 __builtin_ctz and __builtin_ffs families of functions. 16364 * There is a new __builtin___clear_cache function for flushing the 16365 instruction cache. GCC expands this function inline on MIPS32 16366 revision 2 targets, otherwise it calls the function specified by 16367 -mcache-flush-func. 16368 16369 MIPS16 improvements 16370 16371 * GCC can now compile objects that contain a mixture of MIPS16 and 16372 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 16373 for specifying which mode a function should use. 16374 * A new option called -minterlink-mips16 makes non-MIPS16 code 16375 link-compatible with MIPS16 code. 16376 * After many bug fixes, the long-standing MIPS16 -mhard-float support 16377 should now work fairly reliably. 16378 * GCC can now use the MIPS16e save and restore instructions. 16379 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 16380 with -G0 -fsection-anchors is often smaller than code compiled with 16381 -G8. However, please note that you must usually compile all objects 16382 in your application with the same -G option; see the documentation 16383 of -G for details. 16384 * A new option called-mcode-readable specifies which instructions are 16385 allowed to load from the code segment. -mcode-readable=yes is the 16386 default and says that any instruction may load from the code 16387 segment. The other alternatives are -mcode-readable=pcrel, which 16388 says that only PC-relative MIPS16 instructions may load from the 16389 code segment, and -mcode-readable=no, which says that no 16390 instruction may do so. Please see the documentation for more 16391 details, including example uses. 16392 16393 Small-data improvements 16394 16395 There are three new options for controlling small data: 16396 * -mno-extern-sdata, which disables small-data accesses for 16397 externally-defined variables. Code compiled with -Gn 16398 -mno-extern-sdata will be link-compatible with any -G setting 16399 between -G0 and -Gn inclusive. 16400 * -mno-local-sdata, which disables the use of small-data sections for 16401 data that is not externally visible. This option can be a useful 16402 way of reducing small-data usage in less performance-critical parts 16403 of an application. 16404 * -mno-gpopt, which disables the use of the $gp register while still 16405 honoring the -G limit when placing externally-visible data. This 16406 option implies -mno-extern-sdata and -mno-local-sdata and it can be 16407 useful in situations where $gp does not necessarily hold the 16408 expected value. 16409 16410 Miscellaneous improvements 16411 16412 * There is a new option called -mbranch-cost for tweaking the 16413 perceived cost of branches. 16414 * If GCC is configured to use a version of GAS that supports the 16415 .gnu_attribute directive, it will use that directive to record 16416 certain properties of the output code. .gnu_attribute is new to GAS 16417 2.18. 16418 * There are two new function attributes, near and far, for overriding 16419 the command-line setting of -mlong-calls on a function-by-function 16420 basis. 16421 * -mfp64, which previously required a 64-bit target, now works with 16422 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 16423 mipsisa*-sde-elf* configurations provide suitable library support. 16424 * GCC now recognizes the -mdmx and -mmt options and passes them down 16425 to the assembler. It does nothing else with the options at present. 16426 16427 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 16428 (BEA) 16429 16430 * Support has been added for this new architecture. 16431 16432 RS6000 (POWER/PowerPC) 16433 16434 * Support for the PowerPC 750CL paired-single instructions has been 16435 added with a new powerpc-*-linux*paired* target configuration. It 16436 is enabled by an associated -mpaired option and can be accessed 16437 using new built-in functions. 16438 * Support for auto-detecting architecture and system configuration to 16439 auto-select processor optimization tuning. 16440 * Support for VMX on AIX 5.3 has been added. 16441 * Support for AIX Version 6.1 has been added. 16442 16443 S/390, zSeries and System z9 16444 16445 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 16446 added. When using the -march=z9-ec option, the compiler will 16447 generate code making use of instructions provided by the decimal 16448 floating point facility and the floating point conversion facility 16449 (pfpo). Besides the instructions used to implement decimal floating 16450 point operations these facilities also contain instructions to move 16451 between general purpose and floating point registers and to modify 16452 and copy the sign-bit of floating point values. 16453 * When the -march=z9-ec option is used the new 16454 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 16455 decimal floating point hardware instructions will be used or not. 16456 If none of them is given the hardware support is enabled by 16457 default. 16458 * The -mstack-guard option can now be omitted when using stack 16459 checking via -mstack-size in order to let GCC choose a sensible 16460 stack guard value according to the frame size of each function. 16461 * Various changes to improve performance of generated code have been 16462 implemented, including: 16463 + The condition code set by an add logical with carry 16464 instruction is now available for overflow checks like: a + b + 16465 carry < b. 16466 + The test data class instruction is now used to implement 16467 sign-bit and infinity checks of binary and decimal floating 16468 point numbers. 16469 16470 SPARC 16471 16472 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 16473 added. 16474 16475 Xtensa 16476 16477 * Stack unwinding for exception handling now uses by default a 16478 specialized version of DWARF unwinding. This is not 16479 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 16480 Xtensa with previous versions of GCC. 16481 * For Xtensa processors that include the Conditional Store option, 16482 the built-in functions for atomic memory access are now implemented 16483 using S32C1I instructions. 16484 * If the Xtensa NSA option is available, GCC will use it to implement 16485 the __builtin_ctz and __builtin_clz functions. 16486 16487 Documentation improvements 16488 16489 * Existing libstdc++ documentation has been edited and restructured 16490 into a single DocBook XML manual. The results can be viewed online 16491 [21]here. 16492 16493 Other significant improvements 16494 16495 * The compiler's --help command-line option has been extended so that 16496 it now takes an optional set of arguments. These arguments restrict 16497 the information displayed to specific classes of command-line 16498 options, and possibly only a subset of those options. It is also 16499 now possible to replace the descriptive text associated with each 16500 displayed option with an indication of its current value, or for 16501 binary options, whether it has been enabled or disabled. 16502 Here are some examples. The following will display all the options 16503 controlling warning messages: 16504 --help=warnings 16505 16506 Whereas this will display all the undocumented, target specific 16507 options: 16508 --help=target,undocumented 16509 16510 This sequence of commands will display the binary optimizations 16511 that are enabled by -O3: 16512 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 16513 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 16514 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 16515 16516 * The configure options --with-pkgversion and --with-bugurl have been 16517 added. These allow distributors of GCC to include a 16518 distributor-specific string in manuals and --version output and to 16519 specify the URL for reporting bugs in their versions of GCC. 16520 16521 [22]GCC 4.3.1 16522 16523 This is the [23]list of problem reports (PRs) from GCC's bug tracking 16524 system that are known to be fixed in the 4.3.1 release. This list might 16525 not be complete (that is, it is possible that some PRs that have been 16526 fixed are not listed here). 16527 16528 Target Specific Changes 16529 16530 IA-32/x86-64 16531 16532 ABI changes 16533 16534 * Starting with GCC 4.3.1, decimal floating point variables are 16535 aligned to their natural boundaries when they are passed on the 16536 stack for i386. 16537 16538 Command-line changes 16539 16540 * Starting with GCC 4.3.1, the -mcld option has been added to 16541 automatically generate a cld instruction in the prologue of 16542 functions that use string instructions. This option is used for 16543 backward compatibility on some operating systems and can be enabled 16544 by default for 32-bit x86 targets by configuring GCC with the 16545 --enable-cld configure option. 16546 16547 [24]GCC 4.3.2 16548 16549 This is the [25]list of problem reports (PRs) from GCC's bug tracking 16550 system that are known to be fixed in the 4.3.2 release. This list might 16551 not be complete (that is, it is possible that some PRs that have been 16552 fixed are not listed here). 16553 16554 [26]GCC 4.3.3 16555 16556 This is the [27]list of problem reports (PRs) from GCC's bug tracking 16557 system that are known to be fixed in the 4.3.3 release. This list might 16558 not be complete (that is, it is possible that some PRs that have been 16559 fixed are not listed here). 16560 16561 [28]GCC 4.3.4 16562 16563 This is the [29]list of problem reports (PRs) from GCC's bug tracking 16564 system that are known to be fixed in the 4.3.4 release. This list might 16565 not be complete (that is, it is possible that some PRs that have been 16566 fixed are not listed here). 16567 16568 [30]GCC 4.3.5 16569 16570 This is the [31]list of problem reports (PRs) from GCC's bug tracking 16571 system that are known to be fixed in the 4.3.5 release. This list might 16572 not be complete (that is, it is possible that some PRs that have been 16573 fixed are not listed here). 16574 16575 [32]GCC 4.3.6 16576 16577 This is the [33]list of problem reports (PRs) from GCC's bug tracking 16578 system that are known to be fixed in the 4.3.6 release. This list might 16579 not be complete (that is, it is possible that some PRs that have been 16580 fixed are not listed here). 16581 16582 16583 For questions related to the use of GCC, please consult these web 16584 pages and the [34]GCC manuals. If that fails, the 16585 [35]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16586 web pages and the development of GCC are welcome on our developer 16587 list at [36]gcc (a] gcc.gnu.org. All of [37]our lists have public 16588 archives. 16589 16590 Copyright (C) [38]Free Software Foundation, Inc. Verbatim copying and 16591 distribution of this entire article is permitted in any medium, 16592 provided this notice is preserved. 16593 16594 These pages are [39]maintained by the GCC team. Last modified 16595 2025-01-31. 16596 16597 References 16598 16599 1. https://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 16600 2. https://gmplib.org/ 16601 3. https://www.mpfr.org/ 16602 4. https://gcc.gnu.org/install/prerequisites.html 16603 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 16604 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 16605 7. https://gcc.gnu.org/gcc-4.3/porting_to.html 16606 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 16607 9. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 16608 10. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 16609 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 16610 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 16611 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 16612 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 16613 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 16614 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 16615 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 16616 18. https://gcc.gnu.org/gcc-4.3/changes.html#targets 16617 19. https://gcc.gnu.org/gcc-4.3/changes.html#cris 16618 20. https://gcc.gnu.org/gcc-4.3/changes.html#m68k 16619 21. https://gcc.gnu.org/onlinedocs/libstdc++/ 16620 22. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.1 16621 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 16622 24. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.2 16623 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 16624 26. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.3 16625 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 16626 28. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.4 16627 29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 16628 30. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.5 16629 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 16630 32. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.6 16631 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 16632 34. https://gcc.gnu.org/onlinedocs/ 16633 35. mailto:gcc-help (a] gcc.gnu.org 16634 36. mailto:gcc (a] gcc.gnu.org 16635 37. https://gcc.gnu.org/lists.html 16636 38. https://www.fsf.org/ 16637 39. https://gcc.gnu.org/about.html 16638 ====================================================================== 16639 http://gcc.gnu.org/gcc-4.2/index.html 16640 GCC 4.2 Release Series 16641 16642 (This release series is no longer supported.) 16643 16644 May 19, 2008 16645 16646 The [1]GNU project and the GCC developers are pleased to announce the 16647 release of GCC 4.2.4. 16648 16649 This release is a bug-fix release, containing fixes for regressions in 16650 GCC 4.2.3 relative to previous releases of GCC. 16651 16652 Release History 16653 16654 GCC 4.2.4 16655 May 19, 2008 ([2]changes) 16656 16657 GCC 4.2.3 16658 February 1, 2008 ([3]changes) 16659 16660 GCC 4.2.2 16661 October 7, 2007 ([4]changes) 16662 16663 GCC 4.2.1 16664 July 18, 2007 ([5]changes) 16665 16666 GCC 4.2.0 16667 May 13, 2007 ([6]changes) 16668 16669 References and Acknowledgements 16670 16671 GCC used to stand for the GNU C Compiler, but since the compiler 16672 supports several other languages aside from C, it now stands for the 16673 GNU Compiler Collection. 16674 16675 A list of [7]successful builds is updated as new information becomes 16676 available. 16677 16678 The GCC developers would like to thank the numerous people that have 16679 contributed new features, improvements, bug fixes, and other changes as 16680 well as test results to GCC. This [8]amazing group of volunteers is 16681 what makes GCC successful. 16682 16683 For additional information about GCC please refer to the [9]GCC project 16684 web site or contact the [10]GCC development mailing list. 16685 16686 To obtain GCC please use [11]our mirror sites or [12]our version 16687 control system. 16688 16689 16690 For questions related to the use of GCC, please consult these web 16691 pages and the [13]GCC manuals. If that fails, the 16692 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16693 web pages and the development of GCC are welcome on our developer 16694 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 16695 archives. 16696 16697 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 16698 distribution of this entire article is permitted in any medium, 16699 provided this notice is preserved. 16700 16701 These pages are [18]maintained by the GCC team. Last modified 16702 2022-10-26. 16703 16704 References 16705 16706 1. http://www.gnu.org/ 16707 2. https://gcc.gnu.org/gcc-4.2/changes.html 16708 3. https://gcc.gnu.org/gcc-4.2/changes.html 16709 4. https://gcc.gnu.org/gcc-4.2/changes.html 16710 5. https://gcc.gnu.org/gcc-4.2/changes.html 16711 6. https://gcc.gnu.org/gcc-4.2/changes.html 16712 7. https://gcc.gnu.org/gcc-4.2/buildstat.html 16713 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16714 9. https://gcc.gnu.org/index.html 16715 10. mailto:gcc (a] gcc.gnu.org 16716 11. https://gcc.gnu.org/mirrors.html 16717 12. https://gcc.gnu.org/git.html 16718 13. https://gcc.gnu.org/onlinedocs/ 16719 14. mailto:gcc-help (a] gcc.gnu.org 16720 15. mailto:gcc (a] gcc.gnu.org 16721 16. https://gcc.gnu.org/lists.html 16722 17. https://www.fsf.org/ 16723 18. https://gcc.gnu.org/about.html 16724 ====================================================================== 16725 http://gcc.gnu.org/gcc-4.2/changes.html 16726 GCC 4.2 Release Series 16727 Changes, New Features, and Fixes 16728 16729 Caveats 16730 16731 * GCC no longer accepts the -fshared-data option. This option has had 16732 no effect in any GCC 4 release; the targets to which the option 16733 used to apply had been removed before GCC 4.0. 16734 16735 General Optimizer Improvements 16736 16737 * New command-line options specify the possible relationships among 16738 parameters and between parameters and global data. For example, 16739 -fargument-noalias-anything specifies that arguments do not alias 16740 any other storage. 16741 Each language will automatically use whatever option is required by 16742 the language standard. You should not need to use these options 16743 yourself. 16744 16745 New Languages and Language specific improvements 16746 16747 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 16748 * New command-line options -fstrict-overflow and -Wstrict-overflow 16749 have been added. -fstrict-overflow tells the compiler that it may 16750 assume that the program follows the strict signed overflow 16751 semantics permitted for the language: for C and C++ this means that 16752 the compiler may assume that signed overflow does not occur. For 16753 example, a loop like 16754 for (i = 1; i > 0; i *= 2) 16755 16756 is presumably intended to continue looping until i overflows. With 16757 -fstrict-overflow, the compiler may assume that signed overflow 16758 will not occur, and transform this into an infinite loop. 16759 -fstrict-overflow is turned on by default at -O2, and may be 16760 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 16761 be used to warn about cases where the compiler assumes that signed 16762 overflow will not occur. It takes five different levels: 16763 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 16764 -Wstrict-overflow=1 is enabled by -Wall. 16765 * The new command-line option -fno-toplevel-reorder directs GCC to 16766 emit top-level functions, variables, and asm statements in the same 16767 order that they appear in the input file. This is intended to 16768 support existing code which relies on a particular ordering (for 16769 example, code which uses top-level asm statements to switch 16770 sections). For new code, it is generally better to use function and 16771 variable attributes. The -fno-toplevel-reorder option may be used 16772 for most cases which currently use -fno-unit-at-a-time. The 16773 -fno-unit-at-a-time option will be removed in some future version 16774 of GCC. If you know of a case which requires -fno-unit-at-a-time 16775 which is not fixed by -fno-toplevel-reorder, please open a bug 16776 report. 16777 16778 C family 16779 16780 * The pragma redefine_extname will now macro expand its tokens for 16781 compatibility with SunPRO. 16782 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 16783 GCC to handle inline functions as specified in the C99 standard. In 16784 preparation for this, GCC 4.2 will warn about any use of non-static 16785 inline functions in gnu99 or c99 mode. This new warning may be 16786 disabled with the new gnu_inline function attribute or the new 16787 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 16788 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 16789 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 16790 in the current compilation. 16791 * A new command-line option -Waddress has been added to warn about 16792 suspicious uses of memory addresses as, for example, using the 16793 address of a function in a conditional expression, and comparisons 16794 against the memory address of a string literal. This warning is 16795 enabled by -Wall. 16796 16797 C++ 16798 16799 * C++ visibility handling has been overhauled. 16800 Restricted visiblity is propagated from classes to members, from 16801 functions to local statics, and from templates and template 16802 arguments to instantiations, unless the latter has explicitly 16803 declared visibility. 16804 The visibility attribute for a class must come between the 16805 class-key and the name, not after the closing brace. 16806 Attributes are now allowed for enums and elaborated-type-specifiers 16807 that only declare a type. 16808 Members of the anonymous namespace are now local to a particular 16809 translation unit, along with any other declarations which use them, 16810 though they are still treated as having external linkage for 16811 language semantics. 16812 * The (undocumented) extension which permitted templates with default 16813 arguments to be bound to template template parameters with fewer 16814 parameters has been removed. For example: 16815 template <template <typename> class C> 16816 void f(C<double>) {} 16817 16818 template <typename T, typename U = int> 16819 struct S {}; 16820 16821 template void f(S<double>); 16822 16823 is no longer accepted by G++. The reason this code is not accepted 16824 is that S is a template with two parameters; therefore, it cannot 16825 be bound to C which has only one parameter. 16826 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 16827 releases, have been removed. 16828 * The command-line option -fconst-strings, deprecated in previous GCC 16829 releases, has been removed. 16830 * The configure variable enable-__cxa_atexit is now enabled by 16831 default for more targets. Enabling this variable is necessary in 16832 order for static destructors to be executed in the correct order, 16833 but it depends upon the presence of a non-standard C library in the 16834 target library in order to work. The variable is now enabled for 16835 more targets which are known to have suitable C libraries. 16836 * -Wextra will produce warnings for if statements with a semicolon as 16837 the only body, to catch code like: 16838 if (a); 16839 return 1; 16840 return 0; 16841 16842 To suppress the warning in valid cases, use { } instead. 16843 * The C++ front end now also produces strict aliasing warnings when 16844 -fstrict-aliasing -Wstrict-aliasing is in effect. 16845 16846 Runtime Library (libstdc++) 16847 16848 * Added support for TR1 <random>, <complex>, and C compatibility 16849 headers. In addition, a lock-free version of shared_ptr was 16850 contributed as part of Phillip Jordan's Google Summer of Code 16851 project on lock-free containers. 16852 * In association with the Summer of Code work on lock-free 16853 containers, the interface for atomic builtins was adjusted, 16854 creating simpler alternatives for non-threaded code paths. Also, 16855 usage was consolidated and all elements were moved from namespace 16856 std to namespace__gnu_cxx. Affected interfaces are the functions 16857 __exchange_and_add, __atomic_add, and the objects __mutex, 16858 __recursive_mutex, and __scoped_lock. 16859 * Support for versioning weak symbol names via namespace association 16860 was added. However, as this changes the names of exported symbols, 16861 this is turned off by default in the current ABI. Intrepid users 16862 can enable this feature by using 16863 --enable-symvers=gnu-versioned-namespace during configuration. 16864 * Revised, simplified, and expanded policy-based associative 16865 containers, including data types for tree and trie forms 16866 (basic_tree, tree, trie), lists (list_update), and both 16867 collision-chaining and probing hash-based containers 16868 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 16869 the [3]documentation. 16870 * The implementation of the debug mode was modified, whereby the 16871 debug namespaces were nested inside of namespace std and namespace 16872 __gnu_cxx in order to resolve some long standing corner cases 16873 involving name lookup. Debug functionality from the policy-based 16874 data structures was consolidated and enabled with the single macro, 16875 _GLIBCXX_DEBUG. See PR 26142 for more information. 16876 * Added extensions for type traits: __conditional_type, 16877 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 16878 * Added a typelist implementation for compile-time meta-programming. 16879 Elements for typelist construction and operation can be found 16880 within namespace __gnu_cxx::typelist. 16881 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 16882 exception-safety. 16883 * Enabled library-wide visibility control, allowing -fvisibility to 16884 be used. 16885 * Consolidated all nested namespaces and the conversion of 16886 __gnu_internal implementation-private details to anonymous 16887 namespaces whenever possible. 16888 * Implemented LWG resolutions DR 431 and DR 538. 16889 16890 Fortran 16891 16892 * Support for allocatable components has been added (TR 15581 and 16893 Fortran 2003). 16894 * Support for the Fortran 2003 streaming IO extension has been added. 16895 * The GNU Fortran compiler now uses 4-byte record markers by default 16896 for unformatted files to be compatible with g77 and most other 16897 compilers. The implementation allows for records greater than 2 GB 16898 and is compatible with several other compilers. Older versions of 16899 gfortran used 8-byte record markers by default (on most systems). 16900 In order to change the length of the record markers, e.g. to read 16901 unformatted files created by older gfortran versions, the 16902 [4]-frecord-marker=8 option can be used. 16903 16904 Java (GCJ) 16905 16906 * A new command-line option -static-libgcj has been added for targets 16907 that use a linker compatible with GNU Binutils. As its name 16908 implies, this causes libgcj to be linked statically. In some cases 16909 this causes the resulting executable to start faster and use less 16910 memory than if the shared version of libgcj were used. However 16911 caution should be used as it can also cause essential parts of the 16912 library to be omitted. Some of these issues are discussed in: 16913 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 16914 * fastjar is no longer bundled with GCC. To build libgcj, you will 16915 need either InfoZIP (both zip and unzip) or an external jar 16916 program. In the former case, the GCC build will install a jar shell 16917 script that is based on InfoZIP and provides the same functionality 16918 as fastjar. 16919 16920 [6]New Targets and Target Specific Improvements 16921 16922 IA-32/x86-64 16923 16924 * -mtune=generic can now be used to generate code running well on 16925 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 16926 Pentium-M, Intel Pentium 4 and Intel Core 2. 16927 * -mtune=native and -march=native will produce code optimized for the 16928 host architecture as detected using the cpuid instruction. 16929 * Added a new command-line option -fstackrealign and and 16930 __attribute__ ((force_align_arg_pointer)) to realign the stack at 16931 runtime. This allows functions compiled with a vector-aligned stack 16932 to be invoked from legacy objects that keep only word-alignment. 16933 16934 SPARC 16935 16936 * The default CPU setting has been changed from V7 to V9 in 32-bit 16937 mode on Solaris 7 and above. This is already the case in 64-bit 16938 mode. It can be overridden by specifying --with-cpu at configure 16939 time. 16940 * Back-end support of built-in functions for atomic memory access has 16941 been implemented. 16942 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 16943 added. 16944 16945 M32C 16946 16947 * Various bug fixes have made some functions (notably, functions 16948 returning structures) incompatible with previous releases. 16949 Recompiling all libraries is recommended. Note that code quality 16950 has considerably improved since 4.1, making a recompile even more 16951 beneficial. 16952 16953 MIPS 16954 16955 * Added support for the Broadcom SB-1A core. 16956 16957 IA-64 16958 16959 * Added support for IA-64 data and control speculation. By default 16960 speculation is enabled only during second scheduler pass. A number 16961 of machine flags was introduced to control the usage of speculation 16962 for both scheduler passes. 16963 16964 HPPA 16965 16966 * Added Java language support (libffi and libjava) for 32-bit HP-UX 16967 11 target. 16968 16969 [7]Obsolete Systems 16970 16971 Documentation improvements 16972 16973 PDF Documentation 16974 16975 * A make pdf target has been added to the top-level makefile, 16976 enabling automated production of PDF documentation files. 16977 (Front-ends external to GCC should modify their Make-lang.in file 16978 to add a lang.pdf: target.) 16979 16980 Other significant improvements 16981 16982 Build system improvements 16983 16984 * All the components of the compiler are now bootstrapped by default. 16985 This improves the resilience to bugs in the system compiler or 16986 binary compatibility problems, as well as providing better testing 16987 of GCC 4.2 itself. In addition, if you build the compiler from a 16988 combined tree, the assembler, linker, etc. will also be 16989 bootstrapped (i.e. built with themselves). 16990 You can disable this behavior, and go back to the pre-GCC 4.2 set 16991 up, by configuring GCC with --disable-bootstrap. 16992 * The rules that configure follows to find target tools resemble more 16993 closely the locations that the built compiler will search. In 16994 addition, you can use the new configure option --with-target-tools 16995 to specify where to find the target tools used during the build, 16996 without affecting what the built compiler will use. 16997 This can be especially useful when building packages of GCC. For 16998 example, you may want to build GCC with GNU as or ld, even if the 16999 resulting compiler to work with the native assembler and linker. To 17000 do so, you can use --with-target-tools to point to the native 17001 tools. 17002 17003 Incompatible changes to the build system 17004 17005 * Front-ends external to GCC should modify their Make-lang.in file to 17006 replace double-colon rules (e.g. dvi::) with normal rules (like 17007 lang.dvi:). Front-end makefile hooks do not use double-colon rules 17008 anymore. 17009 * Up to GCC 4.1, a popular way to specify the target tools used 17010 during the build was to create directories named gas, binutils, 17011 etc. in the build tree, and create links to the tools from there. 17012 This does not work any more when the compiler is bootstrapped. The 17013 new configure option --with-target-tools provides a better way to 17014 achieve the same effect, and works for all native and cross 17015 settings. 17016 17017 17018 For questions related to the use of GCC, please consult these web 17019 pages and the [8]GCC manuals. If that fails, the 17020 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17021 web pages and the development of GCC are welcome on our developer 17022 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 17023 archives. 17024 17025 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 17026 distribution of this entire article is permitted in any medium, 17027 provided this notice is preserved. 17028 17029 These pages are [13]maintained by the GCC team. Last modified 17030 2025-01-31. 17031 17032 References 17033 17034 1. https://gcc.gnu.org/projects/gomp/ 17035 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 17036 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 17037 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 17038 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 17039 6. https://gcc.gnu.org/gcc-4.2/changes.html#targets 17040 7. https://gcc.gnu.org/gcc-4.2/changes.html#obsolete_systems 17041 8. https://gcc.gnu.org/onlinedocs/ 17042 9. mailto:gcc-help (a] gcc.gnu.org 17043 10. mailto:gcc (a] gcc.gnu.org 17044 11. https://gcc.gnu.org/lists.html 17045 12. https://www.fsf.org/ 17046 13. https://gcc.gnu.org/about.html 17047 ====================================================================== 17048 http://gcc.gnu.org/gcc-4.1/index.html 17049 GCC 4.1 Release Series 17050 17051 (This release series is no longer supported.) 17052 17053 February 13, 2007 17054 17055 The [1]GNU project and the GCC developers are pleased to announce the 17056 release of GCC 4.1.2. 17057 17058 This release is a bug-fix release, containing fixes for regressions in 17059 GCC 4.1.1 relative to previous releases of GCC. 17060 17061 Release History 17062 17063 GCC 4.1.2 17064 February 13, 2007 ([2]changes) 17065 17066 GCC 4.1.1 17067 May 24, 2006 ([3]changes) 17068 17069 GCC 4.1.0 17070 February 28, 2006 ([4]changes) 17071 17072 References and Acknowledgements 17073 17074 GCC used to stand for the GNU C Compiler, but since the compiler 17075 supports several other languages aside from C, it now stands for the 17076 GNU Compiler Collection. 17077 17078 A list of [5]successful builds is updated as new information becomes 17079 available. 17080 17081 The GCC developers would like to thank the numerous people that have 17082 contributed new features, improvements, bug fixes, and other changes as 17083 well as test results to GCC. This [6]amazing group of volunteers is 17084 what makes GCC successful. 17085 17086 For additional information about GCC please refer to the [7]GCC project 17087 web site or contact the [8]GCC development mailing list. 17088 17089 To obtain GCC please use [9]our mirror sites or [10]our version control 17090 system. 17091 17092 17093 For questions related to the use of GCC, please consult these web 17094 pages and the [11]GCC manuals. If that fails, the 17095 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17096 web pages and the development of GCC are welcome on our developer 17097 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 17098 archives. 17099 17100 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 17101 distribution of this entire article is permitted in any medium, 17102 provided this notice is preserved. 17103 17104 These pages are [16]maintained by the GCC team. Last modified 17105 2022-10-26. 17106 17107 References 17108 17109 1. http://www.gnu.org/ 17110 2. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 17111 3. https://gcc.gnu.org/gcc-4.1/changes.html 17112 4. https://gcc.gnu.org/gcc-4.1/changes.html 17113 5. https://gcc.gnu.org/gcc-4.1/buildstat.html 17114 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17115 7. https://gcc.gnu.org/index.html 17116 8. mailto:gcc (a] gcc.gnu.org 17117 9. https://gcc.gnu.org/mirrors.html 17118 10. https://gcc.gnu.org/git.html 17119 11. https://gcc.gnu.org/onlinedocs/ 17120 12. mailto:gcc-help (a] gcc.gnu.org 17121 13. mailto:gcc (a] gcc.gnu.org 17122 14. https://gcc.gnu.org/lists.html 17123 15. https://www.fsf.org/ 17124 16. https://gcc.gnu.org/about.html 17125 ====================================================================== 17126 http://gcc.gnu.org/gcc-4.1/changes.html 17127 GCC 4.1 Release Series 17128 Changes, New Features, and Fixes 17129 17130 The latest release in the 4.1 release series is [1]GCC 4.1.2. 17131 17132 Caveats 17133 17134 General Optimizer Improvements 17135 17136 * GCC now has infrastructure for inter-procedural optimizations and 17137 the following inter-procedural optimizations are implemented: 17138 + Profile guided inlining. When doing profile feedback guided 17139 optimization, GCC can now use the profile to make better 17140 informed decisions on whether inlining of a function is 17141 profitable or not. This means that GCC will no longer inline 17142 functions at call sites that are not executed very often, and 17143 that functions at hot call sites are more likely to be 17144 inlined. 17145 A new parameter min-inline-recursive-probability is also now 17146 available to throttle recursive inlining of functions with 17147 small average recursive depths. 17148 + Discovery of pure and const functions, a form of side-effects 17149 analysis. While older GCC releases could also discover such 17150 special functions, the new IPA-based pass runs earlier so that 17151 the results are available to more optimizers. The pass is also 17152 simply more powerful than the old one. 17153 + Analysis of references to static variables and type escape 17154 analysis, also forms of side-effects analysis. The results of 17155 these passes allow the compiler to be less conservative about 17156 call-clobbered variables and references. This results in more 17157 redundant loads being eliminated and in making static 17158 variables candidates for register promotion. 17159 + Improvement of RTL-based alias analysis. The results of type 17160 escape analysis are fed to the RTL type-based alias analyzer, 17161 allowing it to disambiguate more memory references. 17162 + Interprocedural constant propagation and function versioning. 17163 This pass looks for functions that are always called with the 17164 same constant value for one or more of the function arguments, 17165 and propagates those constants into those functions. 17166 + GCC will now eliminate static variables whose usage was 17167 optimized out. 17168 + -fwhole-program --combine can now be used to make all 17169 functions in program static allowing whole program 17170 optimization. As an exception, the main function and all 17171 functions marked with the new externally_visible attribute are 17172 kept global so that programs can link with runtime libraries. 17173 * GCC can now do a form of partial dead code elimination (PDCE) that 17174 allows code motion of expressions to the paths where the result of 17175 the expression is actually needed. This is not always a win, so the 17176 pass has been limited to only consider profitable cases. Here is an 17177 example: 17178 int foo (int *, int *); 17179 int 17180 bar (int d) 17181 { 17182 int a, b, c; 17183 b = d + 1; 17184 c = d + 2; 17185 a = b + c; 17186 if (d) 17187 { 17188 foo (&b, &c); 17189 a = b + c; 17190 } 17191 printf ("%d\n", a); 17192 } 17193 17194 The a = b + c can be sunk to right before the printf. Normal code 17195 sinking will not do this, it will sink the first one above into the 17196 else-branch of the conditional jump, which still gives you two 17197 copies of the code. 17198 * GCC now has a value range propagation pass. This allows the 17199 compiler to eliminate bounds checks and branches. The results of 17200 the pass can also be used to accurately compute branch 17201 probabilities. 17202 * The pass to convert PHI nodes to straight-line code (a form of 17203 if-conversion for GIMPLE) has been improved significantly. The two 17204 most significant improvements are an improved algorithm to 17205 determine the order in which the PHI nodes are considered, and an 17206 improvement that allow the pass to consider if-conversions of basic 17207 blocks with more than two predecessors. 17208 * Alias analysis improvements. GCC can now differentiate between 17209 different fields of structures in Tree-SSA's virtual operands form. 17210 This lets stores/loads from non-overlapping structure fields not 17211 conflict. A new algorithm to compute points-to sets was contributed 17212 that can allows GCC to see now that p->a and p->b, where p is a 17213 pointer to a structure, can never point to the same field. 17214 * Various enhancements to auto-vectorization: 17215 + Incrementally preserve SSA form when vectorizing. 17216 + Incrementally preserve loop-closed form when vectorizing. 17217 + Improvements to peeling for alignment: generate better code 17218 when the misalignment of an access is known at compile time, 17219 or when different accesses are known to have the same 17220 misalignment, even if the misalignment amount itself is 17221 unknown. 17222 + Consider dependence distance in the vectorizer. 17223 + Externalize generic parts of data reference analysis to make 17224 this analysis available to other passes. 17225 + Vectorization of conditional code. 17226 + Reduction support. 17227 * GCC can now partition functions in sections of hot and cold code. 17228 This can significantly improve performance due to better 17229 instruction cache locality. This feature works best together with 17230 profile feedback driven optimization. 17231 * A new pass to avoid saving of unneeded arguments to the stack in 17232 vararg functions if the compiler can prove that they will not be 17233 needed. 17234 * Transition of basic block profiling to tree level implementation 17235 has been completed. The new implementation should be considerably 17236 more reliable (hopefully avoiding profile mismatch errors when 17237 using -fprofile-use or -fbranch-probabilities) and can be used to 17238 drive higher level optimizations, such as inlining. 17239 The -ftree-based-profiling command-line option was removed and 17240 -fprofile-use now implies disabling old RTL level loop optimizer 17241 (-fno-loop-optimize). Speculative prefetching optimization 17242 (originally enabled by -fspeculative-prefetching) was removed. 17243 17244 New Languages and Language specific improvements 17245 17246 C and Objective-C 17247 17248 * The old Bison-based C and Objective-C parser has been replaced by a 17249 new, faster hand-written recursive-descent parser. 17250 17251 Ada 17252 17253 * The build infrastructure for the Ada runtime library and tools has 17254 been changed to be better integrated with the rest of the build 17255 infrastructure of GCC. This should make doing cross builds of Ada a 17256 bit easier. 17257 17258 [2]C++ 17259 17260 * ARM-style name-injection of friend declarations is no longer the 17261 default. For example: 17262 struct S { 17263 friend void f(); 17264 }; 17265 17266 void g() { f(); } 17267 will not be accepted; instead a declaration of f will need to be 17268 present outside of the scope of S. The new -ffriend-injection 17269 option will enable the old behavior. 17270 * The (undocumented) extension which permitted templates with default 17271 arguments to be bound to template template parameters with fewer 17272 parameters has been deprecated, and will be removed in the next 17273 major release of G++. For example: 17274 template <template <typename> class C> 17275 void f(C<double>) {} 17276 17277 template <typename T, typename U = int> 17278 struct S {}; 17279 17280 template void f(S<double>); 17281 17282 makes use of the deprecated extension. The reason this code is not 17283 valid ISO C++ is that S is a template with two parameters; 17284 therefore, it cannot be bound to C which has only one parameter. 17285 17286 Runtime Library (libstdc++) 17287 17288 * Optimization work: 17289 + A new implementation of std::search_n is provided, better 17290 performing in case of random access iterators. 17291 + Added further efficient specializations of istream functions, 17292 i.e., character array and string extractors. 17293 + Other smaller improvements throughout. 17294 * Policy-based associative containers, designed for high-performance, 17295 flexibility and semantic safety are delivered in ext/pb_assoc. 17296 * A versatile string class, __gnu_cxx::__versa_string, providing 17297 facilities conforming to the standard requirements for 17298 basic_string, is delivered in <ext/vstring.h>. In particular: 17299 + Two base classes are provided: the default one avoids 17300 reference counting and is optimized for short strings; the 17301 alternate one, still uses it while improving in a few low 17302 level areas (e.g., alignment). See vstring_fwd.h for some 17303 useful typedefs. 17304 + Various algorithms have been rewritten (e.g., replace), the 17305 code streamlined and simple optimizations added. 17306 + Option 3 of DR 431 is implemented for both available bases, 17307 thus improving the support for stateful allocators. 17308 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 17309 libstdc++/23953) and LWG resolutions put into effect for the first 17310 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 17311 Issue 6.19). The implementation status of TR1 is now tracked in the 17312 docs in tr1.html. 17313 17314 Objective-C++ 17315 17316 * A new language front end for Objective-C++ has been added. This 17317 language allows users to mix the object oriented features of 17318 Objective-C with those of C++. 17319 17320 Java (GCJ) 17321 17322 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 17323 features (plus some 0.20 bug-fixes) 17324 + Networking 17325 o The java.net.HttpURLConnection implementation no longer 17326 buffers the entire response body in memory. This means 17327 that response bodies larger than available memory can now 17328 be handled. 17329 + (N)IO 17330 o NIO FileChannel.map implementation, fast bulk put 17331 implementation for DirectByteBuffer (speeds up this 17332 method 10x). 17333 o FileChannel.lock() and FileChannel.force() implemented. 17334 + XML 17335 o gnu.xml fix for nodes created outside a namespace 17336 context. 17337 o Add support for output indenting and 17338 cdata-section-elements output instruction in 17339 xml.transform. 17340 o xml.xpath corrections for cases where elements/attributes 17341 might have been created in non-namespace-aware mode. 17342 Corrections to handling of XSL variables and minor 17343 conformance updates. 17344 + AWT 17345 o GNU JAWT implementation, the AWT Native Interface, which 17346 allows direct access to native screen resources from 17347 within a Canvas's paint method. GNU Classpath Examples 17348 comes with a Demo, see libjava/classpath/examples/README. 17349 o awt.datatransfer updated to 1.5 with support for 17350 FlavorEvents. The gtk+ awt peers now allow copy/paste of 17351 text, images, URIs/files and serialized objects with 17352 other applications and tracking clipboard change events 17353 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 17354 objects are supported). A GNU Classpath Examples 17355 datatransfer Demo was added to show the new 17356 functionality. 17357 o Split gtk+ awt peers event handling in two threads and 17358 improve gdk lock handling (solves several awt lock ups). 17359 o Speed up awt Image loading. 17360 o Better gtk+ scrollbar peer implementation when using gtk+ 17361 >= 2.6. 17362 o Handle image loading errors correctly for gdkpixbuf and 17363 MediaTracker. 17364 o Better handle GDK lock. Properly prefix gtkpeer native 17365 functions (cp_gtk). 17366 o GdkGraphics2D has been updated to use Cairo 0.5.x or 17367 higher. 17368 o BufferedImage and GtkImage rewrites. All image drawing 17369 operations should now work correctly (flipping requires 17370 gtk+ >= 2.6) 17371 o When gtk+ 2.6 or higher is installed the default log 17372 handler will produce stack traces whenever a WARNING, 17373 CRITICAL or ERROR message is produced. 17374 + Free Swing 17375 o The RepaintManager has been reworked for more efficient 17376 painting, especially for large GUIs. 17377 o The layout manager OverlayLayout has been implemented, 17378 the BoxLayout has been rewritten to make use of the 17379 SizeRequirements utility class and caching for more 17380 efficient layout. 17381 o Improved accessibility support. 17382 o Significant progress has been made in the implementation 17383 of the javax.swing.plaf.metal package, with most UI 17384 delegates in a working state now. Please test this with 17385 your own applications and provide feedback that will help 17386 us to improve this package. 17387 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 17388 extended to highlight various features in our Free Swing 17389 implementation. And it includes a look and feel switcher 17390 for Metal (default), Ocean and GNU themes. 17391 o The javax.swing.plaf.multi package is now implemented. 17392 o Editing and several key actions for JTree and JTable were 17393 implemented. 17394 o Lots of icons and look and feel improvements for Free 17395 Swing basic and metal themes were added. Try running the 17396 GNU Classpath Swing Demo in examples 17397 (gnu.classpath.examples.swing.Demo) with: 17398 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 17399 l or 17400 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 17401 l 17402 o Start of styled text capabilites for java.swing.text. 17403 o DefaultMutableTreeNode pre-order, post-order, depth-first 17404 and breadth-first traversal enumerations implemented. 17405 o JInternalFrame colors and titlebar draw properly. 17406 o JTree is working up to par (icons, selection and keyboard 17407 traversal). 17408 o JMenus were made more compatible in visual and 17409 programmatic behavior. 17410 o JTable changeSelection and multiple selections 17411 implemented. 17412 o JButton and JToggleButton change states work properly 17413 now. 17414 o JFileChooser fixes. 17415 o revalidate() and repaint() fixes which make Free Swing 17416 much more responsive. 17417 o MetalIconFactory implemented. 17418 o Free Swing Top-Level Compatibility. JFrame, JDialog, 17419 JApplet, JInternalFrame, and JWindow are now 1.5 17420 compatible in the sense that you can call add() and 17421 setLayout() directly on them, which will have the same 17422 effect as calling getContentPane().add() and 17423 getContentPane().setLayout(). 17424 o The JTree interface has been completed. JTrees now 17425 recognizes mouse clicks and selections work. 17426 o BoxLayout works properly now. 17427 o Fixed GrayFilter to actually work. 17428 o Metal SplitPane implemented. 17429 o Lots of Free Swing text and editor stuff work now. 17430 + Free RMI and Corba 17431 o Andrew Watson, Vice President and Technical Director of 17432 the Object Management Group, has officially assigned us 17433 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 17434 mark remote classpath-specific system exceptions. 17435 Obtaining the VMCID means that GNU Classpath now is a 17436 recogniseable type of node in a highly interoperable 17437 CORBA world. 17438 o GNU Classpath now includes the first working draft to 17439 support the RMI over IIOP protocol. The current 17440 implementation is capable of remote invocations, 17441 transferring various Serializables and Externalizables 17442 via RMI-IIOP protocol. It can flatten graphs and, at 17443 least for the simple cases, is interoperable with 1.5 17444 JDKs. 17445 o org.omg.PortableInterceptor and related functionality in 17446 other packages is now implemented: 17447 # The sever and client interceptors work as required 17448 since 1.4. 17449 # The IOR interceptor works as needed for 1.5. 17450 o The org.omg.DynamicAny package is completed and passes 17451 the prepared tests. 17452 o The Portable Object Adapter should now support the output 17453 of the recent IDL to java compilers. These compilers now 17454 generate servants and not CORBA objects as before, making 17455 the output depend on the existing POA implementation. 17456 Completing POA means that such code can already be tried 17457 to run on Classpath. Our POA is tested for the following 17458 usager scenarios: 17459 # POA converts servant to the CORBA object. 17460 # Servant provides to the CORBA object. 17461 # POA activates new CORBA object with the given Object 17462 Id (byte array) that is later accessible for the 17463 servant. 17464 # During the first call, the ServantActivator provides 17465 servant for this and all subsequent calls on the 17466 current object. 17467 # During each call, the ServantLocator provides 17468 servant for this call only. 17469 # ServantLocator or ServantActivator forwards call to 17470 another server. 17471 # POA has a single servant, responsible for all 17472 objects. 17473 # POA has a default servant, but some objects are 17474 explicitly connected to they specific servants. 17475 The POA is verified using tests from the former 17476 cost.omg.org. 17477 o The CORBA implementation is now a working prototype that 17478 should support features up to 1.3 inclusive. We invite 17479 groups writing CORBA dependent applications to try 17480 Classpath implementation, reporting any possible bugs. 17481 The CORBA prototype is interoperable with Sun's 17482 implementation v 1.4, transferring object references, 17483 primitive types, narrow and wide strings, arrays, 17484 structures, trees, abstract interfaces and value types 17485 (feature of CORBA 2.3) between these two platforms. 17486 Remote exceptions are transferred and handled correctly. 17487 The stringified object references (IORs) from various 17488 sources are parsed as required. The transient (for 17489 current session) and permanent (till jre restart) 17490 redirections work. Both Little and Big Endian encoded 17491 messages are accepted. The implementation is verified 17492 using tests from the former cost.omg.org. The current 17493 release includes working examples (see the examples 17494 directory), demonstrating the client-server 17495 communication, using either CORBA Request or IDL-based 17496 stub (usually generated by a IDL to java compiler). These 17497 examples also show how to use the Classpath CORBA naming 17498 service. The IDL to java compiler is not yet written, but 17499 as our library must be compatible, it naturally accepts 17500 the output of other idlj implementations. 17501 + Misc 17502 o Updated TimeZone data against Olson tzdata2005l. 17503 o Make zip and jar packages UTF-8 clean. 17504 o "native" code builds and compiles (warning free) on 17505 Darwin and Solaris. 17506 o java.util.logging.FileHandler now rotates files. 17507 o Start of a generic JDWP framework in gnu/classpath/jdwp. 17508 This is unfinished, but feedback (at classpath (a] gnu.org) 17509 from runtime hackers is greatly appreciated. Although 17510 most of the work is currently being done around gcj/gij 17511 we want this framework to be as VM neutral as possible. 17512 Early design is described in: 17513 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 17514 o QT4 AWT peers, enable by giving configure 17515 --enable-qt-peer. Included, but not ready for production 17516 yet. They are explicitly disabled and not supported. But 17517 if you want to help with the development of these new 17518 features we are interested in feedback. You will have to 17519 explicitly enable them to try them out (and they will 17520 most likely contain bugs). 17521 o Documentation fixes all over the place. See 17522 [4]https://developer.classpath.org/doc/ 17523 17524 [5]New Targets and Target Specific Improvements 17525 17526 IA-32/x86-64 17527 17528 * The x86-64 medium model (that allows building applications whose 17529 data segment exceeds 4GB) was redesigned to match latest ABI draft. 17530 New implementation split large datastructures into separate segment 17531 improving performance of accesses to small datastructures and also 17532 allows linking of small model libraries into medium model programs 17533 as long as the libraries are not accessing the large datastructures 17534 directly. Medium model is also supported in position independent 17535 code now. 17536 The ABI change results in partial incompatibility among medium 17537 model objects. Linking medium model libraries (or objects) compiled 17538 with new compiler into medium model program compiled with older 17539 will likely result in exceeding ranges of relocations. 17540 Binutils 2.16.91 or newer are required for compiling medium model 17541 now. 17542 17543 RS6000 (POWER/PowerPC) 17544 17545 * The AltiVec vector primitives in <altivec.h> are now implemented in 17546 a way that puts a smaller burden on the preprocessor, instead 17547 processing the "overloading" in the front ends. This should benefit 17548 compilation speed on AltiVec vector code. 17549 * AltiVec initializers now are generated more efficiently. 17550 * The popcountb instruction available on POWER5 now is generated. 17551 * The floating point round to integer instructions available on 17552 POWER5+ now is generated. 17553 * Floating point divides can be synthesized using the floating point 17554 reciprocal estimate instructions. 17555 * Double precision floating point constants are initialized as single 17556 precision values if they can be represented exactly. 17557 17558 S/390, zSeries and System z9 17559 17560 * Support for the IBM System z9 109 processor has been added. When 17561 using the -march=z9-109 option, the compiler will generate code 17562 making use of instructions provided by the extended immediate 17563 facility. 17564 * Support for 128-bit IEEE floating point has been added. When using 17565 the -mlong-double-128 option, the compiler will map the long double 17566 data type to 128-bit IEEE floating point. Using this option 17567 constitutes an ABI change, and requires glibc support. 17568 * Various changes to improve performance of generated code have been 17569 implemented, including: 17570 + In functions that do not require a literal pool, register %r13 17571 (which is traditionally reserved as literal pool pointer), can 17572 now be freely used for other purposes by the compiler. 17573 + More precise tracking of register use allows the compiler to 17574 generate more efficient function prolog and epilog code in 17575 certain cases. 17576 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 17577 instructions are now used to implement C string functions. 17578 + The MOVE CHARACTER instruction with single byte overlap is now 17579 used to implement the memset function with non-zero fill byte. 17580 + The LOAD ZERO instructions are now used where appropriate. 17581 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 17582 and INSERT IMMEDIATE instructions are now used more frequently 17583 to optimize bitfield operations. 17584 + The BRANCH ON COUNT instruction is now used more frequently. 17585 In particular, the fact that a loop contains a subroutine call 17586 no longer prevents the compiler from using this instruction. 17587 + The compiler is now aware that all shift and rotate 17588 instructions implicitly truncate the shift count to six bits. 17589 * Back-end support for the following generic features has been 17590 implemented: 17591 + The full set of [6]built-in functions for atomic memory 17592 access. 17593 + The -fstack-protector feature. 17594 + The optimization pass avoiding unnecessary stores of incoming 17595 argument registers in functions with variable argument list. 17596 17597 SPARC 17598 17599 * The default code model in 64-bit mode has been changed from 17600 Medium/Anywhere to Medium/Middle on Solaris. 17601 * TLS support is disabled by default on Solaris prior to release 10. 17602 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 17603 and later) by specifying --enable-tls at configure time. 17604 17605 MorphoSys 17606 17607 * Support has been added for this new architecture. 17608 17609 [7]Obsolete Systems 17610 17611 Documentation improvements 17612 17613 Other significant improvements 17614 17615 * GCC can now emit code for protecting applications from 17616 stack-smashing attacks. The protection is realized by buffer 17617 overflow detection and reordering of stack variables to avoid 17618 pointer corruption. 17619 * Some built-in functions have been fortified to protect them against 17620 various buffer overflow (and format string) vulnerabilities. 17621 Compared to the mudflap bounds checking feature, the safe builtins 17622 have far smaller overhead. This means that programs built using 17623 safe builtins should not experience any measurable slowdown. 17624 17625 [8]GCC 4.1.2 17626 17627 This is the [9]list of problem reports (PRs) from GCC's bug tracking 17628 system that are known to be fixed in the 4.1.2 release. This list might 17629 not be complete (that is, it is possible that some PRs that have been 17630 fixed are not listed here). 17631 17632 When generating code for a shared library, GCC now recognizes that 17633 global functions may be replaced when the program runs. Therefore, it 17634 is now more conservative in deducing information from the bodies of 17635 functions. For example, in this example: 17636 void f() {} 17637 void g() { 17638 try { f(); } 17639 catch (...) { 17640 cout << "Exception"; 17641 } 17642 } 17643 17644 G++ would previously have optimized away the catch clause, since it 17645 would have concluded that f cannot throw exceptions. Because users may 17646 replace f with another function in the main body of the program, this 17647 optimization is unsafe, and is no longer performed. If you wish G++ to 17648 continue to optimize as before, you must add a throw() clause to the 17649 declaration of f to make clear that it does not throw exceptions. 17650 17651 17652 For questions related to the use of GCC, please consult these web 17653 pages and the [10]GCC manuals. If that fails, the 17654 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17655 web pages and the development of GCC are welcome on our developer 17656 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 17657 archives. 17658 17659 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 17660 distribution of this entire article is permitted in any medium, 17661 provided this notice is preserved. 17662 17663 These pages are [15]maintained by the GCC team. Last modified 17664 2025-01-31. 17665 17666 References 17667 17668 1. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 17669 2. https://gcc.gnu.org/gcc-4.1/changes.html#cplusplus 17670 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 17671 4. https://developer.classpath.org/doc/ 17672 5. https://gcc.gnu.org/gcc-4.1/changes.html#targets 17673 6. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 17674 7. https://gcc.gnu.org/gcc-4.1/changes.html#obsolete_systems 17675 8. https://gcc.gnu.org/gcc-4.1/changes.html#GCC4.1.2 17676 9. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 17677 10. https://gcc.gnu.org/onlinedocs/ 17678 11. mailto:gcc-help (a] gcc.gnu.org 17679 12. mailto:gcc (a] gcc.gnu.org 17680 13. https://gcc.gnu.org/lists.html 17681 14. https://www.fsf.org/ 17682 15. https://gcc.gnu.org/about.html 17683 ====================================================================== 17684 http://gcc.gnu.org/gcc-4.0/index.html 17685 GCC 4.0 Release Series 17686 17687 (This release series is no longer supported.) 17688 17689 January 31, 2007 17690 17691 The [1]GNU project and the GCC developers are pleased to announce the 17692 release of GCC 4.0.4. 17693 17694 This release is a bug-fix release, containing fixes for regressions in 17695 GCC 4.0.3 relative to previous releases of GCC. 17696 17697 Release History 17698 17699 GCC 4.0.4 17700 January 31, 2007 ([2]changes) 17701 17702 GCC 4.0.3 17703 March 10, 2006 ([3]changes) 17704 17705 GCC 4.0.2 17706 September 28, 2005 ([4]changes) 17707 17708 GCC 4.0.1 17709 July 7, 2005 ([5]changes) 17710 17711 GCC 4.0.0 17712 April 20, 2005 ([6]changes) 17713 17714 References and Acknowledgements 17715 17716 GCC used to stand for the GNU C Compiler, but since the compiler 17717 supports several other languages aside from C, it now stands for the 17718 GNU Compiler Collection. 17719 17720 A list of [7]successful builds is updated as new information becomes 17721 available. 17722 17723 The GCC developers would like to thank the numerous people that have 17724 contributed new features, improvements, bug fixes, and other changes as 17725 well as test results to GCC. This [8]amazing group of volunteers is 17726 what makes GCC successful. 17727 17728 For additional information about GCC please refer to the [9]GCC project 17729 web site or contact the [10]GCC development mailing list. 17730 17731 To obtain GCC please use [11]our mirror sites, or [12]our version 17732 control system. 17733 17734 17735 For questions related to the use of GCC, please consult these web 17736 pages and the [13]GCC manuals. If that fails, the 17737 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17738 web pages and the development of GCC are welcome on our developer 17739 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 17740 archives. 17741 17742 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 17743 distribution of this entire article is permitted in any medium, 17744 provided this notice is preserved. 17745 17746 These pages are [18]maintained by the GCC team. Last modified 17747 2022-10-26. 17748 17749 References 17750 17751 1. http://www.gnu.org/ 17752 2. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 17753 3. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 17754 4. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 17755 5. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 17756 6. https://gcc.gnu.org/gcc-4.0/changes.html 17757 7. https://gcc.gnu.org/gcc-4.0/buildstat.html 17758 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17759 9. https://gcc.gnu.org/index.html 17760 10. mailto:gcc (a] gcc.gnu.org 17761 11. https://gcc.gnu.org/mirrors.html 17762 12. https://gcc.gnu.org/git.html 17763 13. https://gcc.gnu.org/onlinedocs/ 17764 14. mailto:gcc-help (a] gcc.gnu.org 17765 15. mailto:gcc (a] gcc.gnu.org 17766 16. https://gcc.gnu.org/lists.html 17767 17. https://www.fsf.org/ 17768 18. https://gcc.gnu.org/about.html 17769 ====================================================================== 17770 http://gcc.gnu.org/gcc-4.0/changes.html 17771 GCC 4.0 Release Series 17772 Changes, New Features, and Fixes 17773 17774 The latest release in the 4.0 release series is [1]GCC 4.0.4. 17775 17776 Caveats 17777 17778 * GCC now generates location lists by default when compiling with 17779 debug info and optimization. 17780 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 17781 or later is needed to debug binaries containing location 17782 lists. 17783 + When you are trying to view a value of a variable in a part of 17784 a function where it has no location (for example when the 17785 variable is no longer used and thus its location was used for 17786 something else) GDB will say that it is not available. 17787 You can disable generating location lists by -fno-var-tracking. 17788 * GCC no longer accepts the -fwritable-strings option. Use named 17789 character arrays when you need a writable string. 17790 * The options -freduce-all-givs and -fmove-all-movables have been 17791 discontinued. They were used to circumvent a shortcoming in the 17792 heuristics of the old loop optimization code with respect to common 17793 Fortran constructs. The new (tree) loop optimizer works differently 17794 and doesn't need those work-arounds. 17795 * The graph-coloring register allocator, formerly enabled by the 17796 option -fnew-ra, has been discontinued. 17797 * -I- has been deprecated. -iquote is meant to replace the need for 17798 this option. 17799 * The MIPS -membedded-pic and -mrnames options have been removed. 17800 * All MIPS targets now require the GNU assembler. In particular, IRIX 17801 configurations can no longer use the MIPSpro assemblers, although 17802 they do still support the MIPSpro linkers. 17803 * The SPARC option -mflat has been removed. 17804 * English-language diagnostic messages will now use Unicode quotation 17805 marks in UTF-8 locales. (Non-English messages already used the 17806 quotes appropriate for the language in previous releases.) If your 17807 terminal does not support UTF-8 but you are using a UTF-8 locale 17808 (such locales are the default on many GNU/Linux systems) then you 17809 should set LC_CTYPE=C in the environment to disable that locale. 17810 Programs that parse diagnostics and expect plain ASCII 17811 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 17812 explanation of Unicode quotation marks for more information. 17813 * The specs file is no longer installed on most platforms. Most users 17814 will be totally unaffected. However, if you are accustomed to 17815 editing the specs file yourself, you will now have to use the 17816 -dumpspecs option to generate the specs file, and then edit the 17817 resulting file. 17818 17819 General Optimizer Improvements 17820 17821 * The [3]tree ssa branch has been merged. This merge has brought in a 17822 completely new optimization framework based on a higher level 17823 intermediate representation than the existing RTL representation. 17824 Numerous new code transformations based on the new framework are 17825 available in GCC 4.0, including: 17826 + Scalar replacement of aggregates 17827 + Constant propagation 17828 + Value range propagation 17829 + Partial redundancy elimination 17830 + Load and store motion 17831 + Strength reduction 17832 + Dead store elimination 17833 + Dead and unreachable code elimination 17834 + [4]Autovectorization 17835 + Loop interchange 17836 + Tail recursion by accumulation 17837 Many of these passes outperform their counterparts from previous 17838 GCC releases. 17839 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 17840 scheduling optimization intended for loops that perform heavy 17841 computations. 17842 17843 New Languages and Language specific improvements 17844 17845 C family 17846 17847 * The sentinel attribute has been added to GCC. This function 17848 attribute allows GCC to warn when variadic functions such as execl 17849 are not NULL terminated. See the GCC manual for a complete 17850 description of its behavior. 17851 * Given __attribute__((alias("target"))) it is now an error if target 17852 is not a symbol, defined in the same translation unit. This also 17853 applies to aliases created by #pragma weak alias=target. This is 17854 because it's meaningless to define an alias to an undefined symbol. 17855 On Solaris, the native assembler would have caught this error, but 17856 GNU as does not. 17857 17858 C and Objective-C 17859 17860 * The -Wstrict-aliasing=2 option has been added. This warning catches 17861 all unsafe cases, but it may also give a warning for some cases 17862 that are safe. 17863 * The cast-as-lvalue, conditional-expression-as-lvalue and 17864 compound-expression-as-lvalue extensions, which were deprecated in 17865 3.3.4 and 3.4, have been removed. 17866 * The -fwritable-strings option, which was deprecated in 3.4, has 17867 been removed. 17868 * #pragma pack() semantics have been brought closer to those used by 17869 other compilers. This also applies to C++. 17870 * Taking the address of a variable with register storage is invalid 17871 in C. GCC now issues an error instead of a warning. 17872 * Arrays of incomplete element type are invalid in C. GCC now issues 17873 an error for such arrays. Declarations such as extern struct s x[]; 17874 (where struct s has not been defined) can be moved after the 17875 definition of struct s. Function parameters declared as arrays of 17876 incomplete type can instead be declared as pointers. 17877 17878 C++ 17879 17880 * When compiling without optimizations (-O0), the C++ front end is 17881 much faster than in any previous versions of GCC. Independent 17882 testers have measured speed-ups up to 25% in real-world production 17883 code, compared to the 3.4 family (which was already the fastest 17884 version to date). Upgrading from older versions might show even 17885 bigger improvements. 17886 * ELF visibility attributes can now be applied to a class type, so 17887 that it affects every member function of a class at once, without 17888 having to specify each individually: 17889 class __attribute__ ((visibility("hidden"))) Foo 17890 { 17891 int foo1(); 17892 void foo2(); 17893 }; 17894 The syntax is deliberately similar to the __declspec() system used 17895 by Microsoft Windows based compilers, allowing cross-platform 17896 projects to easily reuse their existing macro system for denoting 17897 exports and imports. By explicitly marking internal classes never 17898 used outside a binary as hidden, one can completely avoid PLT 17899 indirection overheads during their usage by the compiler. You can 17900 find out more about the advantages of this at 17901 [6]https://www.akkadia.org/drepper/dsohowto.pdf 17902 * The -fvisibility-inlines-hidden option has been added which marks 17903 all inlineable functions as having hidden ELF visibility, thus 17904 removing their symbol and typeinfo from the exported symbol table 17905 of the output ELF binary. Using this option can reduce the exported 17906 symbol count of template-heavy code by up to 40% with no code 17907 change at all, thus notably improving link and load times for the 17908 binary as well as a reduction in size of up to 10%. Also, check the 17909 new [7]-fvisibility option. 17910 * The compiler now uses the library interface specified by the [8]C++ 17911 ABI for thread-safe initialization of function-scope static 17912 variables. Most users should leave this alone, but embedded 17913 programmers may want to disable this by specifying 17914 -fno-threadsafe-statics for a small savings in code size. 17915 * Taking the address of an explicit register variable is no longer 17916 supported. Note that C++ allows taking the address of variables 17917 with register storage so this will continue to compile with a 17918 warning. For example, assuming that r0 is a machine register: 17919 register int foo asm ("r0"); 17920 register int bar; 17921 &foo; // error, no longer accepted 17922 &bar; // OK, with a warning 17923 * G++ has an undocumented extension to virtual function covariancy 17924 rules that allowed the overrider to return a type that was 17925 implicitly convertable to the overridden function's return type. 17926 For instance a function returning void * could be overridden by a 17927 function returning T *. This is now deprecated and will be removed 17928 in a future release. 17929 * The G++ minimum and maximum operators (<? and >?) and their 17930 compound forms (<?=) and >?=) have been deprecated and will be 17931 removed in a future version. Code using these operators should be 17932 modified to use std::min and std::max instead. 17933 * Declaration of nested classes of class templates as friends are 17934 supported: 17935 template <typename T> struct A { 17936 class B {}; 17937 }; 17938 class C { 17939 template <typename T> friend class A<T>::B; 17940 }; 17941 This complements the feature member functions of class templates as 17942 friends introduced in GCC 3.4.0. 17943 * When declaring a friend class using an unqualified name, classes 17944 outside the innermost non-class scope are not searched: 17945 class A; 17946 namespace N { 17947 class B { 17948 friend class A; // Refer to N::A which has not been declared yet 17949 // because name outside namespace N are not searched 17950 friend class ::A; // Refer to ::A 17951 }; 17952 } 17953 Hiding the friend name until declaration is still not implemented. 17954 * Friends of classes defined outside their namespace are correctly 17955 handled: 17956 namespace N { 17957 class A; 17958 } 17959 class N::A { 17960 friend class B; // Refer to N::B in GCC 4.0.0 17961 // but ::B in earlier versions of GCC 17962 }; 17963 17964 Runtime Library (libstdc++) 17965 17966 * Optimization work: 17967 + Added efficient specializations of istream functions for char 17968 and wchar_t. 17969 + Further performance tuning of strings, in particular wrt 17970 single-char append and getline. 17971 + iter_swap - and therefore most of the mutating algorithms - 17972 now makes an unqualified call to swap when the value_type of 17973 the two iterators is the same. 17974 * A large subset of the features in Technical Report 1 (TR1 for 17975 short) is experimentally delivered (i.e., no guarantees about the 17976 implementation are provided. In particular it is not promised that 17977 the library will remain link-compatible when code using TR1 is 17978 used): 17979 + General utilities such as reference_wrapper and shared_ptr. 17980 + Function objects, i.e., result_of, mem_fn, bind, function. 17981 + Support for metaprogramming. 17982 + New containers such as tuple, array, unordered_set, 17983 unordered_map, unordered_multiset, unordered_multimap. 17984 * As usual, many bugs have been fixed and LWG resolutions implemented 17985 for the first time (e.g., DR 409). 17986 17987 Java 17988 17989 * In order to prevent naming conflicts with other implementations of 17990 these tools, some GCJ binaries have been renamed: 17991 + rmic is now grmic, 17992 + rmiregistry is now grmiregistry, and 17993 + jar is now fastjar. 17994 In particular, these names were problematic for the jpackage.org 17995 packaging conventions which install symlinks in /usr/bin that point 17996 to the preferred versions of these tools. 17997 * The -findirect-dispatch argument to the compiler now works and 17998 generates code following a new "binary compatibility" ABI. Code 17999 compiled this way follows the binary compatibility rules of the 18000 Java Language Specification. 18001 * libgcj now has support for using GCJ as a JIT, using the 18002 gnu.gcj.jit family of system properties. 18003 * libgcj can now find a shared library corresponding to the bytecode 18004 representation of a class. See the documentation for the new 18005 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 18006 property. 18007 * There have been many improvements to the class library. Here are 18008 some highlights: 18009 + Much more of AWT and Swing exist. 18010 + Many new packages and classes were added, including 18011 java.util.regex, java.net.URI, javax.crypto, 18012 javax.crypto.interfaces, javax.crypto.spec, javax.net, 18013 javax.net.ssl, javax.security.auth, 18014 javax.security.auth.callback, javax.security.auth.login, 18015 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 18016 javax.imageio, javax.imageio.event, javax.imageio.spi, 18017 javax.print, javax.print.attribute, 18018 javax.print.attribute.standard, javax.print.event, and 18019 javax.xml 18020 + Updated SAX and DOM, and imported GNU JAXP 18021 18022 Fortran 18023 18024 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 18025 front end. The new front end supports Fortran 90 and Fortran 95. It 18026 may not yet be as stable as the old Fortran front end. 18027 18028 Ada 18029 18030 * Ada (with tasking and Zero Cost Exceptions) is now available on 18031 many more targets, including but not limited to: alpha-linux, 18032 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 18033 s390x-linux, sparc-linux. 18034 * Some of the new Ada 2005 features are now implemented like 18035 Wide_Wide_Character and Ada.Containers. 18036 * Many bugs have been fixed, tools and documentation improved. 18037 * To compile Ada from the sources, install an older working Ada 18038 compiler and then use --enable-languages=ada at configuration time, 18039 since the Ada front end is not currently activated by default. See 18040 the [10]Installing GCC for details. 18041 18042 [11]New Targets and Target Specific Improvements 18043 18044 H8/300 18045 18046 * The frame layout has changed. In the new layout, the prologue of a 18047 function first saves registers and then allocate space for locals, 18048 resulting in an 1% improvement on code size. 18049 18050 IA-32/x86-64 (AMD64) 18051 18052 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 18053 log1p, log2, logb and tan mathematical builtins (and their float 18054 and long double variants) are now implemented as inline x87 18055 intrinsics when using -ffast-math. 18056 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 18057 (and their float and long double variants) are now implemented as 18058 inline x87 intrinsics when using -ffast-math. 18059 * The x87's fsincos instruction is now used automatically with 18060 -ffast-math when calculating both the sin and cos of the same 18061 argument. 18062 * Instruction selection for multiplication and division by constants 18063 has been improved. 18064 18065 IA-64 18066 18067 * Floating point division, integer division and sqrt are now inlined, 18068 resulting in significant performance improvements on some codes. 18069 18070 MIPS 18071 18072 * Division by zero checks now use conditional traps if the target 18073 processor supports them. This decreases code size by one word per 18074 division operation. The old behavior (branch and break) can be 18075 obtained either at configure time by passing --with-divide=breaks 18076 to configure or at runtime by passing -mdivide-breaks to GCC. 18077 * Support for MIPS64 paired-single instructions has been added. It is 18078 enabled by -mpaired-single and can be accessed using both the 18079 target-independent vector extensions and new MIPS-specific built-in 18080 functions. 18081 * Support for the MIPS-3D ASE has been added. It is enabled by 18082 -mips3d and provides new MIPS-3D-specific built-in functions. 18083 * The -mexplicit-relocs option now supports static n64 code (as is 18084 used, for example, in 64-bit linux kernels). -mexplicit-relocs 18085 should now be feature-complete and is enabled by default when GCC 18086 is configured to use a compatible assembler. 18087 * Support for the NEC VR4130 series has been added. This support 18088 includes the use of VR-specific instructions and a new VR4130 18089 scheduler. Full VR4130 support can be selected with -march=vr4130 18090 while code for any ISA can be tuned for the VR4130 using 18091 -mtune=vr4130. There is also a new -mvr4130-align option that 18092 produces better schedules at the cost of increased code size. 18093 * Support for the Broadcom SB-1 has been extended. There is now an 18094 SB-1 scheduler as well as support for the SB-1-specific 18095 paired-single instructions. Full SB-1 support can be selected with 18096 -march=sb1 while code for any ISA can be optimized for the SB-1 18097 using -mtune=sb1. 18098 * The compiler can now work around errata in R4000, R4400, VR4120 and 18099 VR4130 processors. These workarounds are enabled by -mfix-r4000, 18100 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 18101 and VR4130 workarounds need binutils 2.16 or above. 18102 * IRIX shared libraries are now installed into the standard library 18103 directories: o32 libraries go into lib/, n32 libraries go into 18104 lib32/ and n64 libraries go into lib64/. 18105 * The compiler supports a new -msym32 option. It can be used to 18106 optimize n64 code in which all symbols are known to have 32-bit 18107 values. 18108 18109 S/390 and zSeries 18110 18111 * New command-line options help to generate code intended to run in 18112 an environment where stack space is restricted, e.g. Linux kernel 18113 code: 18114 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 18115 warnings for single functions that require large or dynamic 18116 stack frames. 18117 + -mstack-size and -mstack-guard generate code that checks for 18118 stack overflow at run time. 18119 + -mpacked-stack generates code that reduces the stack frame 18120 size of many functions by reusing unneeded parts of the stack 18121 bias area. 18122 * The -msoft-float option now ensures that generated code never 18123 accesses floating point registers. 18124 * The s390x-ibm-tpf target now fully supports C++, including 18125 exceptions and threads. 18126 * Various changes to improve performance of the generated code have 18127 been implemented, including: 18128 + GCC now uses sibling calls where possible. 18129 + Condition code handling has been optimized, allowing GCC to 18130 omit redundant comparisons in certain cases. 18131 + The cost function guiding many optimizations has been refined 18132 to more accurately represent the z900 and z990 processors. 18133 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 18134 instructions are now used to avoid conditional branches in 18135 certain cases. 18136 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 18137 optimize address arithmetic required to access large stack 18138 frames. 18139 + GCC now makes more efficient use of memory-to-memory type 18140 instructions (MVC, CLC, ...). 18141 + More precise tracking of special register use allows better 18142 instruction scheduling, in particular of the function prologue 18143 and epilogue sequences. 18144 + The Java front end now generates inline code to implement 18145 integer division, instead of calling library routines. 18146 18147 SPARC 18148 18149 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 18150 -mf934 have been removed. They have been replaced with -mcpu=xxx. 18151 * The internal model used to estimate the relative cost of each 18152 instruction has been updated. It is expected to give better results 18153 on recent UltraSPARC processors. 18154 * Code generation for function prologues and epilogues has been 18155 improved, resulting in better scheduling and allowing multiple exit 18156 points in functions. 18157 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 18158 It is enabled by -mvis and provides new built-in functions for VIS 18159 instructions on UltraSPARC processors. 18160 * The option -mapp-regs has been turned on by default on Solaris too. 18161 18162 NetWare 18163 18164 * Novell NetWare (on ix86, no other hardware platform was ever really 18165 supported by this OS) has been re-enabled and the ABI supported by 18166 GCC has been brought into sync with that of MetroWerks CodeWarrior 18167 (the ABI previously supported was that of some Unix systems, which 18168 NetWare never tried to support). 18169 18170 [12]Obsolete Systems 18171 18172 Support for a number of older systems has been declared obsolete in GCC 18173 4.0. Unless there is activity to revive them, the next release of GCC 18174 will have their sources permanently removed. 18175 18176 All GCC ports for the following processor architectures have been 18177 declared obsolete: 18178 * Intel i860 18179 * Ubicom IP2022 18180 * National Semiconductor NS32K (ns32k) 18181 * Texas Instruments TMS320C[34]x 18182 18183 Also, those for some individual systems have been obsoleted: 18184 * SPARC family 18185 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 18186 sparc86x-*-elf) 18187 + OpenBSD 32-bit (sparc-*-openbsd*) 18188 18189 Documentation improvements 18190 18191 Other significant improvements 18192 18193 * Location lists are now generated by default when compiling with 18194 debug info and optimization. Location lists provide more accurate 18195 debug info about locations of variables and they allow debugging 18196 code compiled with -fomit-frame-pointer. 18197 * The -fvisibility option has been added which allows the default ELF 18198 visibility of all symbols to be set per compilation and the new 18199 #pragma GCC visibility preprocessor command allows the setting of 18200 default ELF visibility for a region of code. Using 18201 -fvisibility=hidden especially in combination with the new 18202 -fvisibility-inlines-hidden can yield substantial improvements in 18203 output binary quality including avoiding PLT indirection overheads, 18204 reduction of the exported symbol count by up to 60% (with resultant 18205 improvements to link and load times), better scope for the 18206 optimizer to improve code and up to a 20% reduction in binary size. 18207 Using these options correctly yields a binary with a similar symbol 18208 count to a Windows DLL. 18209 Perhaps more importantly, this new feature finally allows (with 18210 careful planning) complete avoidance of symbol clashes when 18211 manually loading shared objects with RTLD_GLOBAL, thus finally 18212 solving problems many projects such as python were forced to use 18213 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 18214 can find more information about using these options at 18215 [13]https://gcc.gnu.org/wiki/Visibility. 18216 __________________________________________________________________ 18217 18218 [14]GCC 4.0.1 18219 18220 This is the [15]list of problem reports (PRs) from GCC's bug tracking 18221 system that are known to be fixed in the 4.0.1 release. This list might 18222 not be complete (that is, it is possible that some PRs that have been 18223 fixed are not listed here). 18224 18225 [16]GCC 4.0.2 18226 18227 This is the [17]list of problem reports (PRs) from GCC's bug tracking 18228 system that are known to be fixed in the 4.0.2 release. This list might 18229 not be complete (that is, it is possible that some PRs that have been 18230 fixed are not listed here). 18231 18232 Unfortunately, due to a release engineering failure, this release has a 18233 regression on Solaris that will affect some C++ programs. We suggest 18234 that Solaris users apply a [18]patch that corrects the problem. Users 18235 who do not wish to apply the patch should explicitly link C++ programs 18236 with the -pthreads option, even if they do not use threads. This 18237 problem has been corrected in the current 4.0 branch sources and will 18238 not be present in GCC 4.0.3. 18239 18240 [19]GCC 4.0.3 18241 18242 Starting with this release, the function getcontext is recognized by 18243 the compiler as having the same semantics as the setjmp function. In 18244 particular, the compiler will ensure that all registers are dead before 18245 calling such a function and will emit a warning about the variables 18246 that may be clobbered after the second return from the function. 18247 18248 [20]GCC 4.0.4 18249 18250 This is the [21]list of problem reports (PRs) from GCC's bug tracking 18251 system that are known to be fixed in the 4.0.4 release. This list might 18252 not be complete (that is, it is possible that some PRs that have been 18253 fixed are not listed here). 18254 18255 The 4.0.4 release is provided for those that require a high degree of 18256 binary compatibility with previous 4.0.x releases. For most users, the 18257 GCC team recommends that version 4.1.1 or later be used instead." 18258 18259 18260 For questions related to the use of GCC, please consult these web 18261 pages and the [22]GCC manuals. If that fails, the 18262 [23]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18263 web pages and the development of GCC are welcome on our developer 18264 list at [24]gcc (a] gcc.gnu.org. All of [25]our lists have public 18265 archives. 18266 18267 Copyright (C) [26]Free Software Foundation, Inc. Verbatim copying and 18268 distribution of this entire article is permitted in any medium, 18269 provided this notice is preserved. 18270 18271 These pages are [27]maintained by the GCC team. Last modified 18272 2025-01-31. 18273 18274 References 18275 18276 1. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 18277 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 18278 3. https://gcc.gnu.org/projects/tree-ssa/ 18279 4. https://gcc.gnu.org/projects/tree-ssa/vectorization.html 18280 5. https://gcc.gnu.org/news/sms.html 18281 6. https://www.akkadia.org/drepper/dsohowto.pdf 18282 7. https://gcc.gnu.org/gcc-4.0/changes.html#visibility 18283 8. https://itanium-cxx-abi.github.io/cxx-abi/ 18284 9. https://gcc.gnu.org/fortran/ 18285 10. https://gcc.gnu.org/install/ 18286 11. https://gcc.gnu.org/gcc-4.0/changes.html#targets 18287 12. https://gcc.gnu.org/gcc-4.0/changes.html#obsolete_systems 18288 13. https://gcc.gnu.org/wiki/Visibility 18289 14. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.1 18290 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 18291 16. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.2 18292 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 18293 18. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 18294 19. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.3 18295 20. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.4 18296 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 18297 22. https://gcc.gnu.org/onlinedocs/ 18298 23. mailto:gcc-help (a] gcc.gnu.org 18299 24. mailto:gcc (a] gcc.gnu.org 18300 25. https://gcc.gnu.org/lists.html 18301 26. https://www.fsf.org/ 18302 27. https://gcc.gnu.org/about.html 18303 ====================================================================== 18304 http://gcc.gnu.org/gcc-3.4/index.html 18305 GCC 3.4 Release Series 18306 18307 (This release series is no longer supported.) 18308 18309 May 26, 2006 18310 18311 The [1]GNU project and the GCC developers are pleased to announce the 18312 release of GCC 3.4.6. 18313 18314 This release is a bug-fix release, containing fixes for regressions in 18315 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 18316 3.4.x series. 18317 18318 The GCC 3.4 release series includes numerous [2]new features, 18319 improvements, bug fixes, and other changes, thanks to an [3]amazing 18320 group of volunteers. 18321 18322 Release History 18323 18324 GCC 3.4.6 18325 March 6, 2006 ([4]changes) 18326 18327 GCC 3.4.5 18328 November 30, 2005 ([5]changes) 18329 18330 GCC 3.4.4 18331 May 18, 2005 ([6]changes) 18332 18333 GCC 3.4.3 18334 November 4, 2004 ([7]changes) 18335 18336 GCC 3.4.2 18337 September 6, 2004 ([8]changes) 18338 18339 GCC 3.4.1 18340 July 1, 2004 ([9]changes) 18341 18342 GCC 3.4.0 18343 April 18, 2004 ([10]changes) 18344 18345 References and Acknowledgements 18346 18347 GCC used to stand for the GNU C Compiler, but since the compiler 18348 supports several other languages aside from C, it now stands for the 18349 GNU Compiler Collection. 18350 18351 A list of [11]successful builds is updated as new information becomes 18352 available. 18353 18354 The GCC developers would like to thank the numerous people that have 18355 contributed new features, improvements, bug fixes, and other changes as 18356 well as test results to GCC. This [12]amazing group of volunteers is 18357 what makes GCC successful. 18358 18359 For additional information about GCC please refer to the [13]GCC 18360 project web site or contact the [14]GCC development mailing list. 18361 18362 To obtain GCC please use [15]our mirror sites, or [16]our version 18363 control system. 18364 18365 18366 For questions related to the use of GCC, please consult these web 18367 pages and the [17]GCC manuals. If that fails, the 18368 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18369 web pages and the development of GCC are welcome on our developer 18370 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 18371 archives. 18372 18373 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 18374 distribution of this entire article is permitted in any medium, 18375 provided this notice is preserved. 18376 18377 These pages are [22]maintained by the GCC team. Last modified 18378 2022-10-26. 18379 18380 References 18381 18382 1. http://www.gnu.org/ 18383 2. https://gcc.gnu.org/gcc-3.4/changes.html 18384 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18385 4. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 18386 5. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 18387 6. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 18388 7. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 18389 8. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 18390 9. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 18391 10. https://gcc.gnu.org/gcc-3.4/changes.html 18392 11. https://gcc.gnu.org/gcc-3.4/buildstat.html 18393 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18394 13. https://gcc.gnu.org/index.html 18395 14. mailto:gcc (a] gcc.gnu.org 18396 15. https://gcc.gnu.org/mirrors.html 18397 16. https://gcc.gnu.org/git.html 18398 17. https://gcc.gnu.org/onlinedocs/ 18399 18. mailto:gcc-help (a] gcc.gnu.org 18400 19. mailto:gcc (a] gcc.gnu.org 18401 20. https://gcc.gnu.org/lists.html 18402 21. https://www.fsf.org/ 18403 22. https://gcc.gnu.org/about.html 18404 ====================================================================== 18405 http://gcc.gnu.org/gcc-3.4/changes.html 18406 GCC 3.4 Release Series 18407 Changes, New Features, and Fixes 18408 18409 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 18410 is now closed. 18411 18412 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 18413 a bug, please make sure it's really GCC, and not your code, that is 18414 broken. 18415 18416 Caveats 18417 18418 * GNU Make is now required to build GCC. 18419 * With -nostdinc the preprocessor used to ignore both standard 18420 include paths and include paths contained in environment variables. 18421 It was neither documented nor intended that environment variable 18422 paths be ignored, so this has been corrected. 18423 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 18424 -fvolatile-static. It is unlikely that they worked correctly in any 18425 3.x release. 18426 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 18427 * Support for all the systems [3]obsoleted in GCC 3.3 has been 18428 removed from GCC 3.4. See below for a [4]list of systems which are 18429 obsoleted in this release. 18430 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 18431 compilers will not work. 18432 * The implementation of the [5]MIPS ABIs has changed. As a result, 18433 the code generated for certain MIPS targets will not be binary 18434 compatible with earlier releases. 18435 * In previous releases, the MIPS port had a fake "hilo" register with 18436 the user-visible name accum. This register has been removed. 18437 * The implementation of the [6]SPARC ABIs has changed. As a result, 18438 the code generated will not be binary compatible with earlier 18439 releases in certain cases. 18440 * The configure option --enable-threads=pthreads has been removed; 18441 use --enable-threads=posix instead, which should have the same 18442 effect. 18443 * Code size estimates used by inlining heuristics for C, Objective-C, 18444 C++ and Java have been redesigned significantly. As a result the 18445 parameters of -finline-insns, --param max-inline-insns-single and 18446 --param max-inline-insns-auto need to be reconsidered. 18447 * --param max-inline-slope and --param min-inline-insns have been 18448 removed; they are not needed for the new bottom-up inlining 18449 heuristics. 18450 * The new unit-at-a-time compilation scheme has several compatibility 18451 issues: 18452 + The order in which functions, variables, and top-level asm 18453 statements are emitted may have changed. Code relying on some 18454 particular ordering needs to be updated. The majority of such 18455 top-level asm statements can be replaced by section 18456 attributes. 18457 + Unreferenced static variables and functions are removed. This 18458 may result in undefined references when an asm statement 18459 refers to the variable/function directly. In that case either 18460 the variable/function shall be listed in asm statement operand 18461 or in the case of top-level asm statements the attribute used 18462 shall be used to force function/variable to be always output 18463 and considered as a possibly used by unknown code. 18464 For variables the attribute is accepted only by GCC 3.4 and 18465 newer, while for earlier versions it is sufficient to use 18466 unused to silence warnings about the variables not being 18467 referenced. To keep code portable across different GCC 18468 versions, you can use appropriate preprocessor conditionals. 18469 + Static functions now can use non-standard passing conventions 18470 that may break asm statements calling functions directly. 18471 Again the attribute used shall be used to prevent this 18472 behavior. 18473 As a temporary workaround, -fno-unit-at-a-time can be used, but 18474 this scheme may not be supported by future releases of GCC. 18475 * GCC 3.4 automatically places zero-initialized variables in the .bss 18476 section on some operating systems. Versions of GNU Emacs up to (and 18477 including) 21.3 will not work correctly when using this 18478 optimization; you can use -fno-zero-initialized-in-bss to disable 18479 it. 18480 * If GCC 3.4 is configured with --enable-threads=posix (the default 18481 on most targets that support pthreads) then _REENTRANT will be 18482 defined unconditionally by some libstdc++ headers. C++ code which 18483 relies on that macro to detect whether multi-threaded code is being 18484 compiled might change in meaning, possibly resulting in linker 18485 errors for single-threaded programs. Affected users of [7]Boost 18486 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 18487 See Bugzilla for [8]more information. 18488 18489 General Optimizer Improvements 18490 18491 * Usability of the profile feedback and coverage testing has been 18492 improved. 18493 + Performance of profiled programs has been improved by faster 18494 profile merging code. 18495 + Better use of the profile feedback for optimization (loop 18496 unrolling and loop peeling). 18497 + File locking support allowing fork() calls and parallel runs 18498 of profiled programs. 18499 + Coverage file format has been redesigned. 18500 + gcov coverage tool has been improved. 18501 + make profiledbootstrap available to build a faster compiler. 18502 Experiments made on i386 hardware showed an 11% speedup on -O0 18503 and a 7.5% speedup on -O2 compilation of a [9]large C++ 18504 testcase. 18505 + New value profiling pass enabled via -fprofile-values 18506 + New value profile transformations pass enabled via -fvpt aims 18507 to optimize some code sequences by exploiting knowledge about 18508 value ranges or other properties of the operands. At the 18509 moment a conversion of expensive divisions into cheaper 18510 operations has been implemented. 18511 + New -fprofile-generate and -fprofile-use command-line options 18512 to simplify the use of profile feedback. 18513 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 18514 Java which is enabled via -funit-at-a-time (and implied by -O2). In 18515 this scheme a whole file is parsed first and optimized later. The 18516 following basic inter-procedural optimizations are implemented: 18517 + Removal of unreachable functions and variables 18518 + Discovery of local functions (functions with static linkage 18519 whose address is never taken) 18520 + On i386, these local functions use register parameter passing 18521 conventions. 18522 + Reordering of functions in topological order of the call graph 18523 to enable better propagation of optimizing hints (such as the 18524 stack alignments needed by functions) in the back end. 18525 + Call graph based out-of-order inlining heuristics which allows 18526 to limit overall compilation unit growth (--param 18527 inline-unit-growth). 18528 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 18529 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 18530 CPU). 18531 * More realistic code size estimates used by inlining for C, 18532 Objective-C, C++ and Java. The growth of large functions can now be 18533 limited via --param large-function-insns and --param 18534 large-function-growth. 18535 * A new cfg-level loop optimizer pass replaces the old loop unrolling 18536 pass and adds two other loop transformations -- loop peeling and 18537 loop unswitching -- and also uses the profile feedback to limit 18538 code growth. (The three optimizations are enabled by 18539 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 18540 respectively). 18541 The old loop unroller still can be enabled by -fold-unroll-loops 18542 and may produce better code in some cases, especially when the 18543 webizer optimization pass is not run. 18544 * A new web construction pass enabled via -fweb (and implied by -O3) 18545 improves the quality of register allocation, CSE, first scheduling 18546 pass and some other optimization passes by avoiding re-use of 18547 pseudo registers with non-overlapping live ranges. The pass almost 18548 always improves code quality but does make debugging difficult and 18549 thus is not enabled by default by -O2 18550 The pass is especially effective as cleanup after code duplication 18551 passes, such as the loop unroller or the tracer. 18552 * Experimental implementations of superblock or trace scheduling in 18553 the second scheduling pass can be enabled via 18554 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 18555 18556 New Languages and Language specific improvements 18557 18558 Ada 18559 18560 * The Ada front end has been updated to include numerous bug fixes 18561 and enhancements. These include: 18562 + Improved project file support 18563 + Additional set of warnings about potential wrong code 18564 + Improved error messages 18565 + Improved code generation 18566 + Improved cross reference information 18567 + Improved inlining 18568 + Better run-time check elimination 18569 + Better error recovery 18570 + More efficient implementation of unbounded strings 18571 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 18572 ... 18573 + New GNAT.xxxx packages (e.g. GNAT.Strings, 18574 GNAT.Exception_Action) 18575 + New pragmas 18576 + New -gnatS switch replacing gnatpsta 18577 + Implementation of new Ada features (in particular limited 18578 with, limited aggregates) 18579 18580 C/Objective-C/C++ 18581 18582 * Precompiled headers are now supported. Precompiled headers can 18583 dramatically speed up compilation of some projects. There are some 18584 known defects in the current precompiled header implementation that 18585 will result in compiler crashes in relatively rare situations. 18586 Therefore, precompiled headers should be considered a "technology 18587 preview" in this release. Read the manual for details about how to 18588 use precompiled headers. 18589 * File handling in the preprocessor has been rewritten. GCC no longer 18590 gets confused by symlinks and hardlinks, and now has a correct 18591 implementation of #import and #pragma once. These two directives 18592 have therefore been un-deprecated. 18593 * The undocumented extension that allowed C programs to have a label 18594 at the end of a compound statement, which has been deprecated since 18595 GCC 3.0, has been removed. 18596 * The cast-as-lvalue extension has been removed for C++ and 18597 deprecated for C and Objective-C. In particular, code like this: 18598 int i; 18599 (char) i = 5; 18600 18601 or this: 18602 char *p; 18603 ((int *) p)++; 18604 18605 is no longer accepted for C++ and will not be accepted for C and 18606 Objective-C in a future version. 18607 * The conditional-expression-as-lvalue extension has been deprecated 18608 for C and Objective-C. In particular, code like this: 18609 int a, b, c; 18610 (a ? b : c) = 2; 18611 18612 will not be accepted for C and Objective-C in a future version. 18613 * The compound-expression-as-lvalue extension has been deprecated for 18614 C and Objective-C. In particular, code like this: 18615 int a, b; 18616 (a, b) = 2; 18617 18618 will not be accepted for C and Objective-C in a future version. A 18619 possible non-intrusive workaround is the following: 18620 (*(a, &b)) = 2; 18621 18622 * Several [10]built-in functions such as __builtin_popcount for 18623 counting bits, finding the highest and lowest bit in a word, and 18624 parity have been added. 18625 * The -fwritable-strings option has been deprecated and will be 18626 removed. 18627 * Many C math library functions are now recognized as built-ins and 18628 optimized. 18629 * The C, C++, and Objective-C compilers can now handle source files 18630 written in any character encoding supported by the host C library. 18631 The default input character set is taken from the current locale, 18632 and may be overridden with the -finput-charset command line option. 18633 In the future we will add support for inline encoding markers. 18634 18635 [11]C++ 18636 18637 * G++ is now much closer to full conformance to the ISO/ANSI C++ 18638 standard. This means, among other things, that a lot of invalid 18639 constructs which used to be accepted in previous versions will now 18640 be rejected. It is very likely that existing C++ code will need to 18641 be fixed. This document lists some of the most common issues. 18642 * A hand-written recursive-descent C++ parser has replaced the 18643 YACC-derived C++ parser from previous GCC releases. The new parser 18644 contains much improved infrastructure needed for better parsing of 18645 C++ source codes, handling of extensions, and clean separation 18646 (where possible) between proper semantics analysis and parsing. The 18647 new parser fixes many bugs that were found in the old parser. 18648 * You must now use the typename and template keywords to disambiguate 18649 dependent names, as required by the C++ standard. 18650 struct K { 18651 typedef int mytype_t; 18652 }; 18653 18654 template <class T1> struct A { 18655 template <class T2> struct B { 18656 void callme(void); 18657 }; 18658 18659 template <int N> void bar(void) 18660 { 18661 // Use 'typename' to tell the parser that T1::mytype_t names 18662 // a type. This is needed because the name is dependent (in 18663 // this case, on template parameter T1). 18664 typename T1::mytype_t x; 18665 x = 0; 18666 } 18667 }; 18668 18669 template <class T> void template_func(void) 18670 { 18671 // Use 'template' to prefix member templates within 18672 // dependent types (a has type A<T>, which depends on 18673 // the template parameter T). 18674 A<T> a; 18675 a.template bar<0>(); 18676 18677 // Use 'template' to tell the parser that B is a nested 18678 // template class (dependent on template parameter T), and 18679 // 'typename' because the whole A<T>::B<int> is 18680 // the name of a type (again, dependent). 18681 typename A<T>::template B<int> b; 18682 b.callme(); 18683 } 18684 18685 void non_template_func(void) 18686 { 18687 // Outside of any template class or function, no names can be 18688 // dependent, so the use of the keyword 'typename' and 'template' 18689 // is not needed (and actually forbidden). 18690 A<K> a; 18691 a.bar<0>(); 18692 A<K>::B<float> b; 18693 b.callme(); 18694 } 18695 * In a template definition, unqualified names will no longer find 18696 members of a dependent base (as specified by [temp.dep]/3 in the 18697 C++ standard). For example, 18698 template <typename T> struct B { 18699 int m; 18700 int n; 18701 int f (); 18702 int g (); 18703 }; 18704 int n; 18705 int g (); 18706 template <typename T> struct C : B<T> { 18707 void h () 18708 { 18709 m = 0; // error 18710 f (); // error 18711 n = 0; // ::n is modified 18712 g (); // ::g is called 18713 } 18714 }; 18715 You must make the names dependent, e.g. by prefixing them with 18716 this->. Here is the corrected definition of C<T>::h, 18717 template <typename T> void C<T>::h () 18718 { 18719 this->m = 0; 18720 this->f (); 18721 this->n = 0 18722 this->g (); 18723 } 18724 As an alternative solution (unfortunately not backwards compatible 18725 with GCC 3.3), you may use using declarations instead of this->: 18726 template <typename T> struct C : B<T> { 18727 using B<T>::m; 18728 using B<T>::f; 18729 using B<T>::n; 18730 using B<T>::g; 18731 void h () 18732 { 18733 m = 0; 18734 f (); 18735 n = 0; 18736 g (); 18737 } 18738 }; 18739 * In templates, all non-dependent names are now looked up and bound 18740 at definition time (while parsing the code), instead of later when 18741 the template is instantiated. For instance: 18742 void foo(int); 18743 18744 template <int> struct A { 18745 static void bar(void){ 18746 foo('a'); 18747 } 18748 }; 18749 18750 void foo(char); 18751 18752 int main() 18753 { 18754 A<0>::bar(); // Calls foo(int), used to call foo(char). 18755 } 18756 18757 * In an explicit instantiation of a class template, you must use 18758 class or struct before the template-id: 18759 template <int N> 18760 class A {}; 18761 18762 template A<0>; // error, not accepted anymore 18763 template class A<0>; // OK 18764 * The "named return value" and "implicit typename" extensions have 18765 been removed. 18766 * Default arguments in function types have been deprecated and will 18767 be removed. 18768 * ARM-style name-injection of friend declarations has been deprecated 18769 and will be removed. For example: struct S { friend void f(); }; 18770 void g() { f(); } will not be accepted by future versions of G++; 18771 instead a declaration of "f" will need to be present outside of the 18772 scope of "S". 18773 * Covariant returns are implemented for all but varadic functions 18774 that require an adjustment. 18775 * When -pedantic is used, G++ now issues errors about spurious 18776 semicolons. For example, 18777 namespace N {}; // Invalid semicolon. 18778 void f() {}; // Invalid semicolon. 18779 * G++ no longer accepts attributes for a declarator after the 18780 initializer associated with that declarator. For example, 18781 X x(1) __attribute__((...)); 18782 is no longer accepted. Instead, use: 18783 X x __attribute__((...)) (1); 18784 * Inside the scope of a template class, the name of the class itself 18785 can be treated as either a class or a template. So GCC used to 18786 accept the class name as argument of type template, and template 18787 template parameter. However this is not C++ standard compliant. Now 18788 the name is not treated as a valid template template argument 18789 unless you qualify the name by its scope. For example, the code 18790 below no longer compiles. 18791 template <template <class> class TT> class X {}; 18792 template <class T> class Y { 18793 X<Y> x; // Invalid, Y is always a type template parameter. 18794 }; 18795 The valid code for the above example is 18796 X< ::Y> x; // Valid. 18797 (Notice the space between < and : to prevent GCC to interpret this 18798 as a digraph for [.) 18799 * Friend declarations that refer to template specializations are 18800 rejected if the template has not already been declared. For 18801 example, 18802 template <typename T> 18803 class C { 18804 friend void f<> (C&); 18805 }; 18806 is rejected. You must first declare f as a template, 18807 template <typename T> 18808 void f(T); 18809 * In case of friend declarations, every name used in the friend 18810 declaration must be accessible at the point of that declaration. 18811 Previous versions of G++ used to be less strict about this and 18812 allowed friend declarations for private class members, for example. 18813 See the ISO C++ Standard Committee's [12]defect report #209 for 18814 details. 18815 * Declaration of member functions of class templates as friends are 18816 supported. For example, 18817 template <typename T> struct A { 18818 void f(); 18819 }; 18820 class C { 18821 template <typename T> friend void A<T>::f(); 18822 }; 18823 * You must use template <> to introduce template specializations, as 18824 required by the standard. For example, 18825 template <typename T> 18826 struct S; 18827 18828 struct S<int> { }; 18829 is rejected. You must write, 18830 template <> struct S<int> {}; 18831 * G++ used to accept code like this, 18832 struct S { 18833 int h(); 18834 void f(int i = g()); 18835 int g(int i = h()); 18836 }; 18837 This behavior is not mandated by the standard. Now G++ issues an 18838 error about this code. To avoid the error, you must move the 18839 declaration of g before the declaration of f. The default arguments 18840 for g must be visible at the point where it is called. 18841 * The C++ ABI Section 3.3.3 specifications for the array construction 18842 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 18843 NULL when the allocator argument returns NULL. These changes are 18844 incorporated into the libstdc++ runtime library. 18845 * Using a name introduced by a typedef in a friend declaration or in 18846 an explicit instantiation is now rejected, as specified by the ISO 18847 C++ standard. 18848 class A; 18849 typedef A B; 18850 class C { 18851 friend class B; // error, no typedef name here 18852 friend B; // error, friend always needs class/struct/enum 18853 friend class A; // OK 18854 }; 18855 18856 template <int> class Q {}; 18857 typedef Q<0> R; 18858 template class R; // error, no typedef name here 18859 template class Q<0>; // OK 18860 * When allocating an array with a new expression, GCC used to allow 18861 parentheses around the type name. This is actually ill-formed and 18862 it is now rejected: 18863 int* a = new (int)[10]; // error, not accepted anymore 18864 int* a = new int[10]; // OK 18865 * When binding an rvalue of class type to a reference, the copy 18866 constructor of the class must be accessible. For instance, consider 18867 the following code: 18868 class A 18869 { 18870 public: 18871 A(); 18872 18873 private: 18874 A(const A&); // private copy ctor 18875 }; 18876 18877 A makeA(void); 18878 void foo(const A&); 18879 18880 void bar(void) 18881 { 18882 foo(A()); // error, copy ctor is not accessible 18883 foo(makeA()); // error, copy ctor is not accessible 18884 18885 A a1; 18886 foo(a1); // OK, a1 is a lvalue 18887 } 18888 This might be surprising at first sight, especially since most 18889 popular compilers do not correctly implement this rule ([13]further 18890 details). 18891 * When forming a pointer to member or a pointer to member function, 18892 access checks for class visibility (public, protected, private) are 18893 now performed using the qualifying scope of the name itself. This 18894 is better explained with an example: 18895 class A 18896 { 18897 public: 18898 void pub_func(); 18899 protected: 18900 void prot_func(); 18901 private: 18902 void priv_func(); 18903 }; 18904 18905 class B : public A 18906 { 18907 public: 18908 void foo() 18909 { 18910 &A::pub_func; // OK, pub_func is accessible through A 18911 &A::prot_func; // error, cannot access prot_func through A 18912 &A::priv_func; // error, cannot access priv_func through A 18913 18914 &B::pub_func; // OK, pub_func is accessible through B 18915 &B::prot_func; // OK, can access prot_func through B (within B) 18916 &B::priv_func; // error, cannot access priv_func through B 18917 } 18918 }; 18919 18920 Runtime Library (libstdc++) 18921 18922 * Optimization work: 18923 + Streamlined streambuf, filebuf, separate synched with C 18924 Standard I/O streambuf. 18925 + All formatted I/O now uses cached locale information. 18926 + STL optimizations (memory/speed for list, red-black trees as 18927 used by sets and maps). 18928 + More use of GCC builtins. 18929 + String optimizations (avoid contention on 18930 increment/decrement-and-test of the reference count in the 18931 empty-string object, constructor from input_iterators 18932 speedup). 18933 * Static linkage size reductions. 18934 * Large File Support (files larger than 2 GB on 32-bit systems). 18935 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 18936 * Generic character traits. 18937 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 18938 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 18939 * The allocator class is now standard-conformant, and two additional 18940 extension allocators have been added, mt_alloc and 18941 bitmap_allocator. 18942 * PCH support: -include bits/stdc++.h (2x compile speedup). 18943 * Rewrote __cxa_demangle with support for C++ style allocators. 18944 * New debug modes for STL containers and iterators. 18945 * Testsuite rewrite: five times as many tests, plus increasingly 18946 sophisticated tests, including I/O, MT, multi-locale, wide and 18947 narrow characters. 18948 * Use current versions of GNU "autotools" for build/configuration. 18949 18950 Objective-C 18951 18952 * The Objective-C front end has been updated to include the numerous 18953 bug fixes and enhancements previously available only in Apple's 18954 version of GCC. These include: 18955 + Structured exception (@try... @catch... @finally, @throw) and 18956 synchronization (@synchronized) support. These are accessible 18957 via the -fobjc-exceptions switch; as of this writing, they may 18958 only be used in conjunction with -fnext-runtime on Mac OS X 18959 10.3 and later. See [14]Options Controlling Objective-C 18960 Dialect for more information. 18961 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 18962 may now be encoded as 'B'. In addition, the back-end/codegen 18963 dependencies have been removed. 18964 + An overhaul of message dispatch construction, ensuring that 18965 the various receiver types (and casts thereof) are handled 18966 properly, and that correct diagnostics are issued. 18967 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 18968 (-freplace-objc-classes) debugging modes, currently available 18969 on Mac OS X 10.3 and later. See [15]Options Controlling 18970 Objective-C Dialect for more information. 18971 + Access to optimized runtime entry points (-fno-nil-receivers ) 18972 on the assumption that message receivers are never nil. This 18973 is currently available on Mac OS X 10.3 and later. See 18974 [16]Options Controlling Objective-C Dialect for more 18975 information. 18976 18977 Java 18978 18979 * Compiling a .jar file will now cause non-.class entries to be 18980 automatically compiled as resources. 18981 * libgcj has been ported to Darwin. 18982 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 18983 to gcj. 18984 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 18985 code from shared libraries. 18986 * libgcj has been much more completely merged with [17]GNU Classpath. 18987 * Class loading is now much more correct; in particular the caller's 18988 class loader is now used when that is required. 18989 * [18]Eclipse 2.x will run out of the box using gij. 18990 * Parts of java.nio have been implemented. Direct and indirect 18991 buffers work, as do fundamental file and socket operations. 18992 * java.awt has been improved, though it is still not ready for 18993 general use. 18994 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 18995 method. 18996 * The MinGW port has matured. Enhancements include socket timeout 18997 support, thread interruption, improved Runtime.exec() handling and 18998 support for accented characters in filenames. 18999 19000 Fortran 19001 19002 * Fortran improvements are listed in the [19]Fortran documentation. 19003 19004 New Targets and Target Specific Improvements 19005 19006 Alpha 19007 19008 * Several [20]built-in functions have been added such as 19009 __builtin_alpha_zap to allow utilizing the more obscure 19010 instructions of the CPU. 19011 * Parameter passing of complex arguments has changed to match the 19012 ABI. This change is incompatible with previous GCC versions, but 19013 does fix compatibility with the Tru64 compiler and several corner 19014 cases where GCC was incompatible with itself. 19015 19016 ARM 19017 19018 * Nicolas Pitre has contributed his hand-coded floating-point support 19019 code for ARM. It is both significantly smaller and faster than the 19020 existing C-based implementation, even when building applications 19021 for Thumb. The arm-elf configuration has been converted to use the 19022 new code. 19023 * Support for the Intel's iWMMXt architecture, a second generation 19024 XScale processor, has been added. Enabled at run time with the 19025 -mcpu=iwmmxt command line switch. 19026 * A new ARM target has been added: arm-wince-pe. This is similar to 19027 the arm-pe target, but it defaults to using the APCS32 ABI. 19028 * The existing ARM pipeline description has been converted to the use 19029 the [21]DFA processor pipeline model. There is not much change in 19030 code performance, but the description is now [22]easier to 19031 understand. 19032 * Support for the Cirrus EP9312 Maverick floating point co-processor 19033 added. Enabled at run time with the -mcpu=ep9312 command line 19034 switch. Note however that the multilibs to support this chip are 19035 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 19036 enable their production you will have to uncomment the entries in 19037 that file. 19038 19039 H8/300 19040 19041 * Support for long long has been added. 19042 * Support for saveall attribute has been added. 19043 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 19044 for H8/300H and H8S, which is much faster than the previous 19045 implementation. 19046 * A lot of small performance improvements. 19047 19048 IA-32/AMD64 (x86-64) 19049 19050 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 19051 -march=k8 and -mcpu=k8. 19052 * Scalar SSE code generation carefully avoids reformatting penalties, 19053 hidden dependencies and minimizes the number of uops generated on 19054 both Intel and AMD CPUs. 19055 * Vector MMX and SSE operands are now passed in registers to improve 19056 performance and match the argument passing convention used by the 19057 Intel C++ Compiler. As a result it is not possible to call 19058 functions accepting vector arguments compiled by older GCC version. 19059 * Conditional jump elimination is now more aggressive on modern CPUs. 19060 * The Athlon ports has been converted to use the DFA processor 19061 pipeline description. 19062 * Optimization of indirect tail calls is now possible in a similar 19063 fashion as direct sibcall optimization. 19064 * Further small performance improvements. 19065 * -m128bit-long-double is now less buggy. 19066 * __float128 support in 64-bit compilation. 19067 * Support for data structures exceeding 2GB in 64-bit mode. 19068 * -mcpu has been renamed to -mtune. 19069 19070 IA-64 19071 19072 * Tuning code for the Itanium 2 processor has been added. The 19073 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 19074 enabled by default now. To generate code tuned for Itanium 1 the 19075 option -mtune=itanium1 should be used. 19076 * [23]DFA processor pipeline descriptions for the IA-64 processors 19077 have been added. This resulted in about 3% improvement on the 19078 SPECInt2000 benchmark for Itanium 2. 19079 * Instruction bundling for the IA-64 processors has been rewritten 19080 using the DFA pipeline hazard recognizer. It resulted in about 60% 19081 compiler speedup on the SPECInt2000 C programs. 19082 19083 M32R 19084 19085 * Support for the M32R/2 processor has been added by Renesas. 19086 * Support for an M32R GNU/Linux target and PIC code generation has 19087 been added by Renesas. 19088 19089 M68000 19090 19091 * Bernardo Innocenti (Develer S.r.l.) has contributed the 19092 m68k-uclinux target, based on former work done by Paul Dale 19093 (SnapGear Inc.). Code generation for the ColdFire processors family 19094 has been enhanced and extended to support the MCF 53xx and MCF 54xx 19095 cores, integrating former work done by Peter Barada (Motorola). 19096 19097 MIPS 19098 19099 Processor-specific changes 19100 19101 * Support for the RM7000 and RM9000 processors has been added. It can 19102 be selected using the -march compiler option and should work with 19103 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 19104 * Support for revision 2 of the MIPS32 ISA has been added. It can be 19105 selected with the command-line option -march=mips32r2. 19106 * There is a new option, -mfix-sb1, to work around certain SB-1 19107 errata. 19108 19109 Configuration 19110 19111 * It is possible to customize GCC using the following configure-time 19112 options: 19113 + --with-arch, which specifies the default value of the -march 19114 option. 19115 + --with-tune, which specifies the default value of the -mtune 19116 option. 19117 + --with-abi, which specifies the default ABI. 19118 + --with-float=soft, which tells GCC to use software floating 19119 point by default. 19120 + --with-float=hard, which tells GCC to use hardware floating 19121 point by default. 19122 * A 64-bit GNU/Linux port has been added. The associated 19123 configurations are mips64-linux-gnu and mips64el-linux-gnu. 19124 * The 32-bit GNU/Linux port now supports Java. 19125 * The IRIX 6 configuration now supports the o32 ABI and will build 19126 o32 multilibs by default. This support is compatible with both 19127 binutils and the SGI tools, but note that several features, 19128 including debugging information and DWARF2 exception handling, are 19129 only available when using the GNU assembler. Use of the GNU 19130 assembler and linker (version 2.15 or above) is strongly 19131 recommended. 19132 * The IRIX 6 configuration now supports 128-bit long doubles. 19133 * There are two new RTEMS-specific configurations, mips-rtems and 19134 mipsel-rtems. 19135 * There are two new *-elf configurations, mipsisa32r2-elf and 19136 mipsisa32r2el-elf. 19137 19138 General 19139 19140 * Several [24]ABI bugs have been fixed. Unfortunately, these changes 19141 will break binary compatibility with earlier releases. 19142 * GCC can now use explicit relocation operators when generating 19143 -mabicalls code. This behavior is controlled by -mexplicit-relocs 19144 and can have several performance benefits. For example: 19145 + It allows for more optimization of GOT accesses, including 19146 better scheduling and redundancy elimination. 19147 + It allows sibling calls to be implemented as jumps. 19148 + n32 and n64 leaf functions can use a call-clobbered global 19149 pointer instead of $28. 19150 + The code to set up $gp can be removed from functions that 19151 don't need it. 19152 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 19153 option is equivalent to the assembler's -xgot option and should be 19154 used instead of -Wa,-xgot. 19155 * Frame pointer elimination is now supported when generating 64-bit 19156 MIPS16 code. 19157 * Inline block moves have been optimized to take more account of 19158 alignment information. 19159 * Many internal changes have been made to the MIPS port, mostly aimed 19160 at reducing the reliance on assembler macros. 19161 19162 PowerPC 19163 19164 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 19165 [25]ABI incompatibilities regarding the way parameters are passed 19166 during functions calls. These changes may result in incompatibility 19167 between code compiled with GCC 3.3 and GCC 3.4. 19168 19169 PowerPC Darwin 19170 19171 * Support for shared/dylib gcc libraries has been added. It is 19172 enabled by default on powerpc-apple-darwin7.0.0 and up. 19173 * Libgcj is enabled by default. On systems older than 19174 powerpc-apple-darwin7.0.0 you need to install dlcompat. 19175 * 128-bit IBM extended precision format support added for long 19176 double. 19177 19178 PowerPC64 GNU/Linux 19179 19180 * By default, PowerPC64 GNU/Linux now uses natural alignment of 19181 structure elements. The old four byte alignment for double, with 19182 special rules for a struct starting with a double, can be chosen 19183 with -malign-power. This change may result in incompatibility 19184 between code compiled with GCC 3.3 and GCC 3.4. 19185 * -mabi=altivec is now the default rather than -mabi=no-altivec. 19186 * 128-bit IBM extended precision format support added for long 19187 double. 19188 19189 S/390 and zSeries 19190 19191 * New command-line options allow to specify the intended execution 19192 environment for generated code: 19193 + -mesa/-mzarch allows to specify whether to generate code 19194 running in ESA/390 mode or in z/Architecture mode (this is 19195 applicable to 31-bit code only). 19196 + -march allows to specify a minimum processor architecture 19197 level (g5, g6, z900, or z990). 19198 + -mtune allows to specify which processor to tune for. 19199 * It is possible to customize GCC using the following configure-time 19200 options: 19201 + --with-mode, which specifies whether to default to assuming 19202 ESA/390 or z/Architecture mode. 19203 + --with-arch, which specifies the default value of the -march 19204 option. 19205 + --with-tune, which specifies the default value of the -mtune 19206 option. 19207 * Support for the z990 processor has been added, and can be selected 19208 using -march=z990 or -mtune=z990. This includes instruction 19209 scheduling tuned for the superscalar instruction pipeline of the 19210 z990 processor as well as support for all new instructions provided 19211 by the long-displacement facility. 19212 * Support to generate 31-bit code optimized for zSeries processors 19213 (running in ESA/390 or in z/Architecture mode) has been added. This 19214 can be selected using -march=z900 and -mzarch respectively. 19215 * Instruction scheduling for the z900 and z990 processors now uses 19216 the DFA pipeline hazard recognizer. 19217 * GCC no longer generates code to maintain a stack backchain, 19218 previously used to generate stack backtraces for debugging 19219 purposes. As replacement that does not incur runtime overhead, 19220 DWARF-2 call frame information is provided by GCC; this is 19221 supported by GDB 6.1. The old behavior can be restored using the 19222 -mbackchain option. 19223 * The stack frame size of functions may now exceed 2 GB in 64-bit 19224 code. 19225 * A port for the 64-bit IBM TPF operating system has been added; the 19226 configuration is s390x-ibm-tpf. This configuration is supported as 19227 cross-compilation target only. 19228 * Various changes to improve the generated code have been 19229 implemented, including: 19230 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 19231 instructions to significantly speed up many floating-point 19232 applications. 19233 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 19234 WITH BORROW instructions to speed up long long arithmetic. 19235 + GCC now uses the SEARCH STRING instruction to implement 19236 strlen(). 19237 + In many cases, function call overhead for 31-bit code has been 19238 reduced by placing the literal pool after the function code 19239 instead of after the function prolog. 19240 + Register 14 is no longer reserved in 64-bit code. 19241 + Handling of global register variables has been improved. 19242 19243 SPARC 19244 19245 * The option -mflat is deprecated. 19246 * Support for large (> 2GB) frames has been added to the 64-bit port. 19247 * Several [26]ABI bugs have been fixed. Unfortunately, these changes 19248 will break binary compatibility with earlier releases. 19249 * The default debugging format has been switched from STABS to 19250 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 19251 the default debugging format for 64-bit code on Solaris. 19252 19253 SuperH 19254 19255 * Support for the SH2E processor has been added. Enabled at run time 19256 with the -m2e command line switch, or at configure time by 19257 specifying sh2e as the machine part of the target triple. 19258 19259 V850 19260 19261 * Support for the Mitsubishi V850E1 processor has been added. This is 19262 a variant of the V850E processor with some additional debugging 19263 instructions. 19264 19265 Xtensa 19266 19267 * Several ABI bugs have been fixed. Unfortunately, these changes 19268 break binary compatibility with earlier releases. 19269 + For big-endian processors, the padding of aggregate return 19270 values larger than a word has changed. If the size of an 19271 aggregate return value is not a multiple of 32 bits, previous 19272 versions of GCC inserted padding in the most-significant bytes 19273 of the first return value register. Aggregates larger than a 19274 word are now padded in the least-significant bytes of the last 19275 return value register used. Aggregates smaller than a word are 19276 still padded in the most-significant bytes. The return value 19277 padding has not changed for little-endian processors. 19278 + Function arguments with 16-byte alignment are now properly 19279 aligned. 19280 + The implementation of the va_list type has changed. A va_list 19281 value created by va_start from a previous release cannot be 19282 used with va_arg from this release, or vice versa. 19283 * More processor configuration options for Xtensa processors are 19284 supported: 19285 + the ABS instruction is now optional; 19286 + the ADDX* and SUBX* instructions are now optional; 19287 + an experimental CONST16 instruction can be used to synthesize 19288 constants instead of loading them from constant pools. 19289 These and other Xtensa processor configuration options can no 19290 longer be enabled or disabled by command-line options; the 19291 processor configuration must be specified by the xtensa-config.h 19292 header file when building GCC. Additionally, the 19293 -mno-serialize-volatile option is no longer supported. 19294 19295 [27]Obsolete Systems 19296 19297 Support for a number of older systems has been declared obsolete in GCC 19298 3.4. Unless there is activity to revive them, the next release of GCC 19299 will have their sources permanently removed. 19300 19301 All configurations of the following processor architectures have been 19302 declared obsolete: 19303 * Mitsubishi D30V, d30v-* 19304 * AT&T DSP1600 and DSP1610, dsp16xx-* 19305 * Intel 80960, i960 19306 19307 Also, some individual systems have been obsoleted: 19308 * ARM Family 19309 + Support for generating code for operation in APCS/26 mode 19310 (-mapcs-26). 19311 * IBM ESA/390 19312 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 19313 maintained and supported.) 19314 * Intel 386 family 19315 + MOSS, i?86-moss-msdos and i?86-*-moss* 19316 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 19317 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 19318 i?86-*-freebsd2* 19319 + GNU/Linux with a.out object format, i?86-linux*aout* 19320 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 19321 + Interix versions before Interix 3, i?86-*-interix 19322 + Mach microkernel, i?86-mach* 19323 + SCO UnixWare with UDK, i?86-*-udk* 19324 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 19325 + VSTa microkernel, i386-*-vsta 19326 * Motorola M68000 family 19327 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 19328 + NetBSD with a.out object format (before NetBSD 1.4), 19329 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 19330 + Generic System V r.4, m68k-*-sysv4* 19331 * VAX 19332 + Generic VAX, vax-*-* (This is generic VAX only; we have not 19333 obsoleted any VAX triples for specific operating systems.) 19334 19335 Documentation improvements 19336 19337 Other significant improvements 19338 19339 * The build system has undergone several significant cleanups. 19340 Subdirectories will only be configured if they are being built, and 19341 all subdirectory configures are run from the make command. The top 19342 level has been autoconfiscated. 19343 * Building GCC no longer writes to its source directory. This should 19344 help those wishing to share a read-only source directory over NFS 19345 or build from a CD. The exceptions to this feature are if you 19346 configure with either --enable-maintainer-mode or 19347 --enable-generated-files-in-srcdir. 19348 * The -W warning option has been renamed to -Wextra, which is more 19349 easily understood. The older spelling will be retained for 19350 backwards compatibility. 19351 * Substantial improvements in compile time have been made, 19352 particularly for non-optimizing compilations. 19353 __________________________________________________________________ 19354 19355 [28]GCC 3.4.0 19356 19357 Bug Fixes 19358 19359 A vast number of bugs have been fixed in 3.4.0, too many to publish a 19360 complete list here. [29]Follow this link to query the Bugzilla database 19361 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 19362 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 19363 regressions. 19364 __________________________________________________________________ 19365 19366 [30]GCC 3.4.1 19367 19368 Bug Fixes 19369 19370 This section lists the problem reports (PRs) from GCC's bug tracking 19371 system that are known to be fixed in the 3.4.1 release. This list might 19372 not be complete (that is, it is possible that some PRs that have been 19373 fixed are not listed here). 19374 19375 Bootstrap failures 19376 19377 * [31]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 19378 emitted - PIC related 19379 * [32]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 19380 * [33]14760 A bug in configure.in prevents using both 19381 --program-suffix and --program-prefix 19382 * [34]14671 [hppa64] bootstrap fails: ICE in 19383 save_call_clobbered_regs, in caller_save.c 19384 * [35]15093 [alpha][Java] make bootstrap fails to configure libffi on 19385 Alpha 19386 * [36]15178 Solaris 9/x86 fails linking after stage 3 19387 19388 Multi-platform internal compiler errors (ICEs) 19389 19390 * [37]12753 (preprocessor) Memory corruption in preprocessor on bad 19391 input 19392 * [38]13985 ICE in gcc.c-torture/compile/930621-1.c 19393 * [39]14810 (c++) tree check failures with invalid code involving 19394 templates 19395 * [40]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 19396 cp/parser.c 19397 * [41]15044 (c++) ICE on syntax error, template header 19398 * [42]15057 (c++) Compiling of conditional value throw constructs 19399 cause a segmentation violation 19400 * [43]15064 (c++) typeid of template parameter gives ICE 19401 * [44]15142 (c++) ICE when passing a string where a char* is expected 19402 in a throw statement 19403 * [45]15159 ICE in rtl_verify_flow_info_1 19404 * [46]15165 (c++) ICE in instantiate_template 19405 * [47]15193 Unary minus using pointer to V4SF vector causes 19406 -fforce-mem to exhaust all memory 19407 * [48]15209 (c++) Runs out of memory with packed structs 19408 * [49]15227 (c++) Trouble with invalid function definition 19409 * [50]15285 (c++) instantiate_type ICE when forming pointer to 19410 template function 19411 * [51]15299 (c++) ICE in resolve_overloaded_unification 19412 * [52]15329 (c++) ICE on constructor of member template 19413 * [53]15550 ICE in extract_insn, in recog.c 19414 * [54]15554 (c++) ICE in tsubst_copy, in cp/pt.c 19415 * [55]15640 (c++) ICE on invalid code in arg_assoc, in 19416 cp/name-lookup.c 19417 * [56]15666 [unit-at-a-time] Gcc abort on valid code 19418 * [57]15696 (c++) ICE with bad pointer-to-member code 19419 * [58]15701 (c++) ICE with friends and template template parameter 19420 * [59]15761 ICE in do_SUBST, in combine.c 19421 * [60]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 19422 19423 Ada 19424 19425 * [61]14538 All RTEMS targets broken for gnat 19426 19427 C front end 19428 19429 * [62]12391 missing warning about assigning to an incomplete type 19430 * [63]14649 atan(1.0) should not be a constant expression 19431 * [64]15004 [unit-at-a-time] no warning for unused paramater in 19432 static function 19433 * [65]15749 --pedantic-errors behaves differently from --pedantic 19434 with C-compiler on GNU/Linux 19435 19436 C++ compiler and library 19437 19438 * [66]10646 non-const reference is incorrectly matched in a "const T" 19439 partial specialization 19440 * [67]12077 wcin.rdbuf()->in_avail() return value too high 19441 * [68]13598 enc_filebuf doesn't work 19442 * [69]14211 const_cast returns lvalue but should be rvalue 19443 * [70]14220 num_put::do_put() undesired float/double behavior 19444 * [71]14245 problem with user-defined allocators in std::basic_string 19445 * [72]14340 libstdc++ Debug mode: failure to convert iterator to 19446 const_iterator 19447 * [73]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 19448 FILE* 19449 * [74]14668 no warning anymore for reevaluation of declaration 19450 * [75]14775 LFS (large file support) tests missing 19451 * [76]14821 Duplicate namespace alias declaration should not conflict 19452 * [77]14930 Friend declaration ignored 19453 * [78]14932 cannot use offsetof to get offsets of array elements in 19454 g++ 3.4.0 19455 * [79]14950 [non unit-at-a-time] always_inline does not mix with 19456 templates and -O0 19457 * [80]14962 g++ ignores #pragma redefine_extname 19458 * [81]14975 Segfault on low-level write error during imbue 19459 * [82]15002 Linewise stream input is unusably slow (std::string slow) 19460 * [83]15025 compiler accepts redeclaration of template as 19461 non-template 19462 * [84]15046 [arm] Math functions misdetected by cross configuration 19463 * [85]15069 a bit test on a variable of enum type is miscompiled 19464 * [86]15074 g++ -lsupc++ still links against libstdc++ 19465 * [87]15083 spurious "statement has no effect" warning 19466 * [88]15096 parse error with templates and pointer to const member 19467 * [89]15287 combination of operator[] and operator .* fails in 19468 templates 19469 * [90]15317 __attribute__ unused in first parameter of constructor 19470 gives error 19471 * [91]15337 sizeof on incomplete type diagnostic 19472 * [92]15361 bitset<>::_Find_next fails 19473 * [93]15412 _GLIBCXX_ symbols symbols defined and used in different 19474 namespaces 19475 * [94]15427 valid code results in incomplete type error 19476 * [95]15471 Incorrect member pointer offsets in anonymous 19477 structs/unions 19478 * [96]15503 nested template problem 19479 * [97]15507 compiler hangs while laying out union 19480 * [98]15542 operator & and template definitions 19481 * [99]15565 SLES9: leading + sign for unsigned int with showpos 19482 * [100]15625 friend defined inside a template fails to find static 19483 function 19484 * [101]15629 Function templates, overloads, and friend name injection 19485 * [102]15742 'noreturn' attribute ignored in method of template 19486 functions. 19487 * [103]15775 Allocator::pointer consistently ignored 19488 * [104]15821 Duplicate namespace alias within namespace rejected 19489 * [105]15862 'enum yn' fails (confict with undeclared builtin) 19490 * [106]15875 rejects pointer to member in template 19491 * [107]15877 valid code using templates and anonymous enums is 19492 rejected 19493 * [108]15947 Puzzling error message for wrong destructor declaration 19494 in template class 19495 * [109]16020 cannot copy __gnu_debug::bitset 19496 * [110]16154 input iterator concept too restrictive 19497 * [111]16174 deducing top-level consts 19498 19499 Java 19500 19501 * [112]14315 Java compiler is not parallel make safe 19502 19503 Fortran 19504 19505 * [113]15151 [g77] incorrect logical i/o in 64-bit mode 19506 19507 Objective-C 19508 19509 * [114]7993 private variables cannot be shadowed in subclasses 19510 19511 Optimization bugs 19512 19513 * [115]15228 useless copies of floating point operands 19514 * [116]15345 [non-unit-at-a-time] unreferenced nested inline 19515 functions not optimized away 19516 * [117]15945 Incorrect floating point optimization 19517 * [118]15526 ftrapv aborts on 0 * (-1) 19518 * [119]14690 Miscompiled POOMA tests 19519 * [120]15112 GCC generates code to write to unchanging memory 19520 19521 Preprocessor 19522 19523 * [121]15067 Minor glitch in the source of cpp 19524 19525 Main driver program bugs 19526 19527 * [122]1963 collect2 interprets -oldstyle_liblookup as -o 19528 ldstyle_liblookup 19529 19530 x86-specific (Intel/AMD) 19531 19532 * [123]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 19533 section} 19534 19535 HPPA-specific 19536 19537 * [124]14782 GCC produces an unaligned data access at -O2 19538 * [125]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 19539 * [126]15202 ICE in reload_cse_simplify_operands, in postreload.c 19540 19541 IA64-specific 19542 19543 * [127]14610 __float80 constants incorrectly emitted 19544 * [128]14813 init_array sections are initialized in the wrong order 19545 * [129]14857 GCC segfault on duplicated asm statement 19546 * [130]15598 Gcc 3.4 ICE on valid code 19547 * [131]15653 Gcc 3.4 ICE on valid code 19548 19549 MIPS-specific 19550 19551 * [132]15189 wrong filling of delay slot with -march=mips1 -G0 19552 -mno-split-addresses -mno-explicit-relocs 19553 * [133]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 19554 2.14.91 19555 * [134]16144 Bogus reference to __divdf3 when -O1 19556 * [135]16176 Miscompilation of unaligned data in MIPS backend 19557 19558 PowerPC-specific 19559 19560 * [136]11591 ICE in gcc.dg/altivec-5.c 19561 * [137]12028 powerpc-eabispe produces bad sCOND operation 19562 * [138]14478 rs6000 geu/ltu patterns generate incorrect code 19563 * [139]14567 long double and va_arg complex args 19564 * [140]14715 Altivec stack layout may overlap gpr save with stack 19565 temps 19566 * [141]14902 (libstdc++) Stream checking functions fail when -pthread 19567 option is used. 19568 * [142]14924 Compiler ICE on valid code 19569 * [143]14960 -maltivec affects vector return with -mabi=no-altivec 19570 * [144]15106 vector varargs failure passing from altivec to 19571 non-altivec code for -m32 19572 * [145]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 19573 half-word operation 19574 * [146]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 19575 and stvx 19576 * [147]15662 Segmentation fault when an exception is thrown - even if 19577 try and catch are specified 19578 19579 s390-specific 19580 19581 * [148]15054 Bad code due to overlapping stack temporaries 19582 19583 SPARC-specific 19584 19585 * [149]15783 ICE with union assignment in 64-bit mode 19586 * [150]15626 GCC 3.4 emits "ld: warning: relocation error: 19587 R_SPARC_UA32" 19588 19589 x86-64-specific 19590 19591 * [151]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 19592 * [152]14723 Backported -march=nocona from mainline 19593 * [153]15290 __float128 failed to pass to function properly 19594 19595 Cygwin/Mingw32-specific 19596 19597 * [154]15250 Option -mms-bitfields support on GCC 3.4 is not 19598 conformant to MS layout 19599 * [155]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 19600 worker on windows32 targets 19601 19602 Bugs specific to embedded processors 19603 19604 * [156]8309 [m68k] -m5200 produces erroneous SImode set of short 19605 varaible on stack 19606 * [157]13250 [SH] Gcc code for rotation clobbers the register, but 19607 gcc continues to use the register as if it was not clobbered 19608 * [158]13803 [coldfire] movqi operand constraints too restrictivefor 19609 TARGET_COLDFIRE 19610 * [159]14093 [SH] ICE for code when using -mhitachi option in SH 19611 * [160]14457 [m6811hc] ICE with simple c++ source 19612 * [161]14542 [m6811hc] ICE on simple source 19613 * [162]15100 [SH] cc1plus got hang-up on 19614 libstdc++-v3/testsuite/abi_check.cc 19615 * [163]15296 [CRIS] Delayed branch scheduling causing invalid code on 19616 cris-* 19617 * [164]15396 [SH] ICE with -O2 -fPIC 19618 * [165]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 19619 ColdFire 19620 19621 Testsuite problems (compiler not affected) 19622 19623 * [166]11610 libstdc++ testcases 27_io/* don't work properly remotely 19624 * [167]15488 (libstdc++) possibly insufficient file permissions for 19625 executing test suite 19626 * [168]15489 (libstdc++) testsuite_files determined incorrectly 19627 19628 Documentation bugs 19629 19630 * [169]13928 (libstdc++) no whatis info in some man pages generated 19631 by doxygen 19632 * [170]14150 Ada documentation out of date 19633 * [171]14949 (c++) Need to document method visibility changes 19634 * [172]15123 libstdc++-doc: Allocators.3 manpage is empty 19635 __________________________________________________________________ 19636 19637 [173]GCC 3.4.2 19638 19639 Bug Fixes 19640 19641 This section lists the problem reports (PRs) from GCC's bug tracking 19642 system that are known to be fixed in the 3.4.2 release. This list might 19643 not be complete (that is, it is possible that some PRs that have been 19644 fixed are not listed here). 19645 19646 Bootstrap failures and issues 19647 19648 * [174]16469 [mips-sgi-irix5.3] bootstrap fails in 19649 libstdc++-v3/testsuite 19650 * [175]16344 [hppa-linux-gnu] libstdc++'s PCH built by 19651 profiledbootstrap does not work with the built compiler 19652 * [176]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 19653 19654 Multi-platform internal compiler errors (ICEs) 19655 19656 * [177]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 19657 cp_parser_class_specifier, in cp/parser.c 19658 * [178]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 19659 * [179]15461 (c++) ICE due to NRV and inlining 19660 * [180]15890 (c++) ICE in c_expand_expr, in c-common.c 19661 * [181]16180 ICE: segmentation fault in RTL optimization 19662 * [182]16224 (c++) ICE in write_unscoped_name (template/namespace) 19663 * [183]16408 ICE: in delete_insn, in cfgrtl.c 19664 * [184]16529 (c++) ICE for: namespace-alias shall not be declared as 19665 the name of any other entity 19666 * [185]16698 (c++) ICE with exceptions and declaration of __cxa_throw 19667 * [186]16706 (c++) ICE in finish_member_declaration, in 19668 cp/semantics.c 19669 * [187]16810 (c++) Legal C++ program with cast gives ICE in 19670 build_ptrmemfunc 19671 * [188]16851 (c++) ICE when throwing a comma expression 19672 * [189]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 19673 * [190]16904 (c++) ICE in finish_class_member_access_expr, in 19674 cp/typeck.c 19675 * [191]16905 (c++) ICE (segfault) with exceptions 19676 * [192]16964 (c++) ICE in cp_parser_class_specifier due to 19677 redefinition 19678 * [193]17068 (c++) ICE: tree check: expected class 'd', have 'x' 19679 (identifier_node) in dependent_template_p, in cp/pt.c 19680 19681 Preprocessor bugs 19682 19683 * [194]16366 Preprocessor option -remap causes memory corruption 19684 19685 Optimization 19686 19687 * [195]15345 unreferenced nested inline functions not optimized away 19688 * [196]16590 Incorrect execution when compiling with -O2 19689 * [197]16693 Bitwise AND is lost when used within a cast to an enum 19690 of the same precision 19691 * [198]17078 Jump into if(0) substatement fails 19692 19693 Problems in generated debug information 19694 19695 * [199]13956 incorrect stabs for nested local variables 19696 19697 C front end bugs 19698 19699 * [200]16684 GCC should not warn about redundant redeclarations of 19700 built-ins 19701 19702 C++ compiler and library 19703 19704 * [201]12658 Thread safety problems in locale::global() and 19705 locale::locale() 19706 * [202]13092 g++ accepts invalid pointer-to-member conversion 19707 * [203]15320 Excessive memory consumption 19708 * [204]16246 Incorrect template argument deduction 19709 * [205]16273 Memory exhausted when using nested classes and virtual 19710 functions 19711 * [206]16401 ostringstream in gcc 3.4.x very slow for big data 19712 * [207]16411 undefined reference to 19713 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 19714 >::file() 19715 * [208]16489 G++ incorrectly rejects use of a null constant integral 19716 expression as a null constant pointer 19717 * [209]16618 offsetof fails with constant member 19718 * [210]16637 syntax error reported for valid input code 19719 * [211]16717 __attribute__((constructor)) broken in C++ 19720 * [212]16813 compiler error in DEBUG version of range insertion 19721 std::map::insert 19722 * [213]16853 pointer-to-member initialization from incompatible one 19723 accepted 19724 * [214]16889 ambiguity is not detected 19725 * [215]16959 Segmentation fault in ios_base::sync_with_stdio 19726 19727 Java compiler and library 19728 19729 * [216]7587 direct threaded interpreter not thread-safe 19730 * [217]16473 ServerSocket accept() leaks file descriptors 19731 * [218]16478 Hash synchronization deadlock with finalizers 19732 19733 Alpha-specific 19734 19735 * [219]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 19736 * [220]16974 could not split insn (ice in final_scan_insn, in 19737 final.c) 19738 19739 x86-specific 19740 19741 * [221]16298 ICE in output_operand 19742 * [222]17113 ICE with SSE2 intrinsics 19743 19744 x86-64 specific 19745 19746 * [223]14697 libstdc++ couldn't find 32bit libgcc_s 19747 19748 MIPS-specific 19749 19750 * [224]15869 [mips64] No NOP after LW (with -mips1 -O0) 19751 * [225]16325 [mips64] value profiling clobbers gp on mips 19752 * [226]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 19753 char[]s 19754 * [227]16380 [mips64] Use of uninitialised register after dbra 19755 conversion 19756 * [228]16407 [mips64] Unaligned access to local variables 19757 * [229]16643 [mips64] verify_local_live_at_start ICE after 19758 crossjumping & cfgcleanup 19759 19760 ARM-specific 19761 19762 * [230]15927 THUMB -O2: strength-reduced iteration variable ends up 19763 off by 1 19764 * [231]15948 THUMB: ICE with non-commutative cbranch 19765 * [232]17019 THUMB: bad switch statement in md code for 19766 addsi3_cbranch_scratch 19767 19768 IA64-specific 19769 19770 * [233]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 19771 (-mtune=merced) 19772 * [234]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 19773 (-mtune=itanium) 19774 * [235]16278 Gcc failed to build Linux kernel with -mtune=merced 19775 * [236]16414 ICE on valid code: typo in comparison of asm_noperands 19776 result 19777 * [237]16445 ICE on valid code: don't count ignored insns 19778 * [238]16490 ICE (segfault) while compiling with -fprofile-use 19779 * [239]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 19780 19781 PowerPC-specific 19782 19783 * [240]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 19784 * [241]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 19785 issue) 19786 19787 SPARC-specific 19788 19789 * [242]16199 ICE while compiling apache 2.0.49 19790 * [243]16416 -m64 doesn't imply -mcpu=v9 anymore 19791 * [244]16430 ICE when returning non-C aggregates larger than 16 bytes 19792 19793 Bugs specific to embedded processors 19794 19795 * [245]16379 [m32r] can't output large model function call of memcpy 19796 * [246]17093 [m32r] ICE with -msdata=use -O0 19797 * [247]17119 [m32r] ICE at switch case 0x8000 19798 19799 DJGPP-specific 19800 19801 * [248]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 19802 19803 Alpha Tru64-specific 19804 19805 * [249]16210 libstdc++ gratuitously omits "long long" I/O 19806 19807 Testsuite, documentation issues (compiler is not affected): 19808 19809 * [250]15488 (libstdc++) possibly insufficient file permissions for 19810 executing test suite 19811 * [251]16250 ada/doctools runs makeinfo even in release tarball 19812 __________________________________________________________________ 19813 19814 [252]GCC 3.4.3 19815 19816 This is the [253]list of problem reports (PRs) from GCC's bug tracking 19817 system that are known to be fixed in the 3.4.3 release. This list might 19818 not be complete (that is, it is possible that some PRs that have been 19819 fixed are not listed here). 19820 19821 Bootstrap failures 19822 19823 * [254]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 19824 * [255]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 19825 when undeclared 19826 19827 Internal compiler errors (ICEs) affecting multiple platforms 19828 19829 * [256]13948 (java) GCJ segmentation fault while compiling GL4Java 19830 .class files 19831 * [257]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 19832 * [258]16301 (c++) ICE when "strong" attribute is attached to a using 19833 directive 19834 * [259]16566 ICE with flexible arrays 19835 * [260]17023 ICE with nested functions in parameter declaration 19836 * [261]17027 ICE with noreturn function in loop at -O2 19837 * [262]17524 ICE in grokdeclarator, in cp/decl.c 19838 * [263]17826 (c++) ICE in cp_tree_equal 19839 19840 C and optimization bugs 19841 19842 * [264]15526 -ftrapv aborts on 0 * (-1) 19843 * [265]16999 #ident stopped working 19844 * [266]17503 quadratic behaviour in invalid_mode_change_p 19845 * [267]17581 Long long arithmetic fails inside a switch/case 19846 statement when compiled with -O2 19847 * [268]18129 -fwritable-strings doesn't work 19848 19849 C++ compiler and library bugs 19850 19851 * [269]10975 incorrect initial ostringstream::tellp() 19852 * [270]11722 Unbuffered filebuf::sgetn is slow 19853 * [271]14534 Unrecognizing static function as a template parameter 19854 when its return value is also templated 19855 * [272]15172 Copy constructor optimization in aggregate 19856 initialization 19857 * [273]15786 Bad error message for frequently occuring error. 19858 * [274]16162 Rejects valid member-template-definition 19859 * [275]16612 empty basic_strings can't live in shared memory 19860 * [276]16715 std::basic_iostream is instantiated when used, even 19861 though instantiations are already contained in libstdc++ 19862 * [277]16848 code in /ext/demangle.h appears broken 19863 * [278]17132 GCC fails to eliminate function template specialization 19864 when argument deduction fails 19865 * [279]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 19866 in ropeimpl.h 19867 * [280]17327 use of `enumeral_type' in template type unification 19868 * [281]17393 "unused variable '._0'" warning with -Wall 19869 * [282]17501 Confusion with member templates 19870 * [283]17537 g++ not passing -lstdc++ to linker when all command line 19871 arguments are libraries 19872 * [284]17585 usage of unqualified name of static member from within 19873 class not allowed 19874 * [285]17821 Poor diagnostic for using "." instead of "->" 19875 * [286]17829 wrong error: call of overloaded function is ambiguous 19876 * [287]17851 Misleading diagnostic for invalid function declarations 19877 with undeclared types 19878 * [288]17976 Destructor is called twice 19879 * [289]18020 rejects valid definition of enum value in template 19880 * [290]18093 bogus conflict in namespace aliasing 19881 * [291]18140 C++ parser bug when using >> in templates 19882 19883 Fortran 19884 19885 * [292]17541 data statements with double precision constants fail 19886 19887 x86-specific 19888 19889 * [293]17853 -O2 ICE for MMX testcase 19890 19891 SPARC-specific 19892 19893 * [294]17245 ICE compiling gsl-1.5 statistics/lag1.c 19894 19895 Darwin-specific 19896 19897 * [295]17167 FATAL:Symbol L_foo$stub already defined. 19898 19899 AIX-specific 19900 19901 * [296]17277 could not catch an exception when specified -maix64 19902 19903 Solaris-specific 19904 19905 * [297]17505 <cmath> calls acosf(), ceilf(), and other functions 19906 missing from system libraries 19907 19908 HP/UX specific: 19909 19910 * [298]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 19911 19912 ARM-specific 19913 19914 * [299]17384 ICE with mode attribute on structures 19915 19916 MIPS-specific 19917 19918 * [300]17770 No NOP after LWL with -mips1 19919 19920 Other embedded target specific 19921 19922 * [301]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 19923 * [302]14064 [avr-elf] -fdata-sections triggers ICE 19924 * [303]14678 [m68hc11-elf] gcc ICE 19925 * [304]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 19926 * [305]15790 [i686-coff] Alignment error building gcc with i686-coff 19927 target 19928 * [306]15886 [SH] Miscompilation with -O2 -fPIC 19929 * [307]16884 [avr-elf] [fweb related] bug while initializing 19930 variables 19931 19932 Bugs relating to debugger support 19933 19934 * [308]13841 missing debug info for _Complex function arguments 19935 * [309]15860 [big-endian targets] No DW_AT_location debug info is 19936 emitted for formal arguments to a function that uses "register" 19937 qualifiers 19938 19939 Testsuite issues (compiler not affected) 19940 19941 * [310]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 19942 * [311]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 19943 * [312]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 19944 testsuite 19945 19946 Documentation 19947 19948 * [313]15498 typo in gcc manual: non-existing locale example en_UK, 19949 should be en_GB 19950 * [314]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 19951 document broken shell 19952 * [315]16406 USE_LD_AS_NEEDED undocumented 19953 __________________________________________________________________ 19954 19955 [316]GCC 3.4.4 19956 19957 This is the [317]list of problem reports (PRs) from GCC's bug tracking 19958 system that are known to be fixed in the 3.4.4 release. This list might 19959 not be complete (that is, it is possible that some PRs that have been 19960 fixed are not listed here). 19961 __________________________________________________________________ 19962 19963 [318]GCC 3.4.5 19964 19965 This is the [319]list of problem reports (PRs) from GCC's bug tracking 19966 system that are known to be fixed in the 3.4.5 release. This list might 19967 not be complete (that is, it is possible that some PRs that have been 19968 fixed are not listed here). 19969 19970 Bootstrap issues 19971 19972 * [320]24688 sco_math fixincl breaks math.h 19973 19974 C compiler bugs 19975 19976 * [321]17188 struct Foo { } redefinition 19977 * [322]20187 wrong code for ((unsigned char)(unsigned long 19978 long)((a?a:1)&(a*b)))?0:1) 19979 * [323]21873 infinite warning loop on bad array initializer 19980 * [324]21899 enum definition accepts values to be overriden 19981 * [325]22061 ICE in find_function_data, in function.c 19982 * [326]22308 Failure to diagnose violation of constraint 6.516p2 19983 * [327]22458 ICE on missing brace 19984 * [328]22589 ICE casting to long long 19985 * [329]24101 Segfault with preprocessed source 19986 19987 C++ compiler and library bugs 19988 19989 * [330]10611 operations on vector mode not recognized in C++ 19990 * [331]13377 unexpected behavior of namespace usage directive 19991 * [332]16002 Strange error message with new parser 19992 * [333]17413 local classes as template argument 19993 * [334]17609 spurious error message after using keyword 19994 * [335]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 19995 * [336]18124 ICE with invalid template template parameter 19996 * [337]18155 typedef in template declaration not rejected 19997 * [338]18177 ICE with const_cast for undeclared variable 19998 * [339]18368 C++ error message regression 19999 * [340]16378 ICE when returning a copy of a packed member 20000 * [341]18466 int ::i; accepted 20001 * [342]18512 ICE on invalid usage of template base class 20002 * [343]18454 ICE when returning undefined type 20003 * [344]18738 typename not allowed with non-dependent qualified name 20004 * [345]18803 rejects access to operator() in template 20005 * [346]19004 ICE in uses_template_parms, in cp/pt.c 20006 * [347]19208 Spurious error about variably modified type 20007 * [348]18253 bad error message / ICE for invalid template parameter 20008 * [349]19608 ICE after friend function definition in local class 20009 * [350]19884 ICE on explicit instantiation of a non-template 20010 constructor 20011 * [351]20153 ICE when C++ template function contains anonymous union 20012 * [352]20563 Infinite loop in diagnostic (and ice after error 20013 message) 20014 * [353]20789 ICE with incomplete type in template 20015 * [354]21336 Internal compiler error when using custom new operators 20016 * [355]21768 ICE in error message due to violation of coding 20017 conventions 20018 * [356]21853 constness of pointer to data member ignored 20019 * [357]21903 Default argument of template function causes a 20020 compile-time error 20021 * [358]21983 multiple diagnostics 20022 * [359]21987 New testsuite failure 20023 g++.dg/warn/conversion-function-1.C 20024 * [360]22153 ICE on invalid template specialization 20025 * [361]22172 Internal compiler error, seg fault. 20026 * [362]21286 filebuf::xsgetn vs pipes 20027 * [363]22233 ICE with wrong number of template parameters 20028 * [364]22508 ICE after invalid operator new 20029 * [365]22545 ICE with pointer to class member & user defined 20030 conversion operator 20031 * [366]23528 Wrong default allocator in ext/hash_map 20032 * [367]23550 char_traits requirements/1.cc test bad math 20033 * [368]23586 Bad diagnostic for invalid namespace-name 20034 * [369]23624 ICE in invert_truthvalue, in fold-const.c 20035 * [370]23639 Bad error message: not a member of '<declaration error>' 20036 * [371]23797 ICE on typename outside template 20037 * [372]23965 Bogus error message: no matching function for call to 20038 'foo(<type error>)' 20039 * [373]24052 &#`label_decl' not supported by dump_expr#<expression 20040 error> 20041 * [374]24580 virtual base class cause exception not to be caught 20042 20043 Problems in generated debug information 20044 20045 * [375]24267 Bad DWARF for altivec vectors 20046 20047 Optimizations issues 20048 20049 * [376]17810 ICE in verify_local_live_at_start 20050 * [377]17860 Wrong generated code for loop with varying bound 20051 * [378]21709 ICE on compile-time complex NaN 20052 * [379]21964 broken tail call at -O2 or more 20053 * [380]22167 Strange optimization bug when using -Os 20054 * [381]22619 Compilation failure for real_const_1.f and 20055 real_const_2.f90 20056 * [382]23241 Invalid code generated for comparison of uchar to 255 20057 * [383]23478 Miscompilation due to reloading of a var that is also 20058 used in EH pad 20059 * [384]24470 segmentation fault in cc1plus when compiling with -O 20060 * [385]24950 ICE in operand_subword_force 20061 20062 Precompiled headers problems 20063 20064 * [386]14400 Cannot compile qt-x11-free-3.3.0 20065 * [387]14940 PCH largefile test fails on various platforms 20066 20067 Preprocessor bugs 20068 20069 * [388]20239 ICE on empty preprocessed input 20070 * [389]15220 "gcc -E -MM -MG" reports missing system headers in 20071 source directory 20072 20073 Testsuite issues 20074 20075 * [390]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 20076 i686-pc-linux-gnu 20077 20078 Alpha specific 20079 20080 * [391]21888 bootstrap failure with linker relaxation enabled 20081 20082 ARM specific 20083 20084 * [392]15342 [arm-linux]: ICE in verify_local_live_at_start 20085 * [393]23985 Memory aliasing information incorrect in inlined memcpy 20086 20087 ColdFile specific 20088 20089 * [394]16719 Illegal move of byte into address register causes 20090 compiler to ICE 20091 20092 HPPA specific 20093 20094 * [395]21723 ICE while building libgfortran 20095 * [396]21841 -mhp-ld/-mgnu-ld documentation 20096 20097 IA-64 specific 20098 20099 * [397]23644 IA-64 hardware models and configuration options 20100 documentation error 20101 * [398]24718 Shared libgcc not used for linking by default 20102 20103 M68000 specific 20104 20105 * [399]18421 ICE in reload_cse_simplify_operands, in postreload.c 20106 20107 MIPS specific 20108 20109 * [400]20621 ICE in change_address_1, in emit-rtl.c 20110 20111 PowerPC and PowerPC64 specific 20112 20113 * [401]18583 error on valid code: const 20114 __attribute__((altivec(vector__))) doesn't work in arrays 20115 * [402]20191 ICE in reload_cse_simplify_operands 20116 * [403]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 20117 * [404]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 20118 * [405]23404 gij trashes args of functions with more than 8 fp args 20119 * [406]23539 C & C++ compiler generating misaligned references 20120 regardless of compiler flags 20121 * [407]24102 floatdisf2_internal2 broken 20122 * [408]24465 -mminimal-toc miscompilation of __thread vars 20123 20124 Solaris specific 20125 20126 * [409]19933 Problem with define of HUGE_VAL in math_c99 20127 * [410]21889 Native Solaris assembler cannot grok DTP-relative debug 20128 symbols 20129 20130 SPARC specific 20131 20132 * [411]19300 PCH failures on sparc-linux 20133 * [412]20301 Assembler labels have a leading "-" 20134 * [413]20673 C PCH testsuite assembly comparison failure 20135 20136 x86 and x86_64 specific 20137 20138 * [414]18582 ICE with arrays of type V2DF 20139 * [415]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 20140 -fsched2-use-traces 20141 * [416]21716 ICE in reg-stack.c's swap_rtx_condition 20142 * [417]24315 amd64 fails -fpeephole2 20143 __________________________________________________________________ 20144 20145 [418]GCC 3.4.6 20146 20147 This is the [419]list of problem reports (PRs) from GCC's bug tracking 20148 system that are known to be fixed in the 3.4.6 release. This list might 20149 not be complete (that is, it is possible that some PRs that have been 20150 fixed are not listed here). 20151 20152 20153 For questions related to the use of GCC, please consult these web 20154 pages and the [420]GCC manuals. If that fails, the 20155 [421]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20156 web pages and the development of GCC are welcome on our developer 20157 list at [422]gcc (a] gcc.gnu.org. All of [423]our lists have public 20158 archives. 20159 20160 Copyright (C) [424]Free Software Foundation, Inc. Verbatim copying and 20161 distribution of this entire article is permitted in any medium, 20162 provided this notice is preserved. 20163 20164 These pages are [425]maintained by the GCC team. Last modified 20165 2025-01-31. 20166 20167 References 20168 20169 1. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 20170 2. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 20171 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 20172 4. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 20173 5. https://gcc.gnu.org/gcc-3.4/mips-abi.html 20174 6. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 20175 7. https://www.boost.org/ 20176 8. https://gcc.gnu.org/PR11953 20177 9. https://gcc.gnu.org/PR8361 20178 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins 20179 11. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 20180 12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 20181 13. https://gcc.gnu.org/bugs/#cxx_rvalbind 20182 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20183 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20184 16. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20185 17. https://www.gnu.org/software/classpath/ 20186 18. https://www.eclipse.org/ 20187 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 20188 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 20189 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 20190 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 20191 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 20192 24. https://gcc.gnu.org/gcc-3.4/mips-abi.html 20193 25. https://gcc.gnu.org/gcc-3.4/powerpc-abi.html 20194 26. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 20195 27. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 20196 28. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.0 20197 29. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=\[3\.4.*[Rr]egression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 20198 30. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.1 20199 31. https://gcc.gnu.org/PR10129 20200 32. https://gcc.gnu.org/PR14576 20201 33. https://gcc.gnu.org/PR14760 20202 34. https://gcc.gnu.org/PR14671 20203 35. https://gcc.gnu.org/PR15093 20204 36. https://gcc.gnu.org/PR15178 20205 37. https://gcc.gnu.org/PR12753 20206 38. https://gcc.gnu.org/PR13985 20207 39. https://gcc.gnu.org/PR14810 20208 40. https://gcc.gnu.org/PR14883 20209 41. https://gcc.gnu.org/PR15044 20210 42. https://gcc.gnu.org/PR15057 20211 43. https://gcc.gnu.org/PR15064 20212 44. https://gcc.gnu.org/PR15142 20213 45. https://gcc.gnu.org/PR15159 20214 46. https://gcc.gnu.org/PR15165 20215 47. https://gcc.gnu.org/PR15193 20216 48. https://gcc.gnu.org/PR15209 20217 49. https://gcc.gnu.org/PR15227 20218 50. https://gcc.gnu.org/PR15285 20219 51. https://gcc.gnu.org/PR15299 20220 52. https://gcc.gnu.org/PR15329 20221 53. https://gcc.gnu.org/PR15550 20222 54. https://gcc.gnu.org/PR15554 20223 55. https://gcc.gnu.org/PR15640 20224 56. https://gcc.gnu.org/PR15666 20225 57. https://gcc.gnu.org/PR15696 20226 58. https://gcc.gnu.org/PR15701 20227 59. https://gcc.gnu.org/PR15761 20228 60. https://gcc.gnu.org/PR15829 20229 61. https://gcc.gnu.org/PR14538 20230 62. https://gcc.gnu.org/PR12391 20231 63. https://gcc.gnu.org/PR14649 20232 64. https://gcc.gnu.org/PR15004 20233 65. https://gcc.gnu.org/PR15749 20234 66. https://gcc.gnu.org/PR10646 20235 67. https://gcc.gnu.org/PR12077 20236 68. https://gcc.gnu.org/PR13598 20237 69. https://gcc.gnu.org/PR14211 20238 70. https://gcc.gnu.org/PR14220 20239 71. https://gcc.gnu.org/PR14245 20240 72. https://gcc.gnu.org/PR14340 20241 73. https://gcc.gnu.org/PR14600 20242 74. https://gcc.gnu.org/PR14668 20243 75. https://gcc.gnu.org/PR14775 20244 76. https://gcc.gnu.org/PR14821 20245 77. https://gcc.gnu.org/PR14930 20246 78. https://gcc.gnu.org/PR14932 20247 79. https://gcc.gnu.org/PR14950 20248 80. https://gcc.gnu.org/PR14962 20249 81. https://gcc.gnu.org/PR14975 20250 82. https://gcc.gnu.org/PR15002 20251 83. https://gcc.gnu.org/PR15025 20252 84. https://gcc.gnu.org/PR15046 20253 85. https://gcc.gnu.org/PR15069 20254 86. https://gcc.gnu.org/PR15074 20255 87. https://gcc.gnu.org/PR15083 20256 88. https://gcc.gnu.org/PR15096 20257 89. https://gcc.gnu.org/PR15287 20258 90. https://gcc.gnu.org/PR15317 20259 91. https://gcc.gnu.org/PR15337 20260 92. https://gcc.gnu.org/PR15361 20261 93. https://gcc.gnu.org/PR15412 20262 94. https://gcc.gnu.org/PR15427 20263 95. https://gcc.gnu.org/PR15471 20264 96. https://gcc.gnu.org/PR15503 20265 97. https://gcc.gnu.org/PR15507 20266 98. https://gcc.gnu.org/PR15542 20267 99. https://gcc.gnu.org/PR15565 20268 100. https://gcc.gnu.org/PR15625 20269 101. https://gcc.gnu.org/PR15629 20270 102. https://gcc.gnu.org/PR15742 20271 103. https://gcc.gnu.org/PR15775 20272 104. https://gcc.gnu.org/PR15821 20273 105. https://gcc.gnu.org/PR15862 20274 106. https://gcc.gnu.org/PR15875 20275 107. https://gcc.gnu.org/PR15877 20276 108. https://gcc.gnu.org/PR15947 20277 109. https://gcc.gnu.org/PR16020 20278 110. https://gcc.gnu.org/PR16154 20279 111. https://gcc.gnu.org/PR16174 20280 112. https://gcc.gnu.org/PR14315 20281 113. https://gcc.gnu.org/PR15151 20282 114. https://gcc.gnu.org/PR7993 20283 115. https://gcc.gnu.org/PR15228 20284 116. https://gcc.gnu.org/PR15345 20285 117. https://gcc.gnu.org/PR15945 20286 118. https://gcc.gnu.org/PR15526 20287 119. https://gcc.gnu.org/PR14690 20288 120. https://gcc.gnu.org/PR15112 20289 121. https://gcc.gnu.org/PR15067 20290 122. https://gcc.gnu.org/PR1963 20291 123. https://gcc.gnu.org/PR15717 20292 124. https://gcc.gnu.org/PR14782 20293 125. https://gcc.gnu.org/PR14828 20294 126. https://gcc.gnu.org/PR15202 20295 127. https://gcc.gnu.org/PR14610 20296 128. https://gcc.gnu.org/PR14813 20297 129. https://gcc.gnu.org/PR14857 20298 130. https://gcc.gnu.org/PR15598 20299 131. https://gcc.gnu.org/PR15653 20300 132. https://gcc.gnu.org/PR15189 20301 133. https://gcc.gnu.org/PR15331 20302 134. https://gcc.gnu.org/PR16144 20303 135. https://gcc.gnu.org/PR16176 20304 136. https://gcc.gnu.org/PR11591 20305 137. https://gcc.gnu.org/PR12028 20306 138. https://gcc.gnu.org/PR14478 20307 139. https://gcc.gnu.org/PR14567 20308 140. https://gcc.gnu.org/PR14715 20309 141. https://gcc.gnu.org/PR14902 20310 142. https://gcc.gnu.org/PR14924 20311 143. https://gcc.gnu.org/PR14960 20312 144. https://gcc.gnu.org/PR15106 20313 145. https://gcc.gnu.org/PR16026 20314 146. https://gcc.gnu.org/PR15191 20315 147. https://gcc.gnu.org/PR15662 20316 148. https://gcc.gnu.org/PR15054 20317 149. https://gcc.gnu.org/PR15783 20318 150. https://gcc.gnu.org/PR15626 20319 151. https://gcc.gnu.org/PR14326 20320 152. https://gcc.gnu.org/PR14723 20321 153. https://gcc.gnu.org/PR15290 20322 154. https://gcc.gnu.org/PR15250 20323 155. https://gcc.gnu.org/PR15551 20324 156. https://gcc.gnu.org/PR8309 20325 157. https://gcc.gnu.org/PR13250 20326 158. https://gcc.gnu.org/PR13803 20327 159. https://gcc.gnu.org/PR14093 20328 160. https://gcc.gnu.org/PR14457 20329 161. https://gcc.gnu.org/PR14542 20330 162. https://gcc.gnu.org/PR15100 20331 163. https://gcc.gnu.org/PR15296 20332 164. https://gcc.gnu.org/PR15396 20333 165. https://gcc.gnu.org/PR15782 20334 166. https://gcc.gnu.org/PR11610 20335 167. https://gcc.gnu.org/PR15488 20336 168. https://gcc.gnu.org/PR15489 20337 169. https://gcc.gnu.org/PR13928 20338 170. https://gcc.gnu.org/PR14150 20339 171. https://gcc.gnu.org/PR14949 20340 172. https://gcc.gnu.org/PR15123 20341 173. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.2 20342 174. https://gcc.gnu.org/PR16469 20343 175. https://gcc.gnu.org/PR16344 20344 176. https://gcc.gnu.org/PR16842 20345 177. https://gcc.gnu.org/PR12608 20346 178. https://gcc.gnu.org/PR14492 20347 179. https://gcc.gnu.org/PR15461 20348 180. https://gcc.gnu.org/PR15890 20349 181. https://gcc.gnu.org/PR16180 20350 182. https://gcc.gnu.org/PR16224 20351 183. https://gcc.gnu.org/PR16408 20352 184. https://gcc.gnu.org/PR16529 20353 185. https://gcc.gnu.org/PR16698 20354 186. https://gcc.gnu.org/PR16706 20355 187. https://gcc.gnu.org/PR16810 20356 188. https://gcc.gnu.org/PR16851 20357 189. https://gcc.gnu.org/PR16870 20358 190. https://gcc.gnu.org/PR16904 20359 191. https://gcc.gnu.org/PR16905 20360 192. https://gcc.gnu.org/PR16964 20361 193. https://gcc.gnu.org/PR17068 20362 194. https://gcc.gnu.org/PR16366 20363 195. https://gcc.gnu.org/PR15345 20364 196. https://gcc.gnu.org/PR16590 20365 197. https://gcc.gnu.org/PR16693 20366 198. https://gcc.gnu.org/PR17078 20367 199. https://gcc.gnu.org/PR13956 20368 200. https://gcc.gnu.org/PR16684 20369 201. https://gcc.gnu.org/PR12658 20370 202. https://gcc.gnu.org/PR13092 20371 203. https://gcc.gnu.org/PR15320 20372 204. https://gcc.gnu.org/PR16246 20373 205. https://gcc.gnu.org/PR16273 20374 206. https://gcc.gnu.org/PR16401 20375 207. https://gcc.gnu.org/PR16411 20376 208. https://gcc.gnu.org/PR16489 20377 209. https://gcc.gnu.org/PR16618 20378 210. https://gcc.gnu.org/PR16637 20379 211. https://gcc.gnu.org/PR16717 20380 212. https://gcc.gnu.org/PR16813 20381 213. https://gcc.gnu.org/PR16853 20382 214. https://gcc.gnu.org/PR16889 20383 215. https://gcc.gnu.org/PR16959 20384 216. https://gcc.gnu.org/PR7587 20385 217. https://gcc.gnu.org/PR16473 20386 218. https://gcc.gnu.org/PR16478 20387 219. https://gcc.gnu.org/PR10695 20388 220. https://gcc.gnu.org/PR16974 20389 221. https://gcc.gnu.org/PR16298 20390 222. https://gcc.gnu.org/PR17113 20391 223. https://gcc.gnu.org/PR14697 20392 224. https://gcc.gnu.org/PR15869 20393 225. https://gcc.gnu.org/PR16325 20394 226. https://gcc.gnu.org/PR16357 20395 227. https://gcc.gnu.org/PR16380 20396 228. https://gcc.gnu.org/PR16407 20397 229. https://gcc.gnu.org/PR16643 20398 230. https://gcc.gnu.org/PR15927 20399 231. https://gcc.gnu.org/PR15948 20400 232. https://gcc.gnu.org/PR17019 20401 233. https://gcc.gnu.org/PR16130 20402 234. https://gcc.gnu.org/PR16142 20403 235. https://gcc.gnu.org/PR16278 20404 236. https://gcc.gnu.org/PR16414 20405 237. https://gcc.gnu.org/PR16445 20406 238. https://gcc.gnu.org/PR16490 20407 239. https://gcc.gnu.org/PR16683 20408 240. https://gcc.gnu.org/PR16195 20409 241. https://gcc.gnu.org/PR16239 20410 242. https://gcc.gnu.org/PR16199 20411 243. https://gcc.gnu.org/PR16416 20412 244. https://gcc.gnu.org/PR16430 20413 245. https://gcc.gnu.org/PR16379 20414 246. https://gcc.gnu.org/PR17093 20415 247. https://gcc.gnu.org/PR17119 20416 248. https://gcc.gnu.org/PR15928 20417 249. https://gcc.gnu.org/PR16210 20418 250. https://gcc.gnu.org/PR15488 20419 251. https://gcc.gnu.org/PR16250 20420 252. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.3 20421 253. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 20422 254. https://gcc.gnu.org/PR17369 20423 255. https://gcc.gnu.org/PR17850 20424 256. https://gcc.gnu.org/PR13948 20425 257. https://gcc.gnu.org/PR14492 20426 258. https://gcc.gnu.org/PR16301 20427 259. https://gcc.gnu.org/PR16566 20428 260. https://gcc.gnu.org/PR17023 20429 261. https://gcc.gnu.org/PR17027 20430 262. https://gcc.gnu.org/PR17524 20431 263. https://gcc.gnu.org/PR17826 20432 264. https://gcc.gnu.org/PR15526 20433 265. https://gcc.gnu.org/PR16999 20434 266. https://gcc.gnu.org/PR17503 20435 267. https://gcc.gnu.org/PR17581 20436 268. https://gcc.gnu.org/PR18129 20437 269. https://gcc.gnu.org/PR10975 20438 270. https://gcc.gnu.org/PR11722 20439 271. https://gcc.gnu.org/PR14534 20440 272. https://gcc.gnu.org/PR15172 20441 273. https://gcc.gnu.org/PR15786 20442 274. https://gcc.gnu.org/PR16162 20443 275. https://gcc.gnu.org/PR16612 20444 276. https://gcc.gnu.org/PR16715 20445 277. https://gcc.gnu.org/PR16848 20446 278. https://gcc.gnu.org/PR17132 20447 279. https://gcc.gnu.org/PR17259 20448 280. https://gcc.gnu.org/PR17327 20449 281. https://gcc.gnu.org/PR17393 20450 282. https://gcc.gnu.org/PR17501 20451 283. https://gcc.gnu.org/PR17537 20452 284. https://gcc.gnu.org/PR17585 20453 285. https://gcc.gnu.org/PR17821 20454 286. https://gcc.gnu.org/PR17829 20455 287. https://gcc.gnu.org/PR17851 20456 288. https://gcc.gnu.org/PR17976 20457 289. https://gcc.gnu.org/PR18020 20458 290. https://gcc.gnu.org/PR18093 20459 291. https://gcc.gnu.org/PR18140 20460 292. https://gcc.gnu.org/PR17541 20461 293. https://gcc.gnu.org/PR17853 20462 294. https://gcc.gnu.org/PR17245 20463 295. https://gcc.gnu.org/PR17167 20464 296. https://gcc.gnu.org/PR17277 20465 297. https://gcc.gnu.org/PR17505 20466 298. https://gcc.gnu.org/PR17684 20467 299. https://gcc.gnu.org/PR17384 20468 300. https://gcc.gnu.org/PR17770 20469 301. https://gcc.gnu.org/PR11476 20470 302. https://gcc.gnu.org/PR14064 20471 303. https://gcc.gnu.org/PR14678 20472 304. https://gcc.gnu.org/PR15583 20473 305. https://gcc.gnu.org/PR15790 20474 306. https://gcc.gnu.org/PR15886 20475 307. https://gcc.gnu.org/PR16884 20476 308. https://gcc.gnu.org/PR13841 20477 309. https://gcc.gnu.org/PR15860 20478 310. https://gcc.gnu.org/PR17465 20479 311. https://gcc.gnu.org/PR17469 20480 312. https://gcc.gnu.org/PR18138 20481 313. https://gcc.gnu.org/PR15498 20482 314. https://gcc.gnu.org/PR15747 20483 315. https://gcc.gnu.org/PR16406 20484 316. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.4 20485 317. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 20486 318. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.5 20487 319. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 20488 320. https://gcc.gnu.org/PR24688 20489 321. https://gcc.gnu.org/PR17188 20490 322. https://gcc.gnu.org/PR20187 20491 323. https://gcc.gnu.org/PR21873 20492 324. https://gcc.gnu.org/PR21899 20493 325. https://gcc.gnu.org/PR22061 20494 326. https://gcc.gnu.org/PR22208 20495 327. https://gcc.gnu.org/PR22458 20496 328. https://gcc.gnu.org/PR22589 20497 329. https://gcc.gnu.org/PR24101 20498 330. https://gcc.gnu.org/PR10611 20499 331. https://gcc.gnu.org/PR13377 20500 332. https://gcc.gnu.org/PR16002 20501 333. https://gcc.gnu.org/PR17413 20502 334. https://gcc.gnu.org/PR17609 20503 335. https://gcc.gnu.org/PR17618 20504 336. https://gcc.gnu.org/PR18124 20505 337. https://gcc.gnu.org/PR18155 20506 338. https://gcc.gnu.org/PR18177 20507 339. https://gcc.gnu.org/PR18368 20508 340. https://gcc.gnu.org/PR18378 20509 341. https://gcc.gnu.org/PR18466 20510 342. https://gcc.gnu.org/PR18512 20511 343. https://gcc.gnu.org/PR18545 20512 344. https://gcc.gnu.org/PR18738 20513 345. https://gcc.gnu.org/PR18803 20514 346. https://gcc.gnu.org/PR19004 20515 347. https://gcc.gnu.org/PR19208 20516 348. https://gcc.gnu.org/PR19253 20517 349. https://gcc.gnu.org/PR19608 20518 350. https://gcc.gnu.org/PR19884 20519 351. https://gcc.gnu.org/PR20153 20520 352. https://gcc.gnu.org/PR20563 20521 353. https://gcc.gnu.org/PR20789 20522 354. https://gcc.gnu.org/PR21336 20523 355. https://gcc.gnu.org/PR21768 20524 356. https://gcc.gnu.org/PR21853 20525 357. https://gcc.gnu.org/PR21903 20526 358. https://gcc.gnu.org/PR21983 20527 359. https://gcc.gnu.org/PR21987 20528 360. https://gcc.gnu.org/PR22153 20529 361. https://gcc.gnu.org/PR22172 20530 362. https://gcc.gnu.org/PR21286 20531 363. https://gcc.gnu.org/PR22233 20532 364. https://gcc.gnu.org/PR22508 20533 365. https://gcc.gnu.org/PR22545 20534 366. https://gcc.gnu.org/PR23528 20535 367. https://gcc.gnu.org/PR23550 20536 368. https://gcc.gnu.org/PR23586 20537 369. https://gcc.gnu.org/PR23624 20538 370. https://gcc.gnu.org/PR23639 20539 371. https://gcc.gnu.org/PR23797 20540 372. https://gcc.gnu.org/PR23965 20541 373. https://gcc.gnu.org/PR24052 20542 374. https://gcc.gnu.org/PR24580 20543 375. https://gcc.gnu.org/PR24267 20544 376. https://gcc.gnu.org/PR17810 20545 377. https://gcc.gnu.org/PR17860 20546 378. https://gcc.gnu.org/PR21709 20547 379. https://gcc.gnu.org/PR21964 20548 380. https://gcc.gnu.org/PR22167 20549 381. https://gcc.gnu.org/PR22619 20550 382. https://gcc.gnu.org/PR23241 20551 383. https://gcc.gnu.org/PR23478 20552 384. https://gcc.gnu.org/PR24470 20553 385. https://gcc.gnu.org/PR24950 20554 386. https://gcc.gnu.org/PR14400 20555 387. https://gcc.gnu.org/PR14940 20556 388. https://gcc.gnu.org/PR20239 20557 389. https://gcc.gnu.org/PR15220 20558 390. https://gcc.gnu.org/PR19275 20559 391. https://gcc.gnu.org/PR21888 20560 392. https://gcc.gnu.org/PR15342 20561 393. https://gcc.gnu.org/PR23985 20562 394. https://gcc.gnu.org/PR16719 20563 395. https://gcc.gnu.org/PR21723 20564 396. https://gcc.gnu.org/PR21841 20565 397. https://gcc.gnu.org/PR23644 20566 398. https://gcc.gnu.org/PR24718 20567 399. https://gcc.gnu.org/PR18421 20568 400. https://gcc.gnu.org/PR20621 20569 401. https://gcc.gnu.org/PR18583 20570 402. https://gcc.gnu.org/PR20191 20571 403. https://gcc.gnu.org/PR22083 20572 404. https://gcc.gnu.org/PR23070 20573 405. https://gcc.gnu.org/PR23404 20574 406. https://gcc.gnu.org/PR23539 20575 407. https://gcc.gnu.org/PR24102 20576 408. https://gcc.gnu.org/PR24465 20577 409. https://gcc.gnu.org/PR19933 20578 410. https://gcc.gnu.org/PR21889 20579 411. https://gcc.gnu.org/PR19300 20580 412. https://gcc.gnu.org/PR20301 20581 413. https://gcc.gnu.org/PR20673 20582 414. https://gcc.gnu.org/PR18582 20583 415. https://gcc.gnu.org/PR19340 20584 416. https://gcc.gnu.org/PR21716 20585 417. https://gcc.gnu.org/PR24315 20586 418. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.6 20587 419. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 20588 420. https://gcc.gnu.org/onlinedocs/ 20589 421. mailto:gcc-help (a] gcc.gnu.org 20590 422. mailto:gcc (a] gcc.gnu.org 20591 423. https://gcc.gnu.org/lists.html 20592 424. https://www.fsf.org/ 20593 425. https://gcc.gnu.org/about.html 20594 ====================================================================== 20595 http://gcc.gnu.org/gcc-3.3/index.html 20596 GCC 3.3 Release Series 20597 20598 (This release series is no longer supported.) 20599 20600 May 03, 2005 20601 20602 The [1]GNU project and the GCC developers are pleased to announce the 20603 release of GCC 3.3.6. 20604 20605 This release is a bug-fix release, containing fixes for regressions in 20606 GCC 3.3.5 relative to previous releases of GCC. 20607 20608 This release is the last of the series 3.3.x. 20609 20610 The GCC 3.3 release series includes numerous [2]new features, 20611 improvements, bug fixes, and other changes, thanks to an [3]amazing 20612 group of volunteers. 20613 20614 Release History 20615 20616 GCC 3.3.6 20617 May 3, 2005 ([4]changes) 20618 20619 GCC 3.3.5 20620 September 30, 2004 ([5]changes) 20621 20622 GCC 3.3.4 20623 May 31, 2004 ([6]changes) 20624 20625 GCC 3.3.3 20626 February 14, 2004 ([7]changes) 20627 20628 GCC 3.3.2 20629 October 16, 2003 ([8]changes) 20630 20631 GCC 3.3.1 20632 August 8, 2003 ([9]changes) 20633 20634 GCC 3.3 20635 May 14, 2003 ([10]changes) 20636 20637 References and Acknowledgements 20638 20639 GCC used to stand for the GNU C Compiler, but since the compiler 20640 supports several other languages aside from C, it now stands for the 20641 GNU Compiler Collection. 20642 20643 A list of [11]successful builds is updated as new information becomes 20644 available. 20645 20646 The GCC developers would like to thank the numerous people that have 20647 contributed new features, improvements, bug fixes, and other changes as 20648 well as test results to GCC. This [12]amazing group of volunteers is 20649 what makes GCC successful. 20650 20651 For additional information about GCC please refer to the [13]GCC 20652 project web site or contact the [14]GCC development mailing list. 20653 20654 To obtain GCC please use [15]our mirror sites, or our CVS server. 20655 20656 20657 For questions related to the use of GCC, please consult these web 20658 pages and the [16]GCC manuals. If that fails, the 20659 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20660 web pages and the development of GCC are welcome on our developer 20661 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 20662 archives. 20663 20664 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 20665 distribution of this entire article is permitted in any medium, 20666 provided this notice is preserved. 20667 20668 These pages are [21]maintained by the GCC team. Last modified 20669 2022-10-26. 20670 20671 References 20672 20673 1. http://www.gnu.org/ 20674 2. https://gcc.gnu.org/gcc-3.3/changes.html 20675 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20676 4. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 20677 5. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 20678 6. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 20679 7. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 20680 8. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 20681 9. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 20682 10. https://gcc.gnu.org/gcc-3.3/changes.html 20683 11. https://gcc.gnu.org/gcc-3.3/buildstat.html 20684 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20685 13. https://gcc.gnu.org/index.html 20686 14. mailto:gcc (a] gcc.gnu.org 20687 15. https://gcc.gnu.org/mirrors.html 20688 16. https://gcc.gnu.org/onlinedocs/ 20689 17. mailto:gcc-help (a] gcc.gnu.org 20690 18. mailto:gcc (a] gcc.gnu.org 20691 19. https://gcc.gnu.org/lists.html 20692 20. https://www.fsf.org/ 20693 21. https://gcc.gnu.org/about.html 20694 ====================================================================== 20695 http://gcc.gnu.org/gcc-3.3/changes.html 20696 GCC 3.3 Release Series 20697 Changes, New Features, and Fixes 20698 20699 The latest release in the 3.3 release series is [1]GCC 3.3.6. 20700 20701 Caveats 20702 20703 * The preprocessor no longer accepts multi-line string literals. They 20704 were deprecated in 3.0, 3.1, and 3.2. 20705 * The preprocessor no longer supports the -A- switch when appearing 20706 alone. -A- followed by an assertion is still supported. 20707 * Support for all the systems [2]obsoleted in GCC 3.1 has been 20708 removed from GCC 3.3. See below for a [3]list of systems which are 20709 obsoleted in this release. 20710 * Checking for null format arguments has been decoupled from the rest 20711 of the format checking mechanism. Programs which use the format 20712 attribute may regain this functionality by using the new [4]nonnull 20713 function attribute. Note that all functions for which GCC has a 20714 built-in format attribute, an appropriate built-in nonnull 20715 attribute is also applied. 20716 * The DWARF (version 1) debugging format has been deprecated and will 20717 be removed in a future version of GCC. Version 2 of the DWARF 20718 debugging format will continue to be supported for the foreseeable 20719 future. 20720 * The C and Objective-C compilers no longer accept the "Naming Types" 20721 extension (typedef foo = bar); it was already unavailable in C++. 20722 Code which uses it will need to be changed to use the "typeof" 20723 extension instead: typedef typeof(bar) foo. (We have removed this 20724 extension without a period of deprecation because it has caused the 20725 compiler to crash since version 3.0 and no one noticed until very 20726 recently. Thus we conclude it is not in widespread use.) 20727 * The -traditional C compiler option has been removed. It was 20728 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 20729 available.) The <varargs.h> header, used for writing variadic 20730 functions in traditional C, still exists but will produce an error 20731 message if used. 20732 * GCC 3.3.1 automatically places zero-initialized variables in the 20733 .bss section on some operating systems. Versions of GNU Emacs up to 20734 (and including) 21.3 will not work correctly when using this 20735 optimization; you can use -fno-zero-initialized-in-bss to disable 20736 it. 20737 20738 General Optimizer Improvements 20739 20740 * A new scheme for accurately describing processor pipelines, the 20741 [5]DFA scheduler, has been added. 20742 * Pavel Nejedly, Charles University Prague, has contributed new file 20743 format used by the edge coverage profiler (-fprofile-arcs). 20744 The new format is robust and diagnoses common mistakes where 20745 profiles from different versions (or compilations) of the program 20746 are combined resulting in nonsensical profiles and slow code to 20747 produced with profile feedback. Additionally this format allows 20748 extra data to be gathered. Currently, overall statistics are 20749 produced helping optimizers to identify hot spots of a program 20750 globally replacing the old intra-procedural scheme and resulting in 20751 better code. Note that the gcov tool from older GCC versions will 20752 not be able to parse the profiles generated by GCC 3.3 and vice 20753 versa. 20754 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 20755 pass enabled using -ftracer. This pass simplifies the control flow 20756 of functions allowing other optimizations to do better job. 20757 He also contributed the function reordering pass 20758 (-freorder-functions) to optimize function placement using profile 20759 feedback. 20760 20761 New Languages and Language specific improvements 20762 20763 C/ObjC/C++ 20764 20765 * The preprocessor now accepts directives within macro arguments. It 20766 processes them just as if they had not been within macro arguments. 20767 * The separate ISO and traditional preprocessors have been completely 20768 removed. The front end handles either type of preprocessed output 20769 if necessary. 20770 * In C99 mode preprocessor arithmetic is done in the precision of the 20771 target's intmax_t, as required by that standard. 20772 * The preprocessor can now copy comments inside macros to the output 20773 file when the macro is expanded. This feature, enabled using the 20774 -CC option, is intended for use by applications which place 20775 metadata or directives inside comments, such as lint. 20776 * The method of constructing the list of directories to be searched 20777 for header files has been revised. If a directory named by a -I 20778 option is a standard system include directory, the option is 20779 ignored to ensure that the default search order for system 20780 directories and the special treatment of system header files are 20781 not defeated. 20782 * A few more [6]ISO C99 features now work correctly. 20783 * A new function attribute, nonnull, has been added which allows 20784 pointer arguments to functions to be specified as requiring a 20785 non-null value. The compiler currently uses this information to 20786 issue a warning when it detects a null value passed in such an 20787 argument slot. 20788 * A new type attribute, may_alias, has been added. Accesses to 20789 objects with types with this attribute are not subjected to 20790 type-based alias analysis, but are instead assumed to be able to 20791 alias any other type of objects, just like the char type. 20792 20793 C++ 20794 20795 * Type based alias analysis has been implemented for C++ aggregate 20796 types. 20797 20798 Objective-C 20799 20800 * Generate an error if Objective-C objects are passed by value in 20801 function and method calls. 20802 * When -Wselector is used, check the whole list of selectors at the 20803 end of compilation, and emit a warning if a @selector() is not 20804 known. 20805 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 20806 * No longer need to include objc/objc-class.h to compile self calls 20807 in class methods (NeXT runtime only). 20808 * New -Wundeclared-selector option. 20809 * Removed selector bloating which was causing object files to be 10% 20810 bigger on average (GNU runtime only). 20811 * Using at run time @protocol() objects has been fixed in certain 20812 situations (GNU runtime only). 20813 * Type checking has been fixed and improved in many situations 20814 involving protocols. 20815 20816 Java 20817 20818 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 20819 1.4) API. 20820 * The JDK 1.4 assert facility has been implemented. 20821 * The bytecode interpreter is now direct threaded and thus faster. 20822 20823 Fortran 20824 20825 * Fortran improvements are listed in [7]the Fortran documentation. 20826 20827 Ada 20828 20829 * Ada tasking now works with glibc 2.3.x threading libraries. 20830 20831 New Targets and Target Specific Improvements 20832 20833 * The following changes have been made to the HP-PA port: 20834 + The port now defaults to scheduling for the PA8000 series of 20835 processors. 20836 + Scheduling support for the PA7300 processor has been added. 20837 + The 32-bit port now supports weak symbols under HP-UX 11. 20838 + The handling of initializers and finalizers has been improved 20839 under HP-UX 11. The 64-bit port no longer uses collect2. 20840 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 20841 + ABI fixes to correct the passing of small structures by value. 20842 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 20843 use the DFA processor pipeline description. 20844 * The following NetBSD configurations for the SuperH processor family 20845 have been added: 20846 + SH3, big-endian, sh-*-netbsdelf* 20847 + SH3, little-endian, shle-*-netbsdelf* 20848 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 20849 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 20850 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 20851 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 20852 * The following changes have been made to the IA-32/x86-64 port: 20853 + SSE2 and 3dNOW! intrinsics are now supported. 20854 + Support for thread local storage has been added to the IA-32 20855 and x86-64 ports. 20856 + The x86-64 port has been significantly improved. 20857 * The following changes have been made to the MIPS port: 20858 + All configurations now accept the -mabi switch. Note that you 20859 will need appropriate multilibs for this option to work 20860 properly. 20861 + ELF configurations will always pass an ABI flag to the 20862 assembler, except when the MIPS EABI is selected. 20863 + -mabi=64 no longer selects MIPS IV code. 20864 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 20865 been removed from this release. 20866 + -march now changes the core ISA level. In previous releases, 20867 it would change the use of processor-specific extensions, but 20868 would leave the core ISA unchanged. For example, mips64-elf 20869 -march=r8000 will now generate MIPS IV code. 20870 + Under most configurations, -mipsN now acts as a synonym for 20871 -march. 20872 + There are some new preprocessor macros to describe the -march 20873 and -mtune settings. See the documentation of those options 20874 for details. 20875 + Support for the NEC VR-Series processors has been added. This 20876 includes the 54xx, 5500, and 41xx series. 20877 + Support for the Sandcraft sr71k processor has been added. 20878 * The following changes have been made to the S/390 port: 20879 + Support to build the Java runtime libraries has been added. 20880 Java is now enabled by default on s390-*-linux* and 20881 s390x-*-linux* targets. 20882 + Multilib support for the s390x-*-linux* target has been added; 20883 this allows to build 31-bit binaries using the -m31 option. 20884 + Support for thread local storage has been added. 20885 + Inline assembler code may now use the 'Q' constraint to 20886 specify memory operands without index register. 20887 + Various platform-specific performance improvements have been 20888 implemented; in particular, the compiler now uses the BRANCH 20889 ON COUNT family of instructions and makes more frequent use of 20890 the TEST UNDER MASK family of instructions. 20891 * The following changes have been made to the PowerPC port: 20892 + Support for IBM Power4 processor added. 20893 + Support for Motorola e500 SPE added. 20894 + Support for AIX 5.2 added. 20895 + Function and Data sections now supported on AIX. 20896 + Sibcall optimizations added. 20897 * The support for H8 Tiny is added to the H8/300 port with -mn. 20898 20899 [8]Obsolete Systems 20900 20901 Support for a number of older systems has been declared obsolete in GCC 20902 3.3. Unless there is activity to revive them, the next release of GCC 20903 will have their sources permanently removed. 20904 20905 All configurations of the following processor architectures have been 20906 declared obsolete: 20907 * Matsushita MN10200, mn10200-*-* 20908 * Motorola 88000, m88k-*-* 20909 * IBM ROMP, romp-*-* 20910 20911 Also, some individual systems have been obsoleted: 20912 * Alpha 20913 + Interix, alpha*-*-interix* 20914 + Linux libc1, alpha*-*-linux*libc1* 20915 + Linux ECOFF, alpha*-*-linux*ecoff* 20916 * ARM 20917 + Generic a.out, arm*-*-aout* 20918 + Conix, arm*-*-conix* 20919 + "Old ABI," arm*-*-oabi 20920 + StrongARM/COFF, strongarm-*-coff* 20921 * HPPA (PA-RISC) 20922 + Generic OSF, hppa1.0-*-osf* 20923 + Generic BSD, hppa1.0-*-bsd* 20924 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 20925 + HiUX, hppa*-*-hiux* 20926 + Mach Lites, hppa*-*-lites* 20927 * Intel 386 family 20928 + Windows NT 3.x, i?86-*-win32 20929 * MC68000 family 20930 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 20931 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 20932 m68k-sun-mach* 20933 + AT&T systems, m68000-att-sysv* 20934 + Atari systems, m68k-atari-sysv* 20935 + Motorola systems, m68k-motorola-sysv* 20936 + NCR systems, m68k-ncr-sysv* 20937 + Plexus systems, m68k-plexus-sysv* 20938 + Commodore systems, m68k-cbm-sysv* 20939 + Citicorp TTI, m68k-tti-* 20940 + Unos, m68k-crds-unos* 20941 + Concurrent RTU, m68k-ccur-rtu* 20942 + Linux a.out, m68k-*-linux*aout* 20943 + Linux libc1, m68k-*-linux*libc1* 20944 + pSOS, m68k-*-psos* 20945 * MIPS 20946 + Generic ECOFF, mips*-*-ecoff* 20947 + SINIX, mips-sni-sysv4 20948 + Orion RTEMS, mips64orion-*-rtems* 20949 * National Semiconductor 32000 20950 + OpenBSD, ns32k-*-openbsd* 20951 * POWER (aka RS/6000) and PowerPC 20952 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 20953 + Bull BOSX, rs6000-bull-bosx 20954 + Generic Mach, rs6000-*-mach* 20955 + Generic SysV, powerpc*-*-sysv* 20956 + Linux libc1, powerpc*-*-linux*libc1* 20957 * Sun SPARC 20958 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 20959 sparclite-*-aout*, and sparc86x-*-aout* 20960 + NetBSD a.out, sparc-*-netbsd*aout* 20961 + Generic BSD, sparc-*-bsd* 20962 + ChorusOS, sparc-*-chorusos* 20963 + Linux a.out, sparc-*-linux*aout* 20964 + Linux libc1, sparc-*-linux*libc1* 20965 + LynxOS, sparc-*-lynxos* 20966 + Solaris on HAL hardware, sparc-hal-solaris2* 20967 + SunOS versions 3 and 4, sparc-*-sunos[34]* 20968 * NEC V850 20969 + RTEMS, v850-*-rtems* 20970 * VAX 20971 + VMS, vax-*-vms* 20972 20973 Documentation improvements 20974 20975 Other significant improvements 20976 20977 * Almost all front-end dependencies in the compiler have been 20978 separated out into a set of language hooks. This should make adding 20979 a new front end clearer and easier. 20980 * One effect of removing the separate preprocessor is a small 20981 increase in the robustness of the compiler in general, and the 20982 maintainability of target descriptions. Previously target-specific 20983 built-in macros and others, such as __FAST_MATH__, had to be 20984 handled with so-called specs that were hard to maintain. Often they 20985 would fail to behave properly when conflicting options were 20986 supplied on the command line, and define macros in the user's 20987 namespace even when strict ISO compliance was requested. 20988 Integrating the preprocessor has cleanly solved these issues. 20989 * The Makefile suite now supports redirection of make install by 20990 means of the variable DESTDIR. 20991 __________________________________________________________________ 20992 20993 GCC 3.3 20994 20995 Detailed release notes for the GCC 3.3 release follow. 20996 20997 Bug Fixes 20998 20999 bootstrap failures 21000 21001 * [9]10140 cross compiler build failures: missing __mempcpy (DUP: 21002 [10]10198,[11]10338) 21003 21004 Internal compiler errors (multi-platform) 21005 21006 * [12]3581 large string causes segmentation fault in cc1 21007 * [13]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 21008 * [14]5533 (c++) ICE when processing std::accumulate(begin, end, 21009 init, invalid_op) 21010 * [15]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 21011 * [16]6412 (c++) ICE in retrieve_specialization 21012 * [17]6620 (c++) partial template specialization causes an ICE 21013 (segmentation fault) 21014 * [18]6663 (c++) ICE with attribute aligned 21015 * [19]7068 ICE with incomplete types 21016 * [20]7083 (c++) ICE using -gstabs with dodgy class derivation 21017 * [21]7647 (c++) ICE when data member has the name of the enclosing 21018 class 21019 * [22]7675 ICE in fixup_var_refs_1 21020 * [23]7718 'complex' template instantiation causes ICE 21021 * [24]8116 (c++) ICE in member template function 21022 * [25]8358 (ada) Ada compiler accesses freed memory, crashes 21023 * [26]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 21024 fault 21025 * [27]8564 (c++) ICE in find_function_data, in function.c 21026 * [28]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 21027 * [29]8766 (c++) ICE after failed initialization of static template 21028 variable 21029 * [30]8803 ICE in instantiate_virtual_regs_1, in function.c 21030 * [31]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 21031 * [32]8906 (c++) ICE (Segmentation fault) when parsing nested-class 21032 definition 21033 * [33]9216 (c++) ICE on missing template parameter 21034 * [34]9261 (c++) ICE in arg_assoc, in cp/decl2.c 21035 * [35]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 21036 loop 21037 * [36]9429 (c++) ICE in template instantiation with a pointered new 21038 operator 21039 * [37]9516 Internal error when using a big array 21040 * [38]9600 (c++) ICE with typedefs in template class 21041 * [39]9629 (c++) virtual inheritance segfault 21042 * [40]9672 (c++) ICE: Error reporting routines re-entered 21043 * [41]9749 (c++) ICE in write_expression on invalid function 21044 prototype 21045 * [42]9794 (fortran) ICE: floating point exception during constant 21046 folding 21047 * [43]9829 (c++) Missing colon in nested namespace usage causes ICE 21048 * [44]9916 (c++) ICE with noreturn function in ?: statement 21049 * [45]9936 ICE with local function and variable-length 2d array 21050 * [46]10262 (c++) cc1plus crashes with large generated code 21051 * [47]10278 (c++) ICE in parser for invalid code 21052 * [48]10446 (c++) ICE on definition of nonexistent member function of 21053 nested class in a class template 21054 * [49]10451 (c++) ICE in grokdeclarator on spurious mutable 21055 declaration 21056 * [50]10506 (c++) ICE in build_new at cp/init.c with 21057 -fkeep-inline-functions and multiple inheritance 21058 * [51]10549 (c++) ICE in store_bit_field on bitfields that exceed the 21059 precision of the declared type 21060 21061 Optimization bugs 21062 21063 * [52]2001 Inordinately long compile times in reload CSE regs 21064 * [53]2391 Exponential compilation time explosion in combine 21065 * [54]2960 Duplicate loop conditions even with -Os 21066 * [55]4046 redundant conditional branch 21067 * [56]6405 Loop-unrolling related performance regressions 21068 * [57]6798 very long compile time with large case-statement 21069 * [58]6871 const objects shouldn't be moved to .bss 21070 * [59]6909 problem w/ -Os on modified loop-2c.c test case 21071 * [60]7189 gcc -O2 -Wall does not print ``control reaches end of 21072 non-void function'' warning 21073 * [61]7642 optimization problem with signbit() 21074 * [62]8634 incorrect code for inlining of memcpy under -O2 21075 * [63]8750 Cygwin prolog generation erroneously emitting __alloca as 21076 regular function call 21077 21078 C front end 21079 21080 * [64]2161 long if-else cascade overflows parser stack 21081 * [65]4319 short accepted on typedef'd char 21082 * [66]8602 incorrect line numbers in warning messages when using 21083 inline functions 21084 * [67]9177 -fdump-translation-unit: C front end deletes function_decl 21085 AST nodes and breaks debugging dumps 21086 * [68]9853 miscompilation of non-constant structure initializer 21087 21088 c++ compiler and library 21089 21090 * [69]45 legal template specialization code is rejected (DUP: 21091 [70]3784) 21092 * [71]764 lookup failure: friend operator and dereferencing a pointer 21093 and templates (DUP: [72]5116) 21094 * [73]2862 gcc accepts invalid explicit instantiation syntax (DUP: 21095 2863) 21096 * [74]3663 G++ doesn't check access control during template 21097 instantiation 21098 * [75]3797 gcc fails to emit explicit specialization of a template 21099 member 21100 * [76]3948 Two destructors are called when no copy destructor is 21101 defined (ABI change) 21102 * [77]4137 Conversion operator within template is not accepted 21103 * [78]4361 bogus ambiguity taking the address of a member template 21104 * [79]4802 g++ accepts illegal template code (access to private 21105 member; DUP: [80]5837) 21106 * [81]4803 inline function is used but never defined, and g++ does 21107 not object 21108 * [82]5094 Partial specialization cannot be friend? 21109 * [83]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 21110 * [84]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 21111 time 21112 * [85]7015 certain __asm__ constructs rejected 21113 * [86]7086 compile time regression (quadratic behavior in 21114 fixup_var_refs) 21115 * [87]7099 G++ doesn't set the noreturn attribute on std::exit and 21116 std::abort 21117 * [88]7247 copy constructor missing when inlining enabled (invalid 21118 optimization?) 21119 * [89]7441 string array initialization compilation time regression 21120 from seconds to minutes 21121 * [90]7768 __PRETTY_FUNCTION__ for template destructor is wrong 21122 * [91]7804 bad printing of floating point constant in warning message 21123 * [92]8099 Friend classes and template specializations 21124 * [93]8117 member function pointers and multiple inheritance 21125 * [94]8205 using declaration and multiple inheritance 21126 * [95]8645 unnecessary non-zero checks in stl_tree.h 21127 * [96]8724 explicit destructor call for incomplete class allowed 21128 * [97]8805 compile time regression with many member variables 21129 * [98]8691 -O3 and -fno-implicit-templates are incompatible 21130 * [99]8700 unhelpful error message for binding temp to reference 21131 * [100]8724 explicit destructor call for incomplete class allowed 21132 * [101]8949 numeric_limits<>::denorm_min() and is_iec559 problems 21133 * [102]9016 Failure to consistently constant fold "constant" C++ 21134 objects 21135 * [103]9053 g++ confused about ambiguity of overloaded function 21136 templates 21137 * [104]9152 undefined virtual thunks 21138 * [105]9182 basic_filebuf<> does not report errors in codecvt<>::out 21139 * [106]9297 data corruption due to codegen bug (when copying.) 21140 * [107]9318 i/ostream::operator>>/<<(streambuf*) broken 21141 * [108]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 21142 * [109]9400 bogus -Wshadow warning: shadowed declaration of this in 21143 local classes 21144 * [110]9424 i/ostream::operator>>/<<(streambuf*) drops characters 21145 * [111]9425 filebuf::pbackfail broken (DUP: [112]9439) 21146 * [113]9474 GCC freezes in compiling a weird code mixing <iostream> 21147 and <iostream.h> 21148 * [114]9548 Incorrect results from setf(ios::fixed) and precision(-1) 21149 [115][DR 231] 21150 * [116]9555 ostream inserters fail to set badbit on exception 21151 * [117]9561 ostream inserters rethrow exception of wrong type 21152 * [118]9563 ostream::sentry returns true after a failed preparation 21153 * [119]9582 one-definition rule violation in std::allocator 21154 * [120]9622 __PRETTY_FUNCTION__ incorrect in template destructors 21155 * [121]9683 bug in initialization chains for static const variables 21156 from template classes 21157 * [122]9791 -Woverloaded-virtual reports hiding of destructor 21158 * [123]9817 collate::compare doesn't handle nul characters 21159 * [124]9825 filebuf::sputbackc breaks sbumpc 21160 * [125]9826 operator>>(basic_istream, basic_string) fails to compile 21161 with custom traits 21162 * [126]9924 Multiple using statements for builtin functions not 21163 allowed 21164 * [127]9946 destructor is not called for temporary object 21165 * [128]9964 filebuf::close() sometimes fails to close file 21166 * [129]9988 filebuf::overflow writes EOF to file 21167 * [130]10033 optimization breaks polymorphic references w/ typeid 21168 operator 21169 * [131]10097 filebuf::underflow drops characters 21170 * [132]10132 filebuf destructor can throw exceptions 21171 * [133]10180 gcc fails to warn about non-inlined function 21172 * [134]10199 method parametrized by template does not work everywhere 21173 * [135]10300 use of array-new (nothrow) in segfaults on NULL return 21174 * [136]10427 Stack corruption with variable-length automatic arrays 21175 and virtual destructors 21176 * [137]10503 Compilation never stops in fixed_type_or_null 21177 21178 Objective-C 21179 21180 * [138]5956 selectors aren't matched properly when added to the 21181 selector table 21182 21183 Fortran compiler and library 21184 21185 * [139]1832 list directed i/o overflow hangs, -fbounds-check doesn't 21186 detect 21187 * [140]3924 g77 generates code that is rejected by GAS if COFF debug 21188 info requested 21189 * [141]5634 doc: explain that configure --prefix=~/... does not work 21190 * [142]6367 multiple repeat counts confuse namelist read into array 21191 * [143]6491 Logical operations error on logicals when using 21192 -fugly-logint 21193 * [144]6742 Generation of C++ Prototype for FORTRAN and extern "C" 21194 * [145]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 21195 on irix6.5 21196 * [146]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 21197 assume a direct access file 21198 * [147]7278 g77 "bug"; the executable misbehaves (with -O2 21199 -fno-automatic) 21200 * [148]7384 DATE_AND_TIME milliseconds field inactive on Windows 21201 * [149]7388 Incorrect output with 0-based array of characters 21202 * [150]8587 Double complex zero ** double precision number -> NaN 21203 instead of zero 21204 * [151]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 21205 unknown register name line-length-none 21206 * [152]10197 Direct access files not unformatted by default 21207 21208 Java compiler and library 21209 21210 * [153]6005 gcj fails to build rhug on alpha 21211 * [154]6389 System.getProperty("") should always throw an 21212 IllegalArgumentException 21213 * [155]6576 java.util.ResourceBundle.getResource ignores locale 21214 * [156]6652 new java.io.File("").getCanonicalFile() throws exception 21215 * [157]7060 getMethod() doesn't search super interface 21216 * [158]7073 bytecode interpreter gives wrong answer for interface 21217 getSuperclass() 21218 * [159]7180 possible bug in 21219 javax.naming.spi.NamingManager.getPlusPath() 21220 * [160]7416 java.security startup refs "GNU libgcj.security" 21221 * [161]7570 Runtime.exec with null envp: child doesn't inherit parent 21222 env (DUP: [162]7578) 21223 * [163]7611 Internal error while compiling libjava with -O 21224 * [164]7709 NullPointerException in _Jv_ResolvePoolEntry 21225 * [165]7766 ZipInputStream.available returns 0 immediately after 21226 construction 21227 * [166]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 21228 * [167]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 21229 * [168]8142 '$' in class names vs. dlopen 'dynamic string tokens' 21230 * [169]8234 ZipInputStream chokes when InputStream.read() returns 21231 small chunks 21232 * [170]8415 reflection bug: exception info for Method 21233 * [171]8481 java.Random.nextInt(int) may return negative 21234 * [172]8593 Error reading GZIPped files with BufferedReader 21235 * [173]8759 java.beans.Introspector has no flushCaches() or 21236 flushFromCaches() methods 21237 * [174]8997 spin() calls Thread.sleep 21238 * [175]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 21239 instead of the root content of C: 21240 * [176]9254 java::lang::Object::wait(), threads-win32.cc returns 21241 wrong return codes 21242 * [177]9271 Severe bias in java.security.SecureRandom 21243 21244 Ada compiler and library 21245 21246 * [178]6767 make gnatlib-shared fails on -laddr2line 21247 * [179]9911 gnatmake fails to link when GCC configured with 21248 --with-sjlj-exceptions=yes 21249 * [180]10020 Can't bootstrap gcc on AIX with Ada enabled 21250 * [181]10546 Ada tasking not working on Red Hat 9 21251 21252 preprocessor 21253 21254 * [182]7029 preprocessor should ignore #warning with -M 21255 21256 ARM-specific 21257 21258 * [183]2903 [arm] Optimization bug with long long arithmetic 21259 * [184]7873 arm-linux-gcc fails when assigning address to a bit field 21260 21261 FreeBSD-specific 21262 21263 * [185]7680 float functions undefined in math.h/cmath with #define 21264 _XOPEN_SOURCE 21265 21266 HP-UX or HP-PA-specific 21267 21268 * [186]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 21269 * [187]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 21270 fputc_unlocked 21271 * [188]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 21272 21273 m68hc11-specific 21274 21275 * [189]6744 Bad assembler code generated: reference to pseudo 21276 register z 21277 * [190]7361 Internal compiler error in reload_cse_simplify_operands, 21278 in reload1.c 21279 21280 MIPS-specific 21281 21282 * [191]9496 [mips-linux] bug in optimizer? 21283 21284 PowerPC-specific 21285 21286 * [192]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 21287 space 21288 * [193]8480 reload ICEs for LAPACK code on powerpc64-linux 21289 * [194]8784 [AIX] Internal compiler error in simplify_gen_subreg 21290 * [195]10315 [powerpc] ICE: in extract_insn, in recog.c 21291 21292 SPARC-specific 21293 21294 * [196]10267 (documentation) Wrong build instructions for 21295 *-*-solaris2* 21296 21297 x86-specific (Intel/AMD) 21298 21299 * [197]7916 ICE in instantiate_virtual_register_1 21300 * [198]7926 (c++) i486 instructions in header files make c++ programs 21301 crash on i386 21302 * [199]8555 ICE in gen_split_1231 21303 * [200]8994 ICE with -O -march=pentium4 21304 * [201]9426 ICE with -fssa -funroll-loops -fprofile-arcs 21305 * [202]9806 ICE in inline assembly with -fPIC flag 21306 * [203]10077 gcc -msse2 generates movd to move dwords between xmm 21307 regs 21308 * [204]10233 64-bit comparison only comparing bottom 32-bits 21309 * [205]10286 type-punning doesn't work with __m64 and -O 21310 * [206]10308 [x86] ICE with -O -fgcse or -O2 21311 __________________________________________________________________ 21312 21313 [207]GCC 3.3.1 21314 21315 Bug Fixes 21316 21317 This section lists the problem reports (PRs) from GCC's bug tracking 21318 system that are known to be fixed in the 3.3.1 release. This list might 21319 not be complete (that is, it is possible that some PRs that have been 21320 fixed are not listed here). 21321 21322 Bootstrap failures 21323 21324 * [208]11272 [Solaris] make bootstrap fails while building libstdc++ 21325 21326 Internal compiler errors (multi-platform) 21327 21328 * [209]5754 ICE on invalid nested template class 21329 * [210]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 21330 and --enable-checking 21331 * [211]6949 (c++) ICE in tsubst_decl, in cp/pt.c 21332 * [212]7053 (c++) ICE when declaring a function already defined as a 21333 friend method of a template class 21334 * [213]8164 (c++) ICE when using different const expressions as 21335 template parameter 21336 * [214]8384 (c++) ICE in is_base_type, in dwarf2out.c 21337 * [215]9559 (c++) ICE with invalid initialization of a static const 21338 * [216]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 21339 when redeclaring a static member variable 21340 * [217]9864 (fortran) ICE in add_abstract_origin_attribute, in 21341 dwarfout.c with -g -O -finline-functions 21342 * [218]10432 (c++) ICE in poplevel, in cp/decl.c 21343 * [219]10475 ICE in subreg_highpart_offset for code with long long 21344 * [220]10635 (c++) ICE when dereferencing an incomplete type casted 21345 from a void pointer 21346 * [221]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 21347 instantiating static member variables 21348 * [222]10700 ICE in copy_to_mode_reg on 64-bit targets 21349 * [223]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 21350 * [224]10796 (c++) ICE when defining an enum with two values: -1 and 21351 MAX_INT_64BIT 21352 * [225]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 21353 sched.c 21354 * [226]10939 (c++) ICE with template code 21355 * [227]10956 (c++) ICE when specializing a template member function 21356 of a template class, in tsubst, in cp/pt.c 21357 * [228]11041 (c++) ICE: const myclass &x = *x; (when operator*() 21358 defined) 21359 * [229]11059 (c++) ICE with empty union 21360 * [230]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 21361 -O2 -fnon-call-exceptions 21362 * [231]11105 (c++) ICE in mangle_conv_op_name_for_type 21363 * [232]11149 (c++) ICE on error when instantiation with call function 21364 of a base type 21365 * [233]11228 (c++) ICE on new-expression using array operator new and 21366 default-initialization 21367 * [234]11282 (c++) Infinite memory usage after syntax error 21368 * [235]11301 (fortran) ICE with -fno-globals 21369 * [236]11308 (c++) ICE when using an enum type name as if it were a 21370 class or namespace 21371 * [237]11473 (c++) ICE with -gstabs when empty struct inherits from 21372 an empty struct 21373 * [238]11503 (c++) ICE when instantiating template with ADDR_EXPR 21374 * [239]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 21375 template member functions 21376 21377 Optimization bugs 21378 21379 * [240]11198 -O2 -frename-registers generates wrong code (aliasing 21380 problem) 21381 * [241]11304 Wrong code production with -fomit-frame-pointer 21382 * [242]11381 volatile memory access optimized away 21383 * [243]11536 [strength-reduce] -O2 optimization produces wrong code 21384 * [244]11557 constant folding bug generates wrong code 21385 21386 C front end 21387 21388 * [245]5897 No warning for statement after return 21389 * [246]11279 DWARF-2 output mishandles large enums 21390 21391 Preprocessor bugs 21392 21393 * [247]11022 no warning for non-compatible macro redefinition 21394 21395 C++ compiler and library 21396 21397 * [248]2330 static_cast<>() to a private base is allowed 21398 * [249]5388 Incorrect message "operands to ?: have different types" 21399 * [250]5390 Libiberty fails to demangle multi-digit template 21400 parameters 21401 * [251]7877 Incorrect parameter passing to specializations of member 21402 function templates 21403 * [252]9393 Anonymous namespaces and compiling the same file twice 21404 * [253]10032 -pedantic converts some errors to warnings 21405 * [254]10468 const typeof(x) is non-const, but only in templates 21406 * [255]10527 confused error message with "new int()" parameter 21407 initializer 21408 * [256]10679 parameter MIN_INLINE_INSNS is not honored 21409 * [257]10682 gcc chokes on a typedef for an enum inside a class 21410 template 21411 * [258]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 21412 0. 21413 * [259]10845 template member function (with nested template as 21414 parameter) cannot be called anymore if another unrelated template 21415 member function is defined 21416 * [260]10849 Cannot define an out-of-class specialization of a 21417 private nested template class 21418 * [261]10888 Suppress -Winline warnings for system headers 21419 * [262]10929 -Winline warns about functions for which no definition 21420 is visible 21421 * [263]10931 valid conversion static_cast<const unsigned 21422 int&>(lvalue-of-type-int) is rejected 21423 * [264]10940 Bad code with explicit specialization 21424 * [265]10968 If member function implicitly instantiated, explicit 21425 instantiation of class fails to instantiate it 21426 * [266]10990 Cannot convert with dynamic_cast<> to a private base 21427 class from within a member function 21428 * [267]11039 Bad interaction between implicit typename deprecation 21429 and friendship 21430 * [268]11062 (libstdc++) avoid __attribute__ ((unused)); say 21431 "__unused__" instead 21432 * [269]11095 C++ iostream manipulator causes segfault when called 21433 with negative argument 21434 * [270]11098 g++ doesn't emit complete debugging information for 21435 local variables in destructors 21436 * [271]11137 GNU/Linux shared library constructors not called unless 21437 there's one global object 21438 * [272]11154 spurious ambiguity report for template class 21439 specialization 21440 * [273]11329 Compiler cannot find user defined implicit typecast 21441 * [274]11332 Spurious error with casts in ?: expression 21442 * [275]11431 static_cast behavior with subclasses when default 21443 constructor available 21444 * [276]11528 money_get facet does not accept "$.00" as valid 21445 * [277]11546 Type lookup problems in out-of-line definition of a 21446 class doubly nested from a template class 21447 * [278]11567 C++ code containing templated member function with same 21448 name as pure virtual member function results in linking failure 21449 * [279]11645 Failure to deal with using and private inheritance 21450 21451 Java compiler and library 21452 21453 * [280]5179 Qualified static field access doesn't initialize its 21454 class 21455 * [281]8204 gcj -O2 to native reorders certain instructions 21456 improperly 21457 * [282]10838 java.io.ObjectInputStream syntax error 21458 * [283]10886 The RMI registry that comes with GCJ does not work 21459 correctly 21460 * [284]11349 JNDI URL context factories not located correctly 21461 21462 x86-specific (Intel/AMD) 21463 21464 * [285]4823 ICE on inline assembly code 21465 * [286]8878 miscompilation with -O and SSE 21466 * [287]9815 (c++ library) atomicity.h - fails to compile with -O3 21467 -masm=intel 21468 * [288]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 21469 in reload1.c 21470 * [289]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 21471 * [290]10673 ICE for x86-64 on freebsd libc vfprintf.c source 21472 * [291]11044 [x86] out of range loop instructions for FP code on K6 21473 * [292]11089 ICE: instantiate_virtual_regs_lossage while using SSE 21474 built-ins 21475 * [293]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 21476 is used 21477 21478 SPARC- or Solaris- specific 21479 21480 * [294]9362 solaris 'as' dies when fed .s and "-gstabs" 21481 * [295]10142 [SPARC64] gcc produces wrong code when passing 21482 structures by value 21483 * [296]10663 New configure check aborts with Sun tools. 21484 * [297]10835 combinatorial explosion in scheduler on HyperSPARC 21485 * [298]10876 ICE in calculate_giv_inc when building KDE 21486 * [299]10955 wrong code at -O3 for structure argument in context of 21487 structure return 21488 * [300]11018 -mcpu=ultrasparc busts tar-1.13.25 21489 * [301]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 21490 Linux kernel 21491 21492 ia64 specific 21493 21494 * [302]10907 gcc violates the ia64 ABI (GP must be preserved) 21495 * [303]11320 scheduler bug (in machine depended reorganization pass) 21496 * [304]11599 bug with conditional and __builtin_prefetch 21497 21498 PowerPC specific 21499 21500 * [305]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 21501 during loop) 21502 * [306]10871 error in rs6000_stack_info save_size computation 21503 * [307]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 21504 cures it 21505 21506 m68k-specific 21507 21508 * [308]7594 [m68k] ICE on legal code associated with simplify-rtx 21509 * [309]10557 [m68k] ICE in subreg_offset_representable_p 21510 * [310]11054 [m68k] ICE in reg_overlap_mentioned_p 21511 21512 ARM-specific 21513 21514 * [311]10834 [arm] GCC 3.3 still generates incorrect instructions for 21515 functions with __attribute__ ((interrupt ("IRQ"))) 21516 * [312]10842 [arm] Clobbered link register is copied to pc under 21517 certain circumstances 21518 * [313]11052 [arm] noce_process_if_block() can lose REG_INC notes 21519 * [314]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 21520 (3.4) 21521 21522 MIPS-specific 21523 21524 * [315]11084 ICE in propagate_one_insn, in flow.c 21525 21526 SH-specific 21527 21528 * [316]10331 can't compile c++ part of gcc cross compiler for sh-elf 21529 * [317]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 21530 * [318]11096 i686-linux to sh-linux cross compiler fails to compile 21531 C++ files 21532 21533 GNU/Linux (or Hurd?) specific 21534 21535 * [319]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 21536 21537 UnixWare specific 21538 21539 * [320]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 21540 7.1.1 21541 21542 Cygwin (or mingw) specific 21543 21544 * [321]5287 ICE with dllimport attribute 21545 * [322]10148 [MingW/CygWin] Compiler dumps core 21546 21547 DJGPP specific 21548 21549 * [323]8787 GCC fails to emit .intel_syntax when invoked with 21550 -masm=intel on DJGPP 21551 21552 Darwin (and MacOS X) specific 21553 21554 * [324]10900 trampolines crash 21555 21556 Documentation 21557 21558 * [325]1607 (c++) Format attributes on methods undocumented 21559 * [326]4252 Invalid option `-fdump-translation-unit' 21560 * [327]4490 Clarify restrictions on -m96bit-long-double, 21561 -m128bit-long-double 21562 * [328]10355 document an issue with regparm attribute on some systems 21563 (e.g. Solaris) 21564 * [329]10726 (fortran) Documentation for function "IDate Intrinsic 21565 (Unix)" is wrong 21566 * [330]10805 document bug in old version of Sun assembler 21567 * [331]10815 warn against GNU binutils on AIX 21568 * [332]10877 document need for newer binutils on i?86-*-linux-gnu 21569 * [333]11280 Manual incorrect with respect to -freorder-blocks 21570 * [334]11466 Document -mlittle-endian and its restrictions for the 21571 sparc64 port 21572 21573 Testsuite bugs (compiler itself is not affected) 21574 21575 * [335]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 21576 report failure 21577 * [336]10810 gcc-3.3 fails make check: buffer overrun in 21578 test_demangle.c 21579 __________________________________________________________________ 21580 21581 [337]GCC 3.3.2 21582 21583 Bug Fixes 21584 21585 This section lists the problem reports (PRs) from GCC's bug tracker 21586 that are known to be fixed in the 3.3.2 release. This list might not be 21587 complete (that is, it is possible that some PRs that have been fixed 21588 are not listed here). 21589 21590 Bootstrap failures and problems 21591 21592 * [338]8336 [SCO5] bootstrap config still tries to use COFF options 21593 * [339]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 21594 --enable-threads=posix 21595 * [340]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 21596 * [341]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 21597 7.1.1) 21598 * [342]11687 xstormy16-elf build fails in libf2c 21599 * [343]12263 [SGI IRIX] bootstrap fails during compile of 21600 libf2c/libI77/backspace.c 21601 * [344]12490 buffer overflow in scan-decls.c (during Solaris 9 21602 fix-header processing) 21603 21604 Internal compiler errors (multi-platform) 21605 21606 * [345]7277 Casting integers to vector types causes ICE 21607 * [346]7939 (c++) ICE on invalid function template specialization 21608 * [347]11063 (c++) ICE on parsing initialization list of const array 21609 member 21610 * [348]11207 ICE with negative index in array element designator 21611 * [349]11522 (fortran) g77 dwarf-2 ICE in 21612 add_abstract_origin_attribute 21613 * [350]11595 (c++) ICE on duplicate label definition 21614 * [351]11646 (c++) ICE in commit_one_edge_insertion with 21615 -fnon-call-exceptions -fgcse -O 21616 * [352]11665 ICE in struct initializer when taking address 21617 * [353]11852 (c++) ICE with bad struct initializer. 21618 * [354]11878 (c++) ICE in cp_expr_size 21619 * [355]11883 ICE with any -O on mercury-generated C code 21620 * [356]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 21621 cp/typeck2.c when applying typeid operator to template template 21622 parameter 21623 * [357]12146 ICE in lookup_template_function, in cp/pt.c 21624 * [358]12215 ICE in make_label_edge with -fnon-call-exceptions 21625 -fno-gcse -O2 21626 * [359]12369 (c++) ICE with templates and friends 21627 * [360]12446 ICE in emit_move_insn on complicated array reference 21628 * [361]12510 ICE in final_scan_insn 21629 * [362]12544 ICE with large parameters used in nested functions 21630 21631 C and optimization bugs 21632 21633 * [363]9862 spurious warnings with -W -finline-functions 21634 * [364]10962 lookup_field is a linear search on a linked list (can be 21635 slow if large struct) 21636 * [365]11370 -Wunreachable-code gives false complaints 21637 * [366]11637 invalid assembly with -fnon-call-exceptions 21638 * [367]11885 Problem with bitfields in packed structs 21639 * [368]12082 Inappropriate unreachable code warnings 21640 * [369]12180 Inline optimization fails for variadic function 21641 * [370]12340 loop unroller + gcse produces wrong code 21642 21643 C++ compiler and library 21644 21645 * [371]3907 nested template parameter collides with member name 21646 * [372]5293 confusing message when binding a temporary to a reference 21647 * [373]5296 [DR115] Pointers to functions and to template functions 21648 behave differently in deduction 21649 * [374]7939 ICE on function template specialization 21650 * [375]8656 Unable to assign function with __attribute__ and pointer 21651 return type to an appropriate variable 21652 * [376]10147 Confusing error message for invalid template function 21653 argument 21654 * [377]11400 std::search_n() makes assumptions about Size parameter 21655 * [378]11409 issues with using declarations, overloading, and 21656 built-in functions 21657 * [379]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 21658 multiple bits in mask 21659 * [380]11786 operator() call on variable in other namespace not 21660 recognized 21661 * [381]11867 static_cast ignores ambiguity 21662 * [382]11928 bug with conversion operators that are typedefs 21663 * [383]12114 Uninitialized memory accessed in dtor 21664 * [384]12163 static_cast + explicit constructor regression 21665 * [385]12181 Wrong code with comma operator and c++ 21666 * [386]12236 regparm and fastcall messes up parameters 21667 * [387]12266 incorrect instantiation of unneeded template during 21668 overload resolution 21669 * [388]12296 istream::peek() doesn't set eofbit 21670 * [389]12298 [sjlj exceptions] Stack unwind destroys 21671 not-yet-constructed object 21672 * [390]12369 ICE with templates and friends 21673 * [391]12337 apparently infinite loop in g++ 21674 * [392]12344 stdcall attribute ignored if function returns a pointer 21675 * [393]12451 missing(late) class forward declaration in cxxabi.h 21676 * [394]12486 g++ accepts invalid use of a qualified name 21677 21678 x86 specific (Intel/AMD) 21679 21680 * [395]8869 [x86 MMX] ICE with const variable optimization and MMX 21681 builtins 21682 * [396]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 21683 -O2 21684 * [397]11689 g++3.3 emits un-assembleable code for k6 architecture 21685 * [398]12116 [k6] Invalid assembly output values with X-MAME code 21686 * [399]12070 ICE converting between double and long double with 21687 -msoft-float 21688 21689 ia64-specific 21690 21691 * [400]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 21692 * [401]11535 __builtin_return_address may not work on ia64 21693 * [402]11693 [ia64] ICE in gen_nop_type 21694 * [403]12224 [ia64] Thread-local storage doesn't work 21695 21696 PowerPC-specific 21697 21698 * [404]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 21699 kernel 21700 * [405]11319 loop miscompiled on ppc32 21701 * [406]11949 ICE Compiler segfault with ffmpeg -maltivec code 21702 21703 SPARC-specific 21704 21705 * [407]11662 wrong code for expr. with cast to long long and 21706 exclusive or 21707 * [408]11965 invalid assembler code for a shift < 32 operation 21708 * [409]12301 (c++) stack corruption when a returned expression throws 21709 an exception 21710 21711 Alpha-specific 21712 21713 * [410]11717 [alpha-linux] unrecognizable insn compiling for.c of 21714 kernel 2.4.22-pre8 21715 21716 HPUX-specific 21717 21718 * [411]11313 problem with #pragma weak and static inline functions 21719 * [412]11712 __STDC_EXT__ not defined for C++ by default anymore? 21720 21721 Solaris specific 21722 21723 * [413]12166 Profiled programs crash if PROFDIR is set 21724 21725 Solaris-x86 specific 21726 21727 * [414]12101 i386 Solaris no longer works with GNU as? 21728 21729 Miscellaneous embedded target-specific bugs 21730 21731 * [415]10988 [m32r-elf] wrong blockmove code with -O3 21732 * [416]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 21733 -O2 21734 * [417]11902 [sh4] spec file improperly inserts rpath even when none 21735 needed 21736 * [418]11903 [sh4] -pthread fails to link due to error in spec file 21737 on sh4 21738 __________________________________________________________________ 21739 21740 [419]GCC 3.3.3 21741 21742 Minor features 21743 21744 In addition to the bug fixes documented below, this release contains 21745 few minor features such as: 21746 * Support for --with-sysroot 21747 * Support for automatic detection of executable stacks 21748 * Support for SSE3 instructions 21749 * Support for thread local storage debugging under GDB on S390 21750 21751 Bug Fixes 21752 21753 This section lists the problem reports (PRs) from GCC's bug tracker 21754 that are known to be fixed in the 3.3.3 release. This list might not be 21755 complete (that is, it is possible that some PRs that have been fixed 21756 are not listed here). 21757 21758 Bootstrap failures and issues 21759 21760 * [420]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 21761 * [421]12399 boehm-gc fails (when building a cross compiler): libtool 21762 unable to infer tagged configuration 21763 * [422]13068 mklibgcc.in doesn't handle multi-level multilib 21764 subdirectories properly 21765 21766 Internal compiler errors (multi-platform) 21767 21768 * [423]10060 ICE (stack overflow) on huge file (300k lines) due to 21769 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 21770 * [424]10555 (c++) ICE on undefined template argument 21771 * [425]10706 (c++) ICE in mangle_class_name_for_template 21772 * [426]11496 (fortran) error in flow_loops_find when -funroll-loops 21773 active 21774 * [427]11741 ICE in pre_insert_copy_insn, in gcse.c 21775 * [428]12440 GCC crashes during compilation of quicktime4linux 2.0.0 21776 * [429]12632 (fortran) -fbounds-check ICE 21777 * [430]12712 (c++) ICE on short legit C++ code fragment with gcc 21778 3.3.2 21779 * [431]12726 (c++) ICE (segfault) on trivial code 21780 * [432]12890 (c++) ICE on compilation of class with throwing method 21781 * [433]12900 (c++) ICE in rtl_verify_flow_info_1 21782 * [434]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 21783 correct code with -O2 -fno-force-mem 21784 * [435]13289 (c++) ICE in regenerate_decl_from_template on recursive 21785 template 21786 * [436]13318 ICE: floating point exception in the loop optimizer 21787 * [437]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 21788 except.c 21789 * [438]13574 (c++) invalid array default initializer in class lets 21790 gcc consume all memory and die 21791 * [439]13475 ICE on SIMD variables with partial value initialization 21792 * [440]13797 (c++) ICE on invalid template parameter 21793 * [441]13824 (java) gcj SEGV with simple .java program 21794 21795 C and optimization bugs 21796 21797 * [442]8776 loop invariants are not removed (most likely) 21798 * [443]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 21799 strncmp by memcmp 21800 * [444]11350 undefined labels with -Os -fPIC 21801 * [445]12826 Optimizer removes reference through volatile pointer 21802 * [446]12500 stabs debug info: void no longer a predefined / builtin 21803 type 21804 * [447]12941 builtin-bitops-1.c miscompilation (latent bug) 21805 * [448]12953 tree inliner bug (in inline_forbidden_p) and fix 21806 * [449]13041 linux-2.6/sound/core/oss/rate.c miscompiled 21807 * [450]13507 spurious printf format warning 21808 * [451]13382 Type information for const pointer disappears during 21809 optimization. 21810 * [452]13394 noreturn attribute ignored on recursive invokation 21811 * [453]13400 Compiled code crashes storing to read-only location 21812 * [454]13521 Endless loop in calculate_global_regs_live 21813 21814 C++ compiler and library 21815 21816 Some of the bug fixes in this list were made to implement decisions 21817 that the ISO C++ standards committee has made concerning several defect 21818 reports (DRs). Links in the list below point to detailed discussion of 21819 the relevant defect report. 21820 * [455]2094 unimplemented: use of `ptrmem_cst' in template type 21821 unification 21822 * [456]2294 using declaration confusion 21823 * [457]5050 template instantiation depth exceeds limit: recursion 21824 problem? 21825 * [458]9371 Bad exception handling in 21826 i/ostream::operator>>/<<(streambuf*) 21827 * [459]9546 bad exception handling in ostream members 21828 * [460]10081 basic_ios::_M_cache_locale leaves NULL members in the 21829 face of unknown locales 21830 * [461]10093 [462][DR 61] Setting failbit in exceptions doesn't work 21831 * [463]10095 istream::operator>>(int&) sets ios::badbit when 21832 ios::failbit is set. 21833 * [464]11554 Warning about reordering of initializers doesn't mention 21834 location of constructor 21835 * [465]12297 istream::sentry::sentry() handles eof() incorrectly. 21836 * [466]12352 Exception safety problems in src/localename.cc 21837 * [467]12438 Memory leak in locale::combine() 21838 * [468]12540 Memory leak in locale::locale(const char*) 21839 * [469]12594 DRs [470]60 [TC] and [471]63 [TC] not implemented 21840 * [472]12657 Resolution of [473]DR 292 (WP) still unimplemented 21841 * [474]12696 memory eating infinite loop in diagnostics (error 21842 recovery problem) 21843 * [475]12815 Code compiled with optimization behaves unexpectedly 21844 * [476]12862 Conflicts between typedefs/enums and namespace member 21845 declarations 21846 * [477]12926 Wrong value after assignment in initialize list using 21847 bit-fields 21848 * [478]12967 Resolution of [479]DR 300 [WP] still unimplemented 21849 * [480]12971 Resolution of [481]DR 328 [WP] still unimplemented 21850 * [482]13007 basic_streambuf::pubimbue, imbue wrong 21851 * [483]13009 Implicitly-defined assignment operator writes to wrong 21852 memory 21853 * [484]13057 regparm attribute not applied to destructor 21854 * [485]13070 -Wformat option ignored in g++ 21855 * [486]13081 forward template declarations in <complex> let inlining 21856 fail 21857 * [487]13239 Assertion does not seem to work correctly anymore 21858 * [488]13262 "xxx is private within this context" when initializing a 21859 self-contained template class 21860 * [489]13290 simple typo in concept checking for std::generate_n 21861 * [490]13323 Template code does not compile in presence of typedef 21862 * [491]13369 __verify_grouping (and __add_grouping?) not correct 21863 * [492]13371 infinite loop with packed struct and inlining 21864 * [493]13445 Template argument replacement "dereferences" a typedef 21865 * [494]13461 Fails to access protected-ctor from public constant 21866 * [495]13462 Non-standard-conforming type set::pointer 21867 * [496]13478 gcc uses wrong constructor to initialize a const 21868 reference 21869 * [497]13544 "conflicting types" for enums in different scopes 21870 * [498]13650 string::compare should not (always) use 21871 traits_type::length() 21872 * [499]13683 bogus warning about passing non-PODs through ellipsis 21873 * [500]13688 Derived class is denied access to protected base class 21874 member class 21875 * [501]13774 Member variable cleared in virtual multiple inheritance 21876 class 21877 * [502]13884 Protect sstream.tcc from extern template use 21878 21879 Java compiler and library 21880 21881 * [503]10746 [win32] garbage collection crash in GCJ 21882 21883 Objective-C compiler and library 21884 21885 * [504]11433 Crash due to dereferencing null pointer when querying 21886 protocol 21887 21888 Fortran compiler and library 21889 21890 * [505]12633 logical expression gives incorrect result with 21891 -fugly-logint option 21892 * [506]13037 [gcse-lm] g77 generates incorrect code 21893 * [507]13213 Hex constant problem when compiling with -fugly-logint 21894 and -ftypeless-boz 21895 21896 x86-specific (Intel/AMD) 21897 21898 * [508]4490 ICE with -m128bit-long-double 21899 * [509]12292 [x86_64] ICE: RTL check: expected code `const_int', have 21900 `reg' in make_field_assignment, in combine.c 21901 * [510]12441 ICE: can't find a register to spill 21902 * [511]12943 array static-init failure under -fpic, -fPIC 21903 * [512]13608 Incorrect code with -O3 -ffast-math 21904 21905 PowerPC-specific 21906 21907 * [513]11598 testcase gcc.dg/20020118-1.c fails runtime check of 21908 __attribute__((aligned(16))) 21909 * [514]11793 ICE in extract_insn, in recog.c (const_vector's) 21910 * [515]12467 vmsumubm emitted when vmsummbm appropriate (typo in 21911 altivec.md) 21912 * [516]12537 g++ generates writeable text sections 21913 21914 SPARC-specific 21915 21916 * [517]12496 wrong result for __atomic_add(&value, -1) when using -O0 21917 -m64 21918 * [518]12865 mprotect call to make trampoline executable may fail 21919 * [519]13354 ICE in sparc_emit_set_const32 21920 21921 ARM-specific 21922 21923 * [520]10467 [arm] ICE in pre_insert_copy_insn, 21924 21925 ia64-specific 21926 21927 * [521]11226 ICE passing struct arg with two floats 21928 * [522]11227 ICE for _Complex float, _Complex long double args 21929 * [523]12644 GCC 3.3.2 fails to compile glibc on ia64 21930 * [524]13149 build gcc-3.3.2 1305 error:unrecognizable insn 21931 * Various fixes for libunwind 21932 21933 Alpha-specific 21934 21935 * [525]12654 Incorrect comparison code generated for Alpha 21936 * [526]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 21937 * [527]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 21938 21939 HPPA-specific 21940 21941 * [528]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 21942 * [529]12158 [hppa] compilation does not terminate at -O1 21943 21944 S390-specific 21945 21946 * [530]11992 Wrong built-in code for memcmp with length 1<<24: only 21947 (1<<24)-1 possible for CLCL-Instruction 21948 21949 SH-specific 21950 21951 * [531]9365 segfault in gen_far_branch (config/sh/sh.c) 21952 * [532]10392 optimizer generates faulty array indexing 21953 * [533]11322 SH profiler outputs multiple definitions of symbol 21954 * [534]13069 gcc/config/sh/rtems.h broken 21955 * [535]13302 Putting a va_list in a struct causes seg fault 21956 * [536]13585 Incorrect optimization of call to sfunc 21957 * Fix inappropriately exported libgcc functions from the shared 21958 library 21959 21960 Other embedded target specific 21961 21962 * [537]8916 [mcore] unsigned char assign gets hosed. 21963 * [538]11576 [h8300] ICE in change_address_1, in emit-rtl.c 21964 * [539]13122 [h8300] local variable gets corrupted by function call 21965 when -fomit-frame-pointer is given 21966 * [540]13256 [cris] strict_low_part mistreated in delay slots 21967 * [541]13373 [mcore] optimization with -frerun-cse-after-loop 21968 -fexpensive-optimizations produces wrong code on mcore 21969 21970 GNU HURD-specific 21971 21972 * [542]12561 gcc/config/t-gnu needs updating to work with 21973 --with-sysroot 21974 21975 Tru64 Unix specific 21976 21977 * [543]6243 testsuite fails almost all tests due to no libintl in 21978 LD_LIBRARY_PATH during test. 21979 * [544]11397 weak aliases broken on Tru64 UNIX 21980 21981 AIX-specific 21982 21983 * [545]12505 build failure due to defines of uchar in cpphash.h and 21984 sys/types.h 21985 * [546]13150 WEAK symbols not exported by collect2 21986 21987 IRIX-specific 21988 21989 * [547]12666 fixincludes problem on IRIX 6.5.19m 21990 21991 Solaris-specific 21992 21993 * [548]12969 Including sys/byteorder.h breaks configure checks 21994 21995 Testsuite problems (compiler is not affected) 21996 21997 * [549]10819 testsuite creates CR+LF on compiler version lines in 21998 test summary files 21999 * [550]11612 abi_check not finding correct libgcc_s.so.1 22000 22001 Miscellaneous 22002 22003 * [551]13211 using -###, incorrect warnings about unused linker file 22004 are produced 22005 __________________________________________________________________ 22006 22007 [552]GCC 3.3.4 22008 22009 This is the [553]list of problem reports (PRs) from GCC's bug tracking 22010 system that are known to be fixed in the 3.3.4 release. This list might 22011 not be complete (that is, it is possible that some PRs that have been 22012 fixed are not listed here). 22013 __________________________________________________________________ 22014 22015 [554]GCC 3.3.5 22016 22017 This is the [555]list of problem reports (PRs) from GCC's bug tracking 22018 system that are known to be fixed in the 3.3.5 release. This list might 22019 not be complete (that is, it is possible that some PRs that have been 22020 fixed are not listed here). 22021 __________________________________________________________________ 22022 22023 [556]GCC 3.3.6 22024 22025 This is the [557]list of problem reports (PRs) from GCC's bug tracking 22026 system that are known to be fixed in the 3.3.6 release. This list might 22027 not be complete (that is, it is possible that some PRs that have been 22028 fixed are not listed here). 22029 22030 22031 For questions related to the use of GCC, please consult these web 22032 pages and the [558]GCC manuals. If that fails, the 22033 [559]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22034 web pages and the development of GCC are welcome on our developer 22035 list at [560]gcc (a] gcc.gnu.org. All of [561]our lists have public 22036 archives. 22037 22038 Copyright (C) [562]Free Software Foundation, Inc. Verbatim copying and 22039 distribution of this entire article is permitted in any medium, 22040 provided this notice is preserved. 22041 22042 These pages are [563]maintained by the GCC team. Last modified 22043 2025-06-05. 22044 22045 References 22046 22047 1. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 22048 2. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 22049 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 22050 4. https://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 22051 5. https://gcc.gnu.org/news/dfa.html 22052 6. https://gcc.gnu.org/projects/c-status.html 22053 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 22054 8. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 22055 9. https://gcc.gnu.org/PR10140 22056 10. https://gcc.gnu.org/PR10198 22057 11. https://gcc.gnu.org/PR10338 22058 12. https://gcc.gnu.org/PR3581 22059 13. https://gcc.gnu.org/PR4382 22060 14. https://gcc.gnu.org/PR5533 22061 15. https://gcc.gnu.org/PR6387 22062 16. https://gcc.gnu.org/PR6412 22063 17. https://gcc.gnu.org/PR6620 22064 18. https://gcc.gnu.org/PR6663 22065 19. https://gcc.gnu.org/PR7068 22066 20. https://gcc.gnu.org/PR7083 22067 21. https://gcc.gnu.org/PR7647 22068 22. https://gcc.gnu.org/PR7675 22069 23. https://gcc.gnu.org/PR7718 22070 24. https://gcc.gnu.org/PR8116 22071 25. https://gcc.gnu.org/PR8358 22072 26. https://gcc.gnu.org/PR8511 22073 27. https://gcc.gnu.org/PR8564 22074 28. https://gcc.gnu.org/PR8660 22075 29. https://gcc.gnu.org/PR8766 22076 30. https://gcc.gnu.org/PR8803 22077 31. https://gcc.gnu.org/PR8846 22078 32. https://gcc.gnu.org/PR8906 22079 33. https://gcc.gnu.org/PR9216 22080 34. https://gcc.gnu.org/PR9261 22081 35. https://gcc.gnu.org/PR9263 22082 36. https://gcc.gnu.org/PR9429 22083 37. https://gcc.gnu.org/PR9516 22084 38. https://gcc.gnu.org/PR9600 22085 39. https://gcc.gnu.org/PR9629 22086 40. https://gcc.gnu.org/PR9672 22087 41. https://gcc.gnu.org/PR9749 22088 42. https://gcc.gnu.org/PR9794 22089 43. https://gcc.gnu.org/PR9829 22090 44. https://gcc.gnu.org/PR9916 22091 45. https://gcc.gnu.org/PR9936 22092 46. https://gcc.gnu.org/PR10262 22093 47. https://gcc.gnu.org/PR10278 22094 48. https://gcc.gnu.org/PR10446 22095 49. https://gcc.gnu.org/PR10451 22096 50. https://gcc.gnu.org/PR10506 22097 51. https://gcc.gnu.org/PR10549 22098 52. https://gcc.gnu.org/PR2001 22099 53. https://gcc.gnu.org/PR2391 22100 54. https://gcc.gnu.org/PR2960 22101 55. https://gcc.gnu.org/PR4046 22102 56. https://gcc.gnu.org/PR6405 22103 57. https://gcc.gnu.org/PR6798 22104 58. https://gcc.gnu.org/PR6871 22105 59. https://gcc.gnu.org/PR6909 22106 60. https://gcc.gnu.org/PR7189 22107 61. https://gcc.gnu.org/PR7642 22108 62. https://gcc.gnu.org/PR8634 22109 63. https://gcc.gnu.org/PR8750 22110 64. https://gcc.gnu.org/PR2161 22111 65. https://gcc.gnu.org/PR4319 22112 66. https://gcc.gnu.org/PR8602 22113 67. https://gcc.gnu.org/PR9177 22114 68. https://gcc.gnu.org/PR9853 22115 69. https://gcc.gnu.org/PR45 22116 70. https://gcc.gnu.org/PR3784 22117 71. https://gcc.gnu.org/PR764 22118 72. https://gcc.gnu.org/PR5116 22119 73. https://gcc.gnu.org/PR2862 22120 74. https://gcc.gnu.org/PR3663 22121 75. https://gcc.gnu.org/PR3797 22122 76. https://gcc.gnu.org/PR3948 22123 77. https://gcc.gnu.org/PR4137 22124 78. https://gcc.gnu.org/PR4361 22125 79. https://gcc.gnu.org/PR4802 22126 80. https://gcc.gnu.org/PR5837 22127 81. https://gcc.gnu.org/PR4803 22128 82. https://gcc.gnu.org/PR5094 22129 83. https://gcc.gnu.org/PR5730 22130 84. https://gcc.gnu.org/PR6713 22131 85. https://gcc.gnu.org/PR7015 22132 86. https://gcc.gnu.org/PR7086 22133 87. https://gcc.gnu.org/PR7099 22134 88. https://gcc.gnu.org/PR7247 22135 89. https://gcc.gnu.org/PR7441 22136 90. https://gcc.gnu.org/PR7768 22137 91. https://gcc.gnu.org/PR7804 22138 92. https://gcc.gnu.org/PR8099 22139 93. https://gcc.gnu.org/PR8117 22140 94. https://gcc.gnu.org/PR8205 22141 95. https://gcc.gnu.org/PR8645 22142 96. https://gcc.gnu.org/PR8724 22143 97. https://gcc.gnu.org/PR8805 22144 98. https://gcc.gnu.org/PR8691 22145 99. https://gcc.gnu.org/PR8700 22146 100. https://gcc.gnu.org/PR8724 22147 101. https://gcc.gnu.org/PR8949 22148 102. https://gcc.gnu.org/PR9016 22149 103. https://gcc.gnu.org/PR9053 22150 104. https://gcc.gnu.org/PR9152 22151 105. https://gcc.gnu.org/PR9182 22152 106. https://gcc.gnu.org/PR9297 22153 107. https://gcc.gnu.org/PR9318 22154 108. https://gcc.gnu.org/PR9320 22155 109. https://gcc.gnu.org/PR9400 22156 110. https://gcc.gnu.org/PR9424 22157 111. https://gcc.gnu.org/PR9425 22158 112. https://gcc.gnu.org/PR9439 22159 113. https://gcc.gnu.org/PR9474 22160 114. https://gcc.gnu.org/PR9548 22161 115. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 22162 116. https://gcc.gnu.org/PR9555 22163 117. https://gcc.gnu.org/PR9561 22164 118. https://gcc.gnu.org/PR9563 22165 119. https://gcc.gnu.org/PR9582 22166 120. https://gcc.gnu.org/PR9622 22167 121. https://gcc.gnu.org/PR9683 22168 122. https://gcc.gnu.org/PR9791 22169 123. https://gcc.gnu.org/PR9817 22170 124. https://gcc.gnu.org/PR9825 22171 125. https://gcc.gnu.org/PR9826 22172 126. https://gcc.gnu.org/PR9924 22173 127. https://gcc.gnu.org/PR9946 22174 128. https://gcc.gnu.org/PR9964 22175 129. https://gcc.gnu.org/PR9988 22176 130. https://gcc.gnu.org/PR10033 22177 131. https://gcc.gnu.org/PR10097 22178 132. https://gcc.gnu.org/PR10132 22179 133. https://gcc.gnu.org/PR10180 22180 134. https://gcc.gnu.org/PR10199 22181 135. https://gcc.gnu.org/PR10300 22182 136. https://gcc.gnu.org/PR10427 22183 137. https://gcc.gnu.org/PR10503 22184 138. https://gcc.gnu.org/PR5956 22185 139. https://gcc.gnu.org/PR1832 22186 140. https://gcc.gnu.org/PR3924 22187 141. https://gcc.gnu.org/PR5634 22188 142. https://gcc.gnu.org/PR6367 22189 143. https://gcc.gnu.org/PR6491 22190 144. https://gcc.gnu.org/PR6742 22191 145. https://gcc.gnu.org/PR7113 22192 146. https://gcc.gnu.org/PR7236 22193 147. https://gcc.gnu.org/PR7278 22194 148. https://gcc.gnu.org/PR7384 22195 149. https://gcc.gnu.org/PR7388 22196 150. https://gcc.gnu.org/PR8587 22197 151. https://gcc.gnu.org/PR9038 22198 152. https://gcc.gnu.org/PR10197 22199 153. https://gcc.gnu.org/PR6005 22200 154. https://gcc.gnu.org/PR6389 22201 155. https://gcc.gnu.org/PR6576 22202 156. https://gcc.gnu.org/PR6652 22203 157. https://gcc.gnu.org/PR7060 22204 158. https://gcc.gnu.org/PR7073 22205 159. https://gcc.gnu.org/PR7180 22206 160. https://gcc.gnu.org/PR7416 22207 161. https://gcc.gnu.org/PR7570 22208 162. https://gcc.gnu.org/PR7578 22209 163. https://gcc.gnu.org/PR7611 22210 164. https://gcc.gnu.org/PR7709 22211 165. https://gcc.gnu.org/PR7766 22212 166. https://gcc.gnu.org/PR7785 22213 167. https://gcc.gnu.org/PR7786 22214 168. https://gcc.gnu.org/PR8142 22215 169. https://gcc.gnu.org/PR8234 22216 170. https://gcc.gnu.org/PR8415 22217 171. https://gcc.gnu.org/PR8481 22218 172. https://gcc.gnu.org/PR8593 22219 173. https://gcc.gnu.org/PR8759 22220 174. https://gcc.gnu.org/PR8997 22221 175. https://gcc.gnu.org/PR9253 22222 176. https://gcc.gnu.org/PR9254 22223 177. https://gcc.gnu.org/PR9271 22224 178. https://gcc.gnu.org/PR6767 22225 179. https://gcc.gnu.org/PR9911 22226 180. https://gcc.gnu.org/PR10020 22227 181. https://gcc.gnu.org/PR10546 22228 182. https://gcc.gnu.org/PR7029 22229 183. https://gcc.gnu.org/PR2903 22230 184. https://gcc.gnu.org/PR7873 22231 185. https://gcc.gnu.org/PR7680 22232 186. https://gcc.gnu.org/PR8705 22233 187. https://gcc.gnu.org/PR9986 22234 188. https://gcc.gnu.org/PR10056 22235 189. https://gcc.gnu.org/PR6744 22236 190. https://gcc.gnu.org/PR7361 22237 191. https://gcc.gnu.org/PR9496 22238 192. https://gcc.gnu.org/PR7067 22239 193. https://gcc.gnu.org/PR8480 22240 194. https://gcc.gnu.org/PR8784 22241 195. https://gcc.gnu.org/PR10315 22242 196. https://gcc.gnu.org/PR10267 22243 197. https://gcc.gnu.org/PR7916 22244 198. https://gcc.gnu.org/PR7926 22245 199. https://gcc.gnu.org/PR8555 22246 200. https://gcc.gnu.org/PR8994 22247 201. https://gcc.gnu.org/PR9426 22248 202. https://gcc.gnu.org/PR9806 22249 203. https://gcc.gnu.org/PR10077 22250 204. https://gcc.gnu.org/PR10233 22251 205. https://gcc.gnu.org/PR10286 22252 206. https://gcc.gnu.org/PR10308 22253 207. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.1 22254 208. https://gcc.gnu.org/PR11272 22255 209. https://gcc.gnu.org/PR5754 22256 210. https://gcc.gnu.org/PR6597 22257 211. https://gcc.gnu.org/PR6949 22258 212. https://gcc.gnu.org/PR7053 22259 213. https://gcc.gnu.org/PR8164 22260 214. https://gcc.gnu.org/PR8384 22261 215. https://gcc.gnu.org/PR9559 22262 216. https://gcc.gnu.org/PR9649 22263 217. https://gcc.gnu.org/PR9864 22264 218. https://gcc.gnu.org/PR10432 22265 219. https://gcc.gnu.org/PR10475 22266 220. https://gcc.gnu.org/PR10635 22267 221. https://gcc.gnu.org/PR10661 22268 222. https://gcc.gnu.org/PR10700 22269 223. https://gcc.gnu.org/PR10712 22270 224. https://gcc.gnu.org/PR10796 22271 225. https://gcc.gnu.org/PR10890 22272 226. https://gcc.gnu.org/PR10939 22273 227. https://gcc.gnu.org/PR10956 22274 228. https://gcc.gnu.org/PR11041 22275 229. https://gcc.gnu.org/PR11059 22276 230. https://gcc.gnu.org/PR11083 22277 231. https://gcc.gnu.org/PR11105 22278 232. https://gcc.gnu.org/PR11149 22279 233. https://gcc.gnu.org/PR11228 22280 234. https://gcc.gnu.org/PR11282 22281 235. https://gcc.gnu.org/PR11301 22282 236. https://gcc.gnu.org/PR11308 22283 237. https://gcc.gnu.org/PR11473 22284 238. https://gcc.gnu.org/PR11503 22285 239. https://gcc.gnu.org/PR11513 22286 240. https://gcc.gnu.org/PR11198 22287 241. https://gcc.gnu.org/PR11304 22288 242. https://gcc.gnu.org/PR11381 22289 243. https://gcc.gnu.org/PR11536 22290 244. https://gcc.gnu.org/PR11557 22291 245. https://gcc.gnu.org/PR5897 22292 246. https://gcc.gnu.org/PR11279 22293 247. https://gcc.gnu.org/PR11022 22294 248. https://gcc.gnu.org/PR2330 22295 249. https://gcc.gnu.org/PR5388 22296 250. https://gcc.gnu.org/PR5390 22297 251. https://gcc.gnu.org/PR7877 22298 252. https://gcc.gnu.org/PR9393 22299 253. https://gcc.gnu.org/PR10032 22300 254. https://gcc.gnu.org/PR10468 22301 255. https://gcc.gnu.org/PR10527 22302 256. https://gcc.gnu.org/PR10679 22303 257. https://gcc.gnu.org/PR10682 22304 258. https://gcc.gnu.org/PR10689 22305 259. https://gcc.gnu.org/PR10845 22306 260. https://gcc.gnu.org/PR10849 22307 261. https://gcc.gnu.org/PR10888 22308 262. https://gcc.gnu.org/PR10929 22309 263. https://gcc.gnu.org/PR10931 22310 264. https://gcc.gnu.org/PR10940 22311 265. https://gcc.gnu.org/PR10968 22312 266. https://gcc.gnu.org/PR10990 22313 267. https://gcc.gnu.org/PR11039 22314 268. https://gcc.gnu.org/PR11062 22315 269. https://gcc.gnu.org/PR11095 22316 270. https://gcc.gnu.org/PR11098 22317 271. https://gcc.gnu.org/PR11137 22318 272. https://gcc.gnu.org/PR11154 22319 273. https://gcc.gnu.org/PR11329 22320 274. https://gcc.gnu.org/PR11332 22321 275. https://gcc.gnu.org/PR11431 22322 276. https://gcc.gnu.org/PR11528 22323 277. https://gcc.gnu.org/PR11546 22324 278. https://gcc.gnu.org/PR11567 22325 279. https://gcc.gnu.org/PR11645 22326 280. https://gcc.gnu.org/PR5179 22327 281. https://gcc.gnu.org/PR8204 22328 282. https://gcc.gnu.org/PR10838 22329 283. https://gcc.gnu.org/PR10886 22330 284. https://gcc.gnu.org/PR11349 22331 285. https://gcc.gnu.org/PR4823 22332 286. https://gcc.gnu.org/PR8878 22333 287. https://gcc.gnu.org/PR9815 22334 288. https://gcc.gnu.org/PR10402 22335 289. https://gcc.gnu.org/PR10504 22336 290. https://gcc.gnu.org/PR10673 22337 291. https://gcc.gnu.org/PR11044 22338 292. https://gcc.gnu.org/PR11089 22339 293. https://gcc.gnu.org/PR11420 22340 294. https://gcc.gnu.org/PR9362 22341 295. https://gcc.gnu.org/PR10142 22342 296. https://gcc.gnu.org/PR10663 22343 297. https://gcc.gnu.org/PR10835 22344 298. https://gcc.gnu.org/PR10876 22345 299. https://gcc.gnu.org/PR10955 22346 300. https://gcc.gnu.org/PR11018 22347 301. https://gcc.gnu.org/PR11556 22348 302. https://gcc.gnu.org/PR10907 22349 303. https://gcc.gnu.org/PR11320 22350 304. https://gcc.gnu.org/PR11599 22351 305. https://gcc.gnu.org/PR9745 22352 306. https://gcc.gnu.org/PR10871 22353 307. https://gcc.gnu.org/PR11440 22354 308. https://gcc.gnu.org/PR7594 22355 309. https://gcc.gnu.org/PR10557 22356 310. https://gcc.gnu.org/PR11054 22357 311. https://gcc.gnu.org/PR10834 22358 312. https://gcc.gnu.org/PR10842 22359 313. https://gcc.gnu.org/PR11052 22360 314. https://gcc.gnu.org/PR11183 22361 315. https://gcc.gnu.org/PR11084 22362 316. https://gcc.gnu.org/PR10331 22363 317. https://gcc.gnu.org/PR10413 22364 318. https://gcc.gnu.org/PR11096 22365 319. https://gcc.gnu.org/PR2873 22366 320. https://gcc.gnu.org/PR3163 22367 321. https://gcc.gnu.org/PR5287 22368 322. https://gcc.gnu.org/PR10148 22369 323. https://gcc.gnu.org/PR8787 22370 324. https://gcc.gnu.org/PR10900 22371 325. https://gcc.gnu.org/PR1607 22372 326. https://gcc.gnu.org/PR4252 22373 327. https://gcc.gnu.org/PR4490 22374 328. https://gcc.gnu.org/PR10355 22375 329. https://gcc.gnu.org/PR10726 22376 330. https://gcc.gnu.org/PR10805 22377 331. https://gcc.gnu.org/PR10815 22378 332. https://gcc.gnu.org/PR10877 22379 333. https://gcc.gnu.org/PR11280 22380 334. https://gcc.gnu.org/PR11466 22381 335. https://gcc.gnu.org/PR10737 22382 336. https://gcc.gnu.org/PR10810 22383 337. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.2 22384 338. https://gcc.gnu.org/PR8336 22385 339. https://gcc.gnu.org/PR9330 22386 340. https://gcc.gnu.org/PR9631 22387 341. https://gcc.gnu.org/PR9877 22388 342. https://gcc.gnu.org/PR11687 22389 343. https://gcc.gnu.org/PR12263 22390 344. https://gcc.gnu.org/PR12490 22391 345. https://gcc.gnu.org/PR7277 22392 346. https://gcc.gnu.org/PR7939 22393 347. https://gcc.gnu.org/PR11063 22394 348. https://gcc.gnu.org/PR11207 22395 349. https://gcc.gnu.org/PR11522 22396 350. https://gcc.gnu.org/PR11595 22397 351. https://gcc.gnu.org/PR11646 22398 352. https://gcc.gnu.org/PR11665 22399 353. https://gcc.gnu.org/PR11852 22400 354. https://gcc.gnu.org/PR11878 22401 355. https://gcc.gnu.org/PR11883 22402 356. https://gcc.gnu.org/PR11991 22403 357. https://gcc.gnu.org/PR12146 22404 358. https://gcc.gnu.org/PR12215 22405 359. https://gcc.gnu.org/PR12369 22406 360. https://gcc.gnu.org/PR12446 22407 361. https://gcc.gnu.org/PR12510 22408 362. https://gcc.gnu.org/PR12544 22409 363. https://gcc.gnu.org/PR9862 22410 364. https://gcc.gnu.org/PR10962 22411 365. https://gcc.gnu.org/PR11370 22412 366. https://gcc.gnu.org/PR11637 22413 367. https://gcc.gnu.org/PR11885 22414 368. https://gcc.gnu.org/PR12082 22415 369. https://gcc.gnu.org/PR12180 22416 370. https://gcc.gnu.org/PR12340 22417 371. https://gcc.gnu.org/PR3907 22418 372. https://gcc.gnu.org/PR5293 22419 373. https://gcc.gnu.org/PR5296 22420 374. https://gcc.gnu.org/PR7939 22421 375. https://gcc.gnu.org/PR8656 22422 376. https://gcc.gnu.org/PR10147 22423 377. https://gcc.gnu.org/PR11400 22424 378. https://gcc.gnu.org/PR11409 22425 379. https://gcc.gnu.org/PR11740 22426 380. https://gcc.gnu.org/PR11786 22427 381. https://gcc.gnu.org/PR11867 22428 382. https://gcc.gnu.org/PR11928 22429 383. https://gcc.gnu.org/PR12114 22430 384. https://gcc.gnu.org/PR12163 22431 385. https://gcc.gnu.org/PR12181 22432 386. https://gcc.gnu.org/PR12236 22433 387. https://gcc.gnu.org/PR12266 22434 388. https://gcc.gnu.org/PR12296 22435 389. https://gcc.gnu.org/PR12298 22436 390. https://gcc.gnu.org/PR12369 22437 391. https://gcc.gnu.org/PR12337 22438 392. https://gcc.gnu.org/PR12344 22439 393. https://gcc.gnu.org/PR12451 22440 394. https://gcc.gnu.org/PR12486 22441 395. https://gcc.gnu.org/PR8869 22442 396. https://gcc.gnu.org/PR9786 22443 397. https://gcc.gnu.org/PR11689 22444 398. https://gcc.gnu.org/PR12116 22445 399. https://gcc.gnu.org/PR12070 22446 400. https://gcc.gnu.org/PR11184 22447 401. https://gcc.gnu.org/PR11535 22448 402. https://gcc.gnu.org/PR11693 22449 403. https://gcc.gnu.org/PR12224 22450 404. https://gcc.gnu.org/PR11087 22451 405. https://gcc.gnu.org/PR11319 22452 406. https://gcc.gnu.org/PR11949 22453 407. https://gcc.gnu.org/PR11662 22454 408. https://gcc.gnu.org/PR11965 22455 409. https://gcc.gnu.org/PR12301 22456 410. https://gcc.gnu.org/PR11717 22457 411. https://gcc.gnu.org/PR11313 22458 412. https://gcc.gnu.org/PR11712 22459 413. https://gcc.gnu.org/PR12166 22460 414. https://gcc.gnu.org/PR12101 22461 415. https://gcc.gnu.org/PR10988 22462 416. https://gcc.gnu.org/PR11805 22463 417. https://gcc.gnu.org/PR11902 22464 418. https://gcc.gnu.org/PR11903 22465 419. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.3 22466 420. https://gcc.gnu.org/PR11890 22467 421. https://gcc.gnu.org/PR12399 22468 422. https://gcc.gnu.org/PR13068 22469 423. https://gcc.gnu.org/PR10060 22470 424. https://gcc.gnu.org/PR10555 22471 425. https://gcc.gnu.org/PR10706 22472 426. https://gcc.gnu.org/PR11496 22473 427. https://gcc.gnu.org/PR11741 22474 428. https://gcc.gnu.org/PR12440 22475 429. https://gcc.gnu.org/PR12632 22476 430. https://gcc.gnu.org/PR12712 22477 431. https://gcc.gnu.org/PR12726 22478 432. https://gcc.gnu.org/PR12890 22479 433. https://gcc.gnu.org/PR12900 22480 434. https://gcc.gnu.org/PR13060 22481 435. https://gcc.gnu.org/PR13289 22482 436. https://gcc.gnu.org/PR13318 22483 437. https://gcc.gnu.org/PR13392 22484 438. https://gcc.gnu.org/PR13574 22485 439. https://gcc.gnu.org/PR13475 22486 440. https://gcc.gnu.org/PR13797 22487 441. https://gcc.gnu.org/PR13824 22488 442. https://gcc.gnu.org/PR8776 22489 443. https://gcc.gnu.org/PR10339 22490 444. https://gcc.gnu.org/PR11350 22491 445. https://gcc.gnu.org/PR12826 22492 446. https://gcc.gnu.org/PR12500 22493 447. https://gcc.gnu.org/PR12941 22494 448. https://gcc.gnu.org/PR12953 22495 449. https://gcc.gnu.org/PR13041 22496 450. https://gcc.gnu.org/PR13507 22497 451. https://gcc.gnu.org/PR13382 22498 452. https://gcc.gnu.org/PR13394 22499 453. https://gcc.gnu.org/PR13400 22500 454. https://gcc.gnu.org/PR13521 22501 455. https://gcc.gnu.org/PR2094 22502 456. https://gcc.gnu.org/PR2294 22503 457. https://gcc.gnu.org/PR5050 22504 458. https://gcc.gnu.org/PR9371 22505 459. https://gcc.gnu.org/PR9546 22506 460. https://gcc.gnu.org/PR10081 22507 461. https://gcc.gnu.org/PR10093 22508 462. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 22509 463. https://gcc.gnu.org/PR10095 22510 464. https://gcc.gnu.org/PR11554 22511 465. https://gcc.gnu.org/PR12297 22512 466. https://gcc.gnu.org/PR12352 22513 467. https://gcc.gnu.org/PR12438 22514 468. https://gcc.gnu.org/PR12540 22515 469. https://gcc.gnu.org/PR12594 22516 470. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 22517 471. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 22518 472. https://gcc.gnu.org/PR12657 22519 473. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 22520 474. https://gcc.gnu.org/PR12696 22521 475. https://gcc.gnu.org/PR12815 22522 476. https://gcc.gnu.org/PR12862 22523 477. https://gcc.gnu.org/PR12926 22524 478. https://gcc.gnu.org/PR12967 22525 479. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 22526 480. https://gcc.gnu.org/PR12971 22527 481. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 22528 482. https://gcc.gnu.org/PR13007 22529 483. https://gcc.gnu.org/PR13009 22530 484. https://gcc.gnu.org/PR13057 22531 485. https://gcc.gnu.org/PR13070 22532 486. https://gcc.gnu.org/PR13081 22533 487. https://gcc.gnu.org/PR13239 22534 488. https://gcc.gnu.org/PR13262 22535 489. https://gcc.gnu.org/PR13290 22536 490. https://gcc.gnu.org/PR13323 22537 491. https://gcc.gnu.org/PR13369 22538 492. https://gcc.gnu.org/PR13371 22539 493. https://gcc.gnu.org/PR13445 22540 494. https://gcc.gnu.org/PR13461 22541 495. https://gcc.gnu.org/PR13462 22542 496. https://gcc.gnu.org/PR13478 22543 497. https://gcc.gnu.org/PR13544 22544 498. https://gcc.gnu.org/PR13650 22545 499. https://gcc.gnu.org/PR13683 22546 500. https://gcc.gnu.org/PR13688 22547 501. https://gcc.gnu.org/PR13774 22548 502. https://gcc.gnu.org/PR13884 22549 503. https://gcc.gnu.org/PR10746 22550 504. https://gcc.gnu.org/PR11433 22551 505. https://gcc.gnu.org/PR12633 22552 506. https://gcc.gnu.org/PR13037 22553 507. https://gcc.gnu.org/PR13213 22554 508. https://gcc.gnu.org/PR4490 22555 509. https://gcc.gnu.org/PR12292 22556 510. https://gcc.gnu.org/PR12441 22557 511. https://gcc.gnu.org/PR12943 22558 512. https://gcc.gnu.org/PR13608 22559 513. https://gcc.gnu.org/PR11598 22560 514. https://gcc.gnu.org/PR11793 22561 515. https://gcc.gnu.org/PR12467 22562 516. https://gcc.gnu.org/PR12537 22563 517. https://gcc.gnu.org/PR12496 22564 518. https://gcc.gnu.org/PR12865 22565 519. https://gcc.gnu.org/PR13354 22566 520. https://gcc.gnu.org/PR10467 22567 521. https://gcc.gnu.org/PR11226 22568 522. https://gcc.gnu.org/PR11227 22569 523. https://gcc.gnu.org/PR12644 22570 524. https://gcc.gnu.org/PR13149 22571 525. https://gcc.gnu.org/PR12654 22572 526. https://gcc.gnu.org/PR12965 22573 527. https://gcc.gnu.org/PR13031 22574 528. https://gcc.gnu.org/PR11634 22575 529. https://gcc.gnu.org/PR12158 22576 530. https://gcc.gnu.org/PR11992 22577 531. https://gcc.gnu.org/PR9365 22578 532. https://gcc.gnu.org/PR10392 22579 533. https://gcc.gnu.org/PR11322 22580 534. https://gcc.gnu.org/PR13069 22581 535. https://gcc.gnu.org/PR13302 22582 536. https://gcc.gnu.org/PR13585 22583 537. https://gcc.gnu.org/PR8916 22584 538. https://gcc.gnu.org/PR11576 22585 539. https://gcc.gnu.org/PR13122 22586 540. https://gcc.gnu.org/PR13256 22587 541. https://gcc.gnu.org/PR13373 22588 542. https://gcc.gnu.org/PR12561 22589 543. https://gcc.gnu.org/PR6243 22590 544. https://gcc.gnu.org/PR11397 22591 545. https://gcc.gnu.org/PR12505 22592 546. https://gcc.gnu.org/PR13150 22593 547. https://gcc.gnu.org/PR12666 22594 548. https://gcc.gnu.org/PR12969 22595 549. https://gcc.gnu.org/PR10819 22596 550. https://gcc.gnu.org/PR11612 22597 551. https://gcc.gnu.org/PR13211 22598 552. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.4 22599 553. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 22600 554. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.5 22601 555. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 22602 556. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.6 22603 557. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 22604 558. https://gcc.gnu.org/onlinedocs/ 22605 559. mailto:gcc-help (a] gcc.gnu.org 22606 560. mailto:gcc (a] gcc.gnu.org 22607 561. https://gcc.gnu.org/lists.html 22608 562. https://www.fsf.org/ 22609 563. https://gcc.gnu.org/about.html 22610 ====================================================================== 22611 http://gcc.gnu.org/gcc-3.2/index.html 22612 GCC 3.2 Release Series 22613 22614 (This release series is no longer supported.) 22615 22616 April 25, 2003 22617 22618 The [1]GNU project and the GCC developers are pleased to announce the 22619 release of GCC 3.2.3. 22620 22621 The purpose of the GCC 3.2 release series is to provide a stable 22622 platform for OS distributors to use building their next releases. A 22623 primary objective was to stabilize the C++ ABI; we believe that the 22624 interface to the compiler and the C++ standard library are now 22625 relatively stable. 22626 22627 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 22628 interoperate with code compiled by GCC 3.1.1 or earlier. 22629 22630 Please refer to our [2]detailed list of news, caveats, and bug-fixes 22631 for further information. 22632 22633 Release History 22634 22635 GCC 3.2.3 22636 April 25, 2003 ([3]changes) 22637 22638 GCC 3.2.2 22639 February 5, 2003 ([4]changes) 22640 22641 GCC 3.2.1 22642 November 19, 2002 ([5]changes) 22643 22644 GCC 3.2 22645 August 14, 2002 ([6]changes) 22646 22647 References and Acknowledgements 22648 22649 GCC used to stand for the GNU C Compiler, but since the compiler 22650 supports several other languages aside from C, it now stands for the 22651 GNU Compiler Collection. 22652 22653 A list of [7]successful builds is updated as new information becomes 22654 available. 22655 22656 The GCC developers would like to thank the numerous people that have 22657 contributed new features, improvements, bug fixes, and other changes as 22658 well as test results to GCC. This [8]amazing group of volunteers is 22659 what makes GCC successful. 22660 22661 For additional information about GCC please refer to the [9]GCC project 22662 web site or contact the [10]GCC development mailing list. 22663 22664 To obtain GCC please use [11]our mirror sites, or our CVS server. 22665 22666 22667 For questions related to the use of GCC, please consult these web 22668 pages and the [12]GCC manuals. If that fails, the 22669 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22670 web pages and the development of GCC are welcome on our developer 22671 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 22672 archives. 22673 22674 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 22675 distribution of this entire article is permitted in any medium, 22676 provided this notice is preserved. 22677 22678 These pages are [17]maintained by the GCC team. Last modified 22679 2022-10-26. 22680 22681 References 22682 22683 1. http://www.gnu.org/ 22684 2. https://gcc.gnu.org/gcc-3.2/changes.html 22685 3. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 22686 4. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 22687 5. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 22688 6. https://gcc.gnu.org/gcc-3.2/changes.html#3.2 22689 7. https://gcc.gnu.org/gcc-3.2/buildstat.html 22690 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 22691 9. https://gcc.gnu.org/index.html 22692 10. mailto:gcc (a] gcc.gnu.org 22693 11. https://gcc.gnu.org/mirrors.html 22694 12. https://gcc.gnu.org/onlinedocs/ 22695 13. mailto:gcc-help (a] gcc.gnu.org 22696 14. mailto:gcc (a] gcc.gnu.org 22697 15. https://gcc.gnu.org/lists.html 22698 16. https://www.fsf.org/ 22699 17. https://gcc.gnu.org/about.html 22700 ====================================================================== 22701 http://gcc.gnu.org/gcc-3.2/changes.html 22702 GCC 3.2 Release Series 22703 Changes, New Features, and Fixes 22704 22705 The latest release in the 3.2 release series is [1]GCC 3.2.3. 22706 22707 Caveats and New Features 22708 22709 Caveats 22710 22711 * The C++ compiler does not correctly zero-initialize 22712 pointers-to-data members. You must explicitly initialize them. For 22713 example: int S::*m(0); will work, but depending on 22714 default-initialization to zero will not work. This bug cannot be 22715 fixed in GCC 3.2 without inducing unacceptable risks. It will be 22716 fixed in GCC 3.3. 22717 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 22718 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 22719 a number of C++ ABI fixes which make its C++ compiler generate 22720 binary code which is incompatible with the C++ compilers found in 22721 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 22722 22723 Frontend Enhancements 22724 22725 C/C++/Objective-C 22726 22727 * The method of constructing the list of directories to be searched 22728 for header files has been revised. If a directory named by a -I 22729 option is a standard system include directory, the option is 22730 ignored to ensure that the default search order for system 22731 directories and the special treatment of system header files are 22732 not defeated. 22733 * The C and Objective-C compilers no longer accept the "Naming Types" 22734 extension (typedef foo = bar); it was already unavailable in C++. 22735 Code which uses it will need to be changed to use the "typeof" 22736 extension instead: typedef typeof(bar) foo. (We have removed this 22737 extension without a period of deprecation because it has caused the 22738 compiler to crash since version 3.0 and no one noticed until very 22739 recently. Thus we conclude it is not in widespread use.) 22740 22741 C++ 22742 22743 * GCC 3.2 fixed several differences between the C++ ABI implemented 22744 in GCC and the multi-vendor standard, but more have been found 22745 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 22746 code which is affected by these bugs. We will fix these bugs in 22747 some future release, once we are confident that all have been 22748 found; until then, it is our intention to make changes to the ABI 22749 only if they are necessary for correct compilation of C++, as 22750 opposed to conformance to the ABI documents. 22751 * For details on how to build an ABI compliant compiler for GNU/Linux 22752 systems, check the [3]common C++ ABI page. 22753 22754 New Targets and Target Specific Improvements 22755 22756 IA-32 22757 22758 * Fixed a number of bugs in SSE and MMX intrinsics. 22759 * Fixed common compiler crashes with SSE instruction set enabled 22760 (implied by -march=pentium3, pentium4, athlon-xp) 22761 * __m128 and __m128i is not 128bit aligned when used in structures. 22762 22763 x86-64 22764 22765 * A bug whereby the compiler could generate bad code for bzero has 22766 been fixed. 22767 * ABI fixes (implying ABI incompatibilities with previous version in 22768 some corner cases) 22769 * Fixed prefetch code generation 22770 __________________________________________________________________ 22771 22772 [4]GCC 3.2.3 22773 22774 3.2.3 is a bug fix release only; there are no new features that were 22775 not present in GCC 3.2.2. 22776 22777 Bug Fixes 22778 22779 This section lists the problem reports (PRs) from GCC's bug tracking 22780 system that are known to be fixed in the 3.2.3 release. This list might 22781 not be complete (that is, it is possible that some PRs that have been 22782 fixed are not listed here), and some of the titles have been changed to 22783 make them more clear. 22784 22785 Internal Compiler Errors (multi-platform) 22786 22787 * [5]3782: (c++) -quiet -fstats produces a segmentation fault in 22788 cc1plus 22789 * [6]6440: (c++) template specializations cause ICE 22790 * [7]7050: (c++) ICE on: (i ? get_string() : throw) 22791 * [8]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 22792 * [9]7982: (c++) ICE due to infinite recursion (using STL set) 22793 * [10]8068: exceedingly high (infinite) memory usage 22794 * [11]8178: ICE with __builtin_ffs 22795 * [12]8396: ICE in copy_to_mode_reg, in explow.c 22796 * [13]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 22797 * [14]9768: ICE when optimizing inline code at -O2 22798 * [15]9798: (c++) Infinite recursion (segfault) in 22799 cp/decl.c:push_using_directive with recursive using directives 22800 * [16]9799: mismatching structure initializer with nested flexible 22801 array member: ICE 22802 * [17]9928: ICE on duplicate enum declaration 22803 * [18]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 22804 sparc, alpha) 22805 * [19]10352: ICE in find_reloads_toplev 22806 * [20]10336: ICE with -Wunreachable-code 22807 22808 C/optimizer bugs: 22809 22810 * [21]8224: Incorrect joining of signed and unsigned division 22811 * [22]8613: -O2 produces wrong code with builtin strlen and 22812 postincrements 22813 * [23]8828: gcc reports some code is unreachable when it is not 22814 * [24]9226: GCSE breaking argument passing 22815 * [25]9853: miscompilation of non-constant structure initializer 22816 * [26]9797: C99-style struct initializers are miscompiled 22817 * [27]9967: Some standard C function calls should not be replaced 22818 when optimizing for size 22819 * [28]10116: ce2: invalid merge of join_bb in the context of switch 22820 statements 22821 * [29]10171: wrong code for inlined function 22822 * [30]10175: -Wunreachable-code doesn't work for single lines 22823 22824 C++ compiler and library: 22825 22826 * [31]8316: Confusing diagnostic for code that misuses conversion 22827 operators 22828 * [32]9169: filebuf output fails if codecvt<>::out returns noconv 22829 * [33]9420: incomplete type incorrectly reported 22830 * [34]9459: typeof in return type specification of template not 22831 supported 22832 * [35]9507: filebuf::open handles ios_base::ate incorrectly 22833 * [36]9538: Out-of-bounds memory access in streambuf::sputbackc 22834 * [37]9602: Total confusion about template/friend/virtual/abstract 22835 * [38]9993: destructor not called for local object created within and 22836 returned from infinite loop 22837 * [39]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 22838 system 22839 22840 Java compiler and library: 22841 22842 * [40]9652: libgcj build fails on irix6.5.1[78] 22843 * [41]10144: gas on solaris complains about bad .stabs lines for 22844 java, native as unaffected 22845 22846 x86-specific (Intel/AMD): 22847 22848 * [42]8746: gcc miscompiles Linux kernel ppa driver on x86 22849 * [43]9888: -mcpu=k6 -Os produces out of range loop instructions 22850 * [44]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 22851 failed 22852 * [45]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 22853 failed 22854 22855 SPARC-specific: 22856 22857 * [46]7784: [Sparc] ICE in extract_insn, in recog.c 22858 * [47]7796: sparc extra failure with -m64 on execute/930921-1.c in 22859 unroll.c 22860 * [48]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 22861 * [49]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 22862 execute/loop-2d.c 22863 * [50]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 22864 * [51]9414: Scheduling bug on Ultrasparc 22865 * [52]10067: GCC-3.2.2 outputs invalid asm on sparc64 22866 22867 m68k-specific: 22868 22869 * [53]7248: broken "inclusive or" code 22870 * [54]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 22871 22872 PowerPC-specific: 22873 22874 * [55]9732: Wrong code with -O2 -fPIC 22875 * [56]10073: ICE: powerpc cannot split insn 22876 22877 Alpha-specific: 22878 22879 * [57]7702: optimization problem on a DEC alpha under OSF1 22880 * [58]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 22881 22882 HP-specific: 22883 22884 * [59]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 22885 * [60]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 22886 (missing symbol) 22887 * [61]10271: Floating point args don't get reloaded across function 22888 calls with -O2 22889 22890 MIPS specific: 22891 22892 * [62]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 22893 compile/920501-4.c 22894 22895 CRIS specific: 22896 22897 * [63]10377: gcc-3.2.2 creates bad assembler code for cris 22898 22899 Miscellaneous and minor bugs: 22900 22901 * [64]6955: collect2 says "core dumped" when there is no core 22902 __________________________________________________________________ 22903 22904 [65]GCC 3.2.2 22905 22906 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 22907 install by means of the DESTDIR variable. Parts of the GCC tree have 22908 featured that support long before, but now it is available even from 22909 the top level. 22910 22911 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 22912 features that were not present in GCC 3.2.1. 22913 22914 Bug Fixes 22915 22916 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 22917 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 22918 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 22919 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 22920 change, and thus restores ABI-compatibility with previous releases 22921 (except GCC 3.2.1) on these platforms. 22922 22923 This section lists the problem reports (PRs) from GCC's bug tracking 22924 system that are known to be fixed in the 3.2.2 release. This list might 22925 not be complete (that is, it is possible that some PRs that have been 22926 fixed are not listed here) and some of the titles have been changed to 22927 make them more clear. 22928 22929 Internal Compiler Errors (multi-platform) 22930 22931 * [66]5919: (c++) ICE when passing variable array to template 22932 function 22933 * [67]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 22934 * [68]7507: ICE with -O2 when address of called function is a 22935 complicated expression 22936 * [69]7622: ICE with nested inline functions if function's address is 22937 taken 22938 * [70]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 22939 [71]9258) 22940 * [72]8031: (c++) ICE in code comparing typeids and casting from 22941 virtual base 22942 * [73]8275: ICE in simplify_subreg 22943 * [74]8332: (c++) builtin strlen/template interaction causes ICE 22944 * [75]8372: (c++) ICE on explicit call of destructor 22945 * [76]8439: (c, not c++) empty struct causes ICE 22946 * [77]8442: (c++) ICE with nested template classes 22947 * [78]8518: ICE when compiling mplayer ("extern inline" issue) 22948 * [79]8615: (c++) ICE with out-of-range character constant template 22949 argument 22950 * [80]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 22951 * [81]8799: (c++) ICE: error reporting routines re-entered 22952 * [82]9328: (c++) ICE with typeof(X) for overloaded X 22953 * [83]9465: (preprocessor) cpp -traditional ICE on null bytes 22954 22955 C++ (compiler and library) bugs 22956 22957 * [84]47: scoping in nested classes is broken 22958 * [85]6745: problems with iostream rdbuf() member function 22959 * [86]8214: conversion from const char* const to char* sometimes 22960 accepted illegally 22961 * [87]8493: builtin strlen and overload resolution (same bug as 22962 [88]8332) 22963 * [89]8503: strange behaviour of function types 22964 * [90]8727: compiler confused by inheritance from an anonymous struct 22965 * [91]7445: poor performance of std::locale::classic() in 22966 multi-threaded applications 22967 * [92]8230: mishandling of overflow in vector<T>::resize 22968 * [93]8399: sync_with_stdio(false) breaks unformatted input 22969 * [94]8662: illegal access of private member of unnamed class is 22970 accepted 22971 * [95]8707: "make distclean" fails in libstdc++-v3 directory 22972 * [96]8708: __USE_MALLOC doesn't work 22973 * [97]8790: Use of non-thread-safe strtok in src/localename.cc 22974 * [98]8887: Bug in date formats with --enable-clocale=generic 22975 * [99]9076: Call Frame Instructions are not handled correctly during 22976 unwind operation 22977 * [100]9151: std::setprecision limited to 16 digits when outputting a 22978 double to a stream 22979 * [101]9168: codecvt<char, char, mbstate_t> overwrites output buffers 22980 * [102]9269: libstdc++ headers: explicit specialization of function 22981 must precede its first use 22982 * [103]9322: return value of basic_streambuf<>::getloc affected by 22983 locale::global 22984 * [104]9433: segfault in runtime support for dynamic_cast 22985 22986 C and optimizer bugs 22987 22988 * [105]8032: GCC incorrectly initializes static structs that have 22989 flexible arrays 22990 * [106]8639: simple arithmetic expression broken 22991 * [107]8794: optimization improperly eliminates certain expressions 22992 * [108]8832: traditional "asm volatile" code is illegally optimized 22993 * [109]8988: loop optimizer bug: with -O2, code is generated that 22994 segfaults (found on i386, bug present for all platforms) 22995 * [110]9492: structure copy clobbers subsequent stores to structure 22996 22997 Objective-C bugs 22998 22999 * [111]9267: Objective-C parser won't build with newer bison versions 23000 (e.g. 1.875) 23001 23002 Ada bugs 23003 23004 * [112]8344: Ada build problem due to conflict between gcc/final.o, 23005 gcc/ada/final.o 23006 23007 Preprocessor bugs 23008 23009 * [113]8524: _Pragma within macros is improperly expanded 23010 * [114]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 23011 -fshort-wchar 23012 23013 ARM-specific 23014 23015 * [115]9090: arm ICE with >= -O2; regression from gcc-2.95 23016 23017 x86-specific (Intel/AMD) 23018 23019 * [116]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 23020 * [117]8599: loop unroll bug with -march=k6-3 23021 * [118]9506: ABI breakage in structure return (affects BSD and 23022 Cygwin, but not GNU/Linux) 23023 23024 FreeBSD 5.0 specific 23025 23026 * [119]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 23027 23028 RTEMS-specific 23029 23030 * [120]9292: hppa1.1-rtems configurery problems 23031 * [121]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 23032 * [122]9295: [mips-rtems] config/mips/rtems.h init/fini issue 23033 * [123]9296: gthr-rtems regression 23034 * [124]9316: powerpc-rtems: extending multilibs 23035 23036 HP-PA specific 23037 23038 * [125]9493: ICE with -O2 when building a simple function 23039 23040 Documentation 23041 23042 * [126]7341: hyperlink to gcov in GCC documentation doesn't work 23043 * [127]8947: Please add a warning about "-malign-double" in docs 23044 * [128]7448, [129]8882: typo cleanups 23045 __________________________________________________________________ 23046 23047 [130]GCC 3.2.1 23048 23049 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 23050 generates code that is known not to be binary-compatible with the 23051 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 23052 in the distribution, for details. 23053 23054 This release also removes an old GCC extension, "naming types", and the 23055 documentation now directs users to use a different GCC extension, 23056 __typeof__, instead. The feature had evidently been broken for a while. 23057 23058 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 23059 the new warning there are no new features that were not present in GCC 23060 3.2. 23061 23062 In addition, the previous fix for [131]PR 7445 (poor performance of 23063 std::locale::classic() in multi-threaded applications) was reverted 23064 ("unfixed"), because the "fix" was not thread-safe. 23065 23066 Bug Fixes 23067 23068 This section lists the problem reports (PRs) from GCC's bug tracking 23069 system that are known to be fixed in the 3.2.1 release. This list might 23070 not be complete (that is, it is possible that some PRs that have been 23071 fixed are not listed here). As you can see, the number of bug fixes is 23072 quite large, so it is strongly recommended that users of earlier GCC 23073 3.x releases upgrade to GCC 3.2.1. 23074 23075 Internal Compiler Errors (multi-platform) 23076 23077 * [132]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 23078 * [133]5661: (c++) ICE instantiating template on array of unknown 23079 size (bad code) 23080 * [134]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 23081 64-bit platforms 23082 * [135]6994: (c++) ICE in find_function_data 23083 * [136]7150: preprocessor: GCC -dM -E gives an ICE 23084 * [137]7160: ICE when optimizing branches without a return value 23085 * [138]7228: (c++) ICE when using member template and template 23086 function 23087 * [139]7266: (c++) ICE with -pedantic on missing typename 23088 * [140]7353: ICE from use of "Naming Types" extension, see above 23089 * [141]7411: ICE in instantiate_virtual_regs_1, in function.c 23090 * [142]7478: (c++) ICE on static_cast inside template 23091 * [143]7526: preprocessor core dump when _Pragma implies #pragma 23092 dependency 23093 * [144]7721: (c++) ICE on simple (but incorrect) template ([145]7803 23094 is a duplicate) 23095 * [146]7754: (c++) ICE on union with template parameter 23096 * [147]7788: (c++) redeclaring a definition as an incomplete class 23097 causes ICE 23098 * [148]8031: (c++) ICE in comptypes, in cp/typeck.c 23099 * [149]8055: preprocessor dies with SIG11 when building FreeBSD 23100 kernel 23101 * [150]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 23102 variables 23103 * [151]8134: (c++) ICE in force_store_init_value on legal code 23104 * [152]8149: (c++) ICE on incomplete type 23105 * [153]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 23106 initialization 23107 23108 C++ (compiler and library) bugs 23109 23110 * [154]5607: No pointer adjustment in covariant return types 23111 * [155]6579: Infinite loop with statement expressions in member 23112 initialization 23113 * [156]6803: Default copy constructor bug in GCC 3.1 23114 * [157]7176: g++ confused by friend and static member with same name 23115 * [158]7188: Segfault with template class and recursive (incorrect) 23116 initializer list 23117 * [159]7306: Regression: GCC 3.x fails to compile code with virtual 23118 inheritance if a method has a variable number of arguments 23119 * [160]7461: ctype<char>::classic_table() returns offset array on 23120 Cygwin 23121 * [161]7524: f(const float arg[3]) fails 23122 * [162]7584: Erroneous ambiguous base error on using declaration 23123 * [163]7676: Member template overloading problem 23124 * [164]7679: infinite loop when a right parenthesis is missing 23125 * [165]7811: default locale not taken from environment 23126 * [166]7961: compare( char *) implemented incorrectly in 23127 basic_string<> 23128 * [167]8071: basic_ostream::operator<<(streambuf*) loops forever if 23129 streambuf::underflow() leaves gptr() NULL (dups: [168]8127, 23130 [169]6745) 23131 * [170]8096: deque::at() throws std::range_error instead of 23132 std::out_of_range 23133 * [171]8127: cout << cin.rdbuf() infinite loop 23134 * [172]8218: Excessively large memory consumed for classes with large 23135 array members 23136 * [173]8287: GCC 3.2: Destructor called for non-constructed local 23137 object 23138 * [174]8347: empty vector range used in string construction causes 23139 core dump 23140 * [175]8348: fail() flag is set in istringstream when eof() flag is 23141 set 23142 * [176]8391: regression: infinite loop in cp/decl2.c(finish_file) 23143 23144 C and optimizer bugs 23145 23146 * [177]6627: -fno-align-functions doesn't seem to disable function 23147 alignment 23148 * [178]6631: life_analysis misoptimizes code to initialize fields of 23149 a structure 23150 * [179]7102: unsigned char division results in floating exception 23151 * [180]7120: Run once loop should *always* be unrolled 23152 (pessimization) 23153 * [181]7209: Bug involving array referencing and ?: operator 23154 * [182]7515: invalid inlining of global function with -O3 23155 * [183]7814: incorrect scheduling for glibc-2.2.92 strcpy test 23156 * [184]8467: bug in sibling call optimization 23157 23158 Preprocessor bugs 23159 23160 * [185]4890: incorrect line markers from the traditional preprocessor 23161 * [186]7357: -M option omits system headers files (making it the same 23162 as -MM) 23163 * [187]7358: Changes to Sun's make Dependencies 23164 * [188]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 23165 C headers 23166 * [189]7862: Interrupting GCC -MD removes .d file but not .o 23167 * [190]8190: Failed compilation deletes -MD dependency file 23168 * [191]8524: _Pragma within macro is improperly expanded 23169 23170 x86 specific (Intel/AMD) 23171 23172 * [192]5351: (i686-only) function pass-by-value structure copy 23173 corrupts stack ([193]7591 is a duplicate) 23174 * [194]6845, [195]7034, [196]7124, [197]7174: ICE's with 23175 -march=pentium3/pentium2/athlon (these are all the same underlying 23176 bug, in MMX register use) 23177 * [198]7134, [199]7375, [200]7390: ICE with -march=athlon (maybe same 23178 as above?) 23179 * [201]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 23180 * [202]6981: wrong code in 64-bit manipulation on x86 23181 * [203]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 23182 macro 23183 * [204]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 23184 intrinsics are broken 23185 * [205]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 23186 -march=pentium4 23187 * [206]7693: Typo in i386 mmintrin.h header 23188 * [207]7723: ICE - Pentium3 sse - GCC 3.2 23189 * [208]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 23190 * [209]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 23191 23192 PowerPC specific 23193 23194 * [210]5967: GCC bug when profiling nested functions on powerpc 23195 * [211]6984: wrong code generated with -O2, -O3, -Os for do-while 23196 loop on PowerPC 23197 * [212]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 23198 * [213]7130: miscompiled code for GCC-3.1 on 23199 powerpc-unknown-linux-gnu with -funroll-all-loops 23200 * [214]7133: PowerPC ICE: unrecognizable insn 23201 * [215]7380: ICE in extract_insn, at recog.c:2148 23202 * [216]8252: ICE on Altivec code with optimization turned on 23203 * [217]8451: Altivec ICE in GCC 3.2 23204 23205 HP/PA specific 23206 23207 * [218]7250: __ashrdi3 returns wrong value on 32 bit hppa 23208 23209 SPARC specific 23210 23211 * [219]6668: when using --disable-multilib, libgcc_s.so is installed 23212 in the wrong place on sparc-solaris 23213 * [220]7151: ICE when compiling for UltraSPARC 23214 * [221]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 23215 double and -O1 23216 * [222]7842: [REGRESSION] SPARC code gen bug 23217 23218 ARM specific 23219 23220 * [223]7856: [arm] invalid offset in constant pool reference 23221 * [224]7967: optimization produces wrong code (ARM) 23222 23223 Alpha specific 23224 23225 * [225]7374: __builtin_fabsl broken on alpha 23226 23227 IBM s390 specific 23228 23229 * [226]7370: ICE in fixup_var_refs_1 on s390x 23230 * [227]7409: loop optimization bug on s390x-linux-gnu 23231 * [228]8232: s390x: ICE when using bcmp with int length argument 23232 23233 SCO specific 23234 23235 * [229]7623: SCO OpenServer build fails with machmode.def: undefined 23236 symbol: BITS_PER_UNIT 23237 23238 m68k/Coldfire specific 23239 23240 * [230]8314: crtbegin, crtend need to be multilib'ed for this 23241 platform 23242 23243 Documentation 23244 23245 * [231]761: Document some undocumented options 23246 * [232]5610: Fix documentation about invoking SSE instructions 23247 (-mfpmath=sse) 23248 * [233]7484: List -Wmissing-declarations as C-only option 23249 * [234]7531: -mcmodel not documented for x86-64 23250 * [235]8120: Update documentation of bad use of ## 23251 __________________________________________________________________ 23252 23253 [236]GCC 3.2 23254 23255 3.2 is a small bug fix release, but there is a change to the 23256 application binary interface (ABI), hence the change to the second part 23257 of the version number. 23258 23259 The main purpose of the 3.2 release is to correct a couple of problems 23260 in the C++ ABI, with the intention of providing a stable interface 23261 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 23262 23263 Bug Fixes 23264 23265 C++ 23266 23267 * [237]7320: g++ 3.2 relocation problem 23268 * [238]7470: vtable: virtual function pointers not in declaration 23269 order 23270 23271 libstdc++ 23272 23273 * [239]6410: Trouble with non-ASCII monetary symbols and wchar_t 23274 * [240]6503, [241]6642, [242]7186: Problems with comparing or 23275 subtracting various types of const and non-const iterators 23276 * [243]7216: ambiguity with basic_iostream::traits_type 23277 * [244]7220: problem with basic_istream::ignore(0,delimiter) 23278 * [245]7222: locale::operator==() doesn't work on std::locale("") 23279 * [246]7286: placement operator delete issue 23280 * [247]7442: cxxabi.h does not match the C++ ABI 23281 * [248]7445: poor performance of std::locale::classic() in 23282 multi-threaded applications 23283 23284 x86-64 specific 23285 23286 * [249]7291: off-by-one in generated inline bzero code for x86-64 23287 23288 23289 For questions related to the use of GCC, please consult these web 23290 pages and the [250]GCC manuals. If that fails, the 23291 [251]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23292 web pages and the development of GCC are welcome on our developer 23293 list at [252]gcc (a] gcc.gnu.org. All of [253]our lists have public 23294 archives. 23295 23296 Copyright (C) [254]Free Software Foundation, Inc. Verbatim copying and 23297 distribution of this entire article is permitted in any medium, 23298 provided this notice is preserved. 23299 23300 These pages are [255]maintained by the GCC team. Last modified 23301 2025-04-08. 23302 23303 References 23304 23305 1. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 23306 2. https://gcc.gnu.org/gcc-3.1/changes.html 23307 3. https://gcc.gnu.org/gcc-3.2/c++-abi.html 23308 4. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.3 23309 5. https://gcc.gnu.org/PR3782 23310 6. https://gcc.gnu.org/PR6440 23311 7. https://gcc.gnu.org/PR7050 23312 8. https://gcc.gnu.org/PR7741 23313 9. https://gcc.gnu.org/PR7982 23314 10. https://gcc.gnu.org/PR8068 23315 11. https://gcc.gnu.org/PR8178 23316 12. https://gcc.gnu.org/PR8396 23317 13. https://gcc.gnu.org/PR8674 23318 14. https://gcc.gnu.org/PR9768 23319 15. https://gcc.gnu.org/PR9798 23320 16. https://gcc.gnu.org/PR9799 23321 17. https://gcc.gnu.org/PR9928 23322 18. https://gcc.gnu.org/PR10114 23323 19. https://gcc.gnu.org/PR10352 23324 20. https://gcc.gnu.org/PR10336 23325 21. https://gcc.gnu.org/PR8224 23326 22. https://gcc.gnu.org/PR8613 23327 23. https://gcc.gnu.org/PR8828 23328 24. https://gcc.gnu.org/PR9226 23329 25. https://gcc.gnu.org/PR9853 23330 26. https://gcc.gnu.org/PR9797 23331 27. https://gcc.gnu.org/PR9967 23332 28. https://gcc.gnu.org/PR10116 23333 29. https://gcc.gnu.org/PR10171 23334 30. https://gcc.gnu.org/PR10175 23335 31. https://gcc.gnu.org/PR8316 23336 32. https://gcc.gnu.org/PR9169 23337 33. https://gcc.gnu.org/PR9420 23338 34. https://gcc.gnu.org/PR9459 23339 35. https://gcc.gnu.org/PR9507 23340 36. https://gcc.gnu.org/PR9538 23341 37. https://gcc.gnu.org/PR9602 23342 38. https://gcc.gnu.org/PR9993 23343 39. https://gcc.gnu.org/PR10167 23344 40. https://gcc.gnu.org/PR9652 23345 41. https://gcc.gnu.org/PR10144 23346 42. https://gcc.gnu.org/PR8746 23347 43. https://gcc.gnu.org/PR9888 23348 44. https://gcc.gnu.org/PR9638 23349 45. https://gcc.gnu.org/PR9954 23350 46. https://gcc.gnu.org/PR7784 23351 47. https://gcc.gnu.org/PR7796 23352 48. https://gcc.gnu.org/PR8281 23353 49. https://gcc.gnu.org/PR8366 23354 50. https://gcc.gnu.org/PR8726 23355 51. https://gcc.gnu.org/PR9414 23356 52. https://gcc.gnu.org/PR10067 23357 53. https://gcc.gnu.org/PR7248 23358 54. https://gcc.gnu.org/PR8343 23359 55. https://gcc.gnu.org/PR9732 23360 56. https://gcc.gnu.org/PR10073 23361 57. https://gcc.gnu.org/PR7702 23362 58. https://gcc.gnu.org/PR9671 23363 59. https://gcc.gnu.org/PR8694 23364 60. https://gcc.gnu.org/PR9953 23365 61. https://gcc.gnu.org/PR10271 23366 62. https://gcc.gnu.org/PR6362 23367 63. https://gcc.gnu.org/PR10377 23368 64. https://gcc.gnu.org/PR6955 23369 65. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.2 23370 66. https://gcc.gnu.org/PR5919 23371 67. https://gcc.gnu.org/PR7129 23372 68. https://gcc.gnu.org/PR7507 23373 69. https://gcc.gnu.org/PR7622 23374 70. https://gcc.gnu.org/PR7681 23375 71. https://gcc.gnu.org/PR9528 23376 72. https://gcc.gnu.org/PR8031 23377 73. https://gcc.gnu.org/PR8275 23378 74. https://gcc.gnu.org/PR8332 23379 75. https://gcc.gnu.org/PR8372 23380 76. https://gcc.gnu.org/PR8439 23381 77. https://gcc.gnu.org/PR8442 23382 78. https://gcc.gnu.org/PR8518 23383 79. https://gcc.gnu.org/PR8615 23384 80. https://gcc.gnu.org/PR8663 23385 81. https://gcc.gnu.org/PR8799 23386 82. https://gcc.gnu.org/PR9328 23387 83. https://gcc.gnu.org/PR9465 23388 84. https://gcc.gnu.org/PR47 23389 85. https://gcc.gnu.org/PR6745 23390 86. https://gcc.gnu.org/PR8214 23391 87. https://gcc.gnu.org/PR8493 23392 88. https://gcc.gnu.org/PR8332 23393 89. https://gcc.gnu.org/PR8503 23394 90. https://gcc.gnu.org/PR8727 23395 91. https://gcc.gnu.org/PR7445 23396 92. https://gcc.gnu.org/PR8230 23397 93. https://gcc.gnu.org/PR8399 23398 94. https://gcc.gnu.org/PR8662 23399 95. https://gcc.gnu.org/PR8707 23400 96. https://gcc.gnu.org/PR8708 23401 97. https://gcc.gnu.org/PR8790 23402 98. https://gcc.gnu.org/PR8887 23403 99. https://gcc.gnu.org/PR9076 23404 100. https://gcc.gnu.org/PR9151 23405 101. https://gcc.gnu.org/PR9168 23406 102. https://gcc.gnu.org/PR9269 23407 103. https://gcc.gnu.org/PR9322 23408 104. https://gcc.gnu.org/PR9433 23409 105. https://gcc.gnu.org/PR8032 23410 106. https://gcc.gnu.org/PR8639 23411 107. https://gcc.gnu.org/PR8794 23412 108. https://gcc.gnu.org/PR8832 23413 109. https://gcc.gnu.org/PR8988 23414 110. https://gcc.gnu.org/PR9492 23415 111. https://gcc.gnu.org/PR9267 23416 112. https://gcc.gnu.org/PR8344 23417 113. https://gcc.gnu.org/PR8524 23418 114. https://gcc.gnu.org/PR8880 23419 115. https://gcc.gnu.org/PR9090 23420 116. https://gcc.gnu.org/PR8588 23421 117. https://gcc.gnu.org/PR8599 23422 118. https://gcc.gnu.org/PR9506 23423 119. https://gcc.gnu.org/PR9484 23424 120. https://gcc.gnu.org/PR9292 23425 121. https://gcc.gnu.org/PR9293 23426 122. https://gcc.gnu.org/PR9295 23427 123. https://gcc.gnu.org/PR9296 23428 124. https://gcc.gnu.org/PR9316 23429 125. https://gcc.gnu.org/PR9493 23430 126. https://gcc.gnu.org/PR7341 23431 127. https://gcc.gnu.org/PR8947 23432 128. https://gcc.gnu.org/PR7448 23433 129. https://gcc.gnu.org/PR8882 23434 130. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.1 23435 131. https://gcc.gnu.org/PR7445 23436 132. https://gcc.gnu.org/PR2521 23437 133. https://gcc.gnu.org/PR5661 23438 134. https://gcc.gnu.org/PR6419 23439 135. https://gcc.gnu.org/PR6994 23440 136. https://gcc.gnu.org/PR7150 23441 137. https://gcc.gnu.org/PR7160 23442 138. https://gcc.gnu.org/PR7228 23443 139. https://gcc.gnu.org/PR7266 23444 140. https://gcc.gnu.org/PR7353 23445 141. https://gcc.gnu.org/PR7411 23446 142. https://gcc.gnu.org/PR7478 23447 143. https://gcc.gnu.org/PR7526 23448 144. https://gcc.gnu.org/PR7721 23449 145. https://gcc.gnu.org/PR7803 23450 146. https://gcc.gnu.org/PR7754 23451 147. https://gcc.gnu.org/PR7788 23452 148. https://gcc.gnu.org/PR8031 23453 149. https://gcc.gnu.org/PR8055 23454 150. https://gcc.gnu.org/PR8067 23455 151. https://gcc.gnu.org/PR8134 23456 152. https://gcc.gnu.org/PR8149 23457 153. https://gcc.gnu.org/PR8160 23458 154. https://gcc.gnu.org/PR5607 23459 155. https://gcc.gnu.org/PR6579 23460 156. https://gcc.gnu.org/PR6803 23461 157. https://gcc.gnu.org/PR7176 23462 158. https://gcc.gnu.org/PR7188 23463 159. https://gcc.gnu.org/PR7306 23464 160. https://gcc.gnu.org/PR7461 23465 161. https://gcc.gnu.org/PR7524 23466 162. https://gcc.gnu.org/PR7584 23467 163. https://gcc.gnu.org/PR7676 23468 164. https://gcc.gnu.org/PR7679 23469 165. https://gcc.gnu.org/PR7811 23470 166. https://gcc.gnu.org/PR7961 23471 167. https://gcc.gnu.org/PR8071 23472 168. https://gcc.gnu.org/PR8127 23473 169. https://gcc.gnu.org/PR6745 23474 170. https://gcc.gnu.org/PR8096 23475 171. https://gcc.gnu.org/PR8127 23476 172. https://gcc.gnu.org/PR8218 23477 173. https://gcc.gnu.org/PR8287 23478 174. https://gcc.gnu.org/PR8347 23479 175. https://gcc.gnu.org/PR8348 23480 176. https://gcc.gnu.org/PR8391 23481 177. https://gcc.gnu.org/PR6627 23482 178. https://gcc.gnu.org/PR6631 23483 179. https://gcc.gnu.org/PR7102 23484 180. https://gcc.gnu.org/PR7120 23485 181. https://gcc.gnu.org/PR7209 23486 182. https://gcc.gnu.org/PR7515 23487 183. https://gcc.gnu.org/PR7814 23488 184. https://gcc.gnu.org/PR8467 23489 185. https://gcc.gnu.org/PR4890 23490 186. https://gcc.gnu.org/PR7357 23491 187. https://gcc.gnu.org/PR7358 23492 188. https://gcc.gnu.org/PR7602 23493 189. https://gcc.gnu.org/PR7862 23494 190. https://gcc.gnu.org/PR8190 23495 191. https://gcc.gnu.org/PR8524 23496 192. https://gcc.gnu.org/PR5351 23497 193. https://gcc.gnu.org/PR7591 23498 194. https://gcc.gnu.org/PR6845 23499 195. https://gcc.gnu.org/PR7034 23500 196. https://gcc.gnu.org/PR7124 23501 197. https://gcc.gnu.org/PR7174 23502 198. https://gcc.gnu.org/PR7134 23503 199. https://gcc.gnu.org/PR7375 23504 200. https://gcc.gnu.org/PR7390 23505 201. https://gcc.gnu.org/PR6890 23506 202. https://gcc.gnu.org/PR6981 23507 203. https://gcc.gnu.org/PR7242 23508 204. https://gcc.gnu.org/PR7396 23509 205. https://gcc.gnu.org/PR7630 23510 206. https://gcc.gnu.org/PR7693 23511 207. https://gcc.gnu.org/PR7723 23512 208. https://gcc.gnu.org/PR7951 23513 209. https://gcc.gnu.org/PR8146 23514 210. https://gcc.gnu.org/PR5967 23515 211. https://gcc.gnu.org/PR6984 23516 212. https://gcc.gnu.org/PR7114 23517 213. https://gcc.gnu.org/PR7130 23518 214. https://gcc.gnu.org/PR7133 23519 215. https://gcc.gnu.org/PR7380 23520 216. https://gcc.gnu.org/PR8252 23521 217. https://gcc.gnu.org/PR8451 23522 218. https://gcc.gnu.org/PR7250 23523 219. https://gcc.gnu.org/PR6668 23524 220. https://gcc.gnu.org/PR7151 23525 221. https://gcc.gnu.org/PR7335 23526 222. https://gcc.gnu.org/PR7842 23527 223. https://gcc.gnu.org/PR7856 23528 224. https://gcc.gnu.org/PR7967 23529 225. https://gcc.gnu.org/PR7374 23530 226. https://gcc.gnu.org/PR7370 23531 227. https://gcc.gnu.org/PR7409 23532 228. https://gcc.gnu.org/PR8232 23533 229. https://gcc.gnu.org/PR7623 23534 230. https://gcc.gnu.org/PR8314 23535 231. https://gcc.gnu.org/PR761 23536 232. https://gcc.gnu.org/PR5610 23537 233. https://gcc.gnu.org/PR7484 23538 234. https://gcc.gnu.org/PR7531 23539 235. https://gcc.gnu.org/PR8120 23540 236. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2 23541 237. https://gcc.gnu.org/PR7320 23542 238. https://gcc.gnu.org/PR7470 23543 239. https://gcc.gnu.org/PR6410 23544 240. https://gcc.gnu.org/PR6503 23545 241. https://gcc.gnu.org/PR6642 23546 242. https://gcc.gnu.org/PR7186 23547 243. https://gcc.gnu.org/PR7216 23548 244. https://gcc.gnu.org/PR7220 23549 245. https://gcc.gnu.org/PR7222 23550 246. https://gcc.gnu.org/PR7286 23551 247. https://gcc.gnu.org/PR7442 23552 248. https://gcc.gnu.org/PR7445 23553 249. https://gcc.gnu.org/PR7291 23554 250. https://gcc.gnu.org/onlinedocs/ 23555 251. mailto:gcc-help (a] gcc.gnu.org 23556 252. mailto:gcc (a] gcc.gnu.org 23557 253. https://gcc.gnu.org/lists.html 23558 254. https://www.fsf.org/ 23559 255. https://gcc.gnu.org/about.html 23560 ====================================================================== 23561 http://gcc.gnu.org/gcc-3.1/index.html 23562 GCC 3.1 23563 23564 (This release series is no longer supported.) 23565 23566 July 27, 2002 23567 23568 The GNU project and the GCC developers are pleased to announce the 23569 release of GCC 3.1.1. 23570 23571 The links below still apply to GCC 3.1.1. 23572 23573 May 15, 2002 23574 23575 The [1]GNU project and the GCC developers are pleased to announce the 23576 release of GCC 3.1. 23577 23578 GCC used to stand for the GNU C Compiler, but since the compiler 23579 supports several other languages aside from C, it now stands for the 23580 GNU Compiler Collection. 23581 23582 A list of [2]successful builds is updated as new information becomes 23583 available. 23584 23585 The GCC developers would like to thank the numerous people that have 23586 contributed [3]new features, improvements, bug fixes, and other changes 23587 as well as test results to GCC. This [4]amazing group of volunteers is 23588 what makes GCC successful. 23589 23590 For additional information about GCC please refer to the [5]GCC project 23591 web site or contact the [6]GCC development mailing list. 23592 23593 To obtain GCC please use [7]our mirror sites, or our CVS server. 23594 __________________________________________________________________ 23595 23596 23597 For questions related to the use of GCC, please consult these web 23598 pages and the [8]GCC manuals. If that fails, the 23599 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23600 web pages and the development of GCC are welcome on our developer 23601 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 23602 archives. 23603 23604 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 23605 distribution of this entire article is permitted in any medium, 23606 provided this notice is preserved. 23607 23608 These pages are [13]maintained by the GCC team. Last modified 23609 2024-07-04. 23610 23611 References 23612 23613 1. https://www.gnu.org/ 23614 2. https://gcc.gnu.org/gcc-3.1/buildstat.html 23615 3. https://gcc.gnu.org/gcc-3.1/changes.html 23616 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 23617 5. https://gcc.gnu.org/index.html 23618 6. mailto:gcc (a] gcc.gnu.org 23619 7. https://gcc.gnu.org/mirrors.html 23620 8. https://gcc.gnu.org/onlinedocs/ 23621 9. mailto:gcc-help (a] gcc.gnu.org 23622 10. mailto:gcc (a] gcc.gnu.org 23623 11. https://gcc.gnu.org/lists.html 23624 12. https://www.fsf.org/ 23625 13. https://gcc.gnu.org/about.html 23626 ====================================================================== 23627 http://gcc.gnu.org/gcc-3.1/changes.html 23628 GCC 3.1 Release Series 23629 Changes, New Features, and Fixes 23630 23631 Additional changes in GCC 3.1.1 23632 23633 * A bug related to how structures and unions are returned has been 23634 fixed for powerpc-*-netbsd*. 23635 * An important bug in the implementation of -fprefetch-loop-arrays 23636 has been fixed. Previously the optimization prefetched random 23637 blocks of memory for most targets except for i386. 23638 * The Java compiler now compiles Java programs much faster and also 23639 works with parallel make. 23640 * Nested functions have been fixed for mips*-*-netbsd*. 23641 * Some missing floating point support routines have beed added for 23642 mips*-*-netbsd*. 23643 * This [1]message gives additional information about the bugs fixed 23644 in this release. 23645 23646 Caveats 23647 23648 * The -traditional C compiler option has been deprecated and will be 23649 removed in GCC 3.3. (It remains possible to preprocess non-C code 23650 with the traditional preprocessor.) 23651 * The default debugging format for most ELF platforms (including 23652 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 23653 from stabs to DWARF2. This requires GDB 5.1.1 or later. 23654 23655 General Optimizer Improvements 23656 23657 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 23658 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 23659 for profile driven optimizations. 23660 Options -fprofile-arcs and -fbranch-probabilities can now be used 23661 to improve speed of the generated code by profiling the actual 23662 program behavior on typical runs. In the absence of profile info 23663 the compiler attempts to guess the profile statically. 23664 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 23665 monitor performance of the generated code. 23666 According to the SPECInt2000 results on an AMD Athlon CPU, the code 23667 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 23668 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 23669 is about 2.1% faster compared to 2.95.3. Tests were done using the 23670 -O2 -march=athlon command-line options. 23671 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 23672 infrastructure developed by CodeSourcery, LLC for the C++ front 23673 end, so that it is now used in the C front end too. Inlining 23674 functions as trees exposes them earlier to the compiler, giving it 23675 more opportunities for optimization. 23676 * Support for data prefetching instructions has been added to the GCC 23677 back end and several targets. A new __builtin_prefetch intrinsic is 23678 available to explicitly insert prefetch instructions and 23679 experimental support for loop array prefetching has been added (see 23680 -fprefetch-loop-array documentation). 23681 * Support for emitting debugging information for macros has been 23682 added for DWARF2. It is activated using -g3. 23683 23684 New Languages and Language specific improvements 23685 23686 C/C++ 23687 23688 * A few more [4]ISO C99 features. 23689 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 23690 * The preprocessor's symbol table has been merged with the symbol 23691 table of the C, C++ and Objective-C front ends. 23692 * The preprocessor consumes less memory than the preprocessor in GCC 23693 3.0, often significantly so. On normal input files, it typically 23694 consumes less memory than pre-3.0 cccp-based GCC, too. 23695 23696 C++ 23697 23698 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 23699 was a workaround to allow std compliant code to work with the 23700 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 23701 * The C++ ABI has been fixed so that void (A::*)() const is mangled 23702 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 23703 affects pointer to cv-qualified member function types. 23704 * The C++ ABI has been changed to correctly handle this code: 23705 struct A { 23706 void operator delete[] (void *, size_t); 23707 }; 23708 23709 struct B : public A { 23710 }; 23711 23712 new B[10]; 23713 23714 The amount of storage allocated for the array will be greater than 23715 it was in 3.0, in order to store the number of elements in the 23716 array, so that the correct size can be passed to operator delete[] 23717 when the array is deleted. Previously, the value passed to operator 23718 delete[] was unpredictable. 23719 This change will only affect code that declares a two-argument 23720 operator delete[] with a second parameter of type size_t in a base 23721 class, and does not override that definition in a derived class. 23722 * The C++ ABI has been changed so that: 23723 struct A { 23724 void operator delete[] (void *, size_t); 23725 void operator delete[] (void *); 23726 }; 23727 23728 does not cause unnecessary storage to be allocated when an array of 23729 A objects is allocated. 23730 This change will only affect code that declares both of these forms 23731 of operator delete[], and declared the two-argument form before the 23732 one-argument form. 23733 * The C++ ABI has been changed so that when a parameter is passed by 23734 value, any cleanup for that parameter is performed in the caller, 23735 as specified by the ia64 C++ ABI, rather than the called function 23736 as before. As a result, classes with a non-trivial destructor but a 23737 trivial copy constructor will be passed and returned by invisible 23738 reference, rather than by bitwise copy as before. 23739 * G++ now supports the "named return value optimization": for code 23740 like 23741 A f () { 23742 A a; 23743 ... 23744 return a; 23745 } 23746 23747 G++ will allocate a in the return value slot, so that the return 23748 becomes a no-op. For this to work, all return statements in the 23749 function must return the same variable. 23750 * Improvements to the C++ library are listed in [5]the libstdc++-v3 23751 FAQ. 23752 23753 Objective-C 23754 23755 * Annoying linker warnings (due to incorrect code being generated) 23756 have been fixed. 23757 * If a class method cannot be found, the compiler no longer issues a 23758 warning if a corresponding instance method exists in the root 23759 class. 23760 * Forward @protocol declarations have been fixed. 23761 * Loading of categories has been fixed in certain situations (GNU run 23762 time only). 23763 * The class lookup in the run-time library has been rewritten so that 23764 class method dispatch is more than twice as fast as it used to be 23765 (GNU run time only). 23766 23767 Java 23768 23769 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 23770 javax.transaction. 23771 * Property files and other system resources can be compiled into 23772 executables which use libgcj using the new gcj --resource feature. 23773 * libgcj has been ported to more platforms. In particular there is 23774 now a mostly-functional mingw32 (Windows) target port. 23775 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 23776 Java code can now be called from a C/C++ application. 23777 * gcj can now use builtin functions for certain known methods, for 23778 instance Math.cos. 23779 * gcj can now automatically remove redundant array-store checks in 23780 some common cases. 23781 * The --no-store-checks optimization option was added. This can be 23782 used to omit runtime store checks for code which is known not to 23783 throw ArrayStoreException 23784 * The following third party interface standards were added to libgcj: 23785 org.w3c.dom and org.xml.sax. 23786 * java.security has been merged with GNU Classpath. The new package 23787 is now JDK 1.2 compliant, and much more complete. 23788 * A bytecode verifier was added to the libgcj interpreter. 23789 * java.lang.Character was rewritten to comply with the Unicode 3.0 23790 standard, and improve performance. 23791 * Partial support for many more locales was added to libgcj. 23792 * Socket timeouts have been implemented. 23793 * libgcj has been merged into a single shared library. There are no 23794 longer separate shared libraries for the garbage collector and 23795 zlib. 23796 * Several performance improvements were made to gcj and libgcj: 23797 + Hash synchronization (thin locks) 23798 + A special allocation path for finalizer-free objects 23799 + Thread-local allocation 23800 + Parallel GC, and other GC tweaks 23801 23802 Fortran 23803 23804 Fortran improvements are listed in [6]the Fortran documentation. 23805 23806 Ada 23807 23808 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 23809 tools. The GNAT compiler fully implements the Ada language as defined 23810 by the ISO/IEC 8652 standard. 23811 23812 Please note that the integration of the Ada front end is still work in 23813 progress. 23814 23815 New Targets and Target Specific Improvements 23816 23817 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 23818 architecture used in new editions of Donald E. Knuth's The Art of 23819 Computer Programming. 23820 * Axis Communications has contributed its port to the CRIS CPU 23821 architecture, used in the ETRAX system-on-a-chip series. 23822 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 23823 SH5 64-bit RISC microprocessor architecture, extending the existing 23824 SH port. 23825 * UltraSPARC is fully supported in 64-bit mode. The option -m64 23826 enables it. 23827 * For compatibility with the Sun compiler #pragma redefine_extname 23828 has been implemented on Solaris. 23829 * The x86 back end has had some noticeable work done to it. 23830 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 23831 Jaeger have contributed a port to the AMD x86-64 architecture. 23832 For more information on x86-64 see http://www.x86-64.org. 23833 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 23834 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 23835 enable the respective instruction sets. Intel C++ compatible 23836 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 23837 will be added in next major release. 23838 + Following those improvements, targets for Pentium MMX, K6-2, 23839 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 23840 added. Refer to the documentation on -march= and -mcpu= 23841 options for details. 23842 + For those targets that support it, -mfpmath=sse will cause the 23843 compiler to generate SSE/SSE2 instructions for floating point 23844 math instead of x87 instructions. Usually, this will lead to 23845 quicker code especially on the Pentium 4. Note that only 23846 scalar floating point instructions are used and GCC does not 23847 exploit SIMD features yet. 23848 + Prefetch support has been added to the Pentium III, Pentium 4, 23849 K6-2, K6-3, and Athlon series. 23850 + Code generated for floating point to integer conversions has 23851 been improved leading to better performance of many 3D 23852 applications. 23853 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 23854 * C++ support for AIX has been improved. 23855 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 23856 PowerPC port supporting the AltiVec programming model (SIMD). The 23857 support, though presently useful, is experimental and is expected 23858 to stabilize for 3.2. The support is written to conform to 23859 Motorola's AltiVec specs. See -maltivec. 23860 23861 [8]Obsolete Systems 23862 23863 Support for a number of older systems has been declared obsolete in GCC 23864 3.1. Unless there is activity to revive them, the next release of GCC 23865 will have their sources permanently removed. 23866 23867 All configurations of the following processor architectures have been 23868 declared obsolete: 23869 * MIL-STD-1750A, 1750a-*-* 23870 * AMD A29k, a29k-*-* 23871 * Convex, c*-convex-* 23872 * Clipper, clipper-*-* 23873 * Elxsi, elxsi-*-* 23874 * Intel i860, i860-*-* 23875 * Sun picoJava, pj-*-* and pjl-*-* 23876 * Western Electric 32000, we32k-*-* 23877 23878 Most configurations of the following processor architectures have been 23879 declared obsolete, but we are preserving a few systems which may have 23880 active developers. It is unlikely that the remaining systems will 23881 survive much longer unless we see definite signs of port activity. 23882 * Motorola 88000 except 23883 + Generic a.out, m88k-*-aout* 23884 + Generic SVR4, m88k-*-sysv4 23885 + OpenBSD, m88k-*-openbsd* 23886 * NS32k except 23887 + NetBSD, ns32k-*-netbsd* 23888 + OpenBSD, ns32k-*-openbsd*. 23889 * ROMP except 23890 + OpenBSD, romp-*-openbsd*. 23891 23892 Finally, only some configurations of these processor architectures are 23893 being obsoleted. 23894 * Alpha: 23895 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 23896 alpha*-*-osf[45], are still supported.) 23897 * ARM: 23898 + RISCiX, arm-*-riscix*. 23899 * i386: 23900 + 386BSD, i?86-*-bsd* 23901 + Chorus, i?86-*-chorusos* 23902 + DG/UX, i?86-*-dgux* 23903 + FreeBSD 1.x, i?86-*-freebsd1.* 23904 + IBM AIX, i?86-*-aix* 23905 + ISC UNIX, i?86-*-isc* 23906 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 23907 + NEXTstep, i?86-next-* 23908 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 23909 + RTEMS/coff, i?86-*-rtemscoff* 23910 + RTEMS/go32, i?86-go32-rtems* 23911 + Sequent/BSD, i?86-sequent-bsd* 23912 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 23913 i?86-sequent-sysv3* 23914 + SunOS, i?86-*-sunos* 23915 * Motorola 68000: 23916 + Altos, m68[k0]*-altos-* 23917 + Apollo, m68[k0]*-apollo-* 23918 + Apple A/UX, m68[k0]*-apple-* 23919 + Bull, m68[k0]*-bull-* 23920 + Convergent, m68[k0]*-convergent-* 23921 + Generic SVR3, m68[k0]*-*-sysv3* 23922 + ISI, m68[k0]*-isi-* 23923 + LynxOS, m68[k0]*-*-lynxos* 23924 + NEXT, m68[k0]*-next-* 23925 + RTEMS/coff, m68[k0]*-*-rtemscoff* 23926 + Sony, m68[k0]*-sony-* 23927 * MIPS: 23928 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 23929 + Generic BSD, mips-*-bsd* 23930 + Generic System V, mips-*-sysv* 23931 + IRIX before version 5, mips-sgi-irix[1234]* 23932 + RiscOS, mips-*-riscos* 23933 + Sony, mips-sony-* 23934 + Tandem, mips-tandem-* 23935 * SPARC: 23936 + RTEMS/a.out, sparc-*-rtemsaout*. 23937 23938 Documentation improvements 23939 23940 * The old manual ("Using and Porting the GNU Compiler Collection") 23941 has been replaced by a users manual ("Using the GNU Compiler 23942 Collection") and a separate internals reference manual ("GNU 23943 Compiler Collection Internals"). 23944 * More complete and much improved documentation about GCC's internal 23945 representation used by the C and C++ front ends. 23946 * Many cleanups and improvements in general. 23947 23948 23949 For questions related to the use of GCC, please consult these web 23950 pages and the [9]GCC manuals. If that fails, the 23951 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23952 web pages and the development of GCC are welcome on our developer 23953 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 23954 archives. 23955 23956 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 23957 distribution of this entire article is permitted in any medium, 23958 provided this notice is preserved. 23959 23960 These pages are [14]maintained by the GCC team. Last modified 23961 2026-06-07. 23962 23963 References 23964 23965 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 23966 2. https://gcc.gnu.org/news/profiledriven.html 23967 3. https://gcc.gnu.org/benchmarks/ 23968 4. https://gcc.gnu.org/projects/c-status.html 23969 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 23970 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 23971 7. https://www.adacore.com/ 23972 8. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 23973 9. https://gcc.gnu.org/onlinedocs/ 23974 10. mailto:gcc-help (a] gcc.gnu.org 23975 11. mailto:gcc (a] gcc.gnu.org 23976 12. https://gcc.gnu.org/lists.html 23977 13. https://www.fsf.org/ 23978 14. https://gcc.gnu.org/about.html 23979 ====================================================================== 23980 http://gcc.gnu.org/gcc-3.0/index.html 23981 GCC 3.0.4 23982 23983 (This release series is no longer supported.) 23984 23985 February 20, 2002 23986 23987 The [1]GNU project and the GCC developers are pleased to announce the 23988 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 23989 series. 23990 23991 GCC used to stand for the GNU C Compiler, but since the compiler 23992 supports several other languages aside from C, it now stands for the 23993 GNU Compiler Collection. 23994 23995 GCC 3.0.x has several new optimizations, new targets, new languages and 23996 many other new features, relative to GCC 2.95.x. See the [2]new 23997 features page for a more complete list. 23998 23999 A list of [3]successful builds is updated as new information becomes 24000 available. 24001 24002 The GCC developers would like to thank the numerous people that have 24003 contributed new features, test results, bug fixes, etc to GCC. This 24004 [4]amazing group of volunteers is what makes GCC successful. 24005 24006 And finally, we can't in good conscience fail to mention some 24007 [5]caveats to using GCC 3.0.x. 24008 24009 For additional information about GCC please refer to the [6]GCC project 24010 web site or contact the [7]GCC development mailing list. 24011 24012 To obtain GCC please use [8]our mirror sites, or our CVS server. 24013 __________________________________________________________________ 24014 24015 Previous 3.0.x Releases 24016 24017 December 20, 2001: GCC 3.0.3 has been released. 24018 October 25, 2001: GCC 3.0.2 has been released. 24019 August 20, 2001: GCC 3.0.1 has been released. 24020 June 18, 2001: GCC 3.0 has been released. 24021 24022 24023 For questions related to the use of GCC, please consult these web 24024 pages and the [9]GCC manuals. If that fails, the 24025 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24026 web pages and the development of GCC are welcome on our developer 24027 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 24028 archives. 24029 24030 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 24031 distribution of this entire article is permitted in any medium, 24032 provided this notice is preserved. 24033 24034 These pages are [14]maintained by the GCC team. Last modified 24035 2024-07-04. 24036 24037 References 24038 24039 1. https://www.gnu.org/ 24040 2. https://gcc.gnu.org/gcc-3.0/features.html 24041 3. https://gcc.gnu.org/gcc-3.0/buildstat.html 24042 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24043 5. https://gcc.gnu.org/gcc-3.0/caveats.html 24044 6. https://gcc.gnu.org/index.html 24045 7. mailto:gcc (a] gcc.gnu.org 24046 8. https://gcc.gnu.org/mirrors.html 24047 9. https://gcc.gnu.org/onlinedocs/ 24048 10. mailto:gcc-help (a] gcc.gnu.org 24049 11. mailto:gcc (a] gcc.gnu.org 24050 12. https://gcc.gnu.org/lists.html 24051 13. https://www.fsf.org/ 24052 14. https://gcc.gnu.org/about.html 24053 ====================================================================== 24054 http://gcc.gnu.org/gcc-3.0/features.html 24055 GCC 3.0 New Features 24056 24057 Additional changes in GCC 3.0.4 24058 24059 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 24060 system, which use the ELF object file format, on x86 processors. 24061 * Correct debugging information is generated from functions that have 24062 lines from multiple files (e.g. yacc output). 24063 * A fix for whitespace handling in the -traditional preprocessor, 24064 which can affect Fortran. 24065 * Fixes to the exception handling runtime. 24066 * More fixes for bad code generation in C++. 24067 * A fix for shared library generation under AIX 4.3. 24068 * Documentation updates. 24069 * Port of GCC to Tensilica's Xtensa processor contributed. 24070 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 24071 24072 Additional changes in GCC 3.0.3 24073 24074 * A fix to correct an accidental change to the PowerPC ABI. 24075 * Fixes for bad code generation on a variety of architectures. 24076 * Improvements to the debugging information generated for C++ 24077 classes. 24078 * Fixes for bad code generation in C++. 24079 * A fix to avoid crashes in the C++ demangler. 24080 * A fix to the C++ standard library to avoid buffer overflows. 24081 * Miscellaneous improvements for a variety of architectures. 24082 24083 Additional changes in GCC 3.0.2 24084 24085 * Fixes for bad code generation during loop unrolling. 24086 * Fixes for bad code generation by the sibling call optimization. 24087 * Minor improvements to x86 code generation. 24088 * Implementation of function descriptors in C++ vtables for IA64. 24089 * Numerous minor bug-fixes. 24090 24091 Additional changes in GCC 3.0.1 24092 24093 * C++ fixes for incorrect code-generation. 24094 * Improved cross-compiling support for the C++ standard library. 24095 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 24096 in GCC 3.0. 24097 * Fixes for various exception-handling bugs. 24098 * A port to the S/390 architecture. 24099 24100 General Optimizer Improvements 24101 24102 * [2]Basic block reordering pass. 24103 * New if-conversion pass with support for conditional (predicated) 24104 execution. 24105 * New tail call and sibling call elimination optimizations. 24106 * New register renaming pass. 24107 * New (experimental) [3]static single assignment (SSA) representation 24108 support. 24109 * New dead-code elimination pass implemented using the SSA 24110 representation. 24111 * [4]Global null pointer test elimination. 24112 * [5]Global code hoisting/unification. 24113 * More builtins and optimizations for stdio.h, string.h and old BSD 24114 functions, as well as for ISO C99 functions. 24115 * New builtin __builtin_expect for giving hints to the branch 24116 predictor. 24117 24118 New Languages and Language specific improvements 24119 24120 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 24121 and supported, including the run-time library containing most 24122 common non-GUI Java classes, a bytecode interpreter, and the Boehm 24123 conservative garbage collector. Many bugs have been fixed. GCJ can 24124 compile Java source or Java bytecodes to either native code or Java 24125 class files, and supports native methods written in either the 24126 standard JNI or the more efficient and convenient CNI. 24127 * Here is a [6]partial list of C++ improvements, both new features 24128 and those no longer supported. 24129 * New C++ ABI. On the IA-64 platform GCC is capable of 24130 inter-operating with other IA-64 compilers. 24131 * The new ABI also significantly reduces the size of symbol and debug 24132 information. 24133 * New C++ support library and many C++ bug fixes, vastly improving 24134 our conformance to the ISO C++ standard. 24135 * New [7]inliner for C++. 24136 * Rewritten C preprocessor, integrated into the C, C++ and 24137 Objective-C compilers, with very many improvements including ISO 24138 C99 support and [8]improvements to dependency generation. 24139 * Support for more [9]ISO C99 features. 24140 * Many improvements to support for checking calls to format functions 24141 such as printf and scanf, including support for ISO C99 format 24142 features, extensions from the Single Unix Specification and GNU 24143 libc 2.2, checking of strfmon formats and features to assist in 24144 auditing for format string security bugs. 24145 * New warnings for C code that may have undefined semantics because 24146 of violations of sequence point rules in the C standard (such as a 24147 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 24148 * Additional warning option -Wfloat-equal. 24149 * Improvements to -Wtraditional. 24150 * Fortran improvements are listed in [10]the Fortran documentation. 24151 24152 New Targets and Target Specific Improvements 24153 24154 * New x86 back end, generating much improved code. 24155 * Support for a generic i386-elf target contributed. 24156 * New option to emit x86 assembly code using Intel style syntax 24157 (-mintel-syntax). 24158 * HPUX 11 support contributed. 24159 * Improved PowerPC code generation, including scheduled prologue and 24160 epilogue. 24161 * Port of GCC to Intel's IA-64 processor contributed. 24162 * Port of GCC to Motorola's MCore 210 and 340 contributed. 24163 * New unified back-end for Arm, Thumb and StrongArm contributed. 24164 * Port of GCC to Intel's XScale processor contributed. 24165 * Port of GCC to Atmel's AVR microcontrollers contributed. 24166 * Port of GCC to Mitsubishi's D30V processor contributed. 24167 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 24168 processor family) contributed. 24169 * Port of GCC to Fujitsu's FR30 processor contributed. 24170 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 24171 * Port of GCC to Sun's picoJava processor core contributed. 24172 24173 Documentation improvements 24174 24175 * Substantially rewritten and improved C preprocessor manual. 24176 * Many improvements to other documentation. 24177 * Manpages for gcc, cpp and gcov are now generated automatically from 24178 the master Texinfo manual, eliminating the problem of manpages 24179 being out of date. (The generated manpages are only extracts from 24180 the full manual, which is provided in Texinfo form, from which 24181 info, HTML, other formats and a printed manual can be generated.) 24182 * Generated info files are included in the release tarballs alongside 24183 their Texinfo sources, avoiding problems on some platforms with 24184 building makeinfo as part of the GCC distribution. 24185 24186 Other significant improvements 24187 24188 * Garbage collection used internally by the compiler for most memory 24189 allocation instead of obstacks. 24190 * Lengauer and Tarjan algorithm used for computing dominators in the 24191 CFG. This algorithm can be significantly faster and more space 24192 efficient than our older algorithm. 24193 * gccbug script provided to assist in submitting bug reports to our 24194 bug tracking system. (Bug reports previously submitted directly to 24195 our mailing lists, for which you received no bug tracking number, 24196 should be submitted again using gccbug if you can reproduce the 24197 problem with GCC 3.0.) 24198 * The internal libgcc library is [11]built as a shared library on 24199 systems that support it. 24200 * Extensive testsuite included with GCC, with many new tests. In 24201 addition to tests for GCC bugs that have been fixed, many tests 24202 have been added for language features, compiler warnings and 24203 builtin functions. 24204 * Additional language-independent warning options -Wpacked, -Wpadded, 24205 -Wunreachable-code and -Wdisabled-optimization. 24206 * Target-independent options -falign-functions, -falign-loops and 24207 -falign-jumps. 24208 24209 Plus a great many bug fixes and almost all the [12]features found in 24210 GCC 2.95. 24211 24212 24213 For questions related to the use of GCC, please consult these web 24214 pages and the [13]GCC manuals. If that fails, the 24215 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24216 web pages and the development of GCC are welcome on our developer 24217 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 24218 archives. 24219 24220 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 24221 distribution of this entire article is permitted in any medium, 24222 provided this notice is preserved. 24223 24224 These pages are [18]maintained by the GCC team. Last modified 24225 2026-01-02. 24226 24227 References 24228 24229 1. http://www.netbsd.org/ 24230 2. https://gcc.gnu.org/news/reorder.html 24231 3. https://gcc.gnu.org/news/ssa.html 24232 4. https://gcc.gnu.org/news/null.html 24233 5. https://gcc.gnu.org/news/unify.html 24234 6. https://gcc.gnu.org/gcc-3.0/c++features.html 24235 7. https://gcc.gnu.org/news/inlining.html 24236 8. https://gcc.gnu.org/news/dependencies.html 24237 9. https://gcc.gnu.org/projects/c-status.html 24238 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 24239 11. https://gcc.gnu.org/gcc-3.0/libgcc.html 24240 12. https://gcc.gnu.org/gcc-2.95/features.html 24241 13. https://gcc.gnu.org/onlinedocs/ 24242 14. mailto:gcc-help (a] gcc.gnu.org 24243 15. mailto:gcc (a] gcc.gnu.org 24244 16. https://gcc.gnu.org/lists.html 24245 17. https://www.fsf.org/ 24246 18. https://gcc.gnu.org/about.html 24247 ====================================================================== 24248 http://gcc.gnu.org/gcc-3.0/caveats.html 24249 GCC 3.0 Caveats 24250 24251 * -fstrict-aliasing is now part of -O2 and higher optimization 24252 levels. This allows the compiler to assume the strictest aliasing 24253 rules applicable to the language being compiled. For C and C++, 24254 this activates optimizations based on the type of expressions. This 24255 optimization may thus break old, non-compliant code. 24256 * Enumerations are now properly promoted to int in function 24257 parameters and function returns. Normally this change is not 24258 visible, but when using -fshort-enums this is an ABI change. 24259 * The undocumented extension that allowed C programs to have a label 24260 at the end of a compound statement has been deprecated and may be 24261 removed in a future version. Programs that now generate a warning 24262 about this may be fixed by adding a null statement (a single 24263 semicolon) after the label. 24264 * The poorly documented extension that allowed string constants in C, 24265 C++ and Objective-C to contain unescaped newlines has been 24266 deprecated and may be removed in a future version. Programs using 24267 this extension may be fixed in several ways: the bare newline may 24268 be replaced by \n, or preceded by \n\, or string concatenation may 24269 be used with the bare newline preceded by \n" and " placed at the 24270 start of the next line. 24271 * The Chill compiler is not included in GCC 3.0, because of the lack 24272 of a volunteer to convert it to use garbage collection. 24273 * Certain non-standard iostream methods from earlier versions of 24274 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 24275 ostream::form, and istream::gets. 24276 * The new C++ ABI is not yet fully supported by current (as of 24277 2001-07-01) releases and development versions of GDB, or any 24278 earlier versions. There is a problem setting breakpoints by line 24279 number, and other related issues that have been fixed in GCC 3.0 24280 but not yet handled in GDB: 24281 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 24282 24283 24284 For questions related to the use of GCC, please consult these web 24285 pages and the [2]GCC manuals. If that fails, the 24286 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24287 web pages and the development of GCC are welcome on our developer 24288 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 24289 24290 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 24291 distribution of this entire article is permitted in any medium, 24292 provided this notice is preserved. 24293 24294 These pages are [7]maintained by the GCC team. Last modified 24295 2026-01-02. 24296 24297 References 24298 24299 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 24300 2. https://gcc.gnu.org/onlinedocs/ 24301 3. mailto:gcc-help (a] gcc.gnu.org 24302 4. mailto:gcc (a] gcc.gnu.org 24303 5. https://gcc.gnu.org/lists.html 24304 6. https://www.fsf.org/ 24305 7. https://gcc.gnu.org/about.html 24306 ====================================================================== 24307 http://gcc.gnu.org/gcc-2.95/index.html 24308 GCC 2.95 24309 24310 (This release series is no longer supported.) 24311 24312 March 16, 2001: The GNU project and the GCC developers are pleased to 24313 announce the release of GCC version 2.95.3. 24314 24315 Release History 24316 24317 GCC 2.95.3 24318 March 16, 2001 24319 24320 GCC 2.95.2 24321 October 27, 1999 24322 24323 GCC 2.95.1 24324 August 19, 1999 24325 24326 GCC 2.95 24327 July 31, 1999. This is the first release of GCC since the April 24328 1999 GCC/EGCS reunification and includes nearly a year's worth 24329 of new development and bugfixes. 24330 24331 References and Acknowledgements 24332 24333 GCC used to stand for the GNU C Compiler, but since the compiler 24334 supports several other languages aside from C, it now stands for the 24335 GNU Compiler Collection. 24336 24337 The whole suite has been extensively [1]regression tested and 24338 [2]package tested. It should be reliable and suitable for widespread 24339 use. 24340 24341 The compiler has several new optimizations, new targets, new languages 24342 and other new features. See the [3]new features page for a more 24343 complete list of new features found in the GCC 2.95 releases. 24344 24345 The sources include installation instructions in both HTML and 24346 plaintext forms in the install directory in the distribution. However, 24347 the most up to date installation instructions and [4]build/test status 24348 are on the web pages. We will update those pages as new information 24349 becomes available. 24350 24351 The GCC developers would like to thank the numerous people that have 24352 contributed new features, test results, bugfixes, etc to GCC. This 24353 [5]amazing group of volunteers is what makes GCC successful. 24354 24355 And finally, we can't in good conscience fail to mention some 24356 [6]caveats to using GCC 2.95. 24357 24358 Download GCC 2.95 from one of our many [7]mirror sites. 24359 24360 For additional information about GCC please see the [8]GCC project web 24361 server or contact the [9]GCC development mailing list. 24362 24363 24364 For questions related to the use of GCC, please consult these web 24365 pages and the [10]GCC manuals. If that fails, the 24366 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24367 web pages and the development of GCC are welcome on our developer 24368 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 24369 archives. 24370 24371 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 24372 distribution of this entire article is permitted in any medium, 24373 provided this notice is preserved. 24374 24375 These pages are [15]maintained by the GCC team. Last modified 24376 2022-10-26. 24377 24378 References 24379 24380 1. https://gcc.gnu.org/gcc-2.95/regress.html 24381 2. https://gcc.gnu.org/gcc-2.95/othertest.html 24382 3. https://gcc.gnu.org/gcc-2.95/features.html 24383 4. https://gcc.gnu.org/gcc-2.95/buildstat.html 24384 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24385 6. https://gcc.gnu.org/gcc-2.95/caveats.html 24386 7. https://gcc.gnu.org/mirrors.html 24387 8. https://gcc.gnu.org/index.html 24388 9. mailto:gcc (a] gcc.gnu.org 24389 10. https://gcc.gnu.org/onlinedocs/ 24390 11. mailto:gcc-help (a] gcc.gnu.org 24391 12. mailto:gcc (a] gcc.gnu.org 24392 13. https://gcc.gnu.org/lists.html 24393 14. https://www.fsf.org/ 24394 15. https://gcc.gnu.org/about.html 24395 ====================================================================== 24396 http://gcc.gnu.org/gcc-2.95/features.html 24397 GCC 2.95 New Features 24398 24399 * General Optimizer Improvements: 24400 + [1]Localized register spilling to improve speed and code 24401 density especially on small register class machines. 24402 + [2]Global CSE using lazy code motion algorithms. 24403 + [3]Improved global constant/copy propagation. 24404 + [4]Improved control flow graph analysis and manipulation. 24405 + [5]Local dead store elimination. 24406 + [6]Memory Load hoisting/store sinking in loops. 24407 + [7]Type based alias analysis is enabled by default. Note this 24408 feature will expose bugs in the Linux kernel. Please refer to 24409 the FAQ (as shipped with GCC 2.95) for additional information 24410 on this issue. 24411 + Major revamp of GIV detection, combination and simplification 24412 to improve loop performance. 24413 + Major improvements to register allocation and reloading. 24414 * New Languages and Language specific improvements 24415 + [8]Many C++ improvements. 24416 + [9]Many Fortran improvements. 24417 + [10]Java front-end has been integrated. A [11]runtime library 24418 is available separately. 24419 + [12]ISO C99 support 24420 + [13]Chill front-end and runtime has been integrated. 24421 + Boehm garbage collector support in libobjc. 24422 + More support for various pragmas which appear in vendor 24423 include files 24424 * New Targets and Target Specific Improvements 24425 + [14]SPARC back end rewrite. 24426 + -mschedule=8000 will optimize code for PA8000 class 24427 processors; -mpa-risc-2-0 will generate code for PA2.0 24428 processors 24429 + Various micro-optimizations for the ia32 port. K6 24430 optimizations 24431 + Compiler will attempt to align doubles in the stack on the 24432 ia32 port 24433 + Alpha EV6 support 24434 + PowerPC 750 24435 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 24436 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 24437 + c3x, c4x 24438 + HyperSPARC 24439 + SparcLite86x 24440 + sh4 24441 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 24442 arm-linux) 24443 + VxWorks targets include support for VxWorks threads 24444 + StrongARM 110 and ARM9 support added. ARM Scheduling 24445 parameters rewritten. 24446 + Various changes to the MIPS port to avoid assembler macros, 24447 which in turn improves performance 24448 + Various performance improvements to the i960 port. 24449 + Major rewrite of ns32k port 24450 * Other significant improvements 24451 + [15]Ability to dump cfg information and display it using vcg. 24452 + The new faster scheme for fixing vendor header files is 24453 enabled by default. 24454 + Experimental internationalization support. 24455 + multibyte character support 24456 + Some compile-time speedups for pathological problems 24457 + Better support for complex types 24458 * Plus the usual mountain of bugfixes 24459 * Core compiler is based on the gcc2 development tree from Sept 30, 24460 1998, so we have all of the [16]features found in GCC 2.8. 24461 24462 Additional Changes in GCC 2.95.1 24463 24464 * Generic bugfixes and improvements 24465 + Various documentation fixes related to the GCC/EGCS merger. 24466 + Fix memory management bug which could lead to spurious aborts, 24467 core dumps or random parsing errors in the compiler. 24468 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 24469 support. 24470 + Fix infinite loop in the CSE optimizer. 24471 + Avoid undefined behavior in compiler FP emulation code 24472 + Fix install problem when prefix is overridden on the make 24473 install command. 24474 + Fix problem with unwanted installation of assert.h on some 24475 systems. 24476 + Fix problem with finding the wrong assembler in a single tree 24477 build. 24478 + Avoid increasing the known alignment of a register that is 24479 already known to be a pointer. 24480 * Platform specific bugfixes and improvements 24481 + Codegen bugfix for prologue/epilogue for cpu32 target. 24482 + Fix long long code generation bug for the Coldfire target. 24483 + Fix various aborts in the SH compiler. 24484 + Fix bugs in libgcc support library for the SH. 24485 + Fix alpha ev6 code generation bug. 24486 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 24487 AIX platforms. 24488 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 24489 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 24490 targets. 24491 + Fix weak symbol handling for rs6000/ppc svr4 targets. 24492 + Fix various problems with 64bit code generation for the 24493 rs6000/ppc port. 24494 + Fix codegen bug which caused tetex to be mis-compiled on the 24495 x86. 24496 + Fix compiler abort in new cfg code exposed by x86 port. 24497 + Fix out of range array reference in code convert flat 24498 registers to the x87 stacked FP register file. 24499 + Fix minor VxWorks configuration bug. 24500 + Fix return type of bsearch for SunOS 4.x. 24501 * Language & Runtime specific fixes. 24502 + The G++ signature extension has been deprecated. It will be 24503 removed in the next major release of G++. Use of signatures 24504 will result in a warning from the compiler. 24505 + Several bugs relating to templates and namespaces were fixed. 24506 + A bug that caused crashes when combining templates with -g on 24507 DWARF1 platforms was fixed. 24508 + Pointers-to-members, virtual functions, and multiple 24509 inheritance should now work together correctly. 24510 + Some code-generation bugs relating to function try blocks were 24511 fixed. 24512 + G++ is a little bit more lenient with certain archaic 24513 constructs than in GCC 2.95. 24514 + Fix to prevent shared library version #s from bring truncated 24515 to 1 digit 24516 + Fix missing std:: in the libstdc++ library. 24517 + Fix stream locking problems in libio. 24518 + Fix problem in java compiler driver. 24519 24520 Additional Changes in GCC 2.95.2 24521 24522 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 24523 the optimizations performed by -fstrict-aliasing are valid according to 24524 the C and C++ standards, the optimization have caused some problems, 24525 particularly with old non-conforming code. 24526 24527 The GCC developers are experimenting with ways to warn users about code 24528 which violates the C/C++ standards, but those warnings are not ready 24529 for widespread use at this time. Rather than wait for those warnings 24530 the GCC developers have chosen to disable -fstrict-aliasing by default 24531 for the GCC 2.95.2 release. 24532 24533 We strongly encourage developers to find and fix code which violates 24534 the C/C++ standards as -fstrict-aliasing may be enabled by default in 24535 future releases. Use the option -fstrict-aliasing to re-enable these 24536 optimizations. 24537 * Generic bugfixes and improvements 24538 + Fix incorrectly optimized memory reference in global common 24539 subexpression elimination (GCSE) optimization pass. 24540 + Fix code generation bug in regmove.c in which it could 24541 incorrectly change a "const" value. 24542 + Fix bug in optimization of conditionals involving volatile 24543 memory references. 24544 + Avoid over-allocation of stack space for some procedures. 24545 + Fixed bug in the compiler which caused incorrect optimization 24546 of an obscure series of bit manipulations, shifts and 24547 arithmetic. 24548 + Fixed register allocator bug which caused teTeX to be 24549 mis-compiled on SPARC targets. 24550 + Avoid incorrect optimization of degenerate case statements for 24551 certain targets such as the ARM. 24552 + Fix out of range memory reference in the jump optimizer. 24553 + Avoid dereferencing null pointer in fix-header. 24554 + Fix test for GCC specific features so that it is possible to 24555 bootstrap with gcc-2.6.2 and older versions of GCC. 24556 + Fix typo in scheduler which could potentially cause out of 24557 range memory accesses. 24558 + Avoid incorrect loop reversal which caused incorrect code for 24559 certain loops on PowerPC targets. 24560 + Avoid incorrect optimization of switch statements on certain 24561 targets (for example the ARM). 24562 * Platform specific bugfixes and improvements 24563 + Work around bug in Sun V5.0 compilers which caused bootstrap 24564 comparison failures on SPARC targets. 24565 + Fix SPARC back end bug which caused aborts in final.c. 24566 + Fix sparc-hal-solaris2* configuration fragments. 24567 + Fix bug in sparc block profiling. 24568 + Fix obscure code generation bug for the PARISC targets. 24569 + Define __STDC_EXT__ for HPUX configurations. 24570 + Various POWERPC64 code generation bugfixes. 24571 + Fix abort for PPC targets using ELF (ex GNU/Linux). 24572 + Fix collect2 problems for AIX targets. 24573 + Correct handling of .file directive for PPC targets. 24574 + Fix bug in fix_trunc x86 patterns. 24575 + Fix x86 port to correctly pop the FP stack for functions that 24576 return structures in memory. 24577 + Fix minor bug in strlen x86 pattern. 24578 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 24579 + Fix template repository code to handle leading underscore in 24580 mangled names. 24581 + Fix weak/weak alias support for OpenBSD. 24582 + GNU/Linux for the ARM has C++ compatible include files. 24583 * Language & Runtime specific fixes. 24584 + Fix handling of constructor attribute in the C front-end which 24585 caused problems building the Chill runtime library on some 24586 targets. 24587 + Fix minor problem merging type qualifiers in the C front-end. 24588 + Fix aliasing bug for pointers and references (C/C++). 24589 + Fix incorrect "non-constant initializer bug" when -traditional 24590 or -fwritable-strings is enabled. 24591 + Fix build error for Chill front-end on SunOS. 24592 + Do not complain about duplicate instantiations when using 24593 -frepo (C++). 24594 + Fix array bounds handling in C++ front-end which caused 24595 problems with dwarf debugging information in some 24596 circumstances. 24597 + Fix minor namespace problem. 24598 + Fix problem linking java programs. 24599 24600 Additional Changes in GCC 2.95.3 24601 24602 * Generic bugfixes and improvements 24603 + Fix numerous problems that caused incorrect optimization in 24604 the register reloading code. 24605 + Fix numerous problems that caused incorrect optimization in 24606 the loop optimizer. 24607 + Fix aborts in the functions build_insn_chain and scan_loops 24608 under some circumstances. 24609 + Fix an alias analysis bug. 24610 + Fix an infinite compilation bug in the combiner. 24611 + A few problems with complex number support have been fixed. 24612 + It is no longer possible for gcc to act as a fork bomb when 24613 installed incorrectly. 24614 + The -fpack-struct option should be recognized now. 24615 + Fixed a bug that caused incorrect code to be generated due to 24616 a lost stack adjustment. 24617 * Platform specific bugfixes and improvements 24618 + Support building ARM toolchains hosted on Windows. 24619 + Fix attribute calculations in ARM toolchains. 24620 + arm-linux support has been improved. 24621 + Fix a PIC failure on sparc targets. 24622 + On ix86 targets, the regparm attribute should now work 24623 reliably. 24624 + Several updates for the h8300 port. 24625 + Fix problem building libio with glibc 2.2. 24626 24627 24628 For questions related to the use of GCC, please consult these web 24629 pages and the [17]GCC manuals. If that fails, the 24630 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24631 web pages and the development of GCC are welcome on our developer 24632 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 24633 archives. 24634 24635 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 24636 distribution of this entire article is permitted in any medium, 24637 provided this notice is preserved. 24638 24639 These pages are [22]maintained by the GCC team. Last modified 24640 2025-10-08. 24641 24642 References 24643 24644 1. https://gcc.gnu.org/news/spill.html 24645 2. https://gcc.gnu.org/news/lcm.html 24646 3. https://gcc.gnu.org/news/cprop.html 24647 4. https://gcc.gnu.org/news/cfg.html 24648 5. https://gcc.gnu.org/news/dse.html 24649 6. https://gcc.gnu.org/news/hoist.html 24650 7. https://gcc.gnu.org/news/alias.html 24651 8. https://gcc.gnu.org/gcc-2.95/c++features.html 24652 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 24653 10. https://gcc.gnu.org/news/gcj-announce.txt 24654 11. https://gcc.gnu.org/news/javaannounce.html 24655 12. https://gcc.gnu.org/projects/c-status.html 24656 13. https://gcc.gnu.org/news/chill.html 24657 14. https://gcc.gnu.org/news/sparc.html 24658 15. https://gcc.gnu.org/news/egcs-vcg.html 24659 16. https://gcc.gnu.org/egcs-1.0/features-2.8.html 24660 17. https://gcc.gnu.org/onlinedocs/ 24661 18. mailto:gcc-help (a] gcc.gnu.org 24662 19. mailto:gcc (a] gcc.gnu.org 24663 20. https://gcc.gnu.org/lists.html 24664 21. https://www.fsf.org/ 24665 22. https://gcc.gnu.org/about.html 24666 ====================================================================== 24667 http://gcc.gnu.org/gcc-2.95/caveats.html 24668 GCC 2.95 Caveats 24669 24670 * GCC 2.95 will issue an error for invalid asm statements that had 24671 been silently accepted by earlier versions of the compiler. This is 24672 particularly noticeable when compiling older versions of the Linux 24673 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 24674 for more information on this issue. 24675 * GCC 2.95 implements type based alias analysis to disambiguate 24676 memory references. Some programs, particularly the Linux kernel 24677 violate ANSI/ISO aliasing rules and therefore may not operate 24678 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 24679 shipped with GCC 2.95) for more information on this issue. 24680 * GCC 2.95 has a known bug in its handling of complex variables for 24681 64bit targets. Instead of silently generating incorrect code, GCC 24682 2.95 will issue a fatal error for situations it can not handle. 24683 This primarily affects the Fortran community as Fortran makes more 24684 use of complex variables than C or C++. 24685 * GCC 2.95 has an integrated libstdc++, but does not have an 24686 integrated libg++. Furthermore old libg++ releases will not work 24687 with GCC 2.95. You can retrieve a recent copy of libg++ from the 24688 [1]GCC ftp server. 24689 Note most C++ programs only need libstdc++. 24690 * Exception handling may not work with shared libraries, particularly 24691 on alphas, hppas, rs6000/powerpc and mips based platforms. 24692 Exception handling is known to work on x86 GNU/Linux platforms with 24693 shared libraries. 24694 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 24695 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 24696 or EGCS 1.1. As a result it may be necessary to fix C++ code before 24697 it will compile with GCC 2.95. 24698 * G++ is also converting toward the ISO C++ standard; as a result 24699 code which was previously valid (and thus accepted by other 24700 compilers and older versions of g++) may no longer be accepted. The 24701 flag -fpermissive may allow some non-conforming code to compile 24702 with GCC 2.95. 24703 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 24704 1.1.x, EGCS 1.0.x or GCC 2.8.x. 24705 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 24706 between Sept 30, 1998 and April 30, 1999 (the official end of the 24707 GCC 2.8 project). Future GCC releases will include all the changes 24708 from the defunct GCC 2.8 sources. 24709 24710 24711 For questions related to the use of GCC, please consult these web 24712 pages and the [2]GCC manuals. If that fails, the 24713 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24714 web pages and the development of GCC are welcome on our developer 24715 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 24716 24717 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 24718 distribution of this entire article is permitted in any medium, 24719 provided this notice is preserved. 24720 24721 These pages are [7]maintained by the GCC team. Last modified 24722 2022-10-26. 24723 24724 References 24725 24726 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 24727 2. https://gcc.gnu.org/onlinedocs/ 24728 3. mailto:gcc-help (a] gcc.gnu.org 24729 4. mailto:gcc (a] gcc.gnu.org 24730 5. https://gcc.gnu.org/lists.html 24731 6. https://www.fsf.org/ 24732 7. https://gcc.gnu.org/about.html 24733 ====================================================================== 24734 http://gcc.gnu.org/egcs-1.1/index.html 24735 EGCS 1.1 24736 24737 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 24738 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 24739 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 24740 24741 EGCS is a free software project to further the development of the GNU 24742 compilers using an open development environment. 24743 24744 EGCS 1.1 is a major new release of the EGCS compiler system. It has 24745 been [1]extensively tested and is believed to be stable and suitable 24746 for widespread use. 24747 24748 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 24749 development sources; it contains all of the new features found in GCC 24750 2.8.1 as well as all new development from GCC up to June 6, 1998. 24751 24752 EGCS 1.1 also contains many improvements and features not found in GCC 24753 or in older versions of EGCS: 24754 * Global common subexpression elimination and global constant/copy 24755 propagation (aka [2]gcse) 24756 * Ongoing improvements to the [3]alias analysis support to allow for 24757 better optimizations throughout the compiler. 24758 * Vastly improved [4]C++ compiler and integrated C++ runtime 24759 libraries. 24760 * Fixes for the /tmp symlink race security problems. 24761 * New targets including mips16, arm-thumb and 64-bit PowerPC. 24762 * Improvements to GNU Fortran (g77) compiler and runtime library made 24763 since g77 version 0.5.23. 24764 24765 See the [5]new features page for a more complete list of new features 24766 found in EGCS 1.1 releases. 24767 24768 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 24769 1.1: 24770 * General improvements and fixes 24771 + Avoid some stack overflows when compiling large functions. 24772 + Avoid incorrect loop invariant code motions. 24773 + Fix some core dumps on Linux kernel code. 24774 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 24775 + Fix code generation problem in gcse. 24776 + Various documentation related fixes. 24777 * g++/libstdc++ improvements and fixes 24778 + MT safe EH fix for setjmp/longjmp based exception handling. 24779 + Fix a few bad interactions between optimization and exception 24780 handling. 24781 + Fixes for demangling of template names starting with "__". 24782 + Fix a bug that would fail to run destructors in some cases 24783 with -O2. 24784 + Fix 'new' of classes with virtual bases. 24785 + Fix crash building Qt on the Alpha. 24786 + Fix failure compiling WIFEXITED macro on GNU/Linux. 24787 + Fix some -frepo failures. 24788 * g77 and libf2c improvements and fixes 24789 + Various documentation fixes. 24790 + Avoid compiler crash on RAND intrinsic. 24791 + Fix minor bugs in makefiles exposed by BSD make programs. 24792 + Define _XOPEN_SOURCE for libI77 build to avoid potential 24793 problems on some 64-bit systems. 24794 + Fix problem with implicit endfile on rewind. 24795 + Fix spurious recursive I/O errors. 24796 * platform specific improvements and fixes 24797 + Match all versions of UnixWare7. 24798 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 24799 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 24800 from unsigned ints to double precision floats. 24801 + Fix ARM ABI issue with NetBSD. 24802 + Fix a few arm code generation bugs. 24803 + Fixincludes will fix additional broken SCO OpenServer header 24804 files. 24805 + Fix a m68k back end bug which caused invalid offsets in reg+d 24806 addresses. 24807 + Fix problems with 64bit AIX 4.3 support. 24808 + Fix handling of long longs for varargs/stdarg functions on the 24809 ppc. 24810 + Minor fixes to CPP predefines for Windows. 24811 + Fix code generation problems with gpr<->fpr copies for 64bit 24812 ppc. 24813 + Fix a few coldfire code generation bugs. 24814 + Fix some more header file problems on SunOS 4.x. 24815 + Fix assert.h handling for RTEMS. 24816 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 24817 + Fix x86 compiler abort in reg-stack pass. 24818 + Fix cygwin/windows problem with section attributes. 24819 + Fix Alpha code generation problem exposed by SMP Linux 24820 kernels. 24821 + Fix typo in m68k 32->64bit integer conversion. 24822 + Make sure target libraries build with -fPIC for PPC & Alpha 24823 targets. 24824 24825 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 24826 1.1.1: 24827 * General improvements and fixes 24828 + Fix bug in loop optimizer which caused the SPARC (and 24829 potentially other) ports to segfault. 24830 + Fix infinite recursion in alias analysis and combiner code. 24831 + Fix bug in regclass preferencing. 24832 + Fix incorrect loop reversal which caused incorrect code to be 24833 generated for several targets. 24834 + Fix return value for builtin memcpy. 24835 + Reduce compile time for certain loops which exposed quadratic 24836 behavior in the loop optimizer. 24837 + Fix bug which caused volatile memory to be written multiple 24838 times when only one write was needed/desired. 24839 + Fix compiler abort in caller-save.c 24840 + Fix combiner bug which caused incorrect code generation for 24841 certain division by constant operations. 24842 + Fix incorrect code generation due to a bug in range check 24843 optimizations. 24844 + Fix incorrect code generation due to mis-handling of clobbered 24845 values in CSE. 24846 + Fix compiler abort/segfault due to incorrect register 24847 splitting when unrolling loops. 24848 + Fix code generation involving autoincremented addresses with 24849 ternary operators. 24850 + Work around bug in the scheduler which caused qt to be 24851 mis-compiled on some platforms. 24852 + Fix code generation problems with -fshort-enums. 24853 + Tighten security for temporary files. 24854 + Improve compile time for codes which make heavy use of 24855 overloaded functions. 24856 + Fix multiply defined constructor/destructor symbol problems. 24857 + Avoid setting bogus RPATH environment variable during 24858 bootstrap. 24859 + Avoid GNU-make dependencies in the texinfo subdir. 24860 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 24861 --enable-cpp=<dirname> can be used to specify an additional 24862 install directory for the cpp wrapper script. 24863 + Fix CSE bug which caused incorrect label-label refs to appear 24864 on some platforms. 24865 + Avoid linking in EH routines from libgcc if they are not 24866 needed. 24867 + Avoid obscure bug in aliasing code. 24868 + Fix bug in weak symbol handling. 24869 * Platform-specific improvements and fixes 24870 + Fix detection of PPro/PII on Unixware 7. 24871 + Fix compiler segfault when building spec99 and other programs 24872 for SPARC targets. 24873 + Fix code-generation bugs for integer and floating point 24874 conditional move instructions on the PPro/PII. 24875 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 24876 + Fix build failure for the arc port. 24877 + Fix floating point format configuration for i?86-gnu port. 24878 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 24879 threads are enabled. 24880 + Fix coldfire code generation bugs. 24881 + Fix "unrecognized insn" problems for Alpha and PPC ports. 24882 + Fix h8/300 code generation problem with floating point values 24883 in memory. 24884 + Fix unrecognized insn problems for the m68k port. 24885 + Fix namespace-pollution problem for the x86 port. 24886 + Fix problems with old assembler on x86 NeXT systems. 24887 + Fix PIC code-generation problems for the SPARC port. 24888 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 24889 + Fix minor ISO namespace violation in Alpha varargs/stdarg 24890 support. 24891 + Fix incorrect "braf" instruction usage for the SH port. 24892 + Fix minor bug in va-sh which prevented its use with -ansi. 24893 + Fix problems recognizing and supporting FreeBSD. 24894 + Handle OpenBSD systems correctly. 24895 + Minor fixincludes fix for Digital UNIX 4.0B. 24896 + Fix problems with ctors/dtors in SCO shared libraries. 24897 + Abort instead of generating incorrect code for PPro/PII 24898 floating point conditional moves. 24899 + Avoid multiply defined symbols on GNU/Linux systems using 24900 libc-5.4.xx. 24901 + Fix abort in alpha compiler. 24902 * Fortran-specific fixes 24903 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 24904 is in the documented, non-Y2K-compliant range of 0-99, instead 24905 of being returned as 100 in the year 2000. 24906 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 24907 milliseconds value properly in Values(8). 24908 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 24909 information properly in SArray(7). 24910 24911 Each release includes installation instructions in both HTML and 24912 plaintext forms (see the INSTALL directory in the toplevel directory of 24913 the distribution). However, we also keep the most up to date 24914 installation instructions and [6]build/test status on our web page. We 24915 will update those pages as new information becomes available. 24916 24917 The EGCS project would like to thank the numerous people that have 24918 contributed new features, test results, bugfixes, etc. This [7]amazing 24919 group of volunteers is what makes EGCS successful. 24920 24921 And finally, we can't in good conscience fail to mention some 24922 [8]caveats to using EGCS 1.1. 24923 24924 Download EGCS from egcs.cygnus.com (USA California). 24925 24926 The EGCS 1.1 release is also available on many mirror sites. 24927 [9]Goto mirror list to find a closer site. 24928 24929 24930 For questions related to the use of GCC, please consult these web 24931 pages and the [10]GCC manuals. If that fails, the 24932 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24933 web pages and the development of GCC are welcome on our developer 24934 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 24935 archives. 24936 24937 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 24938 distribution of this entire article is permitted in any medium, 24939 provided this notice is preserved. 24940 24941 These pages are [15]maintained by the GCC team. Last modified 24942 2024-05-18. 24943 24944 References 24945 24946 1. https://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 24947 2. https://gcc.gnu.org/news/gcse.html 24948 3. https://gcc.gnu.org/news/alias.html 24949 4. https://gcc.gnu.org/egcs-1.1/c++features.html 24950 5. https://gcc.gnu.org/egcs-1.1/features.html 24951 6. https://gcc.gnu.org/egcs-1.1/buildstat.html 24952 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24953 8. https://gcc.gnu.org/egcs-1.1/caveats.html 24954 9. https://gcc.gnu.org/mirrors.html 24955 10. https://gcc.gnu.org/onlinedocs/ 24956 11. mailto:gcc-help (a] gcc.gnu.org 24957 12. mailto:gcc (a] gcc.gnu.org 24958 13. https://gcc.gnu.org/lists.html 24959 14. https://www.fsf.org/ 24960 15. https://gcc.gnu.org/about.html 24961 ====================================================================== 24962 http://gcc.gnu.org/egcs-1.1/features.html 24963 EGCS 1.1 new features 24964 24965 * Integrated GNU Fortran (g77) compiler and runtime library with 24966 improvements, based on g77 version 0.5.23. 24967 * Vast improvements in the C++ compiler; so many they have [1]page of 24968 their own! 24969 * Compiler implements [2]global common subexpression elimination and 24970 global copy/constant propagation. 24971 * More major improvements in the [3]alias analysis code. 24972 * More major improvements in the exception handling code to improve 24973 performance, lower static overhead and provide the infrastructure 24974 for future improvements. 24975 * The infamous /tmp symlink race security problems have been fixed. 24976 * The regmove optimization pass has been nearly completely rewritten 24977 to improve performance of generated code. 24978 * The compiler now recomputes register usage information before local 24979 register allocation. By providing more accurate information to the 24980 priority based allocator, we get better register allocation. 24981 * The register reloading phase of the compiler optimizes spill code 24982 much better than in previous releases. 24983 * Some bad interactions between the register allocator and 24984 instruction scheduler have been fixed, resulting in much better 24985 code for certain programs. Additionally, we have tuned the 24986 scheduler in various ways to improve performance of generated code 24987 for some architectures. 24988 * The compiler's branch shortening algorithms have been significantly 24989 improved to work better on targets which align jump targets. 24990 * The compiler now supports -Os to prefer optimizing for code space 24991 over optimizing for code speed. 24992 * The compiler will now totally eliminate library calls which compute 24993 constant values. This primarily helps targets with no integer 24994 div/mul support and targets without floating point support. 24995 * The compiler now supports an extensive "--help" option. 24996 * cpplib has been greatly improved and may be suitable for limited 24997 use. 24998 * Memory footprint for the compiler has been significantly reduced 24999 for some pathological cases. 25000 * The time to build EGCS has been improved for certain targets 25001 (particularly the alpha and mips platforms). 25002 * Many infrastructure improvements throughout the compiler, plus the 25003 usual mountain of bugfixes and minor improvements. 25004 * Target dependent improvements: 25005 + SPARC port now includes V8 plus and V9 support as well as 25006 performance tuning for Ultra class machines. The SPARC port 25007 now uses the Haifa scheduler. 25008 + Alpha port has been tuned for the EV6 processor and has an 25009 optimized expansion of memcpy/bzero. The Alpha port now uses 25010 the Haifa scheduler. 25011 + RS6000/PowerPC: support for the Power64 architecture and AIX 25012 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 25013 + x86: Alignment of static store data and jump targets is per 25014 Intel recommendations now. Various improvements throughout the 25015 x86 port to improve performance on Pentium processors 25016 (including improved epilogue sequences for Pentium chips and 25017 back end improvements which should help register allocation on 25018 all x86 variants. Conditional move support has been fixed and 25019 enabled for PPro processors. The x86 port also better supports 25020 64bit operations now. Unixware 7, a System V Release 5 target, 25021 is now supported and SCO OpenServer targets can support GAS. 25022 + MIPS has improved multiply/multiply-add support and now 25023 includes mips16 ISA support. 25024 + M68k has many micro-optimizations and Coldfire fixes. 25025 * Core compiler is based on the GCC development tree from June 9, 25026 1998, so we have all of the [4]features found in GCC 2.8. 25027 25028 25029 For questions related to the use of GCC, please consult these web 25030 pages and the [5]GCC manuals. If that fails, the 25031 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25032 web pages and the development of GCC are welcome on our developer 25033 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 25034 25035 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 25036 distribution of this entire article is permitted in any medium, 25037 provided this notice is preserved. 25038 25039 These pages are [10]maintained by the GCC team. Last modified 25040 2023-09-02. 25041 25042 References 25043 25044 1. https://gcc.gnu.org/egcs-1.1/c++features.html 25045 2. https://gcc.gnu.org/news/gcse.html 25046 3. https://gcc.gnu.org/news/alias.html 25047 4. https://gcc.gnu.org/egcs-1.0/features-2.8.html 25048 5. https://gcc.gnu.org/onlinedocs/ 25049 6. mailto:gcc-help (a] gcc.gnu.org 25050 7. mailto:gcc (a] gcc.gnu.org 25051 8. https://gcc.gnu.org/lists.html 25052 9. https://www.fsf.org/ 25053 10. https://gcc.gnu.org/about.html 25054 ====================================================================== 25055 http://gcc.gnu.org/egcs-1.1/caveats.html 25056 EGCS 1.1 Caveats 25057 25058 * EGCS has an integrated libstdc++, but does not have an integrated 25059 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 25060 Lu has made a libg++-2.8.1.2 snapshot available which may work with 25061 EGCS. 25062 Note most C++ programs only need libstdc++. 25063 * Exception handling may not work with shared libraries, particularly 25064 on alphas, hppas, rs6000/powerpc and mips based platforms. 25065 Exception handling is known to work on x86-linux platforms with 25066 shared libraries. 25067 * Some versions of the Linux kernel have bugs which prevent them from 25068 being compiled or from running when compiled by EGCS. See the FAQ 25069 (as shipped with EGCS 1.1) for additional information. 25070 * In general, EGCS is more rigorous about rejecting invalid C++ code 25071 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 25072 a result it may be necessary to fix C++ code before it will compile 25073 with EGCS. 25074 * G++ is also converting toward the ISO C++ standard; as a result 25075 code which was previously valid (and thus accepted by other 25076 compilers and older versions of g++) may no longer be accepted. 25077 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 25078 or GCC 2.8.x due to changes necessary to support thread safe 25079 exception handling. 25080 25081 25082 For questions related to the use of GCC, please consult these web 25083 pages and the [1]GCC manuals. If that fails, the 25084 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25085 web pages and the development of GCC are welcome on our developer 25086 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 25087 25088 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 25089 distribution of this entire article is permitted in any medium, 25090 provided this notice is preserved. 25091 25092 These pages are [6]maintained by the GCC team. Last modified 25093 2022-10-26. 25094 25095 References 25096 25097 1. https://gcc.gnu.org/onlinedocs/ 25098 2. mailto:gcc-help (a] gcc.gnu.org 25099 3. mailto:gcc (a] gcc.gnu.org 25100 4. https://gcc.gnu.org/lists.html 25101 5. https://www.fsf.org/ 25102 6. https://gcc.gnu.org/about.html 25103 ====================================================================== 25104 http://gcc.gnu.org/egcs-1.0/index.html 25105 EGCS 1.0 25106 25107 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 25108 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 25109 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 25110 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 25111 25112 EGCS is a collaborative effort involving several groups of hackers 25113 using an open development model to accelerate development and testing 25114 of GNU compilers and runtime libraries. 25115 25116 An important goal of EGCS is to allow wide scale testing of 25117 experimental features and optimizations; therefore, EGCS contains some 25118 features and optimizations which are still under development. However, 25119 EGCS has been carefully tested and should be comparable in quality to 25120 most GCC releases. 25121 25122 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 25123 development sources; it contains nearly all of the new features found 25124 in GCC 2.8. 25125 25126 EGCS 1.0 also contains many improvements and features not found in GCC 25127 2.7 and even the GCC 2.8 series (which was released after the original 25128 EGCS 1.0 release). 25129 * Integrated C++ runtime libraries, including support for most major 25130 GNU/Linux systems! 25131 * The integrated libstdc++ library includes a verbatim copy of SGI's 25132 STL release. 25133 * Integrated GNU Fortran compiler. 25134 * New instruction scheduler. 25135 * New alias analysis code. 25136 25137 See the [1]new features page for a more complete list of new features. 25138 25139 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 25140 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 25141 EGCS 1.0 release: 25142 * Add support for Red Hat 5.0 Linux and better support for Linux 25143 systems using glibc2. 25144 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 25145 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 25146 fix these problems. 25147 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 25148 handling interfaces. 25149 To avoid future compatibility problems, we strongly urge anyone who 25150 is planning on distributing shared libraries that contain C++ code 25151 to upgrade to EGCS 1.0.1 first. 25152 Soon after EGCS 1.0 was released, the GCC developers made some 25153 incompatible changes in libgcc's exception handling interfaces. 25154 These changes were needed to solve problems on some platforms. This 25155 means that GCC 2.8.0, when released, will not be seamlessly 25156 compatible with shared libraries built by EGCS 1.0. The reason is 25157 that the libgcc.a in GCC 2.8.0 will not contain a function needed 25158 by the old interface. 25159 The result of this is that there may be compatibility problems with 25160 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 25161 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 25162 and libgcc.a has the support routines for both the old and the new 25163 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 25164 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 25165 The maintainers of GCC 2.x have decided against including seamless 25166 support for the old interface in 2.8.0, since it was never 25167 "official", so to avoid future compatibility problems we recommend 25168 against distributing any shared libraries built by EGCS 1.0 that 25169 contain C++ code (upgrade to 1.0.1 and use that). 25170 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 25171 The x86 changes fix code generation errors exposed when building 25172 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 25173 The hppa change fixes a compiler abort when configured for use with 25174 RTEMS. 25175 The MIPS changes fix problems with the definition of LONG_MAX on 25176 newer systems, allow for command line selection of the target ABI, 25177 and fix one code generation problem. 25178 The rs6000/ppc change fixes some problems with passing structures 25179 to varargs/stdarg functions. 25180 * A few machine independent bugfixes, mostly to fix code generation 25181 errors when building Linux kernels or glibc. 25182 * Fix a few critical exception handling and template bugs in the C++ 25183 compiler. 25184 * Fix Fortran namelist bug on alphas. 25185 * Fix build problems on x86-solaris systems. 25186 25187 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 25188 serious problems in EGCS 1.0.1. 25189 * General improvements and fixes 25190 + Memory consumption significantly reduced, especially for 25191 templates and inline functions. 25192 + Fix various problems with glibc2.1. 25193 + Fix loop optimization bug exposed by rs6000/ppc port. 25194 + Fix to avoid potential code generation problems in jump.c. 25195 + Fix some undefined symbol problems in dwarf1 debug support. 25196 * g++/libstdc++ improvements and fixes 25197 + libstdc++ in the EGCS release has been updated and should be 25198 link compatible with libstdc++-2.8. 25199 + Various fixes in libio/libstdc++ to work better on GNU/Linux 25200 systems. 25201 + Fix problems with duplicate symbols on systems that do not 25202 support weak symbols. 25203 + Memory corruption bug and undefined symbols in bastring have 25204 been fixed. 25205 + Various exception handling fixes. 25206 + Fix compiler abort for very long thunk names. 25207 * g77 improvements and fixes 25208 + Fix compiler crash for omitted bound in Fortran CASE 25209 statement. 25210 + Add missing entries to g77 lang-options. 25211 + Fix problem with -fpedantic in the g77 compiler. 25212 + Fix "backspace" problem with g77 on alphas. 25213 + Fix x86 back end problem with Fortran literals and -fpic. 25214 + Fix some of the problems with negative subscripts for g77 on 25215 alphas. 25216 + Fixes for Fortran builds on cygwin32/mingw32. 25217 * platform specific improvements and fixes 25218 + Fix long double problems on x86 (exposed by glibc). 25219 + x86 ports define i386 again to keep imake happy. 25220 + Fix exception handling support on NetBSD ports. 25221 + Several changes to collect2 to fix many problems with AIX. 25222 + Define __ELF__ for GNU/Linux on rs6000. 25223 + Fix -mcall-linux problem on GNU/Linux on rs6000. 25224 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 25225 + Allow autoconf to select a proper install problem on AIX 3.1. 25226 + m68k port support includes -mcpu32 option as well as cpu32 25227 multilibs. 25228 + Fix stdarg bug for irix6. 25229 + Allow EGCS to build on irix5 without the gnu assembler. 25230 + Fix problem with static linking on sco5. 25231 + Fix bootstrap on sco5 with native compiler. 25232 + Fix for abort building newlib on H8 target. 25233 + Fix fixincludes handling of math.h on SunOS. 25234 + Minor fix for Motorola 3300 m68k systems. 25235 25236 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 25237 problems reported by Red Hat for builds of Red Hat 5.1. 25238 * Generic bugfixes: 25239 + Fix a typo in the libio library which resulted in incorrect 25240 behavior of istream::get. 25241 + Fix the Fortran negative array index problem. 25242 + Fix a major problem with the ObjC runtime thread support 25243 exposed by glibc2. 25244 + Reduce memory consumption of the Haifa scheduler. 25245 * Target specific bugfixes: 25246 + Fix one x86 floating point code generation bug exposed by 25247 glibc2 builds. 25248 + Fix one x86 internal compiler error exposed by glibc2 builds. 25249 + Fix profiling bugs on the Alpha. 25250 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 25251 + Fix rs6000/ppc bug when converting values from integer types 25252 to floating point types. 25253 25254 The EGCS 1.0 releases include installation instructions in both HTML 25255 and plaintext forms (see the INSTALL directory in the toplevel 25256 directory of the distribution). However, we also keep the most up to 25257 date installation instructions and [2]build/test status on our web 25258 page. We will update those pages as new information becomes available. 25259 25260 And, we can't in good conscience fail to mention some [3]caveats to 25261 using EGCS. 25262 25263 Update: Big thanks to Stanford for providing a high speed link for 25264 downloading EGCS (go.cygnus.com)! 25265 25266 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 25267 (USA California -- High speed link provided by Stanford). 25268 25269 The EGCS 1.0 release is also available many mirror sites. 25270 [4]Goto mirror list to find a closer site 25271 25272 We'd like to thank the numerous people that have contributed new 25273 features, test results, bugfixes, etc. Unfortunately, they're far too 25274 numerous to mention by name. 25275 25276 25277 For questions related to the use of GCC, please consult these web 25278 pages and the [5]GCC manuals. If that fails, the 25279 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25280 web pages and the development of GCC are welcome on our developer 25281 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 25282 25283 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 25284 distribution of this entire article is permitted in any medium, 25285 provided this notice is preserved. 25286 25287 These pages are [10]maintained by the GCC team. Last modified 25288 2023-09-02. 25289 25290 References 25291 25292 1. https://gcc.gnu.org/egcs-1.0/features.html 25293 2. https://gcc.gnu.org/egcs-1.0/buildstat.html 25294 3. https://gcc.gnu.org/egcs-1.0/caveats.html 25295 4. https://gcc.gnu.org/mirrors.html 25296 5. https://gcc.gnu.org/onlinedocs/ 25297 6. mailto:gcc-help (a] gcc.gnu.org 25298 7. mailto:gcc (a] gcc.gnu.org 25299 8. https://gcc.gnu.org/lists.html 25300 9. https://www.fsf.org/ 25301 10. https://gcc.gnu.org/about.html 25302 ====================================================================== 25303 http://gcc.gnu.org/egcs-1.0/features.html 25304 EGCS 1.0 features 25305 25306 * Core compiler is based on the gcc2 development tree from Aug 2, 25307 1997, so we have most of the [1]features found in GCC 2.8. 25308 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 25309 * Vast improvements in the C++ compiler; so many they have [2]page of 25310 their own! 25311 * Integrated C++ runtime libraries, including support for most major 25312 GNU/Linux systems! 25313 * New instruction scheduler from IBM Haifa which includes support for 25314 function wide instruction scheduling as well as superscalar 25315 scheduling. 25316 * Significantly improved alias analysis code. 25317 * Improved register allocation for two address machines. 25318 * Significant code generation improvements for Fortran code on 25319 Alphas. 25320 * Various optimizations from the g77 project as well as improved loop 25321 optimizations. 25322 * Dwarf2 debug format support for some targets. 25323 * egcs libstdc++ includes the SGI STL implementation without changes. 25324 * As a result of these and other changes, egcs libstc++ is not binary 25325 compatible with previous releases of libstdc++. 25326 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 25327 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 25328 1.1), Support for RTEMS on several embedded targets, Support for 25329 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 25330 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 25331 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 25332 * RS6000/PowerPC ports generate code which can run on all 25333 RS6000/PowerPC variants by default. 25334 * -mcpu= and -march= switches for the x86 port to allow better 25335 control over how the x86 port generates code. 25336 * Includes the template repository patch (aka repo patch); note the 25337 new template code makes repo obsolete for ELF systems using gnu-ld 25338 such as GNU/Linux. 25339 * Plus the usual assortment of bugfixes and improvements. 25340 25341 25342 For questions related to the use of GCC, please consult these web 25343 pages and the [3]GCC manuals. If that fails, the 25344 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25345 web pages and the development of GCC are welcome on our developer 25346 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 25347 25348 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 25349 distribution of this entire article is permitted in any medium, 25350 provided this notice is preserved. 25351 25352 These pages are [8]maintained by the GCC team. Last modified 25353 2022-10-26. 25354 25355 References 25356 25357 1. https://gcc.gnu.org/egcs-1.0/features-2.8.html 25358 2. https://gcc.gnu.org/egcs-1.0/c++features.html 25359 3. https://gcc.gnu.org/onlinedocs/ 25360 4. mailto:gcc-help (a] gcc.gnu.org 25361 5. mailto:gcc (a] gcc.gnu.org 25362 6. https://gcc.gnu.org/lists.html 25363 7. https://www.fsf.org/ 25364 8. https://gcc.gnu.org/about.html 25365 ====================================================================== 25366 http://gcc.gnu.org/egcs-1.0/caveats.html 25367 EGCS 1.0 Caveats 25368 25369 * EGCS has an integrated libstdc++, but does not have an integrated 25370 libg++. Furthermore old libg++ releases will not work with egc; HJ 25371 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 25372 Note most C++ programs only need libstdc++. 25373 * Note that using -pedantic or -Wreturn-type can cause an explosion 25374 in the amount of memory needed for template-heavy C++ code, such as 25375 code that uses STL. Also note that -Wall includes -Wreturn-type, so 25376 if you use -Wall you will need to specify -Wno-return-type to turn 25377 it off. 25378 * Exception handling may not work with shared libraries, particularly 25379 on alphas, hppas, and mips based platforms. Exception handling is 25380 known to work on x86-linux platforms with shared libraries. 25381 * Some versions of the Linux kernel have bugs which prevent them from 25382 being compiled or from running when compiled by EGCS. See the FAQ 25383 (as shipped with EGCS 1.0) for additional information. 25384 * In general, EGCS is more rigorous about rejecting invalid C++ code 25385 or deprecated C++ constructs than G++ 2.7. As a result it may be 25386 necessary to fix C++ code before it will compile with EGCS. 25387 * G++ is also aggressively tracking the C++ standard; as a result 25388 code which was previously valid (and thus accepted by other 25389 compilers and older versions of G++) may no longer be accepted. 25390 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 25391 1.0.x and later releases should work with Red Hat Linux 5.0. 25392 25393 25394 For questions related to the use of GCC, please consult these web 25395 pages and the [1]GCC manuals. If that fails, the 25396 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25397 web pages and the development of GCC are welcome on our developer 25398 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 25399 25400 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 25401 distribution of this entire article is permitted in any medium, 25402 provided this notice is preserved. 25403 25404 These pages are [6]maintained by the GCC team. Last modified 25405 2022-10-26. 25406 25407 References 25408 25409 1. https://gcc.gnu.org/onlinedocs/ 25410 2. mailto:gcc-help (a] gcc.gnu.org 25411 3. mailto:gcc (a] gcc.gnu.org 25412 4. https://gcc.gnu.org/lists.html 25413 5. https://www.fsf.org/ 25414 6. https://gcc.gnu.org/about.html 25415 ====================================================================== 25416
Indexes created Sun Oct 25 00:26:45 UTC 2026