NEWS revision 1.20
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 10 GCC 14 Release Series 11 12 May 23, 2025 13 14 The GCC developers are pleased to announce the release of GCC 14.3. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 14.2 relative to previous releases of GCC. 18 19 Release History 20 21 GCC 14.3 22 May 23, 2025 ([1]changes, [2]documentation) 23 24 GCC 14.2 25 Aug 1, 2024 ([3]changes, [4]documentation) 26 27 GCC 14.1 28 May 7, 2024 ([5]changes, [6]documentation) 29 30 References and Acknowledgements 31 32 GCC used to stand for the GNU C Compiler, but since the compiler 33 supports several other languages aside from C, it now stands for the 34 GNU Compiler Collection. 35 36 The GCC developers would like to thank the numerous people that have 37 contributed new features, improvements, bug fixes, and other changes as 38 well as test results to GCC. This [7]amazing group of volunteers is 39 what makes GCC successful. 40 41 For additional information about GCC please refer to the [8]GCC project 42 web site or contact the [9]GCC development mailing list. 43 44 To obtain GCC please use [10]our mirror sites or [11]our version 45 control system. 46 47 48 For questions related to the use of GCC, please consult these web 49 pages and the [12]GCC manuals. If that fails, the 50 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 51 web pages and the development of GCC are welcome on our developer 52 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 53 archives. 54 55 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 56 distribution of this entire article is permitted in any medium, 57 provided this notice is preserved. 58 59 These pages are [17]maintained by the GCC team. Last modified 60 2025-05-23. 61 62 References 63 64 1. https://gcc.gnu.org/gcc-14/changes.html 65 2. https://gcc.gnu.org/onlinedocs/14.3.0/ 66 3. https://gcc.gnu.org/gcc-14/changes.html 67 4. https://gcc.gnu.org/onlinedocs/14.2.0/ 68 5. https://gcc.gnu.org/gcc-14/changes.html 69 6. https://gcc.gnu.org/onlinedocs/14.1.0/ 70 7. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Contributors.html 71 8. https://gcc.gnu.org/index.html 72 9. mailto:gcc (a] gcc.gnu.org 73 10. https://gcc.gnu.org/mirrors.html 74 11. https://gcc.gnu.org/git.html 75 12. https://gcc.gnu.org/onlinedocs/ 76 13. mailto:gcc-help (a] gcc.gnu.org 77 14. mailto:gcc (a] gcc.gnu.org 78 15. https://gcc.gnu.org/lists.html 79 16. https://www.fsf.org/ 80 17. https://gcc.gnu.org/about.html 81 ====================================================================== 82 http://gcc.gnu.org/gcc-14/changes.html 83 84 GCC 14 Release Series 85 Changes, New Features, and Fixes 86 87 This page is a "brief" summary of some of the huge number of 88 improvements in GCC 14. You may also want to check out our [1]Porting 89 to GCC 14 page and the [2]full GCC documentation. 90 91 Caveats 92 93 * C: Support for the GCC extension, a structure containing a C99 94 flexible array member, or a union containing such a structure, is 95 not the last field of another structure, is deprecated. Refer to 96 [3]Zero Length Arrays. Any code relying on this extension should be 97 modified to ensure that C99 flexible array members only end up at 98 the ends of structures. Please use the warning option 99 [4]-Wflex-array-member-not-at-end to identify all such cases in the 100 source code and modify them. 101 * C: Certain warnings about are now errors, see [5]Porting to GCC 14 102 for details. 103 * [6]-fcf-protection=[full|branch|return|none|check] has been 104 refactored. To override -fcf-protection with a more specific 105 setting, add -fcf-protection=none followed by a specific 106 -fcf-protection=xxx. 107 * Support for the ia64*-*- target ports which have been unmaintained 108 for quite a while has been declared obsolete in GCC 14. The next 109 release of GCC will have their sources permanently removed. 110 * Support for the nios2*-*- target ports has also been declared 111 obsolete in GCC 14, and the sources will also be removed in the 112 next release of GCC. 113 * [7]-fanalyzer is still only suitable for analyzing C code. In 114 particular, using it on C++ is unlikely to give meaningful output. 115 116 [8]General Improvements 117 118 * For offload-device code generated via OpenMP and OpenACC, the math 119 and the Fortran runtime libraries will now automatically be linked, 120 when the user or compiler links them on the host side. Thus, it is 121 no longer required to explicitly pass -lm and/or -lgfortran to the 122 offload-device linker using the [9]-foffload-options= flag. 123 * New configure options: --enable-host-pie, to build the compiler 124 executables as PIE; and --enable-host-bind-now, to link the 125 compiler executables with -Wl,-z,now in order to enable additional 126 hardening. 127 * New option [10]-fhardened, an umbrella option that enables a set of 128 hardening flags. The options it enables can be displayed using the 129 --help=hardened option. 130 * New option [11]-fharden-control-flow-redundancy, to verify, at the 131 end of functions, that the visited basic blocks correspond to a 132 legitimate execution path, so as to detect and prevent attacks that 133 transfer control into the middle of functions. 134 * New type attribute [12]hardbool, for C and Ada. Hardened booleans 135 take user-specified representations for true and false, presumably 136 with higher hamming distance than standard booleans, and get 137 verified at every use, detecting memory corruption and some 138 malicious attacks. 139 * New type attribute [13]strub to control stack scrubbing properties 140 of functions and variables. The stack frame used by functions 141 marked with the attribute gets zeroed-out upon returning or 142 exception escaping. Scalar variables marked with the attribute 143 cause functions containing or accessing them to get stack scrubbing 144 enabled implicitly. 145 * New option [14]-finline-stringops, to force inline expansion of 146 memcmp, memcpy, memmove and memset, even when that is not an 147 optimization, to avoid relying on library implementations. 148 * New option [15]-ftrampoline-impl=, to choose the placement of 149 generated trampolines. The default remains the stack for almost all 150 targets. To override the default, -ftrampoline-impl=heap can be 151 passed on supported ABIs (x86_64 Darwin, x86_64 and aarch64 Linux). 152 * New function attribute [16]null_terminated_string_arg(PARAM_IDX) 153 for indicating parameters that are expected to be null-terminated 154 strings. 155 * The vectorizer now supports vectorizing loops which contain any 156 number of early breaks. This means loops such as: 157 int z[100], y[100], x[100]; 158 int foo (int n) 159 { 160 int res = 0; 161 for (int i = 0; i < n; i++) 162 { 163 y[i] = x[i] * 2; 164 res += x[i] + y[i]; 165 166 if (x[i] > 5) 167 break; 168 169 if (z[i] > 5) 170 break; 171 172 } 173 return res; 174 } 175 176 can now be vectorized on a number of targets. In this first version 177 any input data sources must either have a statically known size at 178 compile time or the vectorizer must be able to determine based on 179 auxiliary information that the accesses are aligned. 180 181 [17]New Languages and Language specific improvements 182 183 * [18]OpenMP 184 + The [19]GNU Offloading and Multi Processing Runtime Library 185 Manual has been updated and extended, improving especially the 186 description of ICVs, memory allocation, environment variables 187 and OpenMP routines. 188 + The requires directive's unified_address requirement is now 189 fulfilled by both AMD GCN and nvptx devices. AMD GCN and nvptx 190 devices now support low-latency allocators as [20]detailed in 191 the manual. Initial support for pinned-memory allocators has 192 been added and, on Linux, [21]libnuma is now used for 193 allocators requesting the nearest-partition trait (both is 194 described in the [22]memory allocation section of the manual). 195 + OpenMP 5.0: The allocate directive is now supported for stack 196 variables in C and Fortran, including the OpenMP 5.1 align 197 modifier. In C and C++ the map clause now accepts lvalue 198 expressions. For Fortran, OpenMP allocators can now be used 199 for allocatables and pointers using the allocate directive and 200 its OpenMP 5.2 replacement, the allocators directive; files 201 using this allocator and all files that might directly or 202 indirectly (intrinsic assignment, intent(out), ...) de- or 203 reallocate such-allocated variables must be compiled with the 204 [23]-fopenmp-allocators option. 205 + OpenMP 5.1: Support was added for collapsing imperfectly 206 nested loops and using present as map-type modifier and in 207 defaultmap. The indirect clause is now supported. The 208 performance of copying strided data from or to nvptx and AMD 209 GPU devices using the OpenMP 5.1 routine 210 omp_target_memcpy_rect has been improved. 211 + OpenMP 5.2: The OMP_TARGET_OFFLOAD=mandatory handling has been 212 updated for the clarifications and changes of the 5.2 213 specification. For Fortran, the list of directives permitted 214 in Fortran pure procedures was extended. Additionally, the 215 spec change has been implemented for default implicit mapping 216 of C/C++ pointers pointing to unmapped storage. The destroy 217 clause now optionally accepts the depend object as argument. 218 + OpenMP 6.0 preview (TR11/TR12): The decl attribute is now 219 supported in C++ 11 and the directive, sequence and decl 220 attributes are now supported in C 23. 221 * [24]OpenACC 222 + OpenACC 2.7: The self clause was added to be used on compute 223 constructs and the default clause for data constructs. 224 Additionally, the readonly modifier is now handled in the 225 copyin clause and cache directive. 226 + OpenACC 3.2: The following API routines are now available in 227 Fortran using the openacc module or the openacc_lib.h header 228 file: acc_malloc, acc_free, acc_map_data, acc_unmap_data, 229 acc_deviceptr, acc_hostptr, acc_memcpy_to_device, 230 acc_memcpy_to_device_async, acc_memcpy_from_device, and 231 acc_memcpy_from_device_async. 232 233 [25]Ada 234 235 * New implementation-defined aspects and pragmas: 236 + [26]Local_Restrictions, which specifies that a particular 237 subprogram does not violate one or more local restrictions, 238 nor can it call a subprogram that is not subject to the same 239 requirements. 240 + [27]User_Aspect_Definition and [28]User_Aspect, which provide 241 a mechanism for avoiding textual duplication if some set of 242 aspect specifications is needed in multiple places. 243 * New implementation-defined aspects and pragmas for verification of 244 the SPARK 2014 subset of Ada: 245 + [29]Always_Terminates, which provides a condition for a 246 subprogram to necessarily complete (either return normally or 247 raise an exception). 248 + [30]Ghost_Predicate, which introduces a subtype predicate that 249 can reference Ghost entities. 250 + [31]Exceptional_Cases, which lists exceptions that might be 251 propagated by the subprogram with side effects in the context 252 of its precondition and associates them with a specific 253 postcondition. 254 + [32]Side_Effects, which indicates that a function should be 255 handled like a procedure with respect to parameter modes, 256 Global contract, exceptional contract and termination: it may 257 have output parameters, write global variables, raise 258 exceptions and not terminate. 259 * The new attributes and contracts have been applied to the relevant 260 parts of the Ada runtime library, which has been subsequently 261 proven to be correct with SPARK 2014. 262 * Support for the LoongArch architecture. 263 * Support for vxWorks 7 Cert RTP has been removed. 264 * Additional hardening improvements. For more information related to 265 hardening options, refer to the [33]GCC Instrumentation Options and 266 the [34]GNAT Reference Manual, Security and Hardening Features. 267 * Improve style checking for redundant parentheses with [35]-gnatyz 268 * New switch [36]-gnateH to force reverse Bit_Order threshold to 64. 269 * Experimental features: 270 + [37]Storage Model: this feature proposes to redesign the 271 concepts of Storage Pools into a more efficient model allowing 272 higher performance and easier integration with low footprint 273 embedded runtimes. 274 + [38]String Interpolation: allows for easier string formatting. 275 * Further clean up and improvements to the GNAT code. 276 277 [39]C family 278 279 * The Clang language extensions __has_feature and __has_extension 280 have been implemented in GCC. These are available from C, C++, and 281 Objective-C(++). This is primarily intended to aid the portability 282 of code written against Clang. 283 284 [40]C 285 286 * Some more C23 features have been implemented: 287 + Bit-precise integer types (_BitInt (N) and unsigned _BitInt 288 (N)): integer types with a specified number of bits. These are 289 only supported on IA-32, x86-64 and AArch64 (little-endian) at 290 present. 291 + Structure, union and enumeration types may be defined more 292 than once in the same scope with the same contents and the 293 same tag; if such types are defined with the same contents and 294 the same tag in different scopes, the types are compatible. 295 + The <stdckdint.h> header for checked integer arithmetic. 296 * In addition to those C23 features, there are new command-line 297 options -std=c23, -std=gnu23 and -Wc11-c23-compat. These are 298 equivalent to the previous options -std=c2x, -std=gnu2x and 299 -Wc11-c2x-compat, which are deprecated but remain supported. 300 * GCC supports a new pragma #pragma GCC novector to indicate to the 301 vectorizer not to vectorize the loop annotated with the pragma. 302 303 [41]C++ 304 305 * Several C++26 features have been implemented: 306 + [42]P1854R4, Making non-encodable string literals ill-formed 307 ([43]PR110341) 308 + [44]P2752R3, Static storage for braced initializers 309 ([45]PR110346) 310 + [46]P2361R6, Unevaluated strings ([47]PR110342) 311 + [48]P2738R1, constexpr cast from void* ([49]PR110344) 312 + [50]P2741R3, User-generated static_assert messages 313 ([51]PR110348) 314 + [52]P2169R4, Placeholder variables with no name ([53]PR110349) 315 + [54]P2864R2, Removing deprecated arithmetic conversion on 316 enumerations 317 + [55]P2748R5, Disallow binding a returned reference to a 318 temporary ([56]PR114455) 319 + [57]P2809R3, Trivial infinite loops are not undefined behavior 320 ([58]PR114462) 321 * Several C++23 features have been implemented: 322 + [59]P0847R7, Deducing this ([60]PR102609) 323 + [61]P2280R4, Using unknown references in constant expressions 324 ([62]PR106650) 325 + [63]P2564R3, consteval needs to propagate up ([64]PR107687) 326 + [65]P2582R1, Class template argument deduction from inherited 327 constructors ([66]PR106653) 328 * Several C++ Defect Reports have been resolved, e.g.: 329 + [67]DR 532, Member/nonmember operator template partial 330 ordering 331 + [68]DR 976, Deduction for const T& conversion operators 332 + [69]DR 2262, Attributes for asm-definition 333 + [70]DR 2359, Unintended copy initialization with designated 334 initializers 335 + [71]DR 2386, tuple_size requirements for structured binding 336 + [72]DR 2406, [[fallthrough]] attribute and iteration 337 statements 338 + [73]DR 2543, constinit and optimized dynamic initialization 339 + [74]DR 2586, Explicit object parameter for assignment and 340 comparison 341 + [75]DR 2735, List-initialization and conversions in overload 342 resolution 343 + [76]DR 2799, Inheriting default constructors 344 * When a diagnostic occurs involving a C++ template, GCC will now 345 quote the source code of the context at which the template is 346 instantiated ("required from here"), rather than just print 347 filename and line/column numbers. 348 * New built-in __type_pack_element to speed up traits such as 349 std::tuple_element ([77]PR100157) 350 * goto can cross the initialization of a trivially initialized object 351 with a non-trivial destructor ([78]DR 2256) 352 * -Wdangling-reference false positives have been reduced. The warning 353 does not warn about std::span-like classes; there is also a new 354 attribute gnu::no_dangling to suppress the warning. See [79]the 355 manual for more info. 356 * noexcept(expr) is now mangled as per the Itanium ABI 357 * the named return value optimization can now be performed even for 358 variables declared in an inner block of a function, see the 359 [80]test 360 * New -Wnrvo warning, to warn if the named return value optimization 361 is not performed although it is allowed by [class.copy.elision]. 362 See [81]the manual for more info. 363 * The backing array for std::initializer_list has been made static, 364 allowing combining multiple equivalent initializer-lists ([82]git) 365 * New -Welaborated-enum-base warning, to warn if an additional 366 enum-base is used in an elaborated-type-specifier 367 * Better #include hints for missing headers ([83]PR110164) 368 * The arguments of a variable template-id are coerced earlier than 369 before, so various problems are detected earlier ([84]PR89442) 370 * -Wmissing-field-initializers is no longer emitted for empty classes 371 ([85]PR110064) 372 * The constexpr code now tracks lifetimes in constant evaluation; 373 this change helps to detect bugs such as accessing a variable whose 374 lifetime has ended ([86]PR70331, [87]PR96630, [88]PR98675) 375 * Array destruction can now be devirtualized 376 * In-class member variable template partial specializations are now 377 accepted ([89]PR71954) 378 * Improved diagnostic for explicit conversion functions: when a 379 conversion doesn't work out only because the conversion function 380 necessary to do the conversion couldn't be used because it was 381 marked explicit, explain that to the user ([90]git) 382 * Corrected mangling of static/thread_local structured bindings at 383 function/block scope ([91]PR111069) 384 * [basic.scope.block]/2 violations are detected even in compound-stmt 385 of function-try-block and for block-scope external variables 386 ([92]PR52953) 387 * Improved "not a constant expression" diagnostic when taking the 388 address of a non-static constexpr variable ([93]PR91483) 389 * Non-dependent simple assignments are checked even in templates 390 ([94]PR18474) 391 * Attributes hot and cold can be applied to classes as well. See 392 [95]the manual for more info. 393 * Function template constraints, as well as CTAD placeholders, are 394 now mangled 395 * Various decltype fixes: [96]PR79620, [97]PR79378, [98]PR83167, 396 [99]PR96917 397 * New option -fdiagnostics-all-candidates to note all candidates 398 during overload resolution failure 399 * -Walloc-size and -Wcalloc-transposed-args warnings are enabled for 400 C++ as well 401 * The DR 2237 code no longer gives an error, it emits a 402 -Wtemplate-id-cdtor warning instead 403 * GCC supports a new pragma #pragma GCC novector to indicate to the 404 vectorizer not to vectorize the loop annotated with the pragma. 405 * C++ module scanning for named modules is now available, based on 406 the format described in [100]P1689R5, Format for describing 407 dependencies of source files. The -fdeps-format=, -fdeps-file=, and 408 -fdeps-target= flags may be used to generate dependency 409 information. In GCC 14 p1689r5 is the only valid argument for 410 -fdeps-format=. 411 412 [101]Runtime Library (libstdc++) 413 414 * The libstdc++exp.a library now includes all the Filesystem TS 415 symbols from the libstdc++fs.a library. The experimental symbols 416 for the C++23 std::stacktrace class are also in libstdc++exp.a, 417 replacing the libstdc++_libbacktrace.a library that GCC 13 418 provides. This means that -lstdc++exp is the only library needed 419 for all experimental libstdc++ features. 420 * Improved experimental support for C++20, including: 421 + std::chrono::parse. 422 + Unicode-aware string handling in std::format. 423 * Improved experimental support for C++23, including: 424 + The std::ranges::to function for converting ranges to 425 containers. 426 + The std::generator view for getting results from coroutines. 427 + The <stacktrace> header is supported by default. 428 + std::print and std::println (requires linking with -lstdc++exp 429 on Windows). 430 + Formatters for std::thread::id and std::stacktrace. 431 + Smart pointer adaptors, std::out_ptr and std::inout_ptr. 432 + Some range adaptors now support move-only types. 433 * Experimental support for C++26, including: 434 + Native handles for filebuf, fstream, etc. 435 + Functions for saturation arithmetic on integers. 436 + std::to_string now uses std::format. 437 + Enhanced formatting of pointers with std::format. 438 + The std::runtime_format function to allow using non-literal 439 format strings with std::format. 440 + Testable result types for <charconv> functions. 441 + The std::text_encoding class for identifying character sets 442 (requires linking with -lstdc++exp for some member functions). 443 * Faster numeric conversions using std::to_string and 444 std::to_wstring. 445 * Updated parallel algorithms that are compatible with oneTBB. 446 * std::numeric_limits<_Float32> and std::numeric_limits<_Float64> are 447 now defined for all standard modes, not only for C++23. 448 * Added missing functions for float and long double to <cmath>. 449 * Using the std::setfill manipulator with std::istream is deprecated. 450 451 [102]D 452 453 * Support for the D programming language has been updated to version 454 2.108.1 of the language and run-time library. Full changelog for 455 this release and previous releases can be found on the 456 [103]dlang.org website. 457 458 [104]Fortran 459 460 * The compiler now accepts the -std=f2023 option, which has been 461 added in preparation of support of Fortran 2023. This option 462 increases the line-length limit for source in free-form to 10000, 463 and statements may have up to 1 million characters. 464 * With the -save-temps option, preprocessed files with the .fii 465 extension will be generated from free-form source files such as 466 .F90 and .fi from fixed-form files such as .F. 467 468 [105]Modula-2 469 470 * The automatic dependency generation options: -M, -MD, -MF, -MMD, 471 -MP, -MQ and -MT have been implemented in the compiler. 472 * The -Wcase-enum and -Wuninit-variable-checking= options have been 473 implemented to provide compile-time warnings against missing case 474 clauses and uninitialized variables respectively. 475 476 [106]Rust 477 478 * Experimental support for the Rust programming language has been 479 added. The compiler is incomplete, but already supports a subset of 480 the Rust programming language. The frontend does not support 481 compiling the Rust standard library, so it cannot be used for most 482 real-world Rust code yet. However, you can experiment with the 483 compiler to run Rust code on #[no_core] targets. 484 485 [107]libgccjit 486 487 * The libgccjit API gained 6 new entry points: 488 + [108]gcc_jit_type_get_restrict for adding restrict to types 489 ([109]LIBGCCJIT_ABI_25). 490 + 4 functions for setting attributes on functions and variables 491 ([110]LIBGCCJIT_ABI_26): 492 o gcc_jit_function_add_attribute 493 o gcc_jit_function_add_string_attribute 494 o gcc_jit_function_add_integer_array_attribute 495 o gcc_jit_lvalue_add_string_attribute 496 + [111]gcc_jit_context_new_sizeof for accessing the size of a 497 type ([112]LIBGCCJIT_ABI_27). 498 499 [113]New Targets and Target Specific Improvements 500 501 [114]AArch64 502 503 * A number of new CPUs are supported through the -mcpu and -mtune 504 options (GCC identifiers in parentheses). 505 + Ampere-1B (ampere1b). 506 + Arm Cortex-A520 (cortex-a520). 507 + Arm Cortex-A720 (cortex-a720). 508 + Arm Cortex-X4 (cortex-x4). 509 + Microsoft Cobalt-100 (cobalt-100). 510 * Additionally, the identifiers generic, generic-armv8-a and 511 generic-armv9-a can be used to optimize code generation for a good 512 blend of CPUs of a particular architecture version. These tunings 513 are also used as the default optimization targets when compiling 514 with the -march=armv8-a or -march=armv9-a options and their point 515 releases e.g. -march=armv8.2-a or -march=armv9.3-a. 516 * New features in the Arm architecture are supported in a number of 517 ways: 518 + Support is added for the Arm Streaming Matrix Extensions SME 519 and SME2 through the +sme and +sme2 extensions to -march=. In 520 particular, this includes support for the Beta state of the 521 [115]SME ACLE in the form of a new intrinsics arm_sme.h 522 intrinsics header and a number of new keyword attributes to 523 manage use of the new Streaming SVE state. For more 524 information please refer to the ACLE documentation. 525 + Libatomic is updated to implement 128-bit atomic operations 526 locklessly on systems with FEAT_LSE2. 527 + Support for FEAT_LRCPC3 is added through ACLE intrinsics in 528 arm_neon.h header and enabled through the +rcpc3 extension to 529 -march=. 530 * As well as numerous AArch64 code generation improvements, the 531 following optimization enhancements are noteworthy: 532 + A new AArch64-specific register allocation pass is added. It 533 runs in addition to standard register allocation. The pass's 534 main purpose is to make use of strided vector register 535 operands in SME instructions. However, it can also remove 536 redundant moves in normal Advanced SIMD and SVE code. The pass 537 is controlled by the new option -mearly-ra= that takes the 538 arguments all, strided, none. -mearly-ra=all is enabled by 539 default at optimization levels -O2 and above. 540 + A new optimization pass to fuse loads and stores to adjacent 541 memory locations into load and store-pair AArch64 542 instructions. The pass is enabled by default when compiling 543 with optimization and runs twice in the optimization pipeline: 544 before and after register allocation. This can be controlled 545 with the options -mearly-ldp-fusion and -mlate-ldp-fusion. 546 * Conformance with the ACLE specification is improved and a number of 547 features aimed at helping developers deploy Arm architecture 548 features are added: 549 + Support for the Beta version of the [116]Function 550 Multiversioning Specification. This feature provides 551 facilities to annotate functions with attributes that allow 552 the compiler to generate multiple versions of the function, 553 selected at runtime based on the architecture features 554 available in the system. Please refer to the ACLE 555 specification for more details. 556 + Support for more ACLE intrinsics in the arm_acle.h header, 557 including the [117]Memory prefetch intrinsics and the 558 [118]Special register intrinsics. This also includes 559 intrinsics for the extension to 128-bit system registers, 560 enabled through the +d128 extension to -march=. 561 + Intrinsics enabled by the +dotprod, +fp16, +fp16fml, +i8mm, 562 +sha3 and +sm4 extensions to -march= no longer require 563 -march=armv8.2-a or higher to be specified. Likewise, the 564 intrinsics enabled by +memtag no longer require 565 -march=armv8.5-a. 566 + Support for the [119]NEON-SVE Bridge intrinsics. These are 567 intrinsics that allow conversions between NEON and SVE 568 vectors, enabled through the inclusion of the 569 arm_neon_sve_bridge.h header. 570 * The option -mtp= is now supported for changing the TPIDR register 571 used for TLS accesses. For more details please refer to the 572 [120]documentation. 573 574 [121]AMD Radeon (GCN) 575 576 * Initial support for the AMD Radeon gfx90c (GCN5), gfx1030, gfx1036 577 (RDNA2), gfx1100 and gfx1103 (RDNA3) devices has been added. LLVM 578 15+ (assembler and linker) is [122]required to support GFX11. 579 * Improved register usage and performance on CDNA Instinct MI100 and 580 MI200 series devices. 581 * The default device architecture is now gfx900 (Vega). 582 * Fiji (gfx803) device support is now deprecated and will be removed 583 from a future release. The default compiler configuration no longer 584 uses Fiji as the default device, and no longer includes the Fiji 585 libraries. Both can be restored by configuring with 586 [123]--with-arch=fiji . 587 588 [124]arm 589 590 * The Cortex-M52 CPU is now supported through the cortex-m52 argument 591 to the -mcpu and -mtune options. 592 593 [125]AVR 594 595 * On AVR64* and AVR128* devices, read-only data is now located in 596 program memory per default and no longer in RAM. 597 + Only a 32 KiB block of program memory can be used to store and 598 access .rodata in that way. Which block is used can be 599 selected by defining the symbol __flmap. As an alternative, 600 the byte address of the block can be specified by the symbol 601 __RODATA_FLASH_START__ which takes precedence over __flmap. 602 For example, linking with 603 -Wl,--defsym,__RODATA_FLASH_START__=32k chooses the second 604 32 KiB block. 605 + The default uses the last 32 KiB block, which is also the 606 hardware default for bit-field NVMCTRL_CTRLB.FLMAP. 607 + When a non-default block is used, then NVMCTRL_CTRLB.FLMAP 608 must be initialized accordingly by hand, or AVR-LibC v2.2 that 609 implements [126]#931 can be used. The latter initializes 610 NVMCTRL_CTRLB.FLMAP in the startup code and according to the 611 value of __flmap or __RODATA_FLASH_START__. 612 + When AVR-LibC with #931 is used, then defining the symbol 613 __flmap_lock to a non-zero value will set bit 614 NVMCTRL_CTRLB.FLMAPLOCK. This will protect NVMCTRL_CTRLB.FLMAP 615 from any further changes -- which would be Undefined Behaviour 616 in C/C++. If you prefer to define the symbol in a C/C++ file, 617 an asm statement can be used: 618 __asm (".global __flmap_lock" "\n\t" 619 "__flmap_lock = 1"); 620 + When you do not want the code from #931, then define a global 621 symbol __do_flmap_init and the linker will not pull in that 622 code from libmcu.a any more. 623 + In order to return to the old placement of read-only data in 624 RAM, the new compiler option -mrodata-in-ram can be used. This 625 is required on devices where the hardware revision is affected 626 by a silicon bug concerning the FLMAP functionality. 627 + Read-only data is located in output section .rodata, whereas 628 it is part of .text when located in RAM. 629 + The feature is only available when the compiler is configured 630 with a version of Binutils that implements [127]PR31124, which 631 is the case for Binutils v2.42 and up. 632 + The implementation consists of two parts: 633 1. Binutils supports new emulations avrxmega2_flmap and 634 avrxmega4_flmap. The sole purpose of these emulations is 635 to provide adjusted default linker description files. 636 Apart from that, these emulations behave exactly the same 637 like avrxmega2 resp. avrxmega4. 638 2. The compiler uses a [128]device-specs file which links 639 the program with -mavrxmega2_flmap or -mavrxmega2 640 depending on -m[no-]rodata-in-ram; and similar for 641 -mavrxmega4[_flmap]. 642 This means the feature can be used with older compiler or 643 Binutils versions; all what's needed is an adjusted linker 644 script and a custom device-specs file. 645 * A new compiler option -m[no-]rodata-in-ram has been added. The 646 default is to locate read-only data in program memory for devices 647 that support it, e.g. for AVR64* and AVR128* devices as explained 648 above, and for devices from the [129]avrxmega3 and [130]avrtiny 649 families. 650 * The new built-in macro __AVR_RODATA_IN_RAM__ is supported on all 651 devices. It's defined to 0 or 1. 652 * A new optimization tries to improve code generation for indirect 653 memory accesses on [131]Reduced Tiny devices. It can be controlled 654 by the new compiler option -mfuse-add=level where level may be 0, 1 655 or 2. 656 * On the Reduced Tiny devices, the meaning of register constraint "w" 657 has been changed. It now constrains the registers R24...R31 as is 658 the case for all the other devices. 659 * Support for the following devices has been added in v14.3: 660 + AVR32SD20, AVR32SD28, AVR32SD32, AVR64SD28, AVR64SD32, 661 AVR64SD48. 662 663 [132]IA-32/x86-64 664 665 * New compiler option -m[no-]evex512 was added. The compiler switch 666 enables/disables 512-bit vector. It will be default on if AVX512F 667 is enabled. 668 * Part of new feature support for Intel APX was added, including 669 EGPR, NDD, PPX and PUSH2POP2. APX support is available via the 670 -mapxf compiler switch. 671 * For inline asm support with APX, by default the EGPR feature was 672 disabled to prevent potential illegal instruction with EGPR occurs. 673 To invoke egpr usage in inline asm, use new compiler option 674 -mapx-inline-asm-use-gpr32 and user should ensure the instruction 675 supports EGPR. 676 * New ISA extension support for Intel AVX10.1 was added. AVX10.1 677 intrinsics are available via the -mavx10.1-256 compiler switch with 678 256-bit vector support. 512-bit vector support for AVX10.1 679 intrinsics are available via the -mavx10.1-512 compiler switch. 680 -mavx10.1 enables AVX10.1 intrinsics with 256-bit vector support in 681 GCC 14.1 and GCC 14.2. Since GCC 14.3, it enables AVX10.1 682 intrinsics with 512-bit vector support (and emits a warning due to 683 this behavior change). 684 * New ISA extension support for Intel AVX-VNNI-INT16 was added. 685 AVX-VNNI-INT16 intrinsics are available via the -mavxvnniint16 686 compiler switch. 687 * New ISA extension support for Intel SHA512 was added. SHA512 688 intrinsics are available via the -msha512 compiler switch. 689 * New ISA extension support for Intel SM3 was added. SM3 intrinsics 690 are available via the -msm3 compiler switch. 691 * New ISA extension support for Intel SM4 was added. SM4 intrinsics 692 are available via the -msm4 compiler switch. 693 * New ISA extension support for Intel USER_MSR was added. USER_MSR 694 intrinsics are available via the -muser_msr compiler switch. 695 * GCC now supports the Intel CPU named Clearwater Forest through 696 -march=clearwaterforest. Based on Sierra Forest, the switch further 697 enables the AVX-VNNI-INT16, PREFETCHI, SHA512, SM3, SM4 and 698 USER_MSR ISA extensions. 699 * GCC now supports the Intel CPU named Gracemont through 700 -march=gracemont. Gracemont is based on Alder Lake. 701 * GCC now supports the Intel CPU named Arrow Lake through 702 -march=arrowlake. Based on Alder Lake, the switch further enables 703 the AVX-IFMA, AVX-NE-CONVERT, AVX-VNNI-INT8 and CMPccXADD ISA 704 extensions. 705 * GCC now supports the Intel CPU named Arrow Lake S through 706 -march=arrowlake-s. Based on Arrow Lake, the switch further enables 707 the AVX-VNNI-INT16, SHA512, SM3 and SM4 ISA extensions. 708 * GCC now supports the Intel CPU named Lunar Lake through 709 -march=lunarlake. Lunar Lake is based on Arrow Lake S. 710 * GCC now supports the Intel CPU named Panther Lake through 711 -march=pantherlake. Based on Arrow Lake S, the switch further 712 enables the PREFETCHI ISA extensions. 713 * Xeon Phi CPUs support (a.k.a. Knight Landing and Knight Mill) are 714 marked as deprecated. GCC will emit a warning when using the 715 -mavx5124fmaps, -mavx5124vnniw, -mavx512er, -mavx512pf, 716 -mprefetchwt1, -march=knl, -march=knm, -mtune=knl or -mtune=knm 717 compiler switches. Support will be removed in GCC 15. 718 * [133]Hardware-assisted AddressSanitizer now works for the x86-64 719 target with LAM_U57. -fsanitize=hwaddress will enable -mlam=u57 by 720 default. 721 * GCC now supports AMD CPUs based on the znver5 core via 722 -march=znver5. In addition to the ISA extensions enabled on a 723 znver4 core, this switch further enables the AVX512VP2INTERSECT, 724 AVXVNNI, MOVDIR64B, MOVDIRI, and PREFETCHI ISA extensions. 725 * The _Float16 and __bf16 type are supported independent of SSE2. W/o 726 SSE2, these types are storage-only, compiler will issue an error 727 when they're used in conversion, unary operation, binary operation, 728 parameter passing or value return. Please use __SSE2__ to detect 729 arithmetic support of these types instead of __FLT16_MAX__(or other 730 similar Macros). 731 732 [134]MCore 733 734 * Bitfields are now signed by default per GCC policy. If you need 735 bitfields to be unsigned, use -funsigned-bitfields. 736 737 [135]LoongArch 738 739 * Support for the following [136]-march parameters has been added: 740 + la64v1.0 741 + la64v1.1 742 + la664 743 It is now recommended to use -march=la64v1.0 as the only compiler 744 option to describe the target ISA when building binaries for 745 distribution. For more information on LoongArch ISA versions, see 746 [137]Toolchain Conventions of the LoongArch(TM) Architecture. 747 * Support for the following [138]-mtune parameters has been added: 748 + generic 749 + la664 750 * New ISA Extension 751 + LSX (Loongson SIMD Extension): Support 128-bit vector 752 instructions and the intrinsics. 753 + LASX (Loongson Advanced SIMD Extension): Support 256-bit 754 vector instructions and the intrinsics. 755 + FRECIPE: Support frecipe.{s/d} and frsqrte.{s/d} instructions 756 and the intrinsics. 757 + DIV32: Support div.w[u] and mod.w[u] instructions with inputs 758 not sign-extended. 759 + LAM_BH: Support am{swap/add}[_db].{b/h} instructions. 760 + LAMCAS: Support amcas[_db].{b/h/w/d} instructions. 761 * New Built-in Macros 762 + __loongarch_arch: Target ISA preset as specified by -march=. 763 For example, compiling with -march=la64v1.0, the value of 764 __loongarch_arch is "la64v1.0". 765 + __loongarch_tune: Processor model as specified by -mtune or 766 its default value. 767 + __loongarch_{simd,sx,asx}: These macros are not defined, or 768 defined as 1. 769 + __loongarch_simd_width: The maximum SIMD bit-width enabled by 770 the compiler. (128 for lsx, and 256 for lasx). 771 + __loongarch_frecipe: It's defined to 1 or undefined. 772 + __loongarch_div32: It's defined to 1 or undefined. 773 + __loongarch_lam_bh: It's defined to 1 or undefined. 774 + __loongarch_lamcas: It's defined to 1 or undefined. 775 + __loongarch_ld_seq_sa: It's defined to 1 or undefined. 776 + __loongarch_version_major: The minimally required LoongArch 777 ISA version (major) to run the compiled program, defined to 1 778 or undefined (iff no such version is known to the compiler). 779 + __loongarch_version_minor: The minimally required LoongArch 780 ISA version (minor) to run the compiled program, defined to 0 781 1 or undefined (iff __loongarch_version_major is undefined). 782 + __FLOAT128_TYPE: It's defined to 1. 783 * New Intrinsics 784 + [139]__builtin_thread_pointer 785 + [140]__lsx_* 786 + [141]__lasx_* 787 + [142]__frecipe_{s/d} and __frsqrte_{s/d} 788 * New Compiler Option 789 + [143]-m[no-]lsx 790 + [144]-m[no-]lasx 791 + [145]-m[no-]frecipe 792 + [146]-m[no-]div32 793 + [147]-m[no-]lam-bh 794 + [148]-m[no-]lamcas 795 + [149]-m[no-]ld-seq-sa 796 + [150]-mrecip= 797 + [151]-m[no-]recip 798 + [152]-mexplicit-relocs={none,always,auto} 799 + [153]-m[no-]relax 800 + [154]-m[no-]pass-mrelax-to-as 801 + [155]-mtls-dialect={trad,desc} 802 * Support for Ada and D. 803 * Support for libffi. 804 * Enable -free by default at -O2 or higher. 805 * Enable -fsched-pressure by default at -O1 or higher. 806 * Support the extreme code model using macro instructions (under 807 -mno-explicit-relocs). 808 * Support call36. 809 * Optimizing built-in functions for memory-model-aware atomic 810 operations using hierarchical dbar instructions. 811 * TLS descriptors support. It is not enabled by default, and can be 812 enabled with [156]-mtls-dialect=desc. The default behavior can be 813 configured with --with-tls=[trad|desc]. 814 815 [157]RISC-V 816 817 * The SLP and loop vectorizer are now enabled for RISC-V when the 818 vector extension is enabled, thanks to Ju-Zhe Zhong from 819 [158]RiVAI, Pan Li from [159]Intel, and Robin Dapp from 820 [160]Ventana Micro for contributing most of the implementation! 821 * The -mrvv-max-lmul= option has been introduced for performance 822 tuning of the loop vectorizer. The default value is 823 -mrvv-max-lmul=m1, which limits the maximum LMUL to 1. The 824 -mrvv-max-lmul=dynamic setting can dynamically select the maximum 825 LMUL value based on register pressure. 826 * Atomic code generation has been improved and is now in conformance 827 with the latest psABI specification, thanks to Patrick O'Neill from 828 [161]Rivos. 829 * Support for the vector intrinsics as specified in [162]version 1.0 830 of the RISC-V vector intrinsic specification. 831 * Support for the experimental vector crypto intrinsics as specified 832 in [163]RISC-V vector intrinsic specification, thanks to Feng Wang 833 et al. from [164]ESWIN Computing 834 * Support for the T-head vector intrinsics. 835 * Support for the scalar bitmanip and scalar crypto intrinsics, 836 thanks to Liao Shihua from [165]PLCT. 837 * Support for the large code model via option -mcmodel=large, thanks 838 to Kuan-Lin Chen from [166]Andes Technology. 839 * Support for the standard vector calling convention variant, thanks 840 to Lehua Ding from [167]RiVAI. 841 * Supports the target attribute, which allows users to compile a 842 function with specific extensions. 843 * -march= option no longer requires the architecture string to be in 844 canonical order, with only a few constraints remaining: the 845 architecture string must start with rv[32|64][i|g|e], and must use 846 an underscore as the separator after a multi-letter extension. 847 * -march=help option has been introduced to dump all supported 848 extensions. 849 * Added experimental support for the -mrvv-vector-bits=zvl option and 850 the riscv_rvv_vector_bits attribute, which specify a fixed length 851 for scalable vector types. This option is optimized for specific 852 vector core implementations; however, the code generated with this 853 option is NOT portable between the core with different VLEN, thanks 854 to Pan Li from [168]Intel. 855 * Support for TLS descriptors has been introduced, which can be 856 enabled by the -mtls-dialect=desc option. The default behavior can 857 be configured with --with-tls=[trad|desc]. 858 * Support for the TLS descriptors, this can be enabled by 859 -mtls-dialect=desc and the default behavior can be configure by 860 --with-tls=[trad|desc], and this feature require glibc 2.40, thanks 861 to Tatsuyuki Ishi from [169]Blue Whale Systems. 862 * Support for the following standard extensions has been added: 863 + Vector crypto extensions: 864 o Zvbb 865 o Zvkb 866 o Zvbc 867 o Zvkg 868 o Zvkned 869 o Zvkhna 870 o Zvkhnb 871 o Zvksed 872 o Zvksh 873 o Zvkn 874 o Zvknc 875 o Zvkng 876 o Zvks 877 o Zvksc 878 o Zvksg 879 o Zvkt 880 + Code size reduction extensions: 881 o Zca 882 o Zcb 883 o Zce 884 o Zcf 885 o Zcd 886 o Zcmp 887 o Zcmt 888 + Zicond 889 + Zfa 890 + Ztso 891 + Zvfbfmin 892 + Zvfhmin 893 + Zvfh 894 + Za64rs 895 + Za128rs 896 + Ziccif 897 + Ziccrse 898 + Ziccamoa 899 + Zicclsm 900 + Zic64b 901 + Smaia 902 + Smepmp 903 + Smstateen 904 + Ssaia 905 + Sscofpmf 906 + Ssstateen 907 + Sstc 908 + Svinval 909 + Svnapot 910 + Svpbmt 911 * Support for the following vendor extensions has been added: 912 + T-Head: 913 o XTheadVector 914 + CORE-V: 915 o XCVmac 916 o XCValu 917 o XCVelw 918 o XCVsimd 919 o XCVbi 920 + Ventana Micro: 921 o XVentanaCondops 922 * The following new CPUs are supported through the -mcpu option (GCC 923 identifiers in parentheses). 924 + SiFive's X280 (sifive-x280). 925 + SiFive's P450 (sifive-p450). 926 + SiFive's P670 (sifive-p670). 927 * The following new CPUs are supported through the -mtune option (GCC 928 identifiers in parentheses). 929 + Generic out-of-order core (generic-ooo). 930 + SiFive's P400 series (sifive-p400-series). 931 + SiFive's P600 series (sifive-p600-series). 932 + XiangShan's Nanhu microarchitecture (xiangshan-nanhu). 933 934 [170]SPARC 935 936 * The implementation of calling conventions for small structures 937 containing arrays of floating-point components has been changed in 938 64-bit mode for the Solaris port to match the implementation of the 939 vendor compiler (and the ABI). As a result, the code generated will 940 not be binary compatible with earlier releases in these cases. 941 942 Documentation improvements 943 944 * GCC's ability to provide clickable hyperlinks to the documentation 945 has been extended, so that whenever GCC refers to a command-line 946 option in quotes in a diagnostic message, the option is a clickable 947 hyperlink (assuming a suitably capable terminal). 948 949 [171]Improvements to Static Analyzer 950 951 * New warnings: 952 + [172]-Wanalyzer-infinite-loop warns about paths through the 953 code which appear to lead to an infinite loop. 954 + [173]-Wanalyzer-overlapping-buffers warns for paths through 955 the code in which overlapping buffers are passed to an API for 956 which the behavior on such buffers is undefined. 957 + [174]-Wanalyzer-undefined-behavior-strtok warns for paths 958 through the code in which a call is made to strtok with 959 undefined behavior. 960 * Previously, the analyzer's "taint" tracking to be explicitly 961 enabled via [175]-fanalyzer-checker=taint (along with 962 [176]-fanalyzer). This is now enabled by default when 963 [177]-fanalyzer is selected, thus also enabling the 6 taint-based 964 warnings: 965 + [178]-Wanalyzer-tainted-allocation-size 966 + [179]-Wanalyzer-tainted-array-index 967 + [180]-Wanalyzer-tainted-assertion 968 + [181]-Wanalyzer-tainted-divisor 969 + [182]-Wanalyzer-tainted-offset 970 + [183]-Wanalyzer-tainted-size 971 * The analyzer will now simulate API calls that expect 972 null-terminated string arguments, and will warn about code paths in 973 which such a call is made with a buffer that isn't properly 974 terminated, either due to a read of an uninitialized byte or an 975 out-of-range accesses seen before any zero byte is seen. This 976 applies to functions that use the new 977 [184]null_terminated_string_arg(PARAM_IDX) attribute, functions 978 that use the [185]format attribute, and to the library functions 979 error (parameter 3), error_at_line (parameter 5), putenv, strchr 980 (parameter 1), and strcpy (parameter 2). 981 * The analyzer now makes use of the function attribute 982 [186]alloc_size allowing [187]-fanalyzer to emit 983 [188]-Wanalyzer-allocation-size, [189]-Wanalyzer-out-of-bounds, and 984 [190]-Wanalyzer-tainted-allocation-size on execution paths 985 involving allocations using such functions. 986 * The analyzer's knowledge about the behavior of the standard library 987 has been extended to cover fopen, strcat, strncpy, and strstr. The 988 analyzer will also more precisely model the behavior of memcpy, 989 memmove, strcpy, strdup, strlen, and of various atomic built-in 990 functions. 991 * The warning [191]-Wanalyzer-out-of-bounds has been extended so 992 that, where possible, it will emit a text-based diagram visualizing 993 the spatial relationship between 994 1. the memory region that the analyzer predicts would be 995 accessed, versus 996 2. the range of memory that is valid to access 997 whether they overlap, are touching, are close or far apart; which 998 one is before or after in memory, the relative sizes involved, the 999 direction of the access (read vs write), and, in some cases, the 1000 values of data involved. 1001 Such "text art" diagrams can be controlled (or suppressed) via a 1002 new [192]-fdiagnostics-text-art-charset= option. 1003 For example, given the out-of-bounds write in strcat in: 1004 void test (void) 1005 { 1006 char buf[10]; 1007 strcpy (buf, "hello"); 1008 strcat (buf, " world!"); 1009 } 1010 1011 it emits: 1012 +----+----+----+----+----++-----+-----+-----+ 1013 |[0] |[1] |[2] |[3] |[4] || [5] | [6] | [7] | 1014 +----+----+----+----+----++-----+-----+-----+ 1015 |' ' |'w' |'o' |'r' |'l' || 'd' | '!' | NUL | 1016 +----+----+----+----+----++-----+-----+-----+ 1017 | string literal (type: 'char[8]') | 1018 +-------------------------------------------+ 1019 | | | | | | | | 1020 | | | | | | | | 1021 v v v v v v v v 1022 +-----+--------------------+----+--------------+----++-----------------+ 1023 | [0] | ... |[5] | ... |[9] || | 1024 +-----+----+----+----+----++----+--------------+----+| | 1025 | 'h' |'e' |'l' |'l' |'o' ||NUL | |after valid range| 1026 +-----+----+----+----+----++----+-------------------+| | 1027 | 'buf' (type: 'char[10]') || | 1028 +---------------------------------------------------++-----------------+ 1029 +-------------------------+-------------------------++--------+--------+ 1030 | | 1031 ---------+-------- ---------+--------- 1032 |capacity: 10 bytes| |overflow of 3 bytes| 1033 ------------------ ------------------- 1034 1035 showing that the overflow occurs partway through the second string 1036 fragment. 1037 * The analyzer will now attempt to track execution paths involving 1038 computed gotos, whereas previously it gave up on such paths. 1039 1040 [193]Improvements to SARIF support 1041 1042 * The SARIF output from [194]-fdiagnostics-format= now adds 1043 indentation and newlines to reflect the logical JSON structure of 1044 the data. The previous compact behavior can be restored via the new 1045 option [195]-fno-diagnostics-json-formatting. This also applies to 1046 the older output format named "json". 1047 * If profiling information about the compiler itself is requested via 1048 [196]-ftime-report, and a SARIF output format is requested via 1049 [197]-fdiagnostics-format=, then the timing and memory usage data 1050 is now written in JSON form into the SARIF output, rather than as 1051 plain text to stderr. 1052 1053 [198]Improvements for plugin authors 1054 1055 * GCC diagnostics have been able to have execution paths associated 1056 with them since GCC 10, but previously these were required to be 1057 single-threaded. As of GCC 14, these execution paths can have 1058 multiple named threads associated with them, with each event being 1059 associated with one of the threads. No existing GCC diagnostics 1060 take advantage of this, but GCC plugins may find this useful for 1061 their own diagnostics; an [199]example is provided in the 1062 testsuite. 1063 * GCC's diagnostics can now optionally add per-diagnostic property 1064 bags to the SARIF output, allowing plugins to capture custom data 1065 as needed with their diagnostics. 1066 1067 [200]GCC 14.1 1068 1069 This is the [201]list of problem reports (PRs) from GCC's bug tracking 1070 system that are known to be fixed in the 14.1 release. This list might 1071 not be complete (that is, it is possible that some PRs that have been 1072 fixed are not listed here). 1073 1074 [202]GCC 14.2 1075 1076 This is the [203]list of problem reports (PRs) from GCC's bug tracking 1077 system that are known to be fixed in the 14.2 release. This list might 1078 not be complete (that is, it is possible that some PRs that have been 1079 fixed are not listed here). 1080 1081 [204]GCC 14.3 1082 1083 This is the [205]list of problem reports (PRs) from GCC's bug tracking 1084 system that are known to be fixed in the 14.3 release. This list might 1085 not be complete (that is, it is possible that some PRs that have been 1086 fixed are not listed here). 1087 1088 1089 For questions related to the use of GCC, please consult these web 1090 pages and the [206]GCC manuals. If that fails, the 1091 [207]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1092 web pages and the development of GCC are welcome on our developer 1093 list at [208]gcc (a] gcc.gnu.org. All of [209]our lists have public 1094 archives. 1095 1096 Copyright (C) [210]Free Software Foundation, Inc. Verbatim copying and 1097 distribution of this entire article is permitted in any medium, 1098 provided this notice is preserved. 1099 1100 These pages are [211]maintained by the GCC team. Last modified 1101 2025-05-23. 1102 1103 References 1104 1105 1. https://gcc.gnu.org/gcc-14/porting_to.html 1106 2. https://gcc.gnu.org/onlinedocs/index.html#current 1107 3. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Zero-Length.html 1108 4. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Warning-Options.html#index-Wflex-array-member-not-at-end 1109 5. https://gcc.gnu.org/gcc-14/porting_to.html#warnings-as-errors 1110 6. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html 1111 7. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html 1112 8. https://gcc.gnu.org/gcc-14/changes.html#general 1113 9. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C-Dialect-Options.html#index-foffload-options 1114 10. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fhardened 1115 11. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fharden-control-flow-redundancy 1116 12. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Type-Attributes.html#index-hardbool-type-attribute 1117 13. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Type-Attributes.html#index-strub-type-attribute 1118 14. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Optimize-Options.html#index-finline-stringops 1119 15. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Code-Gen-Options.html#index-ftrampoline-impl 1120 16. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-null_005fterminated_005fstring_005farg-function-attribute 1121 17. https://gcc.gnu.org/gcc-14/changes.html#languages 1122 18. https://gcc.gnu.org/projects/gomp/ 1123 19. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/ 1124 20. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/Offload-Target-Specifics.html 1125 21. https://github.com/numactl/numactl 1126 22. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/libgomp/Memory-allocation.html 1127 23. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gfortran/Fortran-Dialect-Options.html#index-fopenmp-allocators 1128 24. https://gcc.gnu.org/wiki/OpenACC 1129 25. https://gcc.gnu.org/gcc-14/changes.html#ada 1130 26. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Local_005fRestrictions.html 1131 27. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Pragma-User_005fAspect_005fDefinition.html 1132 28. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-User_005fAspect.html 1133 29. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Always_005fTerminates.html 1134 30. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Ghost_005fPredicate.html 1135 31. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Exceptional_005fCases.html 1136 32. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Aspect-Side_005fEffects.html 1137 33. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Instrumentation-Options.html#index-fharden-compares 1138 34. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Security-Hardening-Features.html 1139 35. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_ugn/Style-Checking.html#index--gnatyz-_0028gcc_0029 1140 36. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_ugn/Alphabetical-List-of-All-Switches.html#index--gnateH-_0028gcc_0029 1141 37. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/Pragma-Storage_005fModel.html 1142 38. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gnat_rm/String-interpolation.html 1143 39. https://gcc.gnu.org/gcc-14/changes.html#c-family 1144 40. https://gcc.gnu.org/gcc-14/changes.html#c 1145 41. https://gcc.gnu.org/gcc-14/changes.html#cxx 1146 42. https://wg21.link/P1854R4 1147 43. https://gcc.gnu.org/PR110341 1148 44. https://wg21.link/P2752R3 1149 45. https://gcc.gnu.org/PR110346 1150 46. https://wg21.link/P2361R6 1151 47. https://gcc.gnu.org/PR110342 1152 48. https://wg21.link/P2738R1 1153 49. https://gcc.gnu.org/PR110344 1154 50. https://wg21.link/P2741R3 1155 51. https://gcc.gnu.org/PR110348 1156 52. https://wg21.link/P2169R4 1157 53. https://gcc.gnu.org/PR110349 1158 54. https://wg21.link/P2864R2 1159 55. https://wg21.link/P2748R5 1160 56. https://gcc.gnu.org/PR114455 1161 57. https://wg21.link/P2809R3 1162 58. https://gcc.gnu.org/PR114462 1163 59. https://wg21.link/P0847R7 1164 60. https://gcc.gnu.org/PR102609 1165 61. https://wg21.link/P2280R4 1166 62. https://gcc.gnu.org/PR106650 1167 63. https://wg21.link/P2564R3 1168 64. https://gcc.gnu.org/PR107687 1169 65. https://wg21.link/P2582R1 1170 66. https://gcc.gnu.org/PR106653 1171 67. https://wg21.link/cwg532 1172 68. https://wg21.link/cwg976 1173 69. https://wg21.link/cwg2262 1174 70. https://wg21.link/cwg2359 1175 71. https://wg21.link/cwg2386 1176 72. https://wg21.link/cwg2406 1177 73. https://wg21.link/cwg2543 1178 74. https://wg21.link/cwg2586 1179 75. https://wg21.link/cwg2735 1180 76. https://wg21.link/cwg2799 1181 77. https://gcc.gnu.org/PR100157 1182 78. https://cplusplus.github.io/CWG/issues/2256.html 1183 79. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdangling-reference 1184 80. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/testsuite/g%2B%2B.dg/opt/nrv23.C;h=9e1253cd830a84ad4de5ff3076a07c543afe344f;hb=7e0b65b239c3a0d68ce94896b236b03de666ffd6 1185 81. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Warning-Options.html#index-Wnrvo 1186 82. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=4d935f52b0d5c00fcc154461b87415ebd8791a94 1187 83. https://gcc.gnu.org/PR110164 1188 84. https://gcc.gnu.org/PR89442 1189 85. https://gcc.gnu.org/PR110064 1190 86. https://gcc.gnu.org/PR70331 1191 87. https://gcc.gnu.org/PR96630 1192 88. https://gcc.gnu.org/PR98675 1193 89. https://gcc.gnu.org/PR71954 1194 90. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=85ad41a494e31311f8a6b2dbe930a128c5e85840 1195 91. https://gcc.gnu.org/PR111069 1196 92. https://gcc.gnu.org/PR52953 1197 93. https://gcc.gnu.org/PR91483 1198 94. https://gcc.gnu.org/PR18474 1199 95. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/C_002b_002b-Attributes.html#index-cold-type-attribute 1200 96. https://gcc.gnu.org/PR79620 1201 97. https://gcc.gnu.org/PR79378 1202 98. https://gcc.gnu.org/PR83167 1203 99. https://gcc.gnu.org/PR96917 1204 100. https://wg21.link/P1689R5 1205 101. https://gcc.gnu.org/gcc-14/changes.html#libstdcxx 1206 102. https://gcc.gnu.org/gcc-14/changes.html#d 1207 103. https://dlang.org/changelog/2.108.1.html 1208 104. https://gcc.gnu.org/gcc-14/changes.html#fortran 1209 105. https://gcc.gnu.org/gcc-14/changes.html#modula2 1210 106. https://gcc.gnu.org/gcc-14/changes.html#rust 1211 107. https://gcc.gnu.org/gcc-14/changes.html#jit 1212 108. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/types.html#c.gcc_jit_type_get_restrict 1213 109. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-25 1214 110. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-26 1215 111. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_sizeof 1216 112. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/jit/topics/compatibility.html#libgccjit-abi-27 1217 113. https://gcc.gnu.org/gcc-14/changes.html#targets 1218 114. https://gcc.gnu.org/gcc-14/changes.html#aarch64 1219 115. https://github.com/ARM-software/acle/blob/main/main/acle.md#sme-language-extensions-and-intrinsics 1220 116. https://github.com/ARM-software/acle/blob/main/main/acle.md#function-multi-versioning 1221 117. https://github.com/ARM-software/acle/blob/main/main/acle.md#memory-prefetch-intrinsics 1222 118. https://github.com/ARM-software/acle/blob/main/main/acle.md#special-register-intrinsics 1223 119. https://github.com/ARM-software/acle/blob/main/main/acle.md#neon-sve-bridge 1224 120. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AArch64-Options.html 1225 121. https://gcc.gnu.org/gcc-14/changes.html#amdgcn 1226 122. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 1227 123. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 1228 124. https://gcc.gnu.org/gcc-14/changes.html#arm 1229 125. https://gcc.gnu.org/gcc-14/changes.html#avr 1230 126. https://github.com/avrdudes/avr-libc/issues/931 1231 127. https://sourceware.org/PR31124 1232 128. https://gcc.gnu.org/wiki/avr-gcc#spec-files 1233 129. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrxmega3 1234 130. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrtiny 1235 131. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/AVR-Options.html#avrtiny 1236 132. https://gcc.gnu.org/gcc-14/changes.html#x86 1237 133. https://gcc.gnu.org/gcc-11/changes.html 1238 134. https://gcc.gnu.org/gcc-14/changes.html#mcore 1239 135. https://gcc.gnu.org/gcc-14/changes.html#loongarch 1240 136. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-march-7 1241 137. https://github.com/loongson/la-toolchain-conventions/ 1242 138. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtune-8 1243 139. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Base-Built-in-Functions.html 1244 140. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-SX-Vector-Intrinsics.html 1245 141. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-ASX-Vector-Intrinsics.html 1246 142. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Base-Built-in-Functions.html 1247 143. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-ml_005ba_005dsx 1248 144. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-ml_005ba_005dsx 1249 145. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mfrecipe 1250 146. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mdiv32 1251 147. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mlam-bh 1252 148. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mlamcas 1253 149. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mld-seq-sa 1254 150. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrecip_003dopt 1255 151. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrecip 1256 152. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mexplicit-relocs-1 1257 153. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mrelax-2 1258 154. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mpass-mrelax-to-as 1259 155. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtls-dialect-1 1260 156. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/LoongArch-Options.html#index-mtls-dialect-1 1261 157. https://gcc.gnu.org/gcc-14/changes.html#riscv 1262 158. https://rivai-ic.com.cn/ 1263 159. https://www.intel.com/ 1264 160. https://www.ventanamicro.com/ 1265 161. https://www.rivosinc.com/ 1266 162. https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/v1.0.x 1267 163. https://github.com/riscv-non-isa/rvv-intrinsic-doc/pull/234 1268 164. https://eswincomputing.com/ 1269 165. https://plctlab.org/ 1270 166. https://www.andestech.com/ 1271 167. https://rivai-ic.com.cn/ 1272 168. https://www.intel.com/ 1273 169. https://bluewhale.systems/ 1274 170. https://gcc.gnu.org/gcc-14/changes.html#sparc 1275 171. https://gcc.gnu.org/gcc-14/changes.html#analyzer 1276 172. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-infinite-loop 1277 173. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-overlapping-buffers 1278 174. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-undefined-behavior-strtok 1279 175. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 1280 176. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1281 177. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1282 178. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 1283 179. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 1284 180. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-assertion 1285 181. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 1286 182. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 1287 183. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 1288 184. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-null_005fterminated_005fstring_005farg-function-attribute 1289 185. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-format-function-attribute 1290 186. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Common-Function-Attributes.html#index-alloc_005fsize-function-attribute 1291 187. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1292 188. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-allocation-size 1293 189. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 1294 190. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 1295 191. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 1296 192. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-text-art-charset 1297 193. https://gcc.gnu.org/gcc-14/changes.html#sarif 1298 194. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1299 195. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-json-formatting 1300 196. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Developer-Options.html#index-ftime-report 1301 197. https://gcc.gnu.org/onlinedocs/gcc-14.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1302 198. https://gcc.gnu.org/gcc-14/changes.html#plugins 1303 199. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a1e9f3ed7aa49adad02190ace0614e0b37fc089 1304 200. https://gcc.gnu.org/gcc-14/changes.html#14.1 1305 201. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.0 1306 202. https://gcc.gnu.org/gcc-14/changes.html#14.2 1307 203. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.2 1308 204. https://gcc.gnu.org/gcc-14/changes.html#14.3 1309 205. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=14.3 1310 206. https://gcc.gnu.org/onlinedocs/ 1311 207. mailto:gcc-help (a] gcc.gnu.org 1312 208. mailto:gcc (a] gcc.gnu.org 1313 209. https://gcc.gnu.org/lists.html 1314 210. https://www.fsf.org/ 1315 211. https://gcc.gnu.org/about.html 1316 ====================================================================== 1317 http://gcc.gnu.org/gcc-13/index.html 1318 1319 GCC 13 Release Series 1320 1321 May 21, 2024 1322 1323 The GCC developers are pleased to announce the release of GCC 13.3. 1324 1325 This release is a bug-fix release, containing fixes for regressions in 1326 GCC 13.2 relative to previous releases of GCC. 1327 1328 Release History 1329 1330 GCC 13.3 1331 May 21, 2024 ([1]changes, [2]documentation) 1332 1333 GCC 13.2 1334 July 27, 2023 ([3]changes, [4]documentation) 1335 1336 GCC 13.1 1337 April 26, 2023 ([5]changes, [6]documentation) 1338 1339 References and Acknowledgements 1340 1341 GCC used to stand for the GNU C Compiler, but since the compiler 1342 supports several other languages aside from C, it now stands for the 1343 GNU Compiler Collection. 1344 1345 The GCC developers would like to thank the numerous people that have 1346 contributed new features, improvements, bug fixes, and other changes as 1347 well as test results to GCC. This [7]amazing group of volunteers is 1348 what makes GCC successful. 1349 1350 For additional information about GCC please refer to the [8]GCC project 1351 web site or contact the [9]GCC development mailing list. 1352 1353 To obtain GCC please use [10]our mirror sites or [11]our version 1354 control system. 1355 1356 1357 For questions related to the use of GCC, please consult these web 1358 pages and the [12]GCC manuals. If that fails, the 1359 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1360 web pages and the development of GCC are welcome on our developer 1361 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 1362 archives. 1363 1364 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 1365 distribution of this entire article is permitted in any medium, 1366 provided this notice is preserved. 1367 1368 These pages are [17]maintained by the GCC team. Last modified 1369 2024-05-30. 1370 1371 References 1372 1373 1. https://gcc.gnu.org/gcc-13/changes.html 1374 2. https://gcc.gnu.org/onlinedocs/13.3.0/ 1375 3. https://gcc.gnu.org/gcc-13/changes.html 1376 4. https://gcc.gnu.org/onlinedocs/13.2.0/ 1377 5. https://gcc.gnu.org/gcc-13/changes.html 1378 6. https://gcc.gnu.org/onlinedocs/13.1.0/ 1379 7. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Contributors.html 1380 8. https://gcc.gnu.org/index.html 1381 9. mailto:gcc (a] gcc.gnu.org 1382 10. https://gcc.gnu.org/mirrors.html 1383 11. https://gcc.gnu.org/git.html 1384 12. https://gcc.gnu.org/onlinedocs/ 1385 13. mailto:gcc-help (a] gcc.gnu.org 1386 14. mailto:gcc (a] gcc.gnu.org 1387 15. https://gcc.gnu.org/lists.html 1388 16. https://www.fsf.org/ 1389 17. https://gcc.gnu.org/about.html 1390 ====================================================================== 1391 http://gcc.gnu.org/gcc-13/changes.html 1392 1393 GCC 13 Release Series 1394 Changes, New Features, and Fixes 1395 1396 This page is a "brief" summary of some of the huge number of 1397 improvements in GCC 13. You may also want to check out our [1]Porting 1398 to GCC 13 page and the [2]full GCC documentation. 1399 1400 Caveats 1401 1402 * OpenMP offloading to Intel MIC has been removed. 1403 * The support for the cr16-elf, tilegx*-linux, tilepro*-linux, 1404 hppa[12]*-*-hpux10*, hppa[12]*-*-hpux11* and m32c-rtems 1405 configurations has been removed. 1406 * Support for Solaris 11.3 (*-*-solaris2.11.3) has been declared 1407 obsolete. The next release of GCC will have corresponding code 1408 permanently removed. Details can be found in the [3]announcement. 1409 * Support for emitting the STABS debugging format (including the 1410 -gstabs and -gxcoff options) has been removed. (This means the dbx 1411 debugger is no longer supported, either.) 1412 * Legacy debug info compression option -gz=zlib-gnu was removed and 1413 the option is ignored right now. 1414 * [4]-Warray-bounds=2 will no longer issue warnings for out of bounds 1415 accesses to trailing struct members of one-element array type 1416 anymore. Instead it diagnoses accesses to trailing arrays according 1417 to [5]-fstrict-flex-arrays. 1418 * [6]-fanalyzer is still only suitable for analyzing C code. In 1419 particular, using it on C++ is unlikely to give meaningful output. 1420 * In the arm port, support for the iWMMXt extensions, enabled through 1421 -mcpu=iwmmxt, has been deprecated and will be removed in a future 1422 release. This includes support for the [7]iWMMXt built-in 1423 functions. 1424 * For C++, construction of the global iostream objects std::cout, 1425 std::cin, etc. is now done inside the standard library, instead of 1426 in every source file that includes the <iostream> header. This 1427 change improves the start-up performance of C++ programs, but it 1428 means that code compiled with GCC 13.1 will crash if the correct 1429 version of libstdc++.so is not used at run time. See the 1430 [8]documentation about using the right libstdc++.so at run time. 1431 Future GCC releases will mitigate the problem so that the program 1432 cannot be run at all with an older libstdc++.so. 1433 1434 [9]General Improvements 1435 1436 * [10]OpenMP 1437 + Reverse offload is now supported with AMD GCN and nvptx 1438 devices. Additionally, the requires handling has been improved 1439 and all clauses are now accepted. If a requirement cannot be 1440 fulfilled for an accessible device, this device is excluded 1441 from the list of available devices. This may imply that the 1442 only device left is the host (the initial device). In 1443 particular, unified_address and unified_shared_memory are 1444 unsupported by all non-host devices. 1445 + OpenMP 5.0: Fortran now supports some non-rectangular loop 1446 nests; for C/C++, the support was added in GCC 11. 1447 + The following OpenMP 5.1 features have been added: the 1448 omp_all_memory reserved locator, the inoutset modifier to the 1449 depend clause, the nowait clause for the taskwait directive 1450 and the omp_target_is_accessible, omp_target_memcpy_async, 1451 omp_target_memcpy_rect_async and omp_get_mapped_ptr API 1452 routines. The assume and assumes directives, the begin/end 1453 declare target syntax in C/C++ and device-specific ICV 1454 settings with environment variables are now supported. 1455 + Initial support for OpenMP 5.2 features has been added: 1456 firstprivate and allocate clauses on the scope construct; the 1457 OpenMP 5.2 syntax of the linear clause; new enum/constants 1458 omp_initial_device and omp_invalid_device; and optionally 1459 omitting the map-type in target enter/exit data. The enter 1460 clause (as alias for to) has been added to the declare target 1461 directive. Also added have been the omp_in_explicit_task 1462 routine and the doacross clause as alias for depend with 1463 source/sink modifier. 1464 + The _ALL suffix to the device-scope environment variables 1465 added in Technical Report (TR11) is already handled. 1466 + For user defined allocators requesting high bandwidth or large 1467 capacity memspaces or interleaved partitioning, the 1468 [11]memkind library is used, if available at run time. 1469 * AddressSanitizer defaults to detect_stack_use_after_return=1 on 1470 GNU/Linux targets. For compatibility, it can be disabled with env 1471 ASAN_OPTIONS=detect_stack_use_after_return=0. 1472 * New debug info compression option value -gz=zstd has been added. 1473 * Link-time optimization improvements: 1474 + LTO supports the newly added jobserver of GNU make jobserver 1475 that uses named pipes (--jobserver-style=fifo) by default. 1476 + If make's jobserver is active, parallel LTO WPA streaming 1477 communicates with it and thus avoids system overcommitting. 1478 * -Ofast, -ffast-math and -funsafe-math-optimizations will no longer 1479 add startup code to alter the floating-point environment when 1480 producing a shared object with -shared. 1481 * GCC can now emit its diagnostics using [12]SARIF. This is a 1482 JSON-based format suited for capturing the results of static 1483 analysis tools (like GCC's [13]-fanalyzer), but it can also be used 1484 to capture other GCC warnings and errors in a machine-readable 1485 format. Specifically, the [14]-fdiagnostics-format= option has been 1486 extended to support these new values: 1487 + -fdiagnostics-format=sarif-stderr 1488 + -fdiagnostics-format=sarif-file 1489 + -fdiagnostics-format=json-stderr, a synonym for the existing 1490 -fdiagnostics-format=json 1491 + -fdiagnostics-format=json-file 1492 where the json-prefixed variants refer to GCC's own JSON diagnostic 1493 format. 1494 * Support for profiling and test coverage in freestanding 1495 environments has been added, see also [15]Profiling and Test 1496 Coverage in Freestanding Environments. 1497 * New options -fharden-compares and -fharden-conditional-branches to 1498 verify compares and conditional branches, to detect some 1499 power-deprivation hardware attacks, using reversed conditions. 1500 1501 [16]New Languages and Language specific improvements 1502 1503 [17]Ada 1504 1505 * Traceback support added in RTEMS for the PPC ELF and ARM 1506 architectures. 1507 * Support for versions older than VxWorks 7 has been removed. 1508 * General improvements to the contracts in the standard libraries. 1509 * Addition of GNAT.Binary_Search. 1510 * Further additions and fixes for the Ada 2022 specification. 1511 * The Pragma SPARK_Mode=>Auto is now accepted. Contract analysis has 1512 been further improved. 1513 * Documentation improvements. 1514 1515 [18]C family 1516 1517 * New warnings: 1518 + [19]-Wxor-used-as-pow warns about uses of ^, the exclusive or 1519 operator, where it appears the user meant exponentiation 1520 ([20]PR90885) 1521 * Three new function attributes for documenting int arguments that 1522 are file descriptors: 1523 + [21]__attribute__((fd_arg(N))) 1524 + [22]__attribute__((fd_arg_read(N))) 1525 + [23]__attribute__((fd_arg_write(N))) 1526 These are used by [24]-fanalyzer to detect misuses of file 1527 descriptors. 1528 * A new statement attribute for C++23 [25]P1774R8 Portable 1529 assumptions support also in C or older C++: 1530 [26]__attribute__((assume(EXPR))); 1531 * GCC can now control when to treat the trailing array of a structure 1532 as a flexible array member for the purpose of accessing the 1533 elements of such an array. By default, all trailing arrays in 1534 aggregates are treated as flexible array members. Use the new 1535 command-line option [27]-fstrict-flex-arrays to control which array 1536 members are treated as flexible arrays. 1537 1538 [28]C 1539 1540 * Several C23 features have been implemented: 1541 + [29]N3042, Introduce the nullptr constant 1542 + [30]N2963, Enhanced Enumerations (fixed underlying types) 1543 + [31]N2975, Relax requirements for variadic parameter lists 1544 + [32]N3007, Type inference for object definitions (auto) 1545 + [33]N3018, The constexpr specifier for object definitions 1546 + [34]N3038, Introduce storage-class specifiers for compound 1547 literals 1548 + typeof (previously supported as an extension) and 1549 typeof_unqual 1550 + New keywords alignas, alignof, bool, false, static_assert, 1551 thread_local, true 1552 + [35]N2764, The noreturn attribute 1553 + Support for empty initializer braces 1554 + __STDC_VERSION_*_H__ header version macros 1555 + Removal of ATOMIC_VAR_INIT 1556 + unreachable macro in <stddef.h> 1557 + Removal of trigraphs 1558 + Removal of unprototyped functions 1559 + printf and scanf format checking with [36]-Wformat for %wN and 1560 %wfN format length modifiers 1561 + [37]N2836, Identifier Syntax using Unicode Standard Annex 31 1562 * In addition to those C23 features, existing features adopted in C23 1563 have been adjusted to follow C23 requirements and are not diagnosed 1564 with -std=c2x -Wpedantic. 1565 * New warnings: 1566 + [38]-Wenum-int-mismatch warns about mismatches between an 1567 enumerated type and an integer type ([39]PR105131) 1568 1569 [40]C++ 1570 1571 * Excess precision support (which has been available in C since GCC 1572 4.5) has been implemented for C++ as well. It is enabled by default 1573 in strict standard modes like -std=c++17, where it defaults to 1574 -fexcess-precision=standard, while in GNU standard modes like 1575 -std=gnu++20 it defaults to -fexcess-precision=fast. The option 1576 mainly affects IA-32/x86-64 using x87 math and in some cases on 1577 Motorola 68000, where float and double expressions are evaluated in 1578 long double precision and S/390, System z, IBM z Systems where 1579 float expressions are evaluated in double precision. Also, on 1580 several architectures where std::float16_t or std::bfloat16_t types 1581 are supported those are evaluated in float precision. 1582 -fexcess-precision=fast restores previous behavior. 1583 * Several C++23 features have been implemented: 1584 + [41]P2324R1, Labels at the end of compound statements 1585 ([42]PR103539) 1586 + [43]P2255R2, A type trait to detect reference binding to 1587 temporary ([44]PR104477) 1588 + [45]P2327R1, De-deprecating volatile compound operations 1589 + [46]P2437R1, Support for #warning ([47]PR106646) 1590 + [48]P2290R3, Delimited escape sequences ([49]PR106645) 1591 + [50]P2071R2, Named universal character escapes ([51]PR106648) 1592 + [52]P2513R3, char8_t Compatibility and Portability Fix 1593 ([53]PR106656) 1594 + [54]P1169R4, static operator() ([55]PR106651) 1595 + [56]P2266R3, Simpler implicit move ([57]PR101165) 1596 + [58]P2468R2, The Equality Operator You Are Looking For 1597 ([59]PR106644) 1598 + [60]P2362R3, Remove non-encodable 1599 wide character literals and multicharacter wide character lite 1600 rals ([61]PR106647) 1601 + [62]P2448R2, Relaxing some constexpr restrictions 1602 ([63]PR106649) 1603 + [64]P1467R9, Extended floating-point types and standard names 1604 ([65]PR106652) 1605 + [66]P1774R8, Portable assumptions ([67]PR106654) 1606 + [68]P2295R6, Support for 1607 UTF-8 as a portable source file encoding ([69]PR106655) 1608 + [70]P2589R1, static operator[] ([71]PR107684) 1609 * New warnings: 1610 + [72]-Wself-move warns when a value is moved to itself with 1611 std::move ([73]PR81159) 1612 + [74]-Wdangling-reference warns when a reference is bound to a 1613 temporary whose lifetime has ended ([75]PR106393) 1614 * The [76]-Wpessimizing-move and [77]-Wredundant-move warnings have 1615 been extended to warn in more contexts. 1616 * The [78]-nostdlib++ option has been added, to enable linking with 1617 g++ without implicitly linking in the C++ standard library. 1618 1619 [79]Runtime Library (libstdc++) 1620 1621 * Improved experimental support for C++20, including: 1622 + <format> header and std::format. 1623 + std::chrono::utc_clock and other clocks, time zones, and 1624 std::format support in the <chrono> header. 1625 * Improved experimental support for C++23, including: 1626 + Additions to the <ranges> header: views::zip, 1627 views::zip_transform, views::adjacent, 1628 views::adjacent_transform views::pairwise, views::slide, 1629 views::chunk, views::chunk_by, views::repeat, views::chunk_by, 1630 views::cartesian_product, views::as_rvalue, views::enumerate, 1631 views::as_const. 1632 + Additions to the <algorithm> header: ranges::contains, 1633 ranges::contains_subrange, ranges::iota, ranges::find_last, 1634 ranges::find_last_if, ranges::find_last_if_not, 1635 ranges::fold_left, ranges::fold_left_first, 1636 ranges::fold_right, ranges::fold_right_last, 1637 ranges::fold_left_with_iter, 1638 ranges::fold_left_first_with_iter. 1639 + Monadic operations for std::expected. 1640 + Constexpr std::bitset, std::to_chars and std::from_chars. 1641 + Library support for extended floating-point types. 1642 * Support for the <experimental/scope> header from v3 of the Library 1643 Fundamentals Technical Specification. 1644 * Support for the <experimental/synchronized_value> header from v2 of 1645 the Concurrency Technical Specification. 1646 * Support for many previously unavailable features in freestanding 1647 mode, thanks to Arsen Arsenovic. For example, std::tuple is now 1648 available for freestanding compilation. The freestanding subset 1649 contains all the components made freestanding by [80]P1642, and 1650 libstdc++ adds more components such as std::array and 1651 std::string_view. Additionally, libstdc++ now respects the 1652 -ffreestanding compiler option and so it is not necessary to build 1653 a separate freestanding installation of libstdc++. Compiling with 1654 -ffreestanding restricts the available features to the freestanding 1655 subset, even if libstdc++ was built as a full, hosted 1656 implementation. 1657 1658 [81]D 1659 1660 * Support for the D programming language has been updated to version 1661 2.103.1 of the language and run-time library. Full changelog for 1662 this release and previous releases can be found on the 1663 [82]dlang.org website. 1664 * The following GCC attributes are now recognized and available from 1665 the gcc.attributes module with short-hand aliases for convenience: 1666 + @attribute("no_sanitize", arguments) or 1667 @no_sanitize(arguments). 1668 + @attribute("register") or @register. 1669 + @attribute("simd") or @simd. 1670 + @attribute("simd_clones", mask) or @simd_clones(mask). 1671 + @attribute("visibility", arguments) or @visibility(arguments). 1672 * New aliases have been added to gcc.attributes for compatibility 1673 with ldc.attributes. 1674 + The @hidden attribute is an alias for @attribute("visibility", 1675 "hidden"). 1676 + The @noSanitize attribute is an alias for 1677 @attribute("no_sanitize"). 1678 * Vector operation intrinsics prefetch, loadUnaligned, 1679 storeUnaligned, shuffle, shufflevector, extractelement, 1680 insertelement, convertvector, and blendvector have been added to 1681 the gcc.simd module. 1682 * New warnings: 1683 + [83]-Wbuiltin-declaration-mismatch= warns when a built-in 1684 function is declared with the wrong signature. 1685 + [84]-Wmismatched-special-enum warns when a special enum is 1686 declared with the wrong base type. 1687 * New version identifier D_Optimized is now predefined when the -O 1688 option, or any higher optimization level is used. 1689 * The predefinition of version D_Exceptions can now by controlled by 1690 the option -fexception. 1691 * The predefinition of version D_TypeInfo can now by controlled by 1692 the option -frtti. 1693 * The -fdebug= and -fversion= compiler switches no longer accept an 1694 integer argument. 1695 1696 [85]Fortran 1697 1698 * Finalization is now fully supported. 1699 1700 [86]Go 1701 1702 * GCC 13, like GCC 12, provides a complete implementation of the Go 1703 1.18 user packages. 1704 * Although Go 1.18 includes support for generic programming, that 1705 support is not yet available in GCC. 1706 1707 [87]Modula-2 1708 1709 * Support for the language Modula-2 has been added. This includes 1710 support for the ISO/IEC 10514-1, PIM2, PIM3, PIM4 dialects together 1711 with a complete set of ISO/IEC 10514-1 and PIM libraries. 1712 * The <* noreturn *> attribute is supported with the -Wreturn-type 1713 [88]option. 1714 1715 [89]New Targets and Target Specific Improvements 1716 1717 [90]AArch64 1718 1719 * The AArch64 target now supports Decimal Floating-point in the BID 1720 format through the libbid library. 1721 * A number of new CPUs are supported through the -mcpu and -mtune 1722 options (GCC identifiers in parentheses). 1723 + Ampere-1A (ampere1a). 1724 + Arm Cortex-A715 (cortex-a715). 1725 + Arm Cortex-X1C (cortex-x1c). 1726 + Arm Cortex-X3 (cortex-x3). 1727 + Arm Neoverse V2 (neoverse-v2). 1728 * Support has been added for the armv9.1-a, armv9.2-a and armv9.3-a 1729 arguments to the -march= option. 1730 * The FEAT_LRCPC feature is now supported by generating the LDAPR 1731 instructions for C and C++ atomic loads with an acquire memory 1732 model. This is enabled when compiling with the +rcpc extension to 1733 -march or a CPU target that supports this feature. 1734 * The FEAT_CSSC feature from the 2022 Arm Architecture extensions is 1735 supported through the +cssc extension option. When enabled, scalar 1736 operations like integer minimum, maximum, absolute value, count 1737 trailing zeroes (__builtin_ctz), population count 1738 (__builtin_popcount) can be implemented in a single instruction. 1739 * The FEAT_LSE2 feature is now supported through libatomic and 1740 provides lockless 16-byte atomics on systems that implement it. 1741 1742 [91]AMD Radeon (GCN) 1743 1744 * Support for the Instinct MI200 series devices ([92] gfx90a) has 1745 been added. 1746 * SIMD vectorization support has been improved; this and 1747 stack-handling changes [93]require newlib 4.3.0 (or newer). 1748 1749 [94]arm 1750 1751 * A number of new CPUs are supported through the -mcpu and -mtune 1752 options (GCC identifiers in parentheses). 1753 + STAR-MC1 (star-mc1). 1754 + Arm Cortex-X1C (cortex-x1c). 1755 + Arm Cortex-M85 (cortex-m85). 1756 * Support has been added for the M-profile PACBTI extension that can 1757 help harden the generated code against return-oriented and 1758 jump-oriented attacks. It can be enabled through the 1759 -mbranch-protection= option. 1760 1761 [95]AVR 1762 1763 * Support for the following devices has been added in v13.3: 1764 + ATtiny102, ATtiny104, ATtiny424, ATtiny426, ATtiny427, 1765 ATtiny824, ATtiny826, ATtiny827, ATtiny1624, ATtiny1626, 1766 ATtiny1627, ATtiny3224, ATtiny3226, ATtiny3227, AVR32DD14, 1767 AVR32DD20, AVR32DD28, AVR32DD32, AVR32DU14, AVR32DU20, 1768 AVR32DU28, AVR32DU32, AVR32EA28, AVR32EA32, AVR32EA48, 1769 AVR64DD14, AVR64DD20, AVR64DD28, AVR64DD32, AVR64DU28, 1770 AVR64DU32, AVR64EA28, AVR64EA32, AVR64EA48, ATA5787, ATA5835, 1771 ATA5700M322. 1772 * Support for the following devices has been added in v13.4: 1773 + AVR32SD20, AVR32SD28, AVR32SD32, AVR64SD28, AVR64SD32, 1774 AVR64SD48. 1775 1776 [96]IA-32/x86-64 1777 1778 * For both C and C++ the __bf16 type is supported on x86 systems with 1779 SSE2 and above enabled. 1780 * Use this __bf16 type for AVX512BF16 intrinsics instead of 1781 __bfloat16 which is typedef for short. __bf16 is now part of the 1782 x86 psABI. Users need to adjust their AVX512BF16-related source 1783 code when upgrading to GCC 13. 1784 * New ISA extension support for Intel AMX-COMPLEX was added. 1785 AMX-COMPLEX intrinsics are available via the -mamx-complex compiler 1786 switch. 1787 * New ISA extension support for Intel AMX-FP16 was added. AMX-FP16 1788 intrinsics are available via the -mamx-fp16 compiler switch. 1789 * New ISA extension support for Intel AVX-IFMA was added. AVX-IFMA 1790 intrinsics are available via the -mavxifma compiler switch. 1791 * New ISA extension support for Intel AVX-NE-CONVERT was added. 1792 AVX-NE-CONVERT intrinsics are available via the -mavxneconvert 1793 compiler switch. 1794 * New ISA extension support for Intel AVX-VNNI-INT8 was added. 1795 AVX-VNNI-INT8 intrinsics are available via the -mavxvnniint8 1796 compiler switch. 1797 * New ISA extension support for Intel CMPccXADD was added. CMPccXADD 1798 intrinsics are available via the -mcmpccxadd compiler switch. 1799 * New ISA extension support for Intel PREFETCHI was added. PREFETCHI 1800 intrinsics are available via the -mprefetchi compiler switch. 1801 * New ISA extension support for Intel RAO-INT was added. RAO-INT 1802 intrinsics are available via the -mraoint compiler switch. 1803 * GCC now supports the Intel CPU named Raptor Lake through 1804 -march=raptorlake. Raptor Lake is based on Alder Lake. 1805 * GCC now supports the Intel CPU named Meteor Lake through 1806 -march=meteorlake. Meteor Lake is based on Alder Lake. 1807 * GCC now supports the Intel CPU named Sierra Forest through 1808 -march=sierraforest. Based on ISA extensions enabled on Alder Lake, 1809 the switch further enables the AVX-IFMA, AVX-NE-CONVERT, 1810 AVX-VNNI-INT8, CMPccXADD, ENQCMD and UINTR ISA extensions. 1811 * GCC now supports the Intel CPU named Grand Ridge through 1812 -march=grandridge. Grand Ridge is based on Sierra Forest. 1813 * GCC now supports the Intel CPU named Emerald Rapids through 1814 -march=emeraldrapids. Emerald Rapids is based on Sapphire Rapids. 1815 * GCC now supports the Intel CPU named Granite Rapids through 1816 -march=graniterapids. Based on Sapphire Rapids, the switch further 1817 enables the AMX-FP16 and PREFETCHI ISA extensions. 1818 * GCC now supports the Intel CPU named Granite Rapids D through 1819 -march=graniterapids-d. Based on Granite Rapids, the switch further 1820 enables the AMX-COMPLEX ISA extensions. 1821 * GCC now supports AMD CPUs based on the znver4 core via 1822 -march=znver4. The switch makes GCC consider using 512-bit vectors 1823 when auto-vectorizing. 1824 1825 [97]LoongArch 1826 1827 * New features 1828 + The new command-line option -mexplicit-relocs decides whether 1829 to use the assembler relocation operator when dealing with 1830 symbolic addresses. It is enabled by default if a compatible 1831 assembler (binutils 2.40 or later) is present at GCC build 1832 time. 1833 + The new command-line option -mdirect-extern-access can be used 1834 to prevent accessing external symbols through GOT. 1835 + The new variable attribute [98]model has been added. 1836 * Built-in functions 1837 + The rint and copysign mathematical builtins (and their float 1838 variants) are now implemented as inline LoongArch intrinsics. 1839 + The lrint, logb, scalbln, scalbn and ldexp mathematical 1840 builtins (and their float variants) are now implemented as 1841 inline LoongArch intrinsics when using -fno-math-errno. 1842 + The lceil and lfloor mathematical builtins (and their float 1843 variants) are now implemented as inline LoongArch intrinsics 1844 when using -ffp-int-builtin-inexact. 1845 * Subprojects Support 1846 + libvtv now supports LoongArch. 1847 + libitm now supports LoongArch. 1848 + Address sanitizers other than HWASan and TSan are now 1849 supported on LoongArch. 1850 1851 [99]NVPTX 1852 1853 * The default value for the [100]-march option can be now changed 1854 when [101]building GCC using the [102]--with-arch= configure 1855 option. GCC's target libraries are then build both with sm_30 and 1856 the specified target architecture. If not specified, GCC defaults 1857 to sm_30. 1858 1859 [103]RISC-V 1860 1861 * Support for vector intrinsics as specified in [104]version 0.11 of 1862 the RISC-V vector intrinsic specification, thanks Ju-Zhe Zhong from 1863 [105]RiVAI for contributing most of implementation. 1864 * Support for the following standard extensions has been added: 1865 + Zawrs 1866 + Zbkb 1867 + Zbkc 1868 + Zbkx 1869 + Zdinx 1870 + Zfinx 1871 + Zfh 1872 + Zfhmin 1873 + Zhinx 1874 + Zhinxmin 1875 + Zicbom 1876 + Zicbop 1877 + Zicboz 1878 + Zknd 1879 + Zkne 1880 + Zksed 1881 + Zksh 1882 + Zmmul 1883 * Support for the following vendor extensions has been added: 1884 + XTheadBa 1885 + XTheadBb 1886 + XTheadBs 1887 + XTheadCmo 1888 + XTheadCondMov 1889 + XTheadFMemIdx 1890 + XTheadFmv 1891 + XTheadInt 1892 + XTheadMac 1893 + XTheadMemIdx 1894 + XTheadMemPair 1895 + XTheadSync 1896 * The following new CPUs are supported through the -mcpu option (GCC 1897 identifiers in parentheses). 1898 + T-Head's XuanTie C906 (thead-c906). 1899 * Improves the multi-lib selection mechanism for the bare-metal 1900 toolchain (riscv*-elf*). GCC will now automatically select the 1901 best-fit multi-lib candidate instead of requiring all possible 1902 reuse rules to be listed at build time. 1903 1904 [106]Operating Systems 1905 1906 [107]Windows 1907 1908 * The GNU threads library used by the win32 thread model has been 1909 reimplemented using direct Win32 API calls, except for the 1910 Objective-C specific subset. It requires Windows XP/Server 2003 or 1911 later. The new implementation also adds the support needed for the 1912 C++11 threads, using again direct Win32 API calls; this additional 1913 layer requires Windows Vista/Server 2008 or later. It is 1914 recommended to use a recent version of MinGW-W64 in conjunction 1915 with the win32 thread model. 1916 1917 [108]Improvements to Static Analyzer 1918 1919 * The analyzer has gained 20 new warnings: 1920 + [109]-Wanalyzer-allocation-size 1921 + [110]-Wanalyzer-deref-before-check 1922 + [111]-Wanalyzer-exposure-through-uninit-copy 1923 + Seven new warnings relating to misuse of file descriptors: 1924 o [112]-Wanalyzer-fd-access-mode-mismatch 1925 o [113]-Wanalyzer-fd-double-close 1926 o [114]-Wanalyzer-fd-leak 1927 o [115]-Wanalyzer-fd-phase-mismatch (e.g. calling accept on 1928 a socket before calling listen on it) 1929 o [116]-Wanalyzer-fd-type-mismatch (e.g. using a stream 1930 socket operation on a datagram socket) 1931 o [117]-Wanalyzer-fd-use-after-close 1932 o [118]-Wanalyzer-fd-use-without-check 1933 along with special-casing handling of the behavior of open, 1934 close, creat, dup, dup2, dup3, pipe, pipe2, read, and write. 1935 + [119]-Wanalyzer-imprecise-fp-arithmetic 1936 + [120]-Wanalyzer-infinite-recursion 1937 + [121]-Wanalyzer-jump-through-null 1938 + [122]-Wanalyzer-out-of-bounds 1939 + [123]-Wanalyzer-putenv-of-auto-var 1940 + [124]-Wanalyzer-tainted-assertion 1941 + Four new warnings for misuses of <stdarg.h>: 1942 o [125]-Wanalyzer-va-list-leak for complaining about 1943 missing va_end after a va_start or va_copy 1944 o [126]-Wanalyzer-va-list-use-after-va-end for complaining 1945 about va_arg or va_copy used on a va_list that's had 1946 va_end called on it 1947 o [127]-Wanalyzer-va-arg-type-mismatch for type-checking of 1948 va_arg usage in interprocedural execution paths against 1949 the types of the parameters that were actually passed to 1950 the variadic call 1951 o [128]-Wanalyzer-va-list-exhausted for complaining in 1952 interprocedural execution paths if va_arg is used too 1953 many times on a va_list 1954 along with numerous other improvements. 1955 1956 [129]Improvements for plugin authors 1957 1958 * GCC diagnostics can now be [130]associated with rules such as from 1959 coding standards documents, or specifications. Such rules have a 1960 code name and can have a URL, which GCC can print in text form or 1961 capture in its [131]SARIF output when emitting diagnostics. 1962 1963 Other significant improvements 1964 1965 [132]GCC 13.1 1966 1967 This is the [133]list of problem reports (PRs) from GCC's bug tracking 1968 system that are known to be fixed in the 13.1 release. This list might 1969 not be complete (that is, it is possible that some PRs that have been 1970 fixed are not listed here). 1971 1972 [134]GCC 13.2 1973 1974 This is the [135]list of problem reports (PRs) from GCC's bug tracking 1975 system that are known to be fixed in the 13.2 release. This list might 1976 not be complete (that is, it is possible that some PRs that have been 1977 fixed are not listed here). 1978 1979 [136]GCC 13.3 1980 1981 This is the [137]list of problem reports (PRs) from GCC's bug tracking 1982 system that are known to be fixed in the 13.3 release. This list might 1983 not be complete (that is, it is possible that some PRs that have been 1984 fixed are not listed here). 1985 1986 Language Specific Changes 1987 1988 C++ 1989 1990 * The libstdc++exp.a library now includes all the Filesystem TS 1991 symbols from the libstdc++fs.a library, and the experimental 1992 symbols for the C++23 std::stacktrace class from the 1993 libstdc++_libbacktrace.a library. This means that -lstdc++exp is 1994 the only library needed for all experimental libstdc++ features. 1995 1996 1997 For questions related to the use of GCC, please consult these web 1998 pages and the [138]GCC manuals. If that fails, the 1999 [139]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2000 web pages and the development of GCC are welcome on our developer 2001 list at [140]gcc (a] gcc.gnu.org. All of [141]our lists have public 2002 archives. 2003 2004 Copyright (C) [142]Free Software Foundation, Inc. Verbatim copying and 2005 distribution of this entire article is permitted in any medium, 2006 provided this notice is preserved. 2007 2008 These pages are [143]maintained by the GCC team. Last modified 2009 2025-04-16. 2010 2011 References 2012 2013 1. https://gcc.gnu.org/gcc-13/porting_to.html 2014 2. https://gcc.gnu.org/onlinedocs/index.html#current 2015 3. https://gcc.gnu.org/pipermail/gcc/2022-December/240322.html 2016 4. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Warray-bounds 2017 5. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C-Dialect-Options.html#index-fstrict-flex-arrays 2018 6. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2019 7. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/ARM-iWMMXt-Built-in-Functions.html 2020 8. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dynamic_or_shared.html#manual.intro.using.linkage.dynamic 2021 9. https://gcc.gnu.org/gcc-13/changes.html#general 2022 10. https://gcc.gnu.org/projects/gomp/ 2023 11. http://memkind.github.io/memkind/ 2024 12. https://sarifweb.azurewebsites.net/ 2025 13. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2026 14. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2027 15. https://gcc.gnu.org/onlinedocs/gcc/Freestanding-Environments.html 2028 16. https://gcc.gnu.org/gcc-13/changes.html#languages 2029 17. https://gcc.gnu.org/gcc-13/changes.html#ada 2030 18. https://gcc.gnu.org/gcc-13/changes.html#c-family 2031 19. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wxor-used-as-pow 2032 20. https://gcc.gnu.org/PR90885 2033 21. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg-function-attribute 2034 22. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg_005fread-function-attribute 2035 23. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html#index-fd_005farg_005fwrite-function-attribute 2036 24. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html 2037 25. https://wg21.link/p1774r8 2038 26. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Statement-Attributes.html#index-assume-statement-attribute 2039 27. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C-Dialect-Options.html#index-fstrict-flex-arrays 2040 28. https://gcc.gnu.org/gcc-13/changes.html#c 2041 29. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3042.htm 2042 30. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2963.htm 2043 31. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2975.pdf 2044 32. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3007.htm 2045 33. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3018.htm 2046 34. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3038.htm 2047 35. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2764.pdf 2048 36. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wformat 2049 37. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2836.pdf 2050 38. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wenum-int-mismatch 2051 39. https://gcc.gnu.org/PR105131 2052 40. https://gcc.gnu.org/gcc-13/changes.html#cxx 2053 41. https://wg21.link/p2324 2054 42. https://gcc.gnu.org/PR103539 2055 43. https://wg21.link/p2255 2056 44. https://gcc.gnu.org/PR104477 2057 45. https://wg21.link/p2327 2058 46. https://wg21.link/p2437 2059 47. https://gcc.gnu.org/PR106646 2060 48. https://wg21.link/p2290 2061 49. https://gcc.gnu.org/PR106645 2062 50. https://wg21.link/p2071 2063 51. https://gcc.gnu.org/PR106648 2064 52. https://wg21.link/p2513 2065 53. https://gcc.gnu.org/PR106656 2066 54. https://wg21.link/p1169r4 2067 55. https://gcc.gnu.org/PR106651 2068 56. https://wg21.link/p2266r3 2069 57. https://gcc.gnu.org/PR101165 2070 58. https://wg21.link/p2468r2 2071 59. https://gcc.gnu.org/PR106644 2072 60. https://wg21.link/p2362r3 2073 61. https://gcc.gnu.org/PR106647 2074 62. https://wg21.link/p2448r2 2075 63. https://gcc.gnu.org/PR106649 2076 64. https://wg21.link/p1467r9 2077 65. https://gcc.gnu.org/PR106652 2078 66. https://wg21.link/p1774r8 2079 67. https://gcc.gnu.org/PR106654 2080 68. https://wg21.link/p2295r6 2081 69. https://gcc.gnu.org/PR106655 2082 70. https://wg21.link/p2589r1 2083 71. https://gcc.gnu.org/PR107684 2084 72. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Warning-Options.html#index-Wself-move 2085 73. https://gcc.gnu.org/PR81159 2086 74. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdangling-reference 2087 75. https://gcc.gnu.org/PR106393 2088 76. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 2089 77. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 2090 78. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Link-Options.html#index-nostdlib_002b_002b 2091 79. https://gcc.gnu.org/gcc-13/changes.html#libstdcxx 2092 80. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p1642r11.html 2093 81. https://gcc.gnu.org/gcc-13/changes.html#d 2094 82. https://dlang.org/changelog/2.103.1.html 2095 83. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gdc/Warnings.html#index-Wno-builtin-declaration-mismatch 2096 84. https://gcc.gnu.org/onlinedocs/gcc-13.2.0/gdc/Warnings.html#index-Wmismatched-special-enum 2097 85. https://gcc.gnu.org/gcc-13/changes.html#fortran 2098 86. https://gcc.gnu.org/gcc-13/changes.html#go 2099 87. https://gcc.gnu.org/gcc-13/changes.html#modula2 2100 88. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gm2/Compiler-options.html 2101 89. https://gcc.gnu.org/gcc-13/changes.html#targets 2102 90. https://gcc.gnu.org/gcc-13/changes.html#aarch64 2103 91. https://gcc.gnu.org/gcc-13/changes.html#amdgcn 2104 92. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/AMD-GCN-Options.html 2105 93. https://gcc.gnu.org/install/specific.html#amdgcn-x-amdhsa 2106 94. https://gcc.gnu.org/gcc-13/changes.html#arm 2107 95. https://gcc.gnu.org/gcc-13/changes.html#avr 2108 96. https://gcc.gnu.org/gcc-13/changes.html#x86 2109 97. https://gcc.gnu.org/gcc-13/changes.html#loongarch 2110 98. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/LoongArch-Variable-Attributes.html#LoongArch-Variable-Attributes 2111 99. https://gcc.gnu.org/gcc-13/changes.html#nvptx 2112 100. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Nvidia-PTX-Options.html 2113 101. https://gcc.gnu.org/install/ 2114 102. https://gcc.gnu.org/install/specific.html#nvptx-x-none 2115 103. https://gcc.gnu.org/gcc-13/changes.html#riscv 2116 104. https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/v0.11.x 2117 105. https://rivai-ic.com.cn/ 2118 106. https://gcc.gnu.org/gcc-13/changes.html#os 2119 107. https://gcc.gnu.org/gcc-13/changes.html#windows 2120 108. https://gcc.gnu.org/gcc-13/changes.html#analyzer 2121 109. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-allocation-size 2122 110. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-deref-before-check 2123 111. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-exposure-through-uninit-copy 2124 112. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-access-mode-mismatch 2125 113. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-double-close 2126 114. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-leak 2127 115. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-phase-mismatch 2128 116. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-type-mismatch 2129 117. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-use-after-close 2130 118. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-fd-use-without-check 2131 119. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-imprecise-fp-arithmetic 2132 120. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-infinite-recursion 2133 121. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-jump-through-null 2134 122. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-out-of-bounds 2135 123. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-putenv-of-auto-var 2136 124. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-assertion 2137 125. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-leak 2138 126. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-use-after-va-end 2139 127. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-arg-type-mismatch 2140 128. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-va-list-exhausted 2141 129. https://gcc.gnu.org/gcc-13/changes.html#plugins 2142 130. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=0b14f590e3e9d95b8211b77d992589d5ab4c25f0 2143 131. https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2144 132. https://gcc.gnu.org/gcc-13/changes.html#13.1 2145 133. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.0 2146 134. https://gcc.gnu.org/gcc-13/changes.html#13.2 2147 135. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.2 2148 136. https://gcc.gnu.org/gcc-13/changes.html#13.3 2149 137. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=13.3 2150 138. https://gcc.gnu.org/onlinedocs/ 2151 139. mailto:gcc-help (a] gcc.gnu.org 2152 140. mailto:gcc (a] gcc.gnu.org 2153 141. https://gcc.gnu.org/lists.html 2154 142. https://www.fsf.org/ 2155 143. https://gcc.gnu.org/about.html 2156 ====================================================================== 2157 http://gcc.gnu.org/gcc-12/index.html 2158 2159 GCC 12 Release Series 2160 2161 Jun 20, 2024 2162 2163 The GCC developers are pleased to announce the release of GCC 12.4. 2164 2165 This release is a bug-fix release, containing fixes for regressions in 2166 GCC 12.3 relative to previous releases of GCC. 2167 2168 Release History 2169 2170 GCC 12.4 2171 Jun 20, 2024 ([1]changes, [2]documentation) 2172 2173 GCC 12.3 2174 May 8, 2023 ([3]changes, [4]documentation) 2175 2176 GCC 12.2 2177 Aug 19, 2022 ([5]changes, [6]documentation) 2178 2179 GCC 12.1 2180 May 6, 2022 ([7]changes, [8]documentation) 2181 2182 References and Acknowledgements 2183 2184 GCC used to stand for the GNU C Compiler, but since the compiler 2185 supports several other languages aside from C, it now stands for the 2186 GNU Compiler Collection. 2187 2188 The GCC developers would like to thank the numerous people that have 2189 contributed new features, improvements, bug fixes, and other changes as 2190 well as test results to GCC. This [9]amazing group of volunteers is 2191 what makes GCC successful. 2192 2193 For additional information about GCC please refer to the [10]GCC 2194 project web site or contact the [11]GCC development mailing list. 2195 2196 To obtain GCC please use [12]our mirror sites or [13]our version 2197 control system. 2198 2199 2200 For questions related to the use of GCC, please consult these web 2201 pages and the [14]GCC manuals. If that fails, the 2202 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2203 web pages and the development of GCC are welcome on our developer 2204 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 2205 archives. 2206 2207 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 2208 distribution of this entire article is permitted in any medium, 2209 provided this notice is preserved. 2210 2211 These pages are [19]maintained by the GCC team. Last modified 2212 2024-06-20. 2213 2214 References 2215 2216 1. https://gcc.gnu.org/gcc-12/changes.html 2217 2. https://gcc.gnu.org/onlinedocs/12.4.0/ 2218 3. https://gcc.gnu.org/gcc-12/changes.html 2219 4. https://gcc.gnu.org/onlinedocs/12.3.0/ 2220 5. https://gcc.gnu.org/gcc-12/changes.html 2221 6. https://gcc.gnu.org/onlinedocs/12.2.0/ 2222 7. https://gcc.gnu.org/gcc-12/changes.html 2223 8. https://gcc.gnu.org/onlinedocs/12.1.0/ 2224 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Contributors.html 2225 10. https://gcc.gnu.org/index.html 2226 11. mailto:gcc (a] gcc.gnu.org 2227 12. https://gcc.gnu.org/mirrors.html 2228 13. https://gcc.gnu.org/git.html 2229 14. https://gcc.gnu.org/onlinedocs/ 2230 15. mailto:gcc-help (a] gcc.gnu.org 2231 16. mailto:gcc (a] gcc.gnu.org 2232 17. https://gcc.gnu.org/lists.html 2233 18. https://www.fsf.org/ 2234 19. https://gcc.gnu.org/about.html 2235 ====================================================================== 2236 http://gcc.gnu.org/gcc-12/changes.html 2237 2238 GCC 12 Release Series 2239 Changes, New Features, and Fixes 2240 2241 This page is a "brief" summary of some of the huge number of 2242 improvements in GCC 12. You may also want to check out our [1]Porting 2243 to GCC 12 page and the [2]full GCC documentation. 2244 2245 Caveats 2246 2247 * An ABI incompatibility between C and C++ when passing or returning 2248 by value certain aggregates containing zero width bit-fields has 2249 been discovered on various targets. As mentioned in [3]PR102024, 2250 since the [4]PR42217 fix in GCC 4.5 the C++ front-end has been 2251 removing zero width bit-fields from the internal representation of 2252 the aggregates after the layout of those aggregates, but the C 2253 front-end kept them, so passing e.g. struct S { float a; int : 0; 2254 float b; } or struct T { float c; int : 0; } by value could differ 2255 between C and C++. Starting with GCC 12 the C++ front-end no longer 2256 removes those bit-fields from the internal representation and per 2257 clarified psABI some targets have been changed, so that they either 2258 ignore those bit-fields in the argument passing by value decisions 2259 in both C and C++, or they always take them into account. x86-64, 2260 ARM and AArch64 will always ignore them (so there is a C ABI 2261 incompatibility between GCC 11 and earlier with GCC 12 or later), 2262 PowerPC64 ELFv2 and S/390 always take them into account (so there 2263 is a C++ ABI incompatibility, GCC 4.4 and earlier compatible with 2264 GCC 12 or later, incompatible with GCC 4.5 through GCC 11). RISC-V 2265 has changed the handling of these already starting with GCC 10. As 2266 the ABI requires, MIPS takes them into account handling function 2267 return values so there is a C++ ABI incompatibility with GCC 4.5 2268 through 11. For function arguments on MIPS, refer to [5]the MIPS 2269 specific entry. GCC 12 on the above targets will report such 2270 incompatibilities as warnings or other diagnostics unless 2271 -Wno-psabi is used. 2272 * C: Computed gotos require a pointer type now. 2273 * C++: Two non-standard std::pair constructors have been deprecated. 2274 These allowed the use of an rvalue and a literal 0 to construct a 2275 pair containing a move-only type and a pointer. The nullptr keyword 2276 should be used to initialize the pointer member instead of a 2277 literal 0, as this is portable to other C++ implementations. 2278 * The configuration option --enable-libstdcxx-allocator no longer 2279 supports the bitmap, mt, and pool arguments. Those configurations 2280 had been broken for some time. 2281 * D: Building and bootstrapping GDC, the D compiler, now requires a 2282 working GDC (GCC version 9.1 or later) and D runtime library, 2283 libphobos, as the D front end is written in D. On some targets, 2284 libphobos isn't enabled by default, but compiles and works if 2285 --enable-libphobos is used. Other targets may require a more recent 2286 version of GCC to bootstrap. Specifics are documented for affected 2287 targets in the [6]Installing GCC manual. 2288 * Fortran: OpenMP code using the omp_lib.h include file can no longer 2289 be compiled with -std=f95 but now requires at least -std=f2003. 2290 Alternatively, use the omp_lib module, which still supports 2291 -std=f95 and is recommended to be used instead in general. 2292 * OpenMP offloading to Intel MIC has been deprecated and will be 2293 removed in a future release. 2294 * The cr16 target with the cr16-*-* configuration has been obsoleted 2295 and will be removed in a future release. 2296 * The hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11* configurations 2297 targeting 32-bit PA-RISC with HP-UX have been obsoleted and will be 2298 removed in a future release. 2299 * The m32c*-*-rtems* configuration has been obsoleted and will be 2300 removed in a future release. 2301 * The support for the m32r-*-linux*, m32rle-*-linux*, 2302 m68k*-*-openbsd* and vax-*-openbsd* configurations has been 2303 removed. 2304 * STABS: Support for emitting the STABS debugging format is 2305 deprecated and will be removed in the next release. All ports now 2306 default to emit DWARF (version 2 or later) debugging info or are 2307 obsoleted. 2308 * The optimization level -Ofast now implies 2309 -fno-semantic-interposition. 2310 2311 [7]General Improvements 2312 2313 * Vectorization is enabled at -O2 which is now equivalent to what 2314 would have been -O2 -ftree-vectorize -fvect-cost-model=very-cheap 2315 in the past. Note that the default vectorizer cost model has been 2316 changed; it used to behave as if -fvect-cost-model=cheap had been 2317 specified. 2318 * GCC now supports the [8]ShadowCallStack sanitizer, which can be 2319 enabled using the command-line option 2320 [9]-fsanitize=shadow-call-stack. This sanitizer currently only 2321 works on AArch64 targets and it requires an environment in which 2322 all code has been compiled with -ffixed-r18. Its primary initial 2323 user is the Linux kernel. 2324 2325 [10]New Languages and Language specific improvements 2326 2327 * OpenMP 2328 + OpenMP 5.0 support has been extended: The close map modifier 2329 and the affinity clause are now supported. In addition, 2330 Fortran gained the following features which were available in 2331 C and C++ before: declare variant is now available, depobj, 2332 mutexinoutset and iterator can now also be used with the 2333 depend clause, defaultmap has been updated for OpenMP 5.0, and 2334 the loop directive and combined directives involving the 2335 master directive have been added. 2336 + The following OpenMP 5.1 features have been added: support for 2337 expressing OpenMP directives as C++ 11 attributes, the masked 2338 and scope constructs, the nothing and error directives, and 2339 using primary with the proc_bind clause and OMP_PROC_BIND 2340 environment variable, the reproducible and unconstrained 2341 modifiers to the order clause, and, for C/C++ only, the align 2342 and allocator modifiers to the allocate clause and the atomic 2343 extensions are now available. The OMP_PLACE environment 2344 variable supports the OpenMP 5.1 features. In addition, the 2345 OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT environment variables 2346 and their associated API routines are now supported as well as 2347 the memory-allocation routines added for Fortran and extended 2348 for C/C++ in OpenMP 5.1. In Fortran code, strictly structured 2349 blocks can be used. 2350 + The [11]OpenMP Implementation Status can be found in the 2351 libgomp manual. 2352 * Version 2.6 of the [12]OpenACC specification continues to be 2353 maintained and improved in the C, C++ and Fortran compilers. See 2354 the [13]implementation status section on the OpenACC wiki page and 2355 the [14]run-time library documentation for further information. In 2356 addition to general performance tuning and bug fixing, new features 2357 include: 2358 + OpenACC worker parallelism for [15]AMD GPUs (already for a 2359 long time supported for [16]Nvidia GPUs). 2360 + Data privatization/sharing at the OpenACC gang level. 2361 + Considerable improvements for the experimental OpenACC 2362 'kernels' decomposition ([17]--param 2363 openacc-kernels=decompose). 2364 + A new warning flag [18]-Wopenacc-parallelism to warn about 2365 potentially suboptimal choices related to OpenACC parallelism. 2366 * The offload target code generation for OpenMP and OpenACC can now 2367 be better adjusted using the new [19]-foffload-options= flag and 2368 the pre-existing but now documented [20]-foffload= flag. 2369 2370 [21]Ada 2371 2372 * Ada 2022 2373 + Added the -gnat2022 flag to indicate strict Ada 2022 2374 compliance. The old -gnat2020 flag is now deprecated. 2375 + Support for Big Numbers (Annex G) has seen continuous 2376 improvements. It is now considered complete. It is compatible 2377 with SPARK, i.e. can be used from SPARK code. 2378 + Continuous improvements to the Ada 2022 standard since GCC 11. 2379 + Greatly improved compile time support. More functions can now 2380 have the with Static aspect and can be used in more contexts. 2381 * Ada 2022 extensions. The use of the -gnatX flag is necessary to 2382 access these features as they are not considered stable or 2383 standard. 2384 + Fixed lower bound for unconstrained arrays. 2385 o type Matrix is array (Natural range 0 .. <>, Natural 2386 range 0 .. <>) of Integer; is now valid. 2387 o Subtypes can also specify a lower bound: subtype String_1 2388 is String (1 .. <>);. Boundaries from slices will "slide" 2389 to the correct lower bound of the subtype. 2390 + Generalized Object.Operand notation. The following code is now 2391 valid V.Add_Element(42);, with V being a vector, for example. 2392 + Additional when constructs. Keywords return, goto and raise 2393 can now use when in addition to the existing exit when. The 2394 following expression is therefore now valid raise 2395 Constraint_Error with "Element is null" when Element = null; 2396 + Pattern matching 2397 o The case statement has been extended to cover records and 2398 arrays as well as finer grained casing on scalar types. 2399 In the future it is expected to provide more compile time 2400 guarantees when accessing discriminated fields. Case 2401 exhaustion is supported for pattern matching. An example 2402 would be 2403 type Sign is (Neg, Zero, Pos); 2404 2405 function Multiply (S1, S2 : Sign) return Sign is 2406 (case (S1, S2) is 2407 when (Neg, Neg) | (Pos, Pos) => Pos, 2408 when (Zero, <>) | (<>, Zero) => Zero, 2409 when (Neg, Pos) | (Pos, Neg) => Neg); 2410 2411 * gnatfind and gnatxref, which were already deprecated, have been 2412 removed. 2413 * Greatly expanded code covered by contracts. Thanks to this work, 2414 there are now several Ada standard libraries fully proven in SPARK 2415 which means they have no runtime nor logical errors. They are 2416 mostly numeric and string handling libraries. 2417 * Enable return-slot optimization for Pure functions. 2418 * General optimizations, improvements and additions to the standard 2419 library. Performance, correctness and in some cases stability was 2420 improved. Memory pools have also seen some minor enhancements. 2421 * Improvements to embedded-RTOS targets such as RTEMS, VxWorks and 2422 QNX. Older targets were removed or cleaned. 2423 * Added some [22]hardening features. 2424 2425 [23]C family 2426 2427 * Support for __builtin_shufflevector compatible with the clang 2428 language extension was added. 2429 * Support for attribute unavailable was added. 2430 * A new built-in function, __builtin_assoc_barrier, was added. It can 2431 be used to inhibit re-association of floating-point expressions. 2432 * Support for __builtin_dynamic_object_size compatible with the clang 2433 language extension was added. 2434 * New warnings: 2435 + [24]-Wbidi-chars warns about potentially misleading UTF-8 2436 bidirectional control characters. The default is 2437 -Wbidi-chars=unpaired ([25]PR103026) 2438 + [26]-Warray-compare warns about comparisons between two 2439 operands of array type ([27]PR97573) 2440 * Enhancements to existing warnings: 2441 + [28]-Wattributes has been extended so that it's possible to 2442 use -Wno-attributes=ns::attr or -Wno-attributes=ns:: to 2443 suppress warnings about unknown scoped attributes (in C++11 2444 and C2X). Similarly, #pragma GCC diagnostic ignored_attributes 2445 "vendor::attr" can be used to achieve the same effect 2446 ([29]PR101940) 2447 2448 [30]C 2449 2450 * Some new features from the upcoming C2X revision of the ISO C 2451 standard are supported with -std=c2x and -std=gnu2x. Some of these 2452 features are also supported as extensions when compiling for older 2453 language versions. In addition to the features listed, some 2454 features previously supported as extensions and now added to the C 2455 standard are enabled by default in C2X mode and not diagnosed with 2456 -std=c2x -Wpedantic. 2457 + Digit separators (as in C++) are supported for C2X. 2458 + The #elifdef and #elifndef preprocessing directives are now 2459 supported. 2460 + The printf and scanf format checking with [31]-Wformat now 2461 supports the %b format specified by C2X for binary integers, 2462 and the %B format recommended by C2X for printf. 2463 2464 [32]C++ 2465 2466 * Several C++23 features have been implemented: 2467 + [33]P1938R3, if consteval ([34]PR100974) 2468 + [35]P0849R8, auto(x): decay-copy in the language 2469 ([36]PR103049) 2470 + [37]P2242R3, Non-literal variables (and labels and gotos) in 2471 constexpr functions ([38]PR102612) 2472 + [39]P2334R1, Support for preprocessing directives elifdef and 2473 elifndef ([40]PR102616) 2474 + [41]P2360R0, Extend init-statement to allow alias-declaration 2475 ([42]PR102617) 2476 + [43]P2128R6, Multidimensional subscript operator 2477 + [44]DR 2397, auto specifier for pointers and references to 2478 arrays ([45]PR100975) 2479 * Several C++ Defect Reports have been resolved, e.g.: 2480 + [46]DR 960, Covariant functions and lvalue/rvalue references 2481 + [47]DR 1227, Mixing immediate and non-immediate contexts in 2482 deduction failure 2483 + [48]DR 1315, Restrictions on non-type template arguments in 2484 partial specializations 2485 + [49]DR 2082, Referring to parameters in unevaluated operands 2486 of default arguments 2487 + [50]DR 2351, void{} 2488 + [51]DR 2374, Overly permissive specification of enum 2489 direct-list-initialization 2490 + [52]DR 2397, auto specifier for pointers and references to 2491 arrays 2492 + [53]DR 2446, Questionable type-dependency of concept-ids 2493 * New command-line option -fimplicit-constexpr can be used to make 2494 inline functions implicitly constexpr ([54]git) 2495 * New command-line option -ffold-simple-inlines can be used to fold 2496 calls to certain trivial inline functions (currently std::move, 2497 std::forward, std::addressof and std::as_const). In contrast to 2498 inlining such calls, folding means that no intermediate code or 2499 debug information will be generated for them; this minimizes the 2500 abstraction penalty incurred for using these functions versus using 2501 the fundamental operations from which they're defined (e.g. 2502 std::move versus static_cast). This flag is enabled by default when 2503 -fno-inline is not active. 2504 * Deduction guides can be declared at class scope ([55]PR79501) 2505 * [56]-Wuninitialized warns about using uninitialized variables in 2506 member initializer lists ([57]PR19808) 2507 * [58]-Wint-in-bool-context is now disabled when instantiating a 2508 template ([59]git) 2509 * Stricter checking of attributes on friend declarations: if a friend 2510 declaration has an attribute, that declaration must be a 2511 definition. Moreover, a C++11 attribute cannot appear in the middle 2512 of the decl-specifier-seq. ([60]PR99032) 2513 * New warning options for C++ language mismatches: 2514 -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions, 2515 -Wc++20-extensions, and -Wc++23-extensions. They are enabled by 2516 default and can be used to control existing pedwarns about 2517 occurrences of new C++ constructs in code using an old C++ standard 2518 dialect. 2519 * New warning [61]-Wmissing-requires warns about missing requires 2520 ([62]git) 2521 * The existing std::is_constant_evaluated in if warning was extended 2522 to warn in more cases ([63]PR100995) 2523 * [64]-Waddress has been enhanced so that it now warns about, for 2524 instance, comparing the address of a nonstatic member function to 2525 null ([65]PR102103) 2526 * Errors about narrowing are no longer hidden if they occur in system 2527 headers 2528 * Ordered comparison of null pointers is now rejected ([66]PR99701) 2529 * Anonymous structs with bases are now rejected ([67]git) 2530 * The compiler rejects taking the address of an immediate member 2531 function ([68]PR102753) 2532 * The compiler has support for C++20 2533 __cpp_lib_is_pointer_interconvertible and 2534 __cpp_lib_is_layout_compatible to help the C++ library implement 2535 [69]P0466, Layout-compatibility and Pointer-interconvertibility 2536 Traits ([70]PR101539) 2537 * Memory usage of constraint subsumption has been improved 2538 ([71]PR100828) 2539 * constinit thread_local variables are optimized better 2540 ([72]PR101786) 2541 * Support for C++17 std::hardware_destructive_interference_size was 2542 added, along with the [73]-Winterference-size warning ([74]git) 2543 * Many bugs in the CTAD handling have been fixed ([75]PR101344, 2544 [76]PR101883, [77]PR89062, [78]PR101233, [79]PR88252, [80]PR86439, 2545 [81]PR98832, [82]PR102933 ...) 2546 * Two-stage name lookup for dependent operator expressions has been 2547 corrected ([83]PR51577) 2548 * Several issues with constrained variable templates have been fixed 2549 ([84]PR98486) 2550 * The compiler performs less instantiating when doing speculative 2551 constant evaluation ([85]git) 2552 * Various diagnostic improvements; e.g., a more precise caret 2553 location for pointer-to-member expressions 2554 * The new -fconstexpr-fp-except flag allows IEC559 floating point 2555 exceptions in constant-expressions. 2556 2557 [86]Runtime Library (libstdc++) 2558 2559 * Improved experimental C++20 support, including: 2560 + std::vector, std::basic_string, std::optional, and 2561 std::variant can be used in constexpr functions. 2562 + std::make_shared for arrays with default initialization, and 2563 std::atomic<std::shared_ptr<T>>. 2564 + Layout-compatibility and pointer-interconvertibility traits. 2565 * Improved experimental C++23 support, including: 2566 + Monadic operations for std::optional. 2567 + std::expected 2568 + std::move_only_function 2569 + <spanstream> 2570 + std::basic_string::resize_and_overwrite 2571 + std::unique_ptr can be used in constexpr functions. 2572 + <stacktrace> (not built by default, requires linking to an 2573 extra library). 2574 + <stdatomic.h> 2575 + std::invoke_r 2576 + constexpr std::type_info::operator== 2577 2578 [87]D 2579 2580 * New features: 2581 + Support for the D programming language has been updated to 2582 version 2.100.1 of the language and run-time library. Full 2583 changelog for this release and previous releases can be found 2584 on the [88]dlang.org website. 2585 + On supported targets, the __traits(compiles) expression can 2586 now be used to determine whether a target-specific built-in is 2587 available without error during CTFE ([89]PR101127). 2588 + Functions annotated with pragma(inline, true) are now compiled 2589 into every module where they are used from ([90]PR106563). 2590 + Partial support for directly importing C99 sources into a D 2591 compilation ([91]ImportC) has been added to the language. A 2592 notable missing feature is support for preprocessing C 2593 imports, which can be worked around by preprocessing all C 2594 sources used for importing ahead of time. 2595 * New language options: 2596 + -fcheck=, enables or disables the code generation of specific 2597 run-time contract checks. 2598 + -fcheckaction=, controls the run-time behavior on an assert, 2599 array bounds check, or final switch contract failure. The 2600 default is -fcheckaction=throw. 2601 + -fdump-c++-spec=, dumps all compiled extern(C++) declarations 2602 as C++ code to the given file. The supplementary option 2603 -fdump-c++-spec-verbose turns on emission of comments for 2604 ignored declarations in the generated spec. 2605 + -fextern-std=, controls which C++ standard extern(C++) 2606 declarations are compiled to be compatible with. The default 2607 is -fextern-std=c++17. 2608 + -fpreview=, added to enable upcoming D language features in 2609 the compiler. 2610 + -frevert=, added to revert D language changes to support older 2611 D codebases that need more time to transition. 2612 + -fsave-mixins=, saves mixins expanded at compile-time to a 2613 file. 2614 * Deprecated and removed features: 2615 + The -Wtemplates compiler switch has been removed, as it had 2616 been superceded by -ftransition=templates, which more 2617 accurately reports on which templates have been instantiated. 2618 + The -ftransition=dip25 and -ftransition=dip1000 compiler 2619 switches have been renamed to -fpreview=dip25 and 2620 -fpreview=dip1000. 2621 2622 [92]Fortran 2623 2624 * WG5/N1942, "TS 29113 Further Interoperability of Fortran with C", 2625 is now fully supported. In addition to implementing previously 2626 missing functionality, such as support for character arguments of 2627 length greater than one in functions marked bind(c) and gaps in the 2628 handling for assumed-rank arrays, numerous other bugs have been 2629 fixed, and an extensive set of new conformance test cases has been 2630 added. 2631 * GCC 12 now uses OPERATION as the name of the function to the 2632 CO_REDUCE intrinsic for the pairwise reduction, thus conforming to 2633 the Fortran 2018 standard. Previous versions used OPERATOR which 2634 conforms to TS 18508. 2635 * On POWER systems which support it, the -mabi=ieeelongdouble option 2636 now selects the IEEE 128-bit floating point format for 2637 REAL(KIND=16). R16_IBM and R16_IEEE have been added to the 2638 -fconvert option, the CONVERT specifier of the OPEN statement and 2639 the GFORTRAN_CONVERT_UNIT environment variable. 2640 2641 [93]Go 2642 2643 * GCC 12 provides a complete implementation of the Go 1.18 user 2644 packages. 2645 * Although Go 1.18 includes support for generic programming, that 2646 support is not yet available in GCC. 2647 2648 [94]libgccjit 2649 2650 * The libgccjit API gained 30 new entry points: 2651 + 17 new "reflection" entry points for querying functions and 2652 types ([95]LIBGCCJIT_ABI_16) 2653 + [96]gcc_jit_lvalue_set_tls_model for supporting thread-local 2654 variables ([97]LIBGCCJIT_ABI_17) 2655 + [98]gcc_jit_lvalue_set_link_section for setting the link 2656 section of global variables, analogous to 2657 [99]__attribute__((section(".section"))) 2658 ([100]LIBGCCJIT_ABI_18) 2659 + 4 new entry points for initializing global variables and 2660 creating constructors for rvalues ([101]LIBGCCJIT_ABI_19) 2661 + Support for sized integer types, including 128-bit integers 2662 and helper functions for such types ([102]LIBGCCJIT_ABI_20) 2663 + [103]gcc_jit_context_new_bitcast for reinterpreting the bits 2664 of an rvalue as a different type ([104]LIBGCCJIT_ABI_21) 2665 + [105]gcc_jit_lvalue_set_register_name for setting a specific 2666 register for a variable ([106]LIBGCCJIT_ABI_22) 2667 + [107]gcc_jit_context_set_bool_print_errors_to_stderr 2668 ([108]LIBGCCJIT_ABI_23) 2669 + 2 new entry points for setting the alignment of a variable 2670 ([109]LIBGCCJIT_ABI_24) 2671 * libgccjit has gained support for the use of various atomic builtins 2672 ([110]PR96066, [111]PR96067) 2673 * [112]gcc_jit_context_new_cast is now able to handle truncation and 2674 extension between different integer types ([113]PR95498) 2675 2676 [114]New Targets and Target Specific Improvements 2677 2678 [115]AArch64 & arm 2679 2680 * Newer revisions of the Arm Architecture are supported as arguments 2681 to the -march option: armv8.7-a, armv8.8-a, armv9-a. 2682 * The Arm Cortex-A510 CPU is now supported through the cortex-a510 2683 argument to the -mcpu and -mtune options. 2684 * GCC can now auto-vectorize operations performing sign-differing 2685 dot-product operations, taking advantage of instructions in the 2686 Advanced SIMD (AArch64/AArch32) and SVE (AArch64) instruction sets. 2687 2688 [116]AArch64 2689 2690 * A number of new CPUs are supported through the -mcpu and -mtune 2691 options (GCC identifiers in parentheses). 2692 + Ampere-1 (ampere1). 2693 + Arm Cortex-A710 (cortex-a710). 2694 + Arm Cortex-X2 (cortex-x2). 2695 * The 64-byte atomic load/store intrinsics to accelerator memory from 2696 the [117]2020 Arm Architecture extensions are supported through the 2697 +ls64 option extension. 2698 * Initial code generation support is supported for hardware 2699 instructions used to accelerate the memcpy,memmove and memset 2700 standard functions. These instructions can be generated when 2701 compiling with the +mopsoption extension. 2702 * The ACLE Advanced SIMD intrinsics accessible through the arm_neon.h 2703 header have been significantly reimplemented and generate 2704 higher-performing code than previous GCC versions. 2705 * The option -mtune=neoverse-512tvb is added to tune for Arm Neoverse 2706 cores that have a total vector bandwidth of 512 bits. Please refer 2707 to the documentation for more details. 2708 2709 [118]AMD Radeon (GCN) 2710 2711 * Debug experience with ROCGDB has been improved. 2712 * Support for the type __int128_t/integer(kind=16) was added. 2713 * For offloading, the limitation of using only one wavefront per 2714 compute unit (CU) has been lifted. Up to 40 workgroups per CU and 2715 16 wavefronts per workgroup are supported (up to a limit of 40 2716 wavefronts in total, per CU). Additionally, the number of used 2717 wavefronts and workgroups was tuned for performance. 2718 2719 [119]arm 2720 2721 * Support is added for accessing the stack canary value via the TLS 2722 register through the -fstack-protector-guard=tls and 2723 -mstack-protector-guard-offset= options. This intended for use in 2724 Linux kernel development. Please refer to the documentation for 2725 more details. 2726 2727 [120]BPF 2728 2729 * Support for CO-RE (compile-once, run-everywhere) has been added to 2730 the BPF back end. CO-RE allows to compile portable BPF programs 2731 that are able to run among different versions of the Linux kernel. 2732 2733 [121]IA-32/x86-64 2734 2735 * New ISA extension support for Intel AVX512-FP16 was added. 2736 AVX512-FP16 intrinsics are available via the -mavx512fp16 compiler 2737 switch. 2738 * For both C and C++ the _Float16 type is supported on x86 systems 2739 with SSE2 enabled. Without {-mavx512fp16}, all operations will be 2740 emulated in software and float instructions. 2741 * Mitigation against straight line speculation (SLS) for function 2742 return and indirect jump is supported via 2743 -mharden-sls=[none|all|return|indirect-jmp]. 2744 * Add CS prefix to call and jmp to indirect thunk with branch target 2745 in r8-r15 registers via -mindirect-branch-cs-prefix. 2746 * Always use global offset table (GOT) to access external data and 2747 function symbols when the new -mno-direct-extern-access 2748 command-line option is specified. 2749 2750 [122]LoongArch 2751 2752 * Support for the LoongArch architecture instruction set has been 2753 added. 2754 * The Loongson CPU codename LA464 and LoongArch 64-bit generic CPU 2755 codename loongarch64 are supported through the -march= and -mtune= 2756 options (GCC identifiers in parentheses). 2757 + Loongson LA464 core (la464). 2758 + LoongArch 64-bit generic core (loongarch64). 2759 2760 [123]MIPS 2761 2762 * The ABI passing arguments containing zero-width fields (for 2763 example, C/C++ zero-width bit-fields, GNU C/C++ zero-length arrays, 2764 and GNU C empty structs) has changed. Now a zero-width field will 2765 not prevent an aligned 64-bit floating-point field next to it from 2766 being passed through FPR. This is compatible with LLVM, but 2767 incompatible with previous GCC releases. GCC 12 on MIPS will report 2768 such incompatibilities as an inform unless -Wno-psabi is used. 2769 * The ABI returning values containing C++17 empty bases has changed. 2770 Now an empty base will not prevent an aggregate containing only one 2771 or two floating-point fields from being returned through FPR. This 2772 is compatible with GCC 6 and earlier, but incompatible with GCC 7 2773 through 11. GCC 12 on MIPS will report such incompatibilities as an 2774 inform unless -Wno-psabi is used. 2775 2776 [124]NVPTX 2777 2778 * The -march flag has been added. The -misa flag is now considered an 2779 alias of the -march flag. 2780 * Support for PTX ISA target architectures sm_53, sm_70, sm_75 and 2781 sm_80 has been added. These can be specified using the -march flag. 2782 * The default PTX ISA target architecture has been set back to sm_30, 2783 to fix support for sm_30 boards. 2784 * The -march-map flag has been added. The -march-map value will be 2785 mapped to an valid -march flag value. For instance, 2786 -march-map=sm_50 maps to -march=sm_35. This can be used to specify 2787 that generated code is to be executed on a board with at least some 2788 specific compute capability, without having to know the valid 2789 values for the -march flag. 2790 * The -mptx flag has been added to specify the PTX ISA version for 2791 the generated code; permitted values are 3.1 (matches previous GCC 2792 versions), 6.0, 6.3, and 7.0. If not specified, the used version is 2793 the minimal version required for -march but at least 6.0. 2794 * An mptx-3.1 multilib was added. This allows using older drivers 2795 which do not support PTX ISA version 6.0. 2796 * The new __PTX_SM__ predefined macro allows code to check the PTX 2797 ISA target architecture being targeted by the compiler. 2798 * The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__ 2799 predefined macros allows code to check the PTX ISA version being 2800 targeted by the compiler. 2801 2802 [125]PowerPC / PowerPC64 / RS6000 2803 2804 * The internal implementation of Power's target-specific built-in 2805 functions has been rewritten to be easier and less error-prone to 2806 maintain. Every attempt has been made to ensure that the new 2807 behavior matches the old behavior, but inevitably some bugs can be 2808 expected. Please [126]report any problems. 2809 * The built-in functions __builtin_get_texasr, __builtin_get_texasru, 2810 __builtin_get_tfhar, __builtin_get_tfiar, __builtin_set_texasr, 2811 __builtin_set_texasru, __builtin_set_tfhar, and __builtin_set_tfiar 2812 now behave as documented in all supported configurations. On prior 2813 releases, the arguments and return values of these functions were 2814 treated as unsigned long long instead of as unsigned long, when the 2815 options -m32 -mpowerpc64 were in effect. 2816 * The overloaded built-in functions vec_cntlz_lsbb and vec_cnttz_lsbb 2817 now behave as documented. On prior releases, these built-in 2818 functions had incorrect semantics on little-endian targets. 2819 2820 [127]PRU 2821 2822 * The [128]__regio_symbol variable qualifier has been added. It 2823 allows easier access in C programs to the __R30 and __R31 CPU I/O 2824 registers. 2825 2826 [129]RISC-V 2827 2828 * The default ISA spec version was bumped to 20191213; see [130]this 2829 announcement for details. 2830 * New ISA extension support for zba, zbb, zbc, zbs was added. 2831 * New ISA extension support for vector and scalar crypto was added, 2832 only support architecture testing macro and -march= parsing. 2833 * The option -mtune=thead-c906 is added to tune for T-HEAD c906 2834 cores. 2835 * libstdc++ no longer attempts to detect built-in atomics. 2836 Distributions that have out-of-tree workarounds for -latomic should 2837 check their ABIs again. 2838 2839 [131]Operating Systems 2840 2841 [132]Improvements to Static Analyzer 2842 2843 * The analyzer has gained a 2844 [133]-Wanalyzer-use-of-uninitialized-value warning, similar to 2845 [134]-Wuninitialized and [135]-Wmaybe-uninitialized, but based on 2846 an interprocedural path-sensitive analysis ([136]PR95006). 2847 Such warnings are not disabled by the new 2848 [137]-ftrivial-auto-var-init (see below), as the latter is 2849 considered a mitigation option. 2850 * [138]-Wanalyzer-write-to-const and 2851 [139]-Wanalyzer-write-to-string-literal will now check for 2852 [140]__attribute__ ((access, ....)) on calls to externally-defined 2853 functions, and complain about read-only regions pointed to by 2854 arguments marked with a write_only or read_write attribute 2855 ([141]PR104793). 2856 * The analyzer's "taint" mode, activated by 2857 [142]-fanalyzer-checker=taint (in addition to [143]-fanalyzer), has 2858 gained four new taint-based warnings: 2859 + [144]-Wanalyzer-tainted-allocation-size for e.g. 2860 attacker-controlled malloc and alloca, 2861 + [145]-Wanalyzer-tainted-divisor for detecting where an 2862 attacker can inject a divide-by-zero, 2863 + [146]-Wanalyzer-tainted-offset for attacker-controlled pointer 2864 offsets, 2865 + [147]-Wanalyzer-tainted-size for attacker-controlled values 2866 being used as a size parameter to calls to memset or to 2867 functions marked with [148]__attribute__ ((access, ....)). 2868 The existing [149]-Wanalyzer-tainted-array-index has been reworded 2869 to talk about "attacker-controlled" rather than "tainted" values, 2870 for consistency with the new warnings. 2871 A new [150]__attribute__ ((tainted_args)) has been added to the C 2872 and C++ front ends, usable on functions, and on function pointer 2873 callback fields in structs. The analyzer's taint mode will treat 2874 all parameters and buffers pointed to by parameters of such 2875 functions as being attacker-controlled, such as for annotating 2876 system calls in an operating system kernel as being an "attack 2877 surface". 2878 * The analyzer now respects [151]__attribute__((const)): it will 2879 treat such functions as returning the same value when given the 2880 same inputs ([152]PR104434), and as having no side effects 2881 ([153]PR104576). 2882 * The analyzer is now able to split its analysis into multiple 2883 execution paths in places where there isn't a split in the control 2884 flow graph. For example, it now handles realloc calls by splitting 2885 the execution path into three possible outcomes for the call: 2886 + failure, returning NULL 2887 + success, growing the buffer in-place without moving it 2888 + success, allocating a new buffer, copying the content of the 2889 old buffer to it, and freeing the old buffer 2890 * The analyzer's interprocedural path exploration logic is now able 2891 to track calls through function pointers. 2892 * The analyzer now makes the assumption that if we know PTR is 2893 non-NULL, then (PTR + OFFSET) is also non-NULL. This isn't strictly 2894 true, but eliminates false positives in practice ([154]PR101962). 2895 * The analyzer has gained some initial support for inline assembler 2896 code. This is extremely limited, and is purely to help suppress 2897 false positives when analyzing the Linux kernel, which makes heavy 2898 use of inline assembler ([155]PR101570). 2899 * The way the analyzer tracks the state of memory along an execution 2900 path has been improved in various ways for GCC 12: 2901 + An optimization for representing bulk updates to memory (e.g. 2902 zero fills) has been removed as it never worked well. In GCC 2903 12 it has been replaced with a simpler and more accurate 2904 approach, eliminating many false positives ([156]PR95006). 2905 + Various optimizations have been added, speeding up the 2906 analysis on a particularly problematic source file from 4 2907 minutes down to 17 seconds ([157]PR104943, [158]PR104954, and 2908 [159]PR104955). 2909 + The analyzer now tracks the sizes of dynamically-allocated 2910 regions, both on the heap (via malloc etc) and stack (via 2911 alloca), though none of the analyzer warnings make use of this 2912 yet in GCC 12. 2913 * The analyzer's handling of switch statements has been rewritten, 2914 fixing various bugs. 2915 2916 Other significant improvements 2917 2918 [160]Eliminating uninitialized variables 2919 2920 * GCC can now [161]initialize all stack variables implicitly, 2921 including padding. This is intended to eliminate all classes of 2922 uninitialized stack variable flaws. Lack of explicit initialization 2923 will still warn when [162]-Wuninitialized is active. For best 2924 debugging, use of the new command-line option 2925 [163]-ftrivial-auto-var-init=pattern can be used to fill variables 2926 with a repeated 0xFE pattern, which tends to illuminate many bugs 2927 (e.g. pointers receive invalid addresses, sizes and indices are 2928 very large). For best production results, the new command-line 2929 option [164]-ftrivial-auto-var-init=zero can be used to fill 2930 variables with 0x00, which tends to provide a safer state for bugs 2931 (e.g. pointers are NULL, strings are NUL filled, and sizes and 2932 indices are 0). 2933 2934 [165]Debugging formats 2935 2936 * GCC can now generate debugging information in [166]CTF, a 2937 lightweight debugging format that provides information about C 2938 types and the association between functions and data symbols and 2939 types. This format is designed to be embedded in ELF files and to 2940 be very compact and simple. A new command-line option -gctf enables 2941 the generation of CTF. 2942 * GCC can now generate debugging information in BTF. This is a 2943 debugging format mainly used in BPF programs and the Linux kernel. 2944 The compiler can generate BTF for any target, when enabled with the 2945 command-line option -gbtf 2946 2947 [167]GCC 12.1 2948 2949 This is the [168]list of problem reports (PRs) from GCC's bug tracking 2950 system that are known to be fixed in the 12.1 release. This list might 2951 not be complete (that is, it is possible that some PRs that have been 2952 fixed are not listed here). 2953 2954 [169]GCC 12.2 2955 2956 This is the [170]list of problem reports (PRs) from GCC's bug tracking 2957 system that are known to be fixed in the 12.2 release. This list might 2958 not be complete (that is, it is possible that some PRs that have been 2959 fixed are not listed here). 2960 2961 Target Specific Changes 2962 2963 LoongArch 2964 2965 * The default setting of -m[check|no-check]-zero-division is changed 2966 for optimized code. Now -mno-check-zero-division is the default for 2967 all optimization levels but -O0 and -Og. The old behavior can be 2968 obtained by explicitly passing -mcheck-zero-division to GCC. 2969 2970 [171]GCC 12.3 2971 2972 Target Specific Changes 2973 2974 x86-64 2975 2976 * GCC now supports AMD CPUs based on the znver4 core via 2977 -march=znver4. The switch makes GCC consider using 512-bit vectors 2978 when auto-vectorizing. 2979 2980 This is the [172]list of problem reports (PRs) from GCC's bug tracking 2981 system that are known to be fixed in the 12.3 release. This list might 2982 not be complete (that is, it is possible that some PRs that have been 2983 fixed are not listed here). 2984 2985 [173]GCC 12.4 2986 2987 This is the [174]list of problem reports (PRs) from GCC's bug tracking 2988 system that are known to be fixed in the 12.4 release. This list might 2989 not be complete (that is, it is possible that some PRs that have been 2990 fixed are not listed here). 2991 2992 2993 For questions related to the use of GCC, please consult these web 2994 pages and the [175]GCC manuals. If that fails, the 2995 [176]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2996 web pages and the development of GCC are welcome on our developer 2997 list at [177]gcc (a] gcc.gnu.org. All of [178]our lists have public 2998 archives. 2999 3000 Copyright (C) [179]Free Software Foundation, Inc. Verbatim copying and 3001 distribution of this entire article is permitted in any medium, 3002 provided this notice is preserved. 3003 3004 These pages are [180]maintained by the GCC team. Last modified 3005 2025-01-31. 3006 3007 References 3008 3009 1. https://gcc.gnu.org/gcc-12/porting_to.html 3010 2. https://gcc.gnu.org/onlinedocs/index.html#current 3011 3. https://gcc.gnu.org/PR102024 3012 4. https://gcc.gnu.org/PR42217 3013 5. https://gcc.gnu.org/gcc-12/changes.html#mips_zero_width_fields 3014 6. https://gcc.gnu.org/install/specific.html 3015 7. https://gcc.gnu.org/gcc-12/changes.html#general 3016 8. https://clang.llvm.org/docs/ShadowCallStack.html 3017 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Instrumentation-Options.html#index-fsanitize_003dshadow-call-stack 3018 10. https://gcc.gnu.org/gcc-12/changes.html#languages 3019 11. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/OpenMP-Implementation-Status.html 3020 12. https://www.openacc.org/ 3021 13. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-12 3022 14. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/Enabling-OpenACC.html 3023 15. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 3024 16. https://gcc.gnu.org/gcc-12/changes.html#nvptx 3025 17. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-param 3026 18. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wopenacc-parallelism 3027 19. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload-options 3028 20. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload 3029 21. https://gcc.gnu.org/gcc-12/changes.html#ada 3030 22. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gnat_rm/Security-Hardening-Features.html#Security-Hardening-Features 3031 23. https://gcc.gnu.org/gcc-12/changes.html#c-family 3032 24. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wbidi-chars 3033 25. https://gcc.gnu.org/PR103026 3034 26. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Warray-compare 3035 27. https://gcc.gnu.org/PR97573 3036 28. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wattributes 3037 29. https://gcc.gnu.org/PR101940 3038 30. https://gcc.gnu.org/gcc-12/changes.html#c 3039 31. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wformat 3040 32. https://gcc.gnu.org/gcc-12/changes.html#cxx 3041 33. https://wg21.link/p1938 3042 34. https://gcc.gnu.org/PR100974 3043 35. https://wg21.link/p0849 3044 36. https://gcc.gnu.org/PR103049 3045 37. https://wg21.link/p2242 3046 38. https://gcc.gnu.org/PR102612 3047 39. https://wg21.link/p2334 3048 40. https://gcc.gnu.org/PR102616 3049 41. https://wg21.link/p2360 3050 42. https://gcc.gnu.org/PR102617 3051 43. https://wg21.link/p2128 3052 44. https://wg21.link/cwg2397 3053 45. https://gcc.gnu.org/PR100975 3054 46. https://wg21.link/cwg960 3055 47. https://wg21.link/cwg1227 3056 48. https://wg21.link/cwg1315 3057 49. https://wg21.link/cwg2082 3058 50. https://wg21.link/cwg2351 3059 51. https://wg21.link/cwg2374 3060 52. https://wg21.link/cwg2397 3061 53. https://wg21.link/cwg2446 3062 54. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=87c2080b 3063 55. https://gcc.gnu.org/PR79501 3064 56. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3065 57. https://gcc.gnu.org/PR19808 3066 58. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wint-in-bool-context 3067 59. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a2b12bc 3068 60. https://gcc.gnu.org/PR99032 3069 61. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmissing-requires 3070 62. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=e18e56c7 3071 63. https://gcc.gnu.org/PR100995 3072 64. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Waddress 3073 65. https://gcc.gnu.org/PR102103 3074 66. https://gcc.gnu.org/PR99701 3075 67. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3ead06c1 3076 68. https://gcc.gnu.org/PR102753 3077 69. https://wg21.link/p0466 3078 70. https://gcc.gnu.org/PR101539 3079 71. https://gcc.gnu.org/PR100828 3080 72. https://gcc.gnu.org/PR101786 3081 73. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Winterference-size 3082 74. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=76b75018 3083 75. https://gcc.gnu.org/PR101344 3084 76. https://gcc.gnu.org/PR101883 3085 77. https://gcc.gnu.org/PR89062 3086 78. https://gcc.gnu.org/PR101233 3087 79. https://gcc.gnu.org/PR88252 3088 80. https://gcc.gnu.org/PR86439 3089 81. https://gcc.gnu.org/PR98832 3090 82. https://gcc.gnu.org/PR102933 3091 83. https://gcc.gnu.org/PR51577 3092 84. https://gcc.gnu.org/PR98486 3093 85. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=1595fe44 3094 86. https://gcc.gnu.org/gcc-12/changes.html#libstdcxx 3095 87. https://gcc.gnu.org/gcc-12/changes.html#d 3096 88. https://dlang.org/changelog/2.100.1.html 3097 89. https://gcc.gnu.org/PR101127 3098 90. https://gcc.gnu.org/PR106563 3099 91. https://dlang.org/spec/importc.html 3100 92. https://gcc.gnu.org/gcc-12/changes.html#fortran 3101 93. https://gcc.gnu.org/gcc-12/changes.html#go 3102 94. https://gcc.gnu.org/gcc-12/changes.html#jit 3103 95. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-16 3104 96. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_tls_model 3105 97. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-17 3106 98. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_link_section 3107 99. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Variable-Attributes.html#index-section-variable-attribute 3108 100. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-18 3109 101. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-19 3110 102. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-20 3111 103. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_bitcast 3112 104. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-21 3113 105. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_register_name 3114 106. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-22 3115 107. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/contexts.html#c.gcc_jit_context_set_bool_print_errors_to_stderr 3116 108. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-23 3117 109. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-24 3118 110. https://gcc.gnu.org/PR96066 3119 111. https://gcc.gnu.org/PR96067 3120 112. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_cast 3121 113. https://gcc.gnu.org/PR95498 3122 114. https://gcc.gnu.org/gcc-12/changes.html#targets 3123 115. https://gcc.gnu.org/gcc-12/changes.html#arm-targets 3124 116. https://gcc.gnu.org/gcc-12/changes.html#aarch64 3125 117. https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/arm-a-profile-architecture-developments-2020 3126 118. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 3127 119. https://gcc.gnu.org/gcc-12/changes.html#arm 3128 120. https://gcc.gnu.org/gcc-12/changes.html#bpf 3129 121. https://gcc.gnu.org/gcc-12/changes.html#x86 3130 122. https://gcc.gnu.org/gcc-12/changes.html#loongarch 3131 123. https://gcc.gnu.org/gcc-12/changes.html#mips 3132 124. https://gcc.gnu.org/gcc-12/changes.html#nvptx 3133 125. https://gcc.gnu.org/gcc-12/changes.html#powerpc 3134 126. https://gcc.gnu.org/bugs/ 3135 127. https://gcc.gnu.org/gcc-12/changes.html#pru 3136 128. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Named-Address-Spaces.html#PRU-Named-Address-Spaces 3137 129. https://gcc.gnu.org/gcc-12/changes.html#riscv 3138 130. https://groups.google.com/a/groups.riscv.org/g/sw-dev/c/aE1ZeHHCYf4 3139 131. https://gcc.gnu.org/gcc-12/changes.html#os 3140 132. https://gcc.gnu.org/gcc-12/changes.html#analyzer 3141 133. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-use-of-uninitialized-value 3142 134. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3143 135. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 3144 136. https://gcc.gnu.org/PR95006 3145 137. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3146 138. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 3147 139. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 3148 140. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 3149 141. https://gcc.gnu.org/PR104793 3150 142. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 3151 143. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 3152 144. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 3153 145. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 3154 146. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 3155 147. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 3156 148. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 3157 149. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 3158 150. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-tainted_005fargs-function-attribute 3159 151. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-const-function-attribute 3160 152. https://gcc.gnu.org/PR104434 3161 153. https://gcc.gnu.org/PR104576 3162 154. https://gcc.gnu.org/PR101962 3163 155. https://gcc.gnu.org/PR101570 3164 156. https://gcc.gnu.org/PR95006 3165 157. https://gcc.gnu.org/PR104943 3166 158. https://gcc.gnu.org/PR104954 3167 159. https://gcc.gnu.org/PR104955 3168 160. https://gcc.gnu.org/gcc-12/changes.html#uninitialized 3169 161. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3170 162. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 3171 163. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3172 164. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 3173 165. https://gcc.gnu.org/gcc-12/changes.html#debug 3174 166. https://ctfstd.org/ 3175 167. https://gcc.gnu.org/gcc-12/changes.html#12.1 3176 168. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.0 3177 169. https://gcc.gnu.org/gcc-12/changes.html#12.2 3178 170. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.2 3179 171. https://gcc.gnu.org/gcc-12/changes.html#12.3 3180 172. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.3 3181 173. https://gcc.gnu.org/gcc-12/changes.html#12.4 3182 174. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.4 3183 175. https://gcc.gnu.org/onlinedocs/ 3184 176. mailto:gcc-help (a] gcc.gnu.org 3185 177. mailto:gcc (a] gcc.gnu.org 3186 178. https://gcc.gnu.org/lists.html 3187 179. https://www.fsf.org/ 3188 180. https://gcc.gnu.org/about.html 3189 ====================================================================== 3190 http://gcc.gnu.org/gcc-11/index.html 3191 3192 GCC 11 Release Series 3193 3194 July 19, 2024 3195 3196 The GCC developers are pleased to announce the release of GCC 11.5. 3197 3198 This release is a bug-fix release, containing fixes for regressions in 3199 GCC 11.4 relative to previous releases of GCC. 3200 3201 Release History 3202 3203 GCC 11.5 3204 July 19, 2024 ([1]changes, [2]documentation) 3205 3206 GCC 11.4 3207 May 29, 2023 ([3]changes, [4]documentation) 3208 3209 GCC 11.3 3210 April 21, 2022 ([5]changes, [6]documentation) 3211 3212 GCC 11.2 3213 July 28, 2021 ([7]changes, [8]documentation) 3214 3215 GCC 11.1 3216 April 27, 2021 ([9]changes, [10]documentation) 3217 3218 References and Acknowledgements 3219 3220 GCC used to stand for the GNU C Compiler, but since the compiler 3221 supports several other languages aside from C, it now stands for the 3222 GNU Compiler Collection. 3223 3224 The GCC developers would like to thank the numerous people that have 3225 contributed new features, improvements, bug fixes, and other changes as 3226 well as test results to GCC. This [11]amazing group of volunteers is 3227 what makes GCC successful. 3228 3229 For additional information about GCC please refer to the [12]GCC 3230 project web site or contact the [13]GCC development mailing list. 3231 3232 To obtain GCC please use [14]our mirror sites or [15]our version 3233 control system. 3234 3235 3236 For questions related to the use of GCC, please consult these web 3237 pages and the [16]GCC manuals. If that fails, the 3238 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3239 web pages and the development of GCC are welcome on our developer 3240 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 3241 archives. 3242 3243 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 3244 distribution of this entire article is permitted in any medium, 3245 provided this notice is preserved. 3246 3247 These pages are [21]maintained by the GCC team. Last modified 3248 2024-07-19. 3249 3250 References 3251 3252 1. https://gcc.gnu.org/gcc-11/changes.html 3253 2. https://gcc.gnu.org/onlinedocs/11.5.0/ 3254 3. https://gcc.gnu.org/gcc-11/changes.html 3255 4. https://gcc.gnu.org/onlinedocs/11.4.0/ 3256 5. https://gcc.gnu.org/gcc-11/changes.html 3257 6. https://gcc.gnu.org/onlinedocs/11.3.0/ 3258 7. https://gcc.gnu.org/gcc-11/changes.html 3259 8. https://gcc.gnu.org/onlinedocs/11.2.0/ 3260 9. https://gcc.gnu.org/gcc-11/changes.html 3261 10. https://gcc.gnu.org/onlinedocs/11.1.0/ 3262 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html 3263 12. https://gcc.gnu.org/index.html 3264 13. mailto:gcc (a] gcc.gnu.org 3265 14. https://gcc.gnu.org/mirrors.html 3266 15. https://gcc.gnu.org/git.html 3267 16. https://gcc.gnu.org/onlinedocs/ 3268 17. mailto:gcc-help (a] gcc.gnu.org 3269 18. mailto:gcc (a] gcc.gnu.org 3270 19. https://gcc.gnu.org/lists.html 3271 20. https://www.fsf.org/ 3272 21. https://gcc.gnu.org/about.html 3273 ====================================================================== 3274 http://gcc.gnu.org/gcc-11/changes.html 3275 3276 GCC 11 Release Series 3277 Changes, New Features, and Fixes 3278 3279 This page is a "brief" summary of some of the huge number of 3280 improvements in GCC 11. You may also want to check out our [1]Porting 3281 to GCC 11 page and the [2]full GCC documentation. 3282 3283 Caveats 3284 3285 * The default mode for C++ is now -std=gnu++17 instead of 3286 -std=gnu++14. Note that [3]C++17 changes to template template 3287 parameter matching can be disabled independently of other features 3288 with -fno-new-ttp-matching. 3289 * When building GCC itself, the host compiler must now support C++11, 3290 rather than C++98. In particular bootstrapping GCC 11 using an 3291 older version of GCC requires a binary of GCC 4.8 or later, rather 3292 than of GCC 3.4 or later as was the case for bootstrapping GCC 10. 3293 * Naming and location of auxiliary and dump output files changed. If 3294 you compile multiple input files in a single command, if you enable 3295 Link Time Optimization, or if you use -dumpbase, -dumpdir, 3296 -save-temps=*, and you expect any file other than the primary 3297 output file(s) to be created as a side effect, watch out for 3298 improvements and a few surprises. See [4]the patch, particularly 3299 its textual description, for more details about the changes. 3300 * -gsplit-dwarf no longer enables debug info generation on its own 3301 but requires a separate -g for this. 3302 * The libstdc++ configure option --enable-cheaders=c_std is 3303 deprecated and will be removed in a future release. It should be 3304 possible to use --enable-cheaders=c_global (the default) with no 3305 change in behaviour. 3306 * The front end for compiling BRIG format of Heterogeneous System 3307 Architecture Intermediate Language (HSAIL) has been deprecated and 3308 will likely be removed in a future release. 3309 * Some short options of the gcov tool have been renamed: -i to -j and 3310 -j to -H. 3311 3312 [5]General Improvements 3313 3314 * [6]ThreadSanitizer improvements to support alternative runtimes and 3315 environments. The [7]Linux Kernel Concurrency Sanitizer (KCSAN) is 3316 now supported. 3317 + Add --param tsan-distinguish-volatile to optionally emit 3318 instrumentation distinguishing volatile accesses. 3319 + Add --param tsan-instrument-func-entry-exit to optionally 3320 control if function entries and exits should be instrumented. 3321 * In previous releases of GCC, the "column numbers" emitted in 3322 diagnostics were actually a count of bytes from the start of the 3323 source line. This could be problematic, both because of: 3324 + multibyte characters (requiring more than one byte to encode), 3325 and 3326 + multicolumn characters (requiring more than one column to 3327 display in a monospace font) 3328 For example, the character p ("GREEK SMALL LETTER PI (U+03C0)") 3329 occupies one column, and its UTF-8 encoding requires two bytes; the 3330 character 🙂 ("SLIGHTLY SMILING FACE (U+1F642)") occupies 3331 two columns, and its UTF-8 encoding requires four bytes. 3332 In GCC 11 the column numbers default to being column numbers, 3333 respecting multi-column characters. The old behavior can be 3334 restored using a new option [8]-fdiagnostics-column-unit=byte. 3335 There is also a new option [9]-fdiagnostics-column-origin=, 3336 allowing the pre-existing default of the left-hand column being 3337 column 1 to be overridden if desired (e.g. for 0-based columns). 3338 The output of [10]-fdiagnostics-format=json has been extended to 3339 supply both byte counts and column numbers for all source 3340 locations. 3341 Additionally, in previous releases of GCC, tab characters in the 3342 source would be emitted verbatim when quoting source code, but be 3343 prefixed with whitespace or line number information, leading to 3344 misalignments in the resulting output when compared with the actual 3345 source. Tab characters are now printed as an appropriate number of 3346 spaces, using the [11]-ftabstop option (which defaults to 8 spaces 3347 per tab stop). 3348 * Introduce [12]Hardware-assisted AddressSanitizer support. This 3349 sanitizer currently only works for the AArch64 target. It helps 3350 debug address problems similarly to [13]AddressSanitizer but is 3351 based on partial hardware assistance and provides probabilistic 3352 protection to use less RAM at run time. [14]Hardware-assisted 3353 AddressSanitizer is not production-ready for user space, and is 3354 provided mainly for use compiling the Linux Kernel. 3355 To use this sanitizer the command line arguments are: 3356 + -fsanitize=hwaddress to instrument userspace code. 3357 + -fsanitize=kernel-hwaddress to instrument kernel code. 3358 * For targets that produce DWARF debugging information GCC now 3359 defaults to [15]DWARF version 5 (with the exception of VxWorks and 3360 Darwin/Mac OS X which default to version 2 and AIX which defaults 3361 to version 4). This can produce up to 25% more compact debug 3362 information compared to earlier versions. 3363 To take full advantage of DWARF version 5 GCC needs to be built 3364 against binutils version 2.35.2 or higher. When GCC is built 3365 against earlier versions of binutils GCC will still emit DWARF 3366 version 5 for most debuginfo data, but will generate version 4 3367 debug line tables (even when explicitly given -gdwarf-5). 3368 The following debug information consumers can process DWARF version 3369 5: 3370 + GDB 8.0, or higher 3371 + valgrind 3.17.0 3372 + elfutils 0.172, or higher (for use with systemtap, 3373 dwarves/pahole, perf and libabigail) 3374 + dwz 0.14 3375 Programs embedding libbacktrace are urged to upgrade to the version 3376 shipping with GCC 11. 3377 To make GCC 11 generate an older DWARF version use -g together with 3378 -gdwarf-2, -gdwarf-3 or -gdwarf-4. 3379 * Vectorizer improvements: 3380 + The straight-line code vectorizer now considers the whole 3381 function when vectorizing and can handle opportunities 3382 crossing CFG merges and backedges. 3383 * A series of conditional expressions that compare the same variable 3384 can be transformed into a switch statement if each of them contains 3385 a comparison expression. Example: 3386 int IsHTMLWhitespace(int aChar) { 3387 return aChar == 0x0009 || aChar == 0x000A || 3388 aChar == 0x000C || aChar == 0x000D || 3389 aChar == 0x0020; 3390 } 3391 3392 This statement can be transformed into a switch statement and then 3393 expanded into a bit-test. 3394 * New command-line options: 3395 + [16]-fbit-tests, enabled by default, can be used to enable or 3396 disable switch expansion using bit-tests. 3397 * Inter-procedural optimization improvements: 3398 + A new IPA-modref pass was added to track side effects of 3399 function calls and improve precision of points-to-analysis. 3400 The pass can be controlled by the [17]-fipa-modref option. 3401 + The identical code folding pass (controlled by [18]-fipa-icf) 3402 was significantly improved to increase the number of unified 3403 functions and to reduce compile-time memory use. 3404 + IPA-CP (Interprocedural constant propagation) heuristics 3405 improved its estimation of potential usefulness of known loop 3406 bounds and strides by taking the estimated frequency of these 3407 loops into account. 3408 * Link-time optimization improvements: 3409 + The LTO bytecode format was optimized for smaller object files 3410 and faster streaming. 3411 + Memory allocation of the linking stage was improved to reduce 3412 peak memory use. 3413 * Profile driven optimization improvements: 3414 + Using [19]-fprofile-values, was improved by tracking more 3415 target values for e.g. indirect calls. 3416 + GCOV data file format outputs smaller files by representing 3417 zero counters in a more compact way. 3418 3419 [20]New Languages and Language specific improvements 3420 3421 * GCC 11 adds support for non-rectangular loop nests in OpenMP 3422 constructs and the allocator routines of [21]OpenMP 5.0, including 3423 initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD 3424 environment variable and the active-levels routines are now 3425 supported. For C/C++, the declare variant and map support has been 3426 extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP 3427 5.0 support has been extended, including the following features 3428 which were before only available in C and C++: order(concurrent), 3429 device_type, memorder-clauses for flush, lastprivate with 3430 conditional modifier, atomic construct and reduction clause 3431 extensions of OpenMP 5.0, if clause with simd and cancel modifiers, 3432 target data without map clause, and limited support for the 3433 requires construct. 3434 * Version 2.6 of the [22]OpenACC specification continues to be 3435 maintained and improved in the C, C++ and Fortran compilers. See 3436 the [23]implementation status section on the OpenACC wiki page and 3437 the [24]run-time library documentation for further information. 3438 3439 [25]C family 3440 3441 * New attributes: 3442 + The [26]no_stack_protector attribute has been added to mark 3443 functions which should not be instrumented with stack 3444 protection (-fstack-protector). 3445 + The existing [27]malloc attribute has been extended so that it 3446 can be used to identify allocator/deallocator API pairs. A 3447 pair of new [28]-Wmismatched-dealloc and 3448 [29]-Wmismatched-new-delete warnings will complain about 3449 mismatched calls, and [30]-Wfree-nonheap-object about 3450 deallocation calls with pointers not obtained from allocation 3451 functions. Additionally, the static analyzer will use these 3452 attributes when checking for leaks, double-frees, 3453 use-after-frees, and similar issues. 3454 * New warnings: 3455 + [31]-Wmismatched-dealloc, enabled by default, warns about 3456 calls to deallocation functions with pointers returned from 3457 mismatched allocation functions. 3458 + [32]-Wsizeof-array-div, enabled by -Wall, warns about 3459 divisions of two sizeof operators when the first one is 3460 applied to an array and the divisor does not equal the size of 3461 the array element. 3462 + [33]-Wstringop-overread, enabled by default, warns about calls 3463 to string functions reading past the end of the arrays passed 3464 to them as arguments. In prior GCC releases most instances of 3465 his warning are diagnosed by -Wstringop-overflow. 3466 + [34]-Wtsan, enabled by default, warns about unsupported 3467 features in ThreadSanitizer (currently 3468 std::atomic_thread_fence). 3469 * Enhancements to existing warnings: 3470 + [35]-Wfree-nonheap-object detects many more instances of calls 3471 to deallocation functions with pointers not returned from a 3472 dynamic memory allocation function. 3473 + [36]-Wmaybe-uninitialized diagnoses passing pointers or 3474 references to uninitialized memory to functions taking 3475 const-qualified arguments. 3476 + [37]-Wuninitialized detects reads from uninitialized 3477 dynamically allocated memory. 3478 * For ELF targets that support the GNU or FreeBSD OSABIs, the used 3479 attribute will now save the symbol declaration it is applied to 3480 from linker garbage collection. 3481 To support this behavior, used symbols that have not been placed in 3482 specific sections (e.g. with the section attribute, or the 3483 -f{function,data}-sections options) will be placed in new, unique 3484 sections. 3485 This functionality requires Binutils version 2.36 or later. 3486 3487 [38]C 3488 3489 * Several new features from the upcoming C2X revision of the ISO C 3490 standard are supported with -std=c2x and -std=gnu2x. Some of these 3491 features are also supported as extensions when compiling for older 3492 language versions. In addition to the features listed, some 3493 features previously supported as extensions and now added to the C 3494 standard are enabled by default in C2X mode and not diagnosed with 3495 -std=c2x -Wpedantic. 3496 + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>. 3497 + As in C++, function definitions no longer need to give names 3498 for unused function parameters. 3499 + The expansions of the true and false macros in <stdbool.h> 3500 have changed so that they have type bool. 3501 + The [[nodiscard]] standard attribute is now supported. 3502 + The __has_c_attribute preprocessor operator is now supported. 3503 + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN, 3504 DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros 3505 for _FloatN, _FloatNx and _DecimalN types, are provided in 3506 <float.h>. There are also corresponding built-in functions 3507 __builtin_nansdN for decimal signaling NaNs. 3508 + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and 3509 LDBL_IS_IEC_60559 are provided in <float.h>. 3510 + The feature test macro __STDC_WANT_IEC_60559_EXT__ is 3511 supported by <float.h>. 3512 + Labels may appear before declarations and at the end of a 3513 compound statement. 3514 * New warnings: 3515 + [39]-Warray-parameter, enabled by -Wall, warns about 3516 redeclarations of functions with ordinary array arguments 3517 declared using inconsistent forms. The warning also enables 3518 the detection of the likely out of bounds accesses in calls to 3519 such functions with smaller arrays. 3520 + [40]-Wvla-parameter, enabled by -Wall, warns redeclarations of 3521 functions with variable length array arguments declared using 3522 inconsistent forms or with mismatched bounds. The warning also 3523 enables the detection of the likely out of bounds accesses in 3524 calls to such functions with smaller arrays. 3525 3526 [41]C++ 3527 3528 * The default mode has been changed to -std=gnu++17. 3529 * Several C++20 features have been implemented: 3530 + the compiler now supports consteval virtual functions 3531 + P2082R1, Fixing CTAD for aggregates 3532 + P0593R6, Pseudo-destructors end object lifetimes 3533 + P1907R1, Inconsistencies with non-type template parameters 3534 (complete implementation) 3535 + P1975R0, Fixing the wording of parenthesized 3536 aggregate-initialization 3537 + P1009R2, Array size deduction in new-expressions 3538 + P1099R5, using enum 3539 + Modules, Requires -fmodules-ts and some aspects are 3540 incomplete. Refer to [42]C++ 20 Status 3541 * The C++ front end has experimental support for some of the upcoming 3542 C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b 3543 or -std=gnu++2b flags, including 3544 + P0330R8, Literal Suffix for (signed) size_t. 3545 For a full list of new features, see [43]the C++ status page. 3546 * Several C++ Defect Reports have been resolved, e.g.: 3547 + DR 625, Use of auto as a template-argument 3548 + DR 1512, Pointer comparison vs qualification conversions 3549 + DR 1722, Should lambda to function pointer conversion function 3550 be noexcept? 3551 + DR 1914, Duplicate standard attributes 3552 + DR 2032, Default template-arguments of variable templates 3553 + DR 2289, Uniqueness of decomposition declaration names 3554 + DR 2237, Can a template-id name a constructor? 3555 + DR 2303, Partial ordering and recursive variadic inheritance 3556 + DR 2369, Ordering between constraints and substitution 3557 + DR 2450, braced-init-list as a template-argument 3558 * G++ now performs better access checking in templates ([44]PR41437). 3559 * reinterpret_casts in constexpr evaluation are now checked more 3560 completely ([45]PR95307). 3561 * The diagnostic for static_assert has been improved: the compiler 3562 now shows the expression including its template arguments (if there 3563 were any), and can point to the failing clause if the condition 3564 comprised of any logical AND operators ([46]PR97518). 3565 * New warnings: 3566 + [47]-Wctad-maybe-unsupported, disabled by default, warns about 3567 performing class template argument deduction on a type with no 3568 deduction guides. 3569 + [48]-Wrange-loop-construct, enabled by -Wall, warns when a 3570 range-based for-loop is creating unnecessary and expensive 3571 copies. 3572 + [49]-Wdeprecated-enum-enum-conversion, enabled by default in 3573 C++20, warns about deprecated arithmetic conversions on 3574 operands of enumeration types, as outlined in 3575 [depr.arith.conv.enum]. 3576 + [50]-Wdeprecated-enum-float-conversion, enabled by default in 3577 C++20, warns about deprecated arithmetic conversions on 3578 operands where one is of enumeration type and the other is of 3579 a floating-point type, as outlined in [depr.arith.conv.enum]. 3580 + [51]-Wmismatched-new-delete, enabled by -Wall, warns about 3581 calls to C++ operator delete with pointers returned from 3582 mismatched forms of operator new or from other mismatched 3583 allocation functions. 3584 + [52]-Wvexing-parse, enabled by default, warns about the most 3585 vexing parse rule: the cases when a declaration looks like a 3586 variable definition, but the C++ language requires it to be 3587 interpreted as a function declaration. 3588 * Enhancements to existing warnings: 3589 + [53]-Wnonnull considers the implicit this argument of every 3590 C++ nonstatic member function to have been implicitly declared 3591 with attribute nonnull and triggers warnings for calls where 3592 the pointer is null. 3593 3594 [54]Runtime Library (libstdc++) 3595 3596 * Improved C++17 support, including: 3597 + std::from_chars and std::to_chars for floating-point types. 3598 * Improved experimental C++20 support, including: 3599 + Calendar additions to <chrono>. Thanks to Cassio Neri for 3600 optimizations. 3601 + std::bit_cast 3602 + std::source_location 3603 + Atomic wait and notify operations. 3604 + <barrier>, <latch>, and <semaphore> 3605 + <syncstream> 3606 + Efficient access to basic_stringbuf's buffer. 3607 + Heterogeneous lookup in unordered containers. 3608 * Experimental C++23 support, including: 3609 + contains member functions for strings, thanks to Paul Fee. 3610 + std::to_underlying, std::is_scoped_enum 3611 * Experimental support for Data-Parallel Types (simd) from the 3612 Parallelism 2 TS, thanks to Matthias Kretz. 3613 * Faster std::uniform_int_distribution, thanks to Daniel Lemire. 3614 3615 [55]D 3616 3617 * New features: 3618 + A new bottom type typeof(*null) has been added to represent 3619 run-time errors and non-terminating functions. This also 3620 introduces a new standard alias for the type named noreturn, 3621 and is implicitly imported into every module. 3622 + Printf-like and scanf-like functions are now detected by 3623 prefixing them with pragma(printf) for printf-like functions 3624 or pragma(scanf) for scanf-like functions. 3625 + The __traits() expression now supports the extensions 3626 isDeprecated, isDisabled, isFuture, isModule, isPackage, 3627 child, isReturnOnStack, isZeroInit, getTargetInfo, 3628 getLocation, hasPostblit, isCopyable, getVisibility, and 3629 totype. 3630 + An expression-based contract syntax has been added to the 3631 language. 3632 + Function literals can now return a value by reference with the 3633 ref keyword. 3634 + A new syntax is available to declare aliases to function types 3635 using the alias syntax based on the assignment operator. 3636 + New types __c_complex_float, __c_complex_double, 3637 __c_complex_real, and __c_wchar_t have been added for 3638 interfacing with C and C++ code, and are available from the 3639 core.stdc.config module. 3640 + User-defined attributes can now be used to annotate enum 3641 members, alias declarations, and function parameters. 3642 + Template alias parameters can now be instantiated with basic 3643 types such as int or void function(). 3644 + The mixin construct can now be used as types in the form 3645 mixin(string) var. 3646 + The mixin construct can now take an argument list, same as 3647 pragma(msg). 3648 * New intrinsics: 3649 + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror 3650 have been added. 3651 + Byte swap intrinsic core.bitop.byteswap for swapping bytes in 3652 a 2-byte ushort has been added. 3653 + Math intrinsics available from core.math now have overloads 3654 for float and double types. 3655 + Volatile intrinsics core.volatile.volatileLoad and 3656 core.volatile.volatileStore have been moved from the 3657 core.bitop module. 3658 * New attributes: 3659 + The following GCC attributes are now recognized and available 3660 from the gcc.attributes module with short-hand aliases for 3661 convenience: 3662 o @attribute("alloc_size", arguments) or 3663 @alloc_size(arguments). 3664 o @attribute("always_inline") or @always_inline. 3665 o @attribute("used") or @used. 3666 o @attribute("optimize", arguments) or 3667 @optimize(arguments). 3668 o @attribute("cold") or @cold. 3669 o @attribute("noplt") or @noplt. 3670 o @attribute("target_clones", arguments) or 3671 @target_clones(arguments). 3672 o @attribute("no_icf") or @no_icf. 3673 o @attribute("noipa") or @noipa. 3674 o @attribute("symver", arguments) or @symver(arguments). 3675 + New aliases have been added to gcc.attributes for 3676 compatibility with ldc.attributes. 3677 o The @allocSize(arguments) attribute is the same as 3678 @alloc_size(arguments), but uses a 0-based index for 3679 function arguments. 3680 o The @assumeUsed attribute is an alias for 3681 @attribute("used"). 3682 o The @fastmath attribute is an alias for 3683 @optimize("Ofast"). 3684 o The @naked attribute is an alias for @attribute("naked"). 3685 This attribute may not be available on all targets. 3686 o The @restrict attribute has been added to specify that a 3687 function parameter is to be restrict-qualified in the C99 3688 sense of the term. 3689 o The @optStrategy(strategy) attribute is an alias for 3690 @optimize("O0") when the strategy is "none", otherwise 3691 @optimize("Os") for the "optsize" and "minsize" 3692 strategies. 3693 o The @polly attribute is an alias for 3694 @optimize("loop-parallelize-all"). 3695 o The @section(name) attribute is an alias for 3696 @attribute("section", name). 3697 o The @target(arguments) attribute is an alias for 3698 attribute("target", arguments). 3699 o The @weak attribute is an alias for @attribute("weak"). 3700 * New language options: 3701 + -fweak-templates, added to control whether declarations that 3702 can be defined in multiple objects should be emitted as weak 3703 symbols. The default is to emit all symbols with extern 3704 linkage as weak, unless the target lacks support for weak 3705 symbols. 3706 + -Wdeprecated, this option is now enabled by default. 3707 + -Wextra, this option now turns on all warnings that are not 3708 part of the core D language front-end - -Waddress, 3709 -Wcast-result, -Wunknown-pragmas. 3710 + -Wvarargs, added to turn on warnings about questionable usage 3711 of the va_start intrinsic. 3712 * Deprecated and removed features: 3713 + Compiler-recognized attributes are now made available from the 3714 gcc.attributes module, the former module gcc.attribute has 3715 been deprecated and will be removed in a future release. 3716 + The @attribute("alias") attribute has been removed, as it had 3717 been superseded by pragma(mangle). 3718 + The @attribute("forceinline") attribute has been removed and 3719 renamed to @always_inline. 3720 + __vector types that are not supported in hardware are now 3721 rejected at compile-time. Previously all vector types were 3722 accepted by the compiler and emulated when target support was 3723 absent. 3724 + The extern(Pascal) linkage attribute has been removed. 3725 + The deprecation phase for -ftransition=import and 3726 -ftransition=checkimports is finished. These switches no 3727 longer have an effect and are now removed. Symbols that are 3728 not visible in a particular scope will no longer be found by 3729 the compiler. 3730 + It is now an error to use private variables selectively 3731 imported from other modules. Due to a bug, some imported 3732 private members were visible from other modules, violating the 3733 specification. 3734 + The -fweak compiler switch has been removed, as it existed 3735 only for testing. 3736 3737 [56]Fortran 3738 3739 * Added DEPRECATED to !GCC$'s attributes directive. 3740 3741 [57]Go 3742 3743 * GCC 11 provides a complete implementation of the Go 1.16.3 user 3744 packages. 3745 3746 [58]libgccjit 3747 3748 * libgccjit was marked as merely "Alpha" quality when [59]originally 3749 added in GCC 5. Given that we have maintained [60]API and ABI 3750 compatibility since then and it is in use by various projects, we 3751 have removed that caveat. 3752 * libgccjit can now be built for MinGW 3753 * The libgccjit API gained 10 new entry points: 3754 + [61]gcc_jit_global_set_initializer 3755 + 9 entrypoints for [62]directly embedding asm statements into a 3756 compile, analogous to inline asm in the C front end 3757 3758 [63]New Targets and Target Specific Improvements 3759 3760 [64]AArch64 & arm 3761 3762 * A number of new CPUs are supported through arguments to the -mcpu 3763 and -mtune options in both the arm and aarch64 back ends (GCC 3764 identifiers in parentheses): 3765 + Arm Cortex-A78 (cortex-a78). 3766 + Arm Cortex-A78AE (cortex-a78ae). 3767 + Arm Cortex-A78C (cortex-a78c). 3768 + Arm Cortex-X1 (cortex-x1). 3769 + Arm Neoverse V1 (neoverse-v1). 3770 + Arm Neoverse N2 (neoverse-n2). 3771 * GCC can now auto-vectorize operations performing addition, 3772 subtraction, multiplication and the accumulate/subtract variants on 3773 complex numbers, taking advantage of the Advanced SIMD instructions 3774 in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64) 3775 and MVE (AArch32 M-profile) instruction sets. 3776 3777 [65]AArch64 3778 3779 * In addition to the above, the following AArch64-only CPUs are now 3780 supported: 3781 + Fujitsu A64FX (a64fx). 3782 + Arm Cortex-R82 (cortex-r82). 3783 * The AArch64 Armv8-R architecture is now supported through the 3784 -march=armv8-r option. 3785 * Mitigation against the [66]Straight-line Speculation vulnerability 3786 is supported with the -mharden-sls= option. Please refer to the 3787 documentation for usage instructions. 3788 * The availability of Advanced SIMD intrinsics available through the 3789 arm_neon.h header is improved and GCC 11 supports the full set of 3790 intrinsics defined by ACLE Q3 2020. 3791 3792 [67]AMD Radeon (GCN) 3793 3794 * Initial support for gfx908 GPUs has been added. 3795 3796 [68]arm 3797 3798 * Initial auto-vectorization is now available when targeting the MVE 3799 instruction set. 3800 * GCC can now make use of the Low Overhead Branch instruction in 3801 Armv8.1-M to optimize loop counters and checks. 3802 * The -mcpu=cortex-m55 option now supports the extensions +nomve and 3803 +nomve.fp to control generation of MVE and MVE floating-point 3804 instructions. 3805 3806 [69]IA-32/x86-64 3807 3808 * New ISA extension support for Intel TSXLDTRK was added to GCC. 3809 TSXLDTRK intrinsics are available via the -mtsxldtrk compiler 3810 switch. 3811 * New ISA extension support for Intel SERIALIZE was added to GCC. 3812 SERIALIZE intrinsics are available via the -mserialize compiler 3813 switch. 3814 * New ISA extension support for Intel HRESET was added to GCC. HRESET 3815 intrinsics are available via the -mhreset compiler switch. 3816 * New ISA extension support for Intel UINTR was added to GCC. UINTR 3817 intrinsics are available via the -muintr compiler switch. 3818 * New ISA extension support for Intel KEYLOCKER was added to GCC. 3819 KEYLOCKER intrinsics are available via the -mkeylocker compiler 3820 switch. 3821 * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16 3822 was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are 3823 available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler 3824 switches. 3825 * New ISA extension support for Intel AVX-VNNI was added to GCC. 3826 AVX-VNNI intrinsics are available via the -mavxvnni compiler 3827 switch. 3828 * GCC now supports the Intel CPU named Sapphire Rapids through 3829 -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B, 3830 AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG, 3831 TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA 3832 extensions. 3833 * GCC now supports the Intel CPU named Alderlake through 3834 -march=alderlake. The switch enables the CLDEMOTE, PTWRITE, 3835 WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions. 3836 * GCC now supports the Intel CPU named Rocketlake through 3837 -march=rocketlake. Rocket Lake is based on Icelake client and minus 3838 SGX. 3839 * GCC now supports AMD CPUs based on the znver3 core via 3840 -march=znver3. 3841 * GCC now supports micro-architecture levels defined in the x86-64 3842 psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4. 3843 3844 [70]Nios II 3845 3846 * The options -mcustom-insn=N no longer produce compiler warnings if 3847 the custom instruction is not generated due to missing optimization 3848 options such as -fno-math-errno, -ffinite-math-only, or 3849 -funsafe-math-optimizations. These warnings were not consistently 3850 emitted for all custom instructions. 3851 * The -mcustom-fpu-cfg=fph2 option has been added to enable the 3852 custom instructions supported by the Nios II Floating Point 3853 Hardware 2 Component. 3854 3855 [71]NVPTX 3856 3857 * The -misa default has changed from sm_30 to sm_35. 3858 * The -m32 compiler switch has been removed. 3859 * The -msoft-stack-reserve-local format has been fixed. Previously, 3860 it accepted -msoft-stack-reserve-local<n>. It now accepts 3861 -msoft-stack-reserve-local=<n>. 3862 3863 [72]S/390, System z, IBM Z Systems 3864 3865 * The behavior when compiling with -fexcess-precision=standard (e.g., 3866 implied by -std=c99) on s390(x) targets can now be controlled at 3867 configure time with the flag --enable-s390-excess-float-precision. 3868 When enabled, GCC will maintain previous behavior and evaluate 3869 float expressions in double precision, which aligns with the 3870 definition of float_t as double. With the flag disabled, GCC will 3871 always evaluate float expressions in single precision. In native 3872 builds and cross compiles that have target libc headers, GCC will 3873 by default match the definition of float_t in the installed glibc. 3874 3875 [73]RISC-V 3876 3877 * Support address sanitizer for RISC-V. 3878 * Support big-endian for RISC-V, thanks to Marcus Comstedt. 3879 * Implement new style of architecture extension test macros: each 3880 architecture extension has a corresponding feature test macro, 3881 which can be used to test its existence and version information. 3882 * Legacy architecture extension test macros like __riscv_atomic are 3883 deprecated, but will still be supported for at least 2 release 3884 cycles. 3885 * Support IFUNC for riscv*-*-linux*. 3886 * Add new option -misa-spec=* to control ISA spec version. This 3887 controls the default version of each extensions. It defaults to 3888 2.2. 3889 * Introduce the --with-multilib-generator configure time option. This 3890 allows for flexible config multi-lib settings. Its syntax is the 3891 same as RISC-V's multilib-generator. 3892 * Extend the sytax for multilib-generator, support expansion operator 3893 * to reduce the complexity of complicated multi-lib re-use rules. 3894 * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the 3895 pipeline model and architecture extensions, like -mtune=* plus 3896 -march=*. 3897 * Support for TLS stack protector canary access, thanks to Cooper Qu. 3898 * Support __builtin_thread_pointer for RISC-V. 3899 * Introduce shorten_memrefs optimization, which reduces the code size 3900 for memory access, thanks to Craig Blackmore. 3901 3902 [74]Operating Systems 3903 3904 [75]AIX 3905 3906 * GCC for AIX can be built as a 64-bit application and the runtime is 3907 built as FAT libraries containing both 32-bit and 64-bit objects. 3908 * Support AIX Vector Extended ABI with -mabi=vec-extabi. 3909 * Thread-Local uninitiated data placed in local common section. 3910 * Use thread-safe access in ctype. 3911 * Link with libc128.a when long-double-128 enabled. 3912 3913 [76]Improvements to Static Analyzer 3914 3915 * The implementation of how program state is tracked within 3916 [77]-fanalyzer has been completely rewritten for GCC 11, fixing 3917 numerous bugs, and allowing for the analyzer to scale to larger C 3918 source files. 3919 * The analysis of allocations and deallocations has been generalized 3920 beyond malloc and free. 3921 + As preliminary work towards eventually supporting C++, the 3922 malloc/free checking will also check new/delete and 3923 new[]/delete[]. However, C++ is not yet properly supported by 3924 [78]-fanalyzer (for example, exception-handling is 3925 unimplemented). 3926 + As noted above, the existing [79]malloc attribute has been 3927 extended so that it can be used to identify 3928 allocator/deallocator API pairs. The analyzer will use these 3929 attributes when checking for leaks, double-frees, 3930 use-after-frees, and similar issues. 3931 + A new [80]-Wanalyzer-mismatching-deallocation warning has been 3932 added, covering such mismatches as using scalar delete rather 3933 vector delete[]. 3934 * The analyzer has gained warnings 3935 [81]-Wanalyzer-shift-count-negative, 3936 [82]-Wanalyzer-shift-count-overflow, [83]-Wanalyzer-write-to-const, 3937 and [84]-Wanalyzer-write-to-string-literal, all enabled by default 3938 when [85]-fanalyzer is enabled. 3939 * The analyzer can now be extended by GCC plugins, allowing for 3940 domain-specific path-sensitive warnings. An example of using a 3941 [86]GCC plugin to check for misuses of CPython's global interpreter 3942 lock can be seen in the test suite 3943 * The analyzer has gained new debugging options 3944 [87]-fdump-analyzer-json and [88]-fno-analyzer-feasibility. 3945 3946 Other significant improvements 3947 3948 * GCC has gained a new environment variable 3949 [89]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to 3950 request machine-readable fix-it hints without needing to adjust 3951 build flags. 3952 3953 [90]GCC 11.1 3954 3955 This is the [91]list of problem reports (PRs) from GCC's bug tracking 3956 system that are known to be fixed in the 11.1 release. This list might 3957 not be complete (that is, it is possible that some PRs that have been 3958 fixed are not listed here). 3959 3960 [92]GCC 11.2 3961 3962 This is the [93]list of problem reports (PRs) from GCC's bug tracking 3963 system that are known to be fixed in the 11.2 release. This list might 3964 not be complete (that is, it is possible that some PRs that have been 3965 fixed are not listed here). 3966 3967 [94]GCC 11.3 3968 3969 This is the [95]list of problem reports (PRs) from GCC's bug tracking 3970 system that are known to be fixed in the 11.3 release. This list might 3971 not be complete (that is, it is possible that some PRs that have been 3972 fixed are not listed here). 3973 3974 [96]GCC 11.4 3975 3976 This is the [97]list of problem reports (PRs) from GCC's bug tracking 3977 system that are known to be fixed in the 11.4 release. This list might 3978 not be complete (that is, it is possible that some PRs that have been 3979 fixed are not listed here). 3980 3981 Target Specific Changes 3982 3983 x86-64 3984 3985 * The x86-64 ABI of passing and returning structure with a 64-bit 3986 single precision vector changed in GCC 11.1 when 3DNOW is disabled. 3987 Disabling 3DNOW no longer changes how they are passed nor returned. 3988 This ABI change is now diagnosed with -Wpsabi. 3989 * Mitigation against straight line speculation (SLS) for function 3990 return and indirect jump is supported via 3991 -mharden-sls=[none|all|return|indirect-jmp]. 3992 * Add CS prefix to call and jmp to indirect thunk with branch target 3993 in r8-r15 registers via -mindirect-branch-cs-prefix. 3994 3995 [98]GCC 11.5 3996 3997 This is the [99]list of problem reports (PRs) from GCC's bug tracking 3998 system that are known to be fixed in the 11.5 release. This list might 3999 not be complete (that is, it is possible that some PRs that have been 4000 fixed are not listed here). 4001 4002 Caveats 4003 4004 aarch64 4005 4006 * Due to a bug introduced a few weeks before the 11.5.0 release if 4007 the compiler is configured without explicit --with-arch=, 4008 --with=cpu= and/or --with-tune= configure options the compiler 4009 without explicit -march= etc. options might act as if asked for 4010 cortex-a34. This can be fixed by applying manually the 4011 [100]r12-8060 commit patch on top of GCC 11.5.0. See [101]PR116029 4012 for more details. GCC 11.4.0 or earlier releases are not affected. 4013 4014 4015 For questions related to the use of GCC, please consult these web 4016 pages and the [102]GCC manuals. If that fails, the 4017 [103]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4018 web pages and the development of GCC are welcome on our developer 4019 list at [104]gcc (a] gcc.gnu.org. All of [105]our lists have public 4020 archives. 4021 4022 Copyright (C) [106]Free Software Foundation, Inc. Verbatim copying and 4023 distribution of this entire article is permitted in any medium, 4024 provided this notice is preserved. 4025 4026 These pages are [107]maintained by the GCC team. Last modified 4027 2025-01-31. 4028 4029 References 4030 4031 1. https://gcc.gnu.org/gcc-11/porting_to.html 4032 2. https://gcc.gnu.org/onlinedocs/index.html#current 4033 3. https://wg21.link/p0522r0 4034 4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html 4035 5. https://gcc.gnu.org/gcc-11/changes.html#general 4036 6. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual 4037 7. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html 4038 8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit 4039 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin 4040 10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 4041 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop 4042 12. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 4043 13. https://github.com/google/sanitizers/wiki/AddressSanitizer 4044 14. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 4045 15. https://dwarfstd.org/doc/DWARF5.pdf 4046 16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests 4047 17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref 4048 18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf 4049 19. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values 4050 20. https://gcc.gnu.org/gcc-11/changes.html#languages 4051 21. https://www.openmp.org/specifications/ 4052 22. https://www.openacc.org/ 4053 23. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-11 4054 24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html 4055 25. https://gcc.gnu.org/gcc-11/changes.html#c-family 4056 26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute 4057 27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 4058 28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 4059 29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 4060 30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 4061 31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 4062 32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div 4063 33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread 4064 34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan 4065 35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 4066 36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 4067 37. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized 4068 38. https://gcc.gnu.org/gcc-11/changes.html#c 4069 39. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter 4070 40. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter 4071 41. https://gcc.gnu.org/gcc-11/changes.html#cxx 4072 42. https://gcc.gnu.org/projects/cxx-status.html#cxx20 4073 43. https://gcc.gnu.org/projects/cxx-status.html#cxx23 4074 44. https://gcc.gnu.org/PR41437 4075 45. https://gcc.gnu.org/PR95307 4076 46. https://gcc.gnu.org/PR97518 4077 47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported 4078 48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct 4079 49. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion 4080 50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion 4081 51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 4082 52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse 4083 53. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull 4084 54. https://gcc.gnu.org/gcc-11/changes.html#libstdcxx 4085 55. https://gcc.gnu.org/gcc-11/changes.html#d 4086 56. https://gcc.gnu.org/gcc-11/changes.html#fortran 4087 57. https://gcc.gnu.org/gcc-11/changes.html#go 4088 58. https://gcc.gnu.org/gcc-11/changes.html#jit 4089 59. https://gcc.gnu.org/gcc-5/changes.html#jit 4090 60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html 4091 61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer 4092 62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html 4093 63. https://gcc.gnu.org/gcc-11/changes.html#targets 4094 64. https://gcc.gnu.org/gcc-11/changes.html#arm-targets 4095 65. https://gcc.gnu.org/gcc-11/changes.html#aarch64 4096 66. https://developer.arm.com/documentation/102587/0102/Straight-line-speculation-frequently-asked-questions 4097 67. https://gcc.gnu.org/gcc-11/changes.html#amdgcn 4098 68. https://gcc.gnu.org/gcc-11/changes.html#arm 4099 69. https://gcc.gnu.org/gcc-11/changes.html#x86 4100 70. https://gcc.gnu.org/gcc-11/changes.html#nios2 4101 71. https://gcc.gnu.org/gcc-11/changes.html#nvptx 4102 72. https://gcc.gnu.org/gcc-11/changes.html#s390 4103 73. https://gcc.gnu.org/gcc-11/changes.html#riscv 4104 74. https://gcc.gnu.org/gcc-11/changes.html#os 4105 75. https://gcc.gnu.org/gcc-11/changes.html#aix 4106 76. https://gcc.gnu.org/gcc-11/changes.html#analyzer 4107 77. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4108 78. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4109 79. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 4110 80. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation 4111 81. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative 4112 82. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow 4113 83. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 4114 84. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 4115 85. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 4116 86. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825 4117 87. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json 4118 88. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility 4119 89. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT 4120 90. https://gcc.gnu.org/gcc-11/changes.html#11.1 4121 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0 4122 92. https://gcc.gnu.org/gcc-11/changes.html#11.2 4123 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2 4124 94. https://gcc.gnu.org/gcc-11/changes.html#11.3 4125 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.3 4126 96. https://gcc.gnu.org/gcc-11/changes.html#11.4 4127 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.4 4128 98. https://gcc.gnu.org/gcc-11/changes.html#11.5 4129 99. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.5 4130 100. https://gcc.gnu.org/r12-8060 4131 101. https://gcc.gnu.org/PR116029 4132 102. https://gcc.gnu.org/onlinedocs/ 4133 103. mailto:gcc-help (a] gcc.gnu.org 4134 104. mailto:gcc (a] gcc.gnu.org 4135 105. https://gcc.gnu.org/lists.html 4136 106. https://www.fsf.org/ 4137 107. https://gcc.gnu.org/about.html 4138 ====================================================================== 4139 http://gcc.gnu.org/gcc-10/index.html 4140 4141 GCC 10 Release Series 4142 4143 (This release series is no longer supported.) 4144 4145 July, 7, 2023 4146 4147 The [1]GNU project and the GCC developers are pleased to announce the 4148 release of GCC 10.5. 4149 4150 This release is a bug-fix release, containing fixes for regressions in 4151 GCC 10.4 relative to previous releases of GCC. 4152 4153 Release History 4154 4155 GCC 10.5 4156 July 7, 2023 ([2]changes, [3]documentation) 4157 4158 GCC 10.4 4159 June 28, 2022 ([4]changes, [5]documentation) 4160 4161 GCC 10.3 4162 April 8, 2021 ([6]changes, [7]documentation) 4163 4164 GCC 10.2 4165 July 23, 2020 ([8]changes, [9]documentation) 4166 4167 GCC 10.1 4168 May 7, 2020 ([10]changes, [11]documentation) 4169 4170 References and Acknowledgements 4171 4172 GCC used to stand for the GNU C Compiler, but since the compiler 4173 supports several other languages aside from C, it now stands for the 4174 GNU Compiler Collection. 4175 4176 The GCC developers would like to thank the numerous people that have 4177 contributed new features, improvements, bug fixes, and other changes as 4178 well as test results to GCC. This [12]amazing group of volunteers is 4179 what makes GCC successful. 4180 4181 For additional information about GCC please refer to the [13]GCC 4182 project web site or contact the [14]GCC development mailing list. 4183 4184 To obtain GCC please use [15]our mirror sites or [16]our version 4185 control system. 4186 4187 4188 For questions related to the use of GCC, please consult these web 4189 pages and the [17]GCC manuals. If that fails, the 4190 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4191 web pages and the development of GCC are welcome on our developer 4192 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 4193 archives. 4194 4195 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 4196 distribution of this entire article is permitted in any medium, 4197 provided this notice is preserved. 4198 4199 These pages are [22]maintained by the GCC team. Last modified 4200 2024-05-30. 4201 4202 References 4203 4204 1. http://www.gnu.org/ 4205 2. https://gcc.gnu.org/gcc-10/changes.html 4206 3. https://gcc.gnu.org/onlinedocs/10.5.0/ 4207 4. https://gcc.gnu.org/gcc-10/changes.html 4208 5. https://gcc.gnu.org/onlinedocs/10.4.0/ 4209 6. https://gcc.gnu.org/gcc-10/changes.html 4210 7. https://gcc.gnu.org/onlinedocs/10.3.0/ 4211 8. https://gcc.gnu.org/gcc-10/changes.html 4212 9. https://gcc.gnu.org/onlinedocs/10.2.0/ 4213 10. https://gcc.gnu.org/gcc-10/changes.html 4214 11. https://gcc.gnu.org/onlinedocs/10.1.0/ 4215 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html 4216 13. https://gcc.gnu.org/index.html 4217 14. mailto:gcc (a] gcc.gnu.org 4218 15. https://gcc.gnu.org/mirrors.html 4219 16. https://gcc.gnu.org/git.html 4220 17. https://gcc.gnu.org/onlinedocs/ 4221 18. mailto:gcc-help (a] gcc.gnu.org 4222 19. mailto:gcc (a] gcc.gnu.org 4223 20. https://gcc.gnu.org/lists.html 4224 21. https://www.fsf.org/ 4225 22. https://gcc.gnu.org/about.html 4226 ====================================================================== 4227 http://gcc.gnu.org/gcc-10/changes.html 4228 4229 GCC 10 Release Series 4230 Changes, New Features, and Fixes 4231 4232 This page is a "brief" summary of some of the huge number of 4233 improvements in GCC 10. You may also want to check out our [1]Porting 4234 to GCC 10 page and the [2]full GCC documentation. 4235 4236 Caveats 4237 4238 * An ABI incompatibility between C++14 and C++17 has been fixed. On 4239 some targets a class with a zero-sized subobject would be passed 4240 incorrectly when compiled as C++17 or C++20. See the [3]C++ notes 4241 below for more details. 4242 * The deprecated Profile Mode and array_allocator extensions have 4243 been removed from libstdc++. 4244 * The non-standard std::__is_nullptr_t type trait is deprecated and 4245 will be removed from libstdc++ in a future release. The standard 4246 trait std::is_null_pointer should be instead. 4247 * The minimum version of the [4]MPFR library required for building 4248 GCC has been increased to version 3.1.0 (released 2011-10-03). 4249 * The automatic template instantiation at link time (-frepo) has been 4250 removed. 4251 * The --param allow-store-data-races internal parameter has been 4252 removed in favor of a new official option -fallow-store-data-races. 4253 While default behavior is unchanged and the new option allows to 4254 correctly maintain a per compilation unit setting across link-time 4255 optimization, alteration of the default via --param 4256 allow-store-data-races will now be diagnosed and build systems have 4257 to be adjusted accordingly. 4258 * Offloading to Heterogeneous System Architecture Intermediate 4259 Language (HSAIL) has been deprecated and will likely be removed in 4260 a future release. 4261 * The type of the std::iterator base class of 4262 std::istreambuf_iterator was changed in C++98 mode to be consistent 4263 with C++11 and later standards. See the [5]libstdc++ notes below 4264 for more details. 4265 * GCC 10.5 does not bootstrap with a C++98 compiler; if you need to 4266 start from C++98, you should build 10.4 or 9.5 instead. 4267 4268 [6]General Improvements 4269 4270 * New built-in functions: 4271 + The [7]__has_builtin built-in preprocessor operator can be 4272 used to query support for built-in functions provided by GCC 4273 and other compilers that support it. 4274 + __builtin_roundeven for the corresponding function from 4275 ISO/IEC TS 18661. 4276 * New command-line options: 4277 + [8]-fallocation-dce removes unneeded pairs of new and delete 4278 operators. 4279 + [9]-fprofile-partial-training can now be used to inform the 4280 compiler that code paths not covered by the training run 4281 should not be optimized for size. 4282 + [10]-fprofile-reproducible controls level of reproducibility 4283 of profile gathered by [11]-fprofile-generate. This makes it 4284 possible to rebuild program with same outcome which is useful, 4285 for example, for distribution packages. 4286 + [12]-fprofile-prefix-path can be used in combination with 4287 -fprofile-generate=profile_dir and -fprofile-use=profile_dir 4288 to inform GCC where the base directory of build source tree is 4289 in case it differs between instrumentation and optimized 4290 builds. 4291 + [13]-fanalyzer enables a new static analysis pass and 4292 associated warnings. This pass performs a time-consuming 4293 exploration of paths through the code in the hope of detecting 4294 various common errors, such as double-free bugs. This option 4295 should be regarded as experimental in this release. In 4296 particular, analysis of non-C code is unlikely to work. 4297 * Inter-procedural optimization improvements: 4298 + The inter-procedural scalar replacement of aggregates 4299 (IPA-SRA) pass was re-implemented to work at link-time and can 4300 now also remove computing and returning unused return values. 4301 + [14]-finline-functions is now enabled at -O2 and was retuned 4302 for better code size versus runtime performance trade-offs. 4303 Inliner heuristics was also significantly sped up to avoid 4304 negative impact to -flto -O2 compile times. 4305 + Inliner heuristics and function cloning can now use 4306 value-range information to predict effectivity of individual 4307 transformations. 4308 + During link-time optimization the C++ One Definition Rule is 4309 used to increase precision of type based alias analysis. 4310 * Link-time optimization improvements: 4311 + A new binary [15]lto-dump has been added. It dumps various 4312 information about LTO bytecode object files. 4313 + The parallel phase of the LTO can automatically detect a 4314 running make's jobserver or fall back to number of available 4315 cores. 4316 + The LTO bytecode can be compressed with the [16]zstd 4317 algorithm. The configure script automatically detects zstd 4318 support. 4319 + Most --param values can now be specified at translation unit 4320 granularity. This includes all parameters controlling the 4321 inliner and other inter-procedural optimizations. Unlike 4322 earlier releases, GCC 10 will ignore parameters controlling 4323 optimizations specified at link-time and apply parameters 4324 specified at compile-time in the same manner as done for 4325 optimization flags. 4326 * Profile driven optimization improvements: 4327 + Profile maintenance during compilation and hot/cold code 4328 partitioning have been improved. 4329 + Using [17]-fprofile-values, an instrumented binary can track 4330 multiple values (up to 4) for e.g. indirect calls and provide 4331 more precise profile information. 4332 4333 [18]New Languages and Language-Specific Improvements 4334 4335 * Version 2.6 of the [19]OpenACC specification is now supported by 4336 the C, C++ and Fortran compilers. See the [20]implementation status 4337 section on the OpenACC wiki page and the [21]run-time library 4338 documentation for further information. 4339 * GCC 10 adds a number of newly implemented [22]OpenMP 5.0 features 4340 such as conditional lastprivate clause, scan and loop directives, 4341 order(concurrent) and use_device_addr clauses support, if clause on 4342 simd construct, and partial support for the declare variant 4343 directive, getting closer to full support of the OpenMP 5.0 4344 standard. 4345 * OpenMP and OpenACC now support [23]offloading to AMD Radeon (GCN) 4346 GPUs; supported are the third-generation Fiji (fiji) and the 4347 fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906). 4348 4349 [24]C family 4350 4351 * New attributes: 4352 + The access function and type attribute has been added to 4353 describe how a function accesses objects passed to it by 4354 pointer or reference, and to associate such arguments with 4355 integer arguments denoting the objects' sizes. The attribute 4356 is used to enable the detection of invalid accesses by 4357 user-defined functions, such as those diagnosed by 4358 -Wstringop-overflow. 4359 + The symver attribute can be used to bind symbols to specific 4360 version nodes on ELF platforms. This is preferred to using 4361 inline assembly with GNU as symver directive because the 4362 latter is not compatible with link-time optimizations. 4363 * New warnings: 4364 + [25]-Wstring-compare, enabled by -Wextra, warns about equality 4365 and inequality expressions between zero and the result of a 4366 call to either strcmp and strncmp that evaluate to a constant 4367 as a result of the length of one argument being greater than 4368 the size of the array pointed to by the other. 4369 + [26]-Wzero-length-bounds, enabled by -Warray-bounds, warns 4370 about accesses to elements of zero-length arrays that might 4371 overlap other members of the same object. 4372 * Enhancements to existing warnings: 4373 + [27]-Warray-bounds detects more out-of-bounds accesses to 4374 member arrays as well as accesses to elements of zero-length 4375 arrays. 4376 + [28]-Wformat-overflow makes full use of string length 4377 information computed by the strlen optimization pass. 4378 + [29]-Wrestrict detects overlapping accesses to dynamically 4379 allocated objects. 4380 + [30]-Wreturn-local-addr diagnoses more instances of return 4381 statements returning addresses of automatic variables. 4382 + [31]-Wstringop-overflow detects more out-of-bounds stores to 4383 member arrays including zero-length arrays, dynamically 4384 allocated objects and variable length arrays, as well as more 4385 instances of reads of unterminated character arrays by string 4386 built-in functions. The warning also detects out-of-bounds 4387 accesses by calls to user-defined functions declared with the 4388 new attribute access. 4389 + [32]-Warith-conversion re-enables warnings from -Wconversion, 4390 -Wfloat-conversion, and -Wsign-conversion that are now off by 4391 default for an expression where the result of an arithmetic 4392 operation will not fit in the target type due to promotion, 4393 but the operands of the expression do fit in the target type. 4394 * Extended characters in identifiers may now be specified directly in 4395 the input encoding (UTF-8, by default), in addition to the UCN 4396 syntax (\uNNNN or \UNNNNNNNN) that is already supported: 4397 4398 static const int p = 3; 4399 int get_nave_pi() { 4400 return p; 4401 } 4402 4403 [33]C 4404 4405 * Several new features from the upcoming C2X revision of the ISO C 4406 standard are supported with -std=c2x and -std=gnu2x. Some of these 4407 features are also supported as extensions when compiling for older 4408 language versions. In addition to the features listed, some 4409 features previously supported as extensions and now added to the C 4410 standard are enabled by default in C2X mode and not diagnosed with 4411 -std=c2x -Wpedantic. 4412 + The [[]] attribute syntax is supported, as in C++. Existing 4413 attributes can be used with this syntax in forms such as 4414 [[gnu::const]]. The standard attributes [[deprecated]], 4415 [[fallthrough]] and [[maybe_unused]] are supported. 4416 + UTF-8 character constants using the u8'' syntax are supported. 4417 + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and 4418 LDBL_NORM_MAX. 4419 + When decimal floating-point arithmetic is supported, <float.h> 4420 defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and 4421 DEC128_TRUE_MIN, in addition to the macros that were 4422 previously only defined if __STDC_WANT_DEC_FP__ was defined 4423 before including <float.h>. 4424 + In C2X mode, empty parentheses in a function definition give 4425 that function a type with a prototype for subsequent calls; 4426 other old-style function definitions are diagnosed by default 4427 in C2X mode. 4428 + The strftime format checking supports the %OB and %Ob formats. 4429 + In C2X mode, -fno-fp-int-builtin-inexact is enabled by 4430 default. 4431 * GCC now defaults to -fno-common. As a result, global variable 4432 accesses are more efficient on various targets. In C, global 4433 variables with multiple tentative definitions now result in linker 4434 errors. With -fcommon such definitions are silently merged during 4435 linking. 4436 4437 [34]C++ 4438 4439 * Several C++20 features have been implemented: 4440 + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, 4441 P0848R3, P1616R1, P1452R2 4442 + P1668R1, Permitting Unevaluated inline-assembly in constexpr 4443 Functions 4444 + P1161R3, Deprecate a[b,c] 4445 + P0848R3, Conditionally Trivial Special Member Functions 4446 + P1091R3, Extending structured bindings 4447 + P1143R2, Adding the constinit keyword 4448 + P1152R4, Deprecating volatile 4449 + P0388R4, Permit conversions to arrays of unknown bound 4450 + P0784R7, constexpr new 4451 + P1301R4, [[nodiscard("with reason")]] 4452 + P1814R0, class template argument deduction for alias templates 4453 + P1816R0, class template argument deduction for aggregates 4454 + P0960R3, Parenthesized initialization of aggregates 4455 + P1331R2, Allow trivial default initialization in constexpr 4456 contexts 4457 + P1327R1, Allowing dynamic_cast and polymorphic typeid in 4458 constexpr contexts 4459 + P0912R5, Coroutines (requires -fcoroutines) 4460 * Several C++ Defect Reports have been resolved, e.g.: 4461 + DR 1560, lvalue-to-rvalue conversion in ?: 4462 + DR 1813, __is_standard_layout for a class with repeated bases 4463 + DR 2094, volatile scalars are trivially copyable, 4464 + DR 2096, constraints on literal unions 4465 + DR 2413, typename in conversion-function-ids 4466 + DR 2352, Similar types and reference binding 4467 + DR 1601, Promotion of enumeration with fixed underlying type 4468 + DR 330, Qualification conversions and pointers to arrays of 4469 pointers 4470 + DR 1307, Overload resolution based on size of array 4471 initializer-list 4472 + DR 1710, Missing template keyword in class-or-decltype 4473 * New warnings: 4474 + [35]-Wmismatched-tags, disabled by default, warns about 4475 declarations of structs, classes, and class templates and 4476 their specializations with a class-key that does not match 4477 either the definition or the first declaration if no 4478 definition is provided. The option is provided to ease 4479 portability to Windows-based compilers. 4480 + [36]-Wredundant-tags, disabled by default, warns about 4481 redundant class-key and enum-key in contexts where the key can 4482 be eliminated without causing an syntactic ambiguity. 4483 * G++ can now detect modifying constant objects in constexpr 4484 evaluation (which is undefined behavior). 4485 * G++ no longer emits bogus -Wsign-conversion warnings with explicit 4486 casts. 4487 * Narrowing is now detected in more contexts (e.g., case values). 4488 * Memory consumption of the compiler has been reduced in constexpr 4489 evaluation. 4490 * The noexcept-specifier is now properly treated as a complete-class 4491 context as per [class.mem]. 4492 * The attribute deprecated can now be used on namespaces too. 4493 * The ABI of passing and returning certain C++ classes by value 4494 changed on several targets in GCC 10, including [37]AArch64, 4495 [38]ARM, [39]PowerPC ELFv2, [40]S/390 and [41]Itanium. These 4496 changes affect classes with a zero-sized subobject (an empty base 4497 class, or data member with the [[no_unique_address]] attribute) 4498 where all other non-static data members have the same type (this is 4499 called a "homogeneous aggregate" in some ABI specifications, or if 4500 there is only one such member, a "single element"). In -std=c++17 4501 and -std=c++20 modes, classes with an empty base class were not 4502 considered to have a single element or to be a homogeneous 4503 aggregate, and so could be passed differently (in the wrong 4504 registers or at the wrong stack address). This could make code 4505 compiled with -std=c++17 and -std=c++14 ABI incompatible. This has 4506 been corrected and the empty bases are ignored in those ABI 4507 decisions, so functions compiled with -std=c++14 and -std=c++17 are 4508 now ABI compatible again. Example: struct empty {}; struct S : 4509 empty { float f; }; void f(S);. Similarly, in classes containing 4510 non-static data members with empty class types using the C++20 4511 [[no_unique_address]] attribute, those members weren't ignored in 4512 the ABI argument passing decisions as they should be. Both of these 4513 ABI changes are now diagnosed with -Wpsabi. 4514 4515 [42]Runtime Library (libstdc++) 4516 4517 * Improved experimental C++2a support, including: 4518 + Library concepts in <concepts> and <iterator>. 4519 + Constrained algorithms in <ranges>, <algorithm>, and <memory> 4520 (thanks to Patrick Palka). 4521 + New algorithms shift_left and shift_right (thanks to Patrick 4522 Palka). 4523 + std::span (thanks to JeanHeyd Meneide). 4524 + Three-way comparisons in <compare> and throughout the library. 4525 + Constexpr support in <algorithm> and elsewhere (thanks to 4526 Edward Smith-Rowland). 4527 + <stop_token> and std::jthread (thanks to Thomas Rodgers). 4528 + std::atomic_ref and std::atomic<floating point>. 4529 + Integer comparison functions (cmp_equal, cmp_less etc.). 4530 + std::ssize, std::to_array. 4531 + std::construct_at, std::destroy, constexpr std::allocator. 4532 + Mathematical constants in <numbers>. 4533 * Support for RDSEED in std::random_device. 4534 * Reduced header dependencies, leading to faster compilation for some 4535 code. 4536 * The std::iterator base class of std::istreambuf_iterator was 4537 changed in C++98 mode to be consistent with C++11 and later 4538 standards. This is expected to have no noticeable effect except in 4539 the unlikely case of a class which has potentially overlapping 4540 subobjects of type std::istreambuf_iterator<C> and another iterator 4541 type with a std::iterator<input_iterator_tag, C, ...> base class. 4542 The layout of such a type might change when compiled as C++98. 4543 [43]Bug 92285 has more details and concrete examples. 4544 4545 [44]D 4546 4547 * Support for static foreach has been implemented. 4548 * Aliases can now be created directly from any __traits that return 4549 symbols or tuples. Previously, an AliasSeq was necessary in order 4550 to alias their return. 4551 * It is now possible to detect the language ABI specified for a 4552 struct, class, or interface using __traits(getLinkage, ...). 4553 * Support for core.math.toPrec intrinsics has been added. These 4554 intrinsics guarantee the rounding to specific floating-point 4555 precisions at specified points in the code. 4556 * Support for pragma(inline) has been implemented. Previously the 4557 pragma was recognized, but had no effect on the compilation. 4558 * Optional parentheses in asm operands are deprecated and will be 4559 removed in a future release. 4560 * All content imported files are now included in the make dependency 4561 list when compiling with -M. 4562 * Compiler recognized attributes provided by the gcc.attribute module 4563 will now take effect when applied to function prototypes as well as 4564 when applied to full function declarations. 4565 * Added a --enable-libphobos-checking configure option to control 4566 whether run-time checks are compiled into the D runtime library. 4567 * Added a --with-libphobos-druntime-only configure option to indicate 4568 whether to build only the core D runtime library, or both the core 4569 and standard libraries into libphobos. 4570 4571 [45]Fortran 4572 4573 * use_device_addr of version 5.0 of the [46]OpenMP specification is 4574 now supported. Note that otherwise OpenMP 4.5 is partially 4575 supported in the Fortran compiler; the largest missing item is 4576 structure element mapping. 4577 * The default buffer size for I/O using unformatted files has been 4578 increased to 1048576. The buffer size for can now be set at runtime 4579 via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and 4580 GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted 4581 files, respectively. 4582 * Mismatches between actual and dummy argument lists in a single file 4583 are now rejected with an error. Use the new option 4584 -fallow-argument-mismatch to turn these errors into warnings; this 4585 option is implied with -std=legacy. -Wargument-mismatch has been 4586 removed. 4587 * The handling of a BOZ literal constant has been reworked to provide 4588 better conformance to the Fortran 2008 and 2018 standards. In these 4589 Fortran standards, a BOZ literal constant is a typeless and 4590 kindless entity. As a part of the rework, documented and 4591 undocumented extensions to the Fortran standard now emit errors 4592 during compilation. Some of these extensions are permitted with the 4593 -fallow-invalid-boz option, which degrades the error to a warning 4594 and the code is compiled as with older gfortran. 4595 * At any optimization level except-Os, gfortran now uses inline 4596 packing for arguments instead of calling a library routine. If the 4597 source contains a large number of arguments that need to be 4598 repacked, code size or time for compilation can become excessive. 4599 If that is the case, -fno-inline-arg-packing can be used to disable 4600 inline argument packing. 4601 * Legacy extensions: 4602 + For formatted input/output, if the explicit widths after the 4603 data-edit descriptors I, F and G have been omitted, default 4604 widths are used. 4605 + A blank format item at the end of a format specification, i.e. 4606 nothing following the final comma, is allowed. Use the option 4607 -fdec-blank-format-item; this option is implied with -fdec. 4608 + The existing support for AUTOMATIC and STATIC attributes has 4609 been extended to allow variables with the AUTOMATIC attribute 4610 to be used in EQUIVALENCE statements. Use -fdec-static; this 4611 option is implied by -fdec. 4612 + Allow character literals in assignments and DATA statements 4613 for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. 4614 Use the option -fdec-char-conversions; this option is implied 4615 with -fdec. 4616 + DEC comparisons, i.e. allow Hollerith constants to be used in 4617 comparisons with INTEGER, REAL, COMPLEX and CHARACTER 4618 expressions. Use the option -fdec. 4619 * Character type names in errors and warnings now include len in 4620 addition to kind; * is used for assumed length. The kind is omitted 4621 if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4). 4622 * CO_BROADCAST now supports derived type variables including objects 4623 with allocatable components. In this case, the optional arguments 4624 STAT= and ERRMSG= are currently ignored. 4625 * The handling of module and submodule names has been reworked to 4626 allow the full 63-character length mandated by the standard. 4627 Previously symbol names were truncated if the combined length of 4628 module, submodule, and function name exceeded 126 characters. This 4629 change therefore breaks the ABI, but only for cases where this 126 4630 character limit was exceeded. 4631 4632 [47]Go 4633 4634 * GCC 10 provides a complete implementation of the Go 1.14.6 user 4635 packages. 4636 4637 [48]libgccjit 4638 4639 * The libgccjit API gained four new entry points: 4640 + [49]gcc_jit_version_major, [50]gcc_jit_version_minor, and 4641 [51]gcc_jit_version_patchlevel for programmatically checking 4642 the libgccjit version from client code, and 4643 + [52]gcc_jit_context_new_bitfield 4644 4645 [53]New Targets and Target Specific Improvements 4646 4647 [54]AArch64 & arm 4648 4649 * The AArch64 and arm ports now support condition flag output 4650 constraints in inline assembly, as indicated by the 4651 __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for 4652 A32 and T32 targets. Please refer to the documentation for more 4653 details. 4654 4655 [55]AArch64 4656 4657 * There have been several improvements related to the Scalable Vector 4658 Extension (SVE): 4659 + The SVE ACLE types and intrinsics are now supported. They can 4660 be accessed using the header file arm_sve.h. 4661 + It is now possible to create fixed-length SVE types using the 4662 arm_sve_vector_bits attribute. For example: 4663 #if __ARM_FEATURE_SVE_BITS==512 4664 typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512))); 4665 typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512))); 4666 #endif 4667 + -mlow-precision-div, -mlow-precision-sqrt and 4668 -mlow-precision-recip-sqrt now work for SVE. 4669 + -msve-vector-bits=128 now generates vector-length-specific 4670 code for little-endian targets. It continues to generate 4671 vector-length-agnostic code for big-endian targets, just as 4672 previous releases did for all targets. 4673 + The vectorizer is now able to use extending loads and 4674 truncating stores, including gather loads and scatter stores. 4675 + The vectorizer now compares the cost of vectorizing with SVE 4676 and vectorizing with Advanced SIMD and tries to pick the best 4677 one. Previously it would always use SVE if possible. 4678 + If a vector loop uses Advanced SIMD rather than SVE, the 4679 vectorizer now considers using SVE to vectorize the left-over 4680 elements (the "scalar tail" or "epilog"). 4681 + Besides these specific points, there have been many general 4682 improvements to the way that the vectorizer uses SVE. 4683 * The -mbranch-protection=pac-ret option now accepts the optional 4684 argument +b-key extension to perform return address signing with 4685 the B-key instead of the A-key. 4686 * The option -moutline-atomics has been added to aid deployment of 4687 the Large System Extensions (LSE) on GNU/Linux systems built with a 4688 baseline architecture targeting Armv8-A. When the option is 4689 specified code is emitted to detect the presence of LSE 4690 instructions at runtime and use them for standard atomic 4691 operations. For more information please refer to the documentation. 4692 * The Transactional Memory Extension is now supported through ACLE 4693 intrinsics. It can be enabled through the +tme option extension 4694 (for example, -march=armv8.5-a+tme). 4695 * A number of features from Armv8.5-A are now supported through ACLE 4696 intrinsics. These include: 4697 + The random number instructions that can be enabled through the 4698 (already present in GCC 9.1) +rng option extension. 4699 + Floating-point intrinsics to round to integer instructions 4700 from Armv8.5-A when targeting -march=armv8.5-a or later. 4701 + Memory Tagging Extension intrinsics enabled through the 4702 +memtag option extension. 4703 * Similarly, the following Armv8.6-A features are now supported 4704 through ACLE intrinsics: 4705 + The bfloat16 extension. This extension is enabled 4706 automatically when Armv8.6-A is selected (such as by 4707 -march=armv8.6-a). It can also be enabled for Armv8.2-A and 4708 later using the +bf16 option extension. 4709 + The Matrix Multiply extension. This extension is split into 4710 three parts, one for each supported data type: 4711 o Support for 8-bit integer matrix multiply instructions. 4712 This extension is enabled automatically when Armv8.6-A is 4713 selected. It can also be enabled for Armv8.2-A and later 4714 using the +i8mm option extension. 4715 o Support for 32-bit floating-point matrix multiply 4716 instructions. This extension can be enabled using the 4717 +f32mm option extension, which also has the effect of 4718 enabling SVE. 4719 o Support for 64-bit floating-point matrix multiply 4720 instructions. This extension can be enabled using the 4721 +f64mm option extension, which likewise has the effect of 4722 enabling SVE. 4723 * SVE2 is now supported through ACLE intrinsics and (to a limited 4724 extent) through autovectorization. It can be enabled through the 4725 +sve2 option extension (for example, -march=armv8.5-a+sve2). 4726 Additional extensions can be enabled through +sve2-sm4, +sve2-aes, 4727 +sve2-sha3 and +sve2-bitperm. 4728 * Support has been added for the following processors (GCC 4729 identifiers in parentheses): 4730 + Arm Cortex-A77 (cortex-a77). 4731 + Arm Cortex-A76AE (cortex-a76ae). 4732 + Arm Cortex-A65 (cortex-a65). 4733 + Arm Cortex-A65AE (cortex-a65ae). 4734 + Arm Cortex-A34 (cortex-a34). 4735 + Marvell ThunderX3 (thunderx3t110). 4736 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4737 options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as 4738 arguments to the equivalent target attributes and pragmas. 4739 4740 [56]arm 4741 4742 * Support for the FDPIC ABI has been added. It uses 64-bit function 4743 descriptors to represent pointers to functions, and enables code 4744 sharing on MMU-less systems. The corresponding target triple is 4745 arm-uclinuxfdpiceabi, and the C library is uclibc-ng. 4746 * Support has been added for the Arm EABI on NetBSD through the 4747 arm*-*-netbsdelf-*eabi* triplet. 4748 * The handling of 64-bit integer operations has been significantly 4749 reworked and improved leading to improved performance and reduced 4750 stack usage when using 64-bit integral data types. The option 4751 -mneon-for-64bits is now deprecated and will be removed in a future 4752 release. 4753 * Support has been added for the following processors (GCC 4754 identifiers in parentheses): 4755 + Arm Cortex-A77 (cortex-a77). 4756 + Arm Cortex-A76AE (cortex-a76ae). 4757 + Arm Cortex-M35P (cortex-m35p). 4758 + Arm Cortex-M55 (cortex-m55). 4759 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4760 options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p. 4761 * Support has been extended for the ACLE [57]data-processing 4762 intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit 4763 multiplication and other related intrinsics aimed at DSP algorithm 4764 optimization. 4765 * Support for -mpure-code in Thumb-1 (v6m) has been added: this 4766 M-profile feature is no longer restricted to targets with MOVT. For 4767 example, -mcpu=cortex-m0 now supports this option. 4768 * Support for the [58]Armv8.1-M Mainline Architecture has been added. 4769 + Armv8.1-M Mainline can be enabled by using the 4770 -march=armv8.1-m.main command-line option. 4771 * Support for the [59]MVE beta ACLE intrinsics has been added. These 4772 intrinsics can be enabled by including the arm_mve.h header file 4773 and passing the +mve or +mve.fp option extensions (for example: 4774 -march=armv8.1-m.main+mve). 4775 * Support for the Custom Datapath Extension beta ACLE [60]intrinsics 4776 has been added. 4777 * Support for Armv8.1-M Mainline Security Extensions architecture has 4778 been added. The -mcmse option, when used in combination with an 4779 Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main 4780 -mcmse), now leads to the generation of improved code sequences 4781 when changing security states. 4782 4783 [61]AMD Radeon (GCN) 4784 4785 * Code generation and in particular vectorization support have been 4786 much improved. 4787 4788 [62]ARC 4789 4790 * The interrupt service routine functions save all used registers, 4791 including extension registers and auxiliary registers used by Zero 4792 Overhead Loops. 4793 * Improve code size by using multiple short instructions instead of a 4794 single long mov or ior instruction when its long immediate constant 4795 is known. 4796 * Fix usage of the accumulator register for ARC600. 4797 * Fix issues with uncached attribute. 4798 * Remove -mq-class option. 4799 * Improve 64-bit integer addition and subtraction operations. 4800 4801 [63]AVR 4802 4803 * Support for the XMEGA-like devices 4804 4805 ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, 4806 ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, 4807 ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, 4808 ATmega4808, ATmega4809 4809 has been added. 4810 * A new command-line option -nodevicespecs has been added. It allows 4811 to provide a custom device-specs file by means of 4812 4813 avr-gcc -nodevicespecs -specs=my-spec-file <options> 4814 and without the need to provide options -B and -mmcu=. See [64]AVR 4815 command-line options for details. This feature is also available in 4816 GCC 9.3+ and GCC 8.4+. 4817 * New command-line options -mdouble=[32,64] and -mlong-double=[32,64] 4818 have been added. They allow to choose the size (in bits) of the 4819 double and long double types, respectively. Whether or not the 4820 mentioned layouts are available, whether the options act as a 4821 multilib option, and the default for either option are controlled 4822 by the new [65]AVR configure options --with-double= and 4823 --with-long-double=. 4824 * A new configure option --with-libf7= has been added. It controls to 4825 which level avr-libgcc provides 64-bit floating point support by 4826 means of [66]Libf7. 4827 * A new configure option --with-double-comparison= has been added. 4828 It's unlikely you need to set this option by hand. 4829 4830 [67]IA-32/x86-64 4831 4832 * Support to expand __builtin_roundeven into the appropriate SSE 4.1 4833 instruction has been added. 4834 * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD 4835 intrinsics are available via the -menqcmd compiler switch. 4836 * GCC now supports the Intel CPU named Cooperlake through 4837 -march=cooperlake. The switch enables the AVX512BF16 ISA 4838 extensions. 4839 * GCC now supports the Intel CPU named Tigerlake through 4840 -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B 4841 AVX512VP2INTERSECT ISA extensions. 4842 4843 [68]MIPS 4844 4845 * The mips*-*-linux* targets now mark object files with appropriate 4846 GNU-stack note, facilitating use of non-executable stack hardening 4847 on GNU/Linux. The soft-float targets have this feature enabled by 4848 default, while for hard-float targets it is required for GCC to be 4849 configured with --with-glibc-version=2.31 against glibc 2.31 or 4850 later. 4851 4852 [69]PowerPC / PowerPC64 / RS6000 4853 4854 * Many vector builtins have been listed as deprecated in the 4855 [70]64-Bit ELF V2 ABI Specification for quite a number of years. 4856 The vector builtins listed in Tables A.8 through A.10 are now 4857 deprecated for GCC 10, and will likely be removed from support in 4858 GCC 11. Note that this does not result in any loss of function. 4859 These deprecated builtins generally provide somewhat nonsensical 4860 argument lists (for example, mixing signed, unsigned, and bool 4861 vector arguments arbitrarily), or are duplicate builtins that are 4862 inconsistent with the expected naming scheme. We expect that this 4863 will be unlikely to affect much if any code, and any required code 4864 changes will be trivial. 4865 4866 [71]PRU 4867 4868 * A new back end targeting TI PRU I/O processors has been contributed 4869 to GCC. 4870 4871 [72]RISC-V 4872 4873 * The riscv*-*-* targets now require GNU binutils version 2.30 or 4874 later, to support new assembly instructions produced by GCC. 4875 4876 [73]V850 4877 4878 * The ABI for V850 nested functions has been changed. Previously the 4879 V850 port used %r20 for the static chain pointer, now the port uses 4880 %r19. This corrects a long standing latent bug in the v850 port 4881 where a call to a nested function would unexpectedly change the 4882 value in %r20. 4883 4884 [74]Operating Systems 4885 4886 [75]Improvements for plugin authors 4887 4888 * GCC diagnostics can now have a chain of events associated with 4889 them, describing a path through the code that triggers the problem. 4890 These can be printed by the diagnostics subsystem in various ways, 4891 controlled by the [76]-fdiagnostics-path-format option, or captured 4892 in JSON form via [77]-fdiagnostics-format=json. 4893 * GCC diagnostics can now be associated with [78]CWE weakness 4894 identifiers, which will appear on the standard error stream, and in 4895 the JSON output from [79]-fdiagnostics-format=json. 4896 4897 Other significant improvements 4898 4899 * To allow inline expansion of both memcpy and memmove, the existing 4900 movmem instruction patterns used for non-overlapping memory copies 4901 have been renamed to cpymem. The movmem name is now used for 4902 overlapping memory moves, consistent with the library functions 4903 memcpy and memmove. 4904 * For many releases, when GCC emits a warning it prints the option 4905 controlling that warning. As of GCC 10, that option text is now a 4906 clickable hyperlink for the documentation of that option (assuming 4907 a [80]sufficiently capable terminal). This behavior can be 4908 controlled via a new [81]-fdiagnostics-urls option (along with 4909 various environment variables and heuristics documented with that 4910 option). 4911 4912 GCC 10.1 4913 4914 This is the [82]list of problem reports (PRs) from GCC's bug tracking 4915 system that are known to be fixed in the 10.1 release. This list might 4916 not be complete (that is, it is possible that some PRs that have been 4917 fixed are not listed here). 4918 4919 [83]GCC 10.2 4920 4921 This is the [84]list of problem reports (PRs) from GCC's bug tracking 4922 system that are known to be fixed in the 10.2 release. This list might 4923 not be complete (that is, it is possible that some PRs that have been 4924 fixed are not listed here). 4925 4926 [85]GCC 10.3 4927 4928 This is the [86]list of problem reports (PRs) from GCC's bug tracking 4929 system that are known to be fixed in the 10.3 release. This list might 4930 not be complete (that is, it is possible that some PRs that have been 4931 fixed are not listed here). 4932 4933 Target Specific Changes 4934 4935 AArch64 4936 4937 * A bug with the Random Number intrinsics in the arm_acle.h header 4938 that resulted in an incorrect status result being returned has been 4939 fixed. 4940 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 4941 options are -mcpu=a64fx and -mtune=a64fx respectively. In 4942 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 4943 tunes the code for the A64FX. This includes tuning the SVE code, 4944 although by default the code is still length-agnostic and so works 4945 for all SVE implementations. Adding -msve-vector-bits=512 makes the 4946 code specific to 512-bit SVE. 4947 4948 x86-64 4949 4950 * GCC 10.3 supports AMD CPUs based on the znver3 core via 4951 -march=znver3. 4952 4953 [87]GCC 10.4 4954 4955 This is the [88]list of problem reports (PRs) from GCC's bug tracking 4956 system that are known to be fixed in the 10.4 release. This list might 4957 not be complete (that is, it is possible that some PRs that have been 4958 fixed are not listed here). 4959 4960 Target Specific Changes 4961 4962 x86-64 4963 4964 * The x86-64 ABI of passing and returning structures with a 64-bit 4965 integer vector changed in GCC 10.1 when MMX is disabled. Disabling 4966 MMX no longer changes how they are passed nor returned. This ABI 4967 change is now diagnosed with -Wpsabi. 4968 4969 [89]GCC 10.5 4970 4971 This is the [90]list of problem reports (PRs) from GCC's bug tracking 4972 system that are known to be fixed in the 10.5 release. This list might 4973 not be complete (that is, it is possible that some PRs that have been 4974 fixed are not listed here). 4975 4976 4977 For questions related to the use of GCC, please consult these web 4978 pages and the [91]GCC manuals. If that fails, the 4979 [92]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4980 web pages and the development of GCC are welcome on our developer 4981 list at [93]gcc (a] gcc.gnu.org. All of [94]our lists have public 4982 archives. 4983 4984 Copyright (C) [95]Free Software Foundation, Inc. Verbatim copying and 4985 distribution of this entire article is permitted in any medium, 4986 provided this notice is preserved. 4987 4988 These pages are [96]maintained by the GCC team. Last modified 4989 2025-03-21. 4990 4991 References 4992 4993 1. https://gcc.gnu.org/gcc-10/porting_to.html 4994 2. https://gcc.gnu.org/onlinedocs/index.html#current 4995 3. https://gcc.gnu.org/gcc-10/changes.html#empty_base 4996 4. https://www.mpfr.org/ 4997 5. https://gcc.gnu.org/gcc-10/changes.html#iterator_base 4998 6. https://gcc.gnu.org/gcc-10/changes.html#general 4999 7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin 5000 8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce 5001 9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training 5002 10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible 5003 11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate 5004 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path 5005 13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html 5006 14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions 5007 15. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html 5008 16. https://facebook.github.io/zstd/ 5009 17. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values 5010 18. https://gcc.gnu.org/gcc-10/changes.html#languages 5011 19. https://www.openacc.org/ 5012 20. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-10 5013 21. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1 5014 22. https://www.openmp.org/specifications/ 5015 23. https://gcc.gnu.org/wiki/Offloading 5016 24. https://gcc.gnu.org/gcc-10/changes.html#c-family 5017 25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare 5018 26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds 5019 27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds 5020 28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow 5021 29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict 5022 30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr 5023 31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow 5024 32. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion 5025 33. https://gcc.gnu.org/gcc-10/changes.html#c 5026 34. https://gcc.gnu.org/gcc-10/changes.html#cxx 5027 35. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags 5028 36. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags 5029 37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383 5030 38. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711 5031 39. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707 5032 40. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704 5033 41. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706 5034 42. https://gcc.gnu.org/gcc-10/changes.html#libstdcxx 5035 43. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285 5036 44. https://gcc.gnu.org/gcc-10/changes.html#d 5037 45. https://gcc.gnu.org/gcc-10/changes.html#fortran 5038 46. https://www.openmp.org/specifications/ 5039 47. https://gcc.gnu.org/gcc-10/changes.html#go 5040 48. https://gcc.gnu.org/gcc-10/changes.html#jit 5041 49. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major 5042 50. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor 5043 51. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel 5044 52. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield 5045 53. https://gcc.gnu.org/gcc-10/changes.html#targets 5046 54. https://gcc.gnu.org/gcc-10/changes.html#arm-targets 5047 55. https://gcc.gnu.org/gcc-10/changes.html#aarch64 5048 56. https://gcc.gnu.org/gcc-10/changes.html#arm 5049 57. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics 5050 58. https://developer.arm.com/Architectures/M-Profile%20Architecture 5051 59. https://developer.arm.com/architectures/instruction-sets/intrinsics/ 5052 60. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension 5053 61. https://gcc.gnu.org/gcc-10/changes.html#amdgcn 5054 62. https://gcc.gnu.org/gcc-10/changes.html#arc 5055 63. https://gcc.gnu.org/gcc-10/changes.html#avr 5056 64. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs 5057 65. https://gcc.gnu.org/install/configure.html#avr 5058 66. https://gcc.gnu.org/wiki/avr-gcc#Libf7 5059 67. https://gcc.gnu.org/gcc-10/changes.html#x86 5060 68. https://gcc.gnu.org/gcc-10/changes.html#mips 5061 69. https://gcc.gnu.org/gcc-10/changes.html#powerpc 5062 70. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture 5063 71. https://gcc.gnu.org/gcc-10/changes.html#pru 5064 72. https://gcc.gnu.org/gcc-10/changes.html#riscv 5065 73. https://gcc.gnu.org/gcc-10/changes.html#v850 5066 74. https://gcc.gnu.org/gcc-10/changes.html#os 5067 75. https://gcc.gnu.org/gcc-10/changes.html#plugins 5068 76. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format 5069 77. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 5070 78. https://cwe.mitre.org/ 5071 79. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 5072 80. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda 5073 81. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls 5074 82. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0 5075 83. https://gcc.gnu.org/gcc-10/changes.html#GCC10.2 5076 84. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2 5077 85. https://gcc.gnu.org/gcc-10/changes.html#GCC10.3 5078 86. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3 5079 87. https://gcc.gnu.org/gcc-10/changes.html#GCC10.4 5080 88. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4 5081 89. https://gcc.gnu.org/gcc-10/changes.html#GCC10.5 5082 90. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.5 5083 91. https://gcc.gnu.org/onlinedocs/ 5084 92. mailto:gcc-help (a] gcc.gnu.org 5085 93. mailto:gcc (a] gcc.gnu.org 5086 94. https://gcc.gnu.org/lists.html 5087 95. https://www.fsf.org/ 5088 96. https://gcc.gnu.org/about.html 5089 ====================================================================== 5090 http://gcc.gnu.org/gcc-9/index.html 5091 5092 GCC 9 Release Series 5093 5094 (This release series is no longer supported.) 5095 5096 May 27, 2022 5097 5098 The [1]GNU project and the GCC developers are pleased to announce the 5099 release of GCC 9.5. 5100 5101 This release is a bug-fix release, containing fixes for regressions in 5102 GCC 9.4 relative to previous releases of GCC. 5103 5104 Release History 5105 5106 GCC 9.5 5107 May 27, 2022 ([2]changes, [3]documentation) 5108 5109 GCC 9.4 5110 June 1, 2021 ([4]changes, [5]documentation) 5111 5112 GCC 9.3 5113 Mar 12, 2020 ([6]changes, [7]documentation) 5114 5115 GCC 9.2 5116 Aug 12, 2019 ([8]changes, [9]documentation) 5117 5118 GCC 9.1 5119 May 3, 2019 ([10]changes, [11]documentation) 5120 5121 References and Acknowledgements 5122 5123 GCC used to stand for the GNU C Compiler, but since the compiler 5124 supports several other languages aside from C, it now stands for the 5125 GNU Compiler Collection. 5126 5127 The GCC developers would like to thank the numerous people that have 5128 contributed new features, improvements, bug fixes, and other changes as 5129 well as test results to GCC. This [12]amazing group of volunteers is 5130 what makes GCC successful. 5131 5132 For additional information about GCC please refer to the [13]GCC 5133 project web site or contact the [14]GCC development mailing list. 5134 5135 To obtain GCC please use [15]our mirror sites or [16]our version 5136 control system. 5137 5138 5139 For questions related to the use of GCC, please consult these web 5140 pages and the [17]GCC manuals. If that fails, the 5141 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5142 web pages and the development of GCC are welcome on our developer 5143 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 5144 archives. 5145 5146 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 5147 distribution of this entire article is permitted in any medium, 5148 provided this notice is preserved. 5149 5150 These pages are [22]maintained by the GCC team. Last modified 5151 2024-05-30. 5152 5153 References 5154 5155 1. http://www.gnu.org/ 5156 2. https://gcc.gnu.org/gcc-9/changes.html 5157 3. https://gcc.gnu.org/onlinedocs/9.5.0/ 5158 4. https://gcc.gnu.org/gcc-9/changes.html 5159 5. https://gcc.gnu.org/onlinedocs/9.4.0/ 5160 6. https://gcc.gnu.org/gcc-9/changes.html 5161 7. https://gcc.gnu.org/onlinedocs/9.3.0/ 5162 8. https://gcc.gnu.org/gcc-9/changes.html 5163 9. https://gcc.gnu.org/onlinedocs/9.2.0/ 5164 10. https://gcc.gnu.org/gcc-9/changes.html 5165 11. https://gcc.gnu.org/onlinedocs/9.1.0/ 5166 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html 5167 13. https://gcc.gnu.org/index.html 5168 14. mailto:gcc (a] gcc.gnu.org 5169 15. https://gcc.gnu.org/mirrors.html 5170 16. https://gcc.gnu.org/git.html 5171 17. https://gcc.gnu.org/onlinedocs/ 5172 18. mailto:gcc-help (a] gcc.gnu.org 5173 19. mailto:gcc (a] gcc.gnu.org 5174 20. https://gcc.gnu.org/lists.html 5175 21. https://www.fsf.org/ 5176 22. https://gcc.gnu.org/about.html 5177 ====================================================================== 5178 http://gcc.gnu.org/gcc-9/changes.html 5179 5180 GCC 9 Release Series 5181 Changes, New Features, and Fixes 5182 5183 This page is a "brief" summary of some of the huge number of 5184 improvements in GCC 9. You may also want to check out our [1]Porting to 5185 GCC 9 page and the [2]full GCC documentation. 5186 5187 Caveats 5188 5189 * On Arm targets (arm*-*-*), [3]a bug in the implementation of the 5190 procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has 5191 been fixed: a structure containing a bit-field based on a 64-bit 5192 integral type and where no other element in a structure required 5193 64-bit alignment could be passed incorrectly to functions. This is 5194 an ABI change. If the option -Wpsabi is enabled (on by default) the 5195 compiler will emit a diagnostic note for code that might be 5196 affected. 5197 * Support for a number of older systems and recently unmaintained or 5198 untested target ports of GCC has been declared obsolete in GCC 9. 5199 Unless there is activity to revive them, the next release of GCC 5200 will have their sources permanently removed. 5201 The following ports for individual systems on particular 5202 architectures have been obsoleted: 5203 + Solaris 10 (*-*-solaris2.10). Details can be found in the 5204 [4]announcement. 5205 + Cell Broadband Engine SPU (spu*-*-*). Details can be found in 5206 the [5]announcement. 5207 * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause 5208 ABI incompatibilities with object files compiled with other 5209 versions of GCC. If the std::rotate algorithm is called with an 5210 empty range then it might cause a divide-by-zero error (as a SIGFPE 5211 signal) and crash. The change has been reverted for GCC 9.2.0 and 5212 future releases. For more details see [6]Bug 90920. The problem can 5213 be avoided by recompiling any objects that might call std::rotate 5214 with an empty range, so that the GCC 9.1.0 definition of 5215 std::rotate is not used. 5216 * The automatic template instantiation at link time ([7]-frepo) has 5217 been deprecated and will be removed in a future release. 5218 * The --with-default-libstdcxx-abi=gcc4-compatible configure option 5219 is broken in the 9.1 and 9.2 releases, producing a shared library 5220 with missing symbols (see [8]Bug 90361). As a workaround, configure 5221 without that option and build GCC as normal, then edit the 5222 installed <bits/c++config.h> headers to define the 5223 _GLIBCXX_USE_CXX11_ABI macro to 0. 5224 5225 [9]General Improvements 5226 5227 The following GCC command line options have been introduced or 5228 improved. 5229 * All command line options that take a byte-size argument accept 5230 64-bit integers as well as standard SI and IEC suffixes such as kb 5231 and KiB, MB and MiB, or GB and GiB denoting the corresponding 5232 multiples of bytes. See [10]Invoking GCC for more. 5233 * A new option [11]-flive-patching=[inline-only-static|inline-clone] 5234 generates code suitable for live patching. At the same time it 5235 provides multiple-level control over IPA optimizations. See the 5236 user guide for more details. 5237 * A new option, --completion, has been added to provide more fine 5238 option completion in a shell. It is intended to be used by 5239 Bash-completion. 5240 * GCC's diagnostics now print source code with a left margin showing 5241 line numbers, configurable with 5242 [12]-fno-diagnostics-show-line-numbers. 5243 GCC's diagnostics can also now label regions of the source code to 5244 show pertinent information, such as the types within an expression. 5245 $ g++ t.cc 5246 t.cc: In function 'int test(const shape&, const shape&)': 5247 t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl 5248 e>' and 'boxed_value<double>') 5249 14 | return (width(s1) * height(s1) 5250 | ~~~~~~~~~~~~~~~~~~~~~~ 5251 | | 5252 | boxed_value<[...]> 5253 15 | + width(s2) * height(s2)); 5254 | ^ ~~~~~~~~~~~~~~~~~~~~~~ 5255 | | 5256 | boxed_value<[...]> 5257 5258 These labels can be disabled via [13]-fno-diagnostics-show-labels. 5259 * A new option [14]-fdiagnostics-format=json has been introduced for 5260 emitting diagnostics in a machine-readable format. 5261 * The alignment-related options [15]-falign-functions, 5262 [16]-falign-labels, [17]-falign-loops, and [18]-falign-jumps 5263 received support for a secondary alignment (e.g. 5264 -falign-loops=n:m:n2:m2). 5265 * New pair of profiling options ([19]-fprofile-filter-files and 5266 [20]-fprofile-exclude-files) has been added. The options help to 5267 filter which source files are instrumented. 5268 * AddressSanitizer generates more compact redzones for automatic 5269 variables. That helps to reduce memory footprint of a sanitized 5270 binary. 5271 * Numerous improvements have been made to the output of 5272 [21]-fopt-info. 5273 Messages are now prefixed with optimized, missed, or note, rather 5274 than the old behavior of all being prefixed with note. 5275 The output from -fopt-info can now contain information on inlining 5276 decisions: 5277 $ g++ -c inline.cc -O2 -fopt-info-inline-all 5278 inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)( 5279 E)) [with T = char**; E = char*]/2. 5280 inline.cc:24:11: optimized: Inlining void foreach(T, T, void (*)(E)) [with T = 5281 char**; E = char*]/2 into int main(int, char**)/1. 5282 inline.cc:19:12: missed: not inlinable: void inline_me(char*)/0 -> int std::pu 5283 ts(const char*)/3, function body not available 5284 inline.cc:13:8: optimized: Inlined void inline_me(char*)/4 into int main(int, c 5285 har**)/1 which now has time 127.363637 and size 11, net change of +0. 5286 Unit growth for small function inlining: 16->16 (0%) 5287 5288 Inlined 2 calls, eliminated 1 functions 5289 5290 5291 The output from the vectorizer has been rationalized so that failed 5292 attempts to vectorize a loop are displayed in the form 5293 [LOOP-LOCATION]: couldn't vectorize this loop 5294 [PROBLEM-LOCATION]: because of [REASON] 5295 5296 rather than an exhaustive log of all decisions made by the 5297 vectorizer. For example: 5298 $ gcc -c v.c -O3 -fopt-info-all-vec 5299 v.c:7:3: missed: couldn't vectorize loop 5300 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 5301 mory"); 5302 v.c:3:6: note: vectorized 0 loops in function. 5303 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 5304 mory"); 5305 5306 The old behavior can be obtained via a new -internals suboption of 5307 -fopt-info. 5308 * A new option, [22]-fsave-optimization-record has been added, which 5309 writes a SRCFILE.opt-record.json.gz file describing the 5310 optimization decisions made by GCC. This is similar to the output 5311 of -fopt-info, but with additional metadata such as the inlining 5312 chain, and profile information (if available). 5313 * Inter-procedural propagation of stack alignment can now be 5314 controlled by [23]-fipa-stack-alignment. 5315 * Propagation of addressability, readonly, and writeonly flags on 5316 static variables can now be controlled by 5317 [24]-fipa-reference-addressable. 5318 5319 The following built-in functions have been introduced. 5320 * [25]__builtin_expect_with_probability to provide branch prediction 5321 probability hints to the optimizer. 5322 * [26]__builtin_has_attribute determines whether a function, type, or 5323 variable has been declared with some attribute. 5324 * [27]__builtin_speculation_safe_value can be used to help mitigate 5325 against unsafe speculative execution. 5326 5327 The following attributes have been introduced. 5328 * The [28]copy function attribute has been added. The attribute can 5329 also be applied to type definitions and to variable declarations. 5330 5331 A large number of improvements to code generation have been made, 5332 including but not limited to the following. 5333 * Switch expansion has been improved by using a different strategy 5334 (jump table, bit test, decision tree) for a subset of switch cases. 5335 * A linear function expression defined as a switch statement can be 5336 transformed by [29]-ftree-switch-conversion. For example: 5337 5338 int 5339 foo (int how) 5340 { 5341 switch (how) { 5342 case 2: how = 205; break; 5343 case 3: how = 305; break; 5344 case 4: how = 405; break; 5345 case 5: how = 505; break; 5346 case 6: how = 605; break; 5347 } 5348 return how; 5349 } 5350 5351 can be transformed into 100 * how + 5 (for values defined in the 5352 switch statement). 5353 * Inter-procedural optimization improvements: 5354 + Inliner defaults were tuned to better suit modern C++ 5355 codebases, especially when built with link time-optimizations. 5356 New parameters max-inline-insns-small, max-inline-insns-size, 5357 uninlined-function-insns, uninlined-function-time, 5358 uninlined-thunk-insns, and uninlined-thunk-time were added. 5359 + Hot/cold partitioning is now more precise and aggressive. 5360 + Improved scalability for very large translation units 5361 (especially when link-time optimizing large programs). 5362 * Profile driven optimization improvements: 5363 + [30]-fprofile-use now enables [31]-fversion-loops-for-strides, 5364 [32]-floop-interchange, [33]-floop-unroll-and-jam, 5365 [34]-ftree-loop-distribution. 5366 + Streaming of counter histograms was removed, which reduces the 5367 size of profile files. Histograms are computed on the fly with 5368 link-time optimization. 5369 + The parameter hot-bb-count-ws-permille was reduced from 999 to 5370 990 to account for more precise histograms. 5371 * Link-time optimization improvements: 5372 + Types are now simplified prior to streaming resulting in 5373 significant reductions of LTO object file sizes and link-time 5374 memory use as well as improvements of link-time parallelism. 5375 + The default number of partitions (--param lto-partitions) was 5376 increased from 32 to 128 enabling effective use of CPUs with 5377 more than 32 hyperthreads. --param 5378 lto-max-streaming-parallelism can now be used to control the 5379 number of streaming processes. 5380 + Warnings on C++ One Decl Rule violations (-Wodr) are now more 5381 informative and produce fewer redundant results. 5382 Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an 5383 8-core machine was reduced by about 5% compared to GCC 8.3, and the 5384 size of LTO object files by 7%. LTO link time improves by 11% on an 5385 8-core machine and scales significantly better for more parallel 5386 build environments. The serial stage of the link-time optimization 5387 is 28% faster consuming 20% less memory. The parallel stage now 5388 scales to up to 128 partitions rather than 32 and reduces memory 5389 use for every worker by 30%. 5390 5391 The following improvements to the gcov command-line utility have been 5392 made. 5393 * The gcov tool received a new option [35]--use-hotness-colors (-q) 5394 that can provide perf-like coloring of hot functions. 5395 * The gcov tool has changed its intermediate format to a new JSON 5396 format. 5397 5398 [36]New Languages and Language specific improvements 5399 5400 [37]OpenACC support in C, C++, and Fortran continues to be maintained 5401 and improved. Most of the OpenACC 2.5 specification is implemented. See 5402 the [38]implementation status section on the OpenACC wiki page for 5403 further information. 5404 5405 [39]C family 5406 5407 * Version 5.0 of the [40]OpenMP specification is now partially 5408 supported in the C and C++ compilers. For details which features of 5409 OpenMP 5.0 are and which are not supported in the GCC 9 release see 5410 [41]this mail. 5411 * New extensions: 5412 + [42]__builtin_convertvector built-in for vector conversions 5413 has been added. 5414 * New warnings: 5415 + [43]-Waddress-of-packed-member, enabled by default, warns 5416 about an unaligned pointer value from the address of a packed 5417 member of a struct or union. 5418 * Enhancements to existing warnings: 5419 + [44]-Warray-bounds detects more instances of out-of-bounds 5420 indices. 5421 + [45]-Wattribute-alias also detects attribute mismatches 5422 between alias declarations and their targets, in addition to 5423 mismatches between their types. 5424 + [46]-Wformat-overflow and [47]-Wformat-truncation have been 5425 extended to all formatted input/output functions (where 5426 applicable) and enhanced to detect a subset of instances of 5427 reading past the end of unterminated constant character arrays 5428 in %s directives. 5429 + [48]-Wmissing-attributes detects instances of missing function 5430 attributes on declarations of aliases and weak references. 5431 + [49]-Wstringop-truncation also detects a subset of instances 5432 of reading past the end of unterminated constant character 5433 arrays, 5434 * If a macro is used with the wrong argument count, the C and C++ 5435 front ends now show the definition of that macro via a note. 5436 * The spelling corrector now considers transposed letters, and the 5437 threshold for similarity has been tightened, to avoid nonsensical 5438 suggestions. 5439 5440 [50]C 5441 5442 * There is now experimental support for -std=c2x, to select support 5443 for the upcoming C2X revision of the ISO C standard. This standard 5444 is in the early stages of development and the only feature 5445 supported in GCC 9 is _Static_assert with a single argument 5446 (support for _Static_assert with two arguments was added in C11 and 5447 GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU 5448 extensions, and -Wc11-c2x-compat, to warn for uses of features 5449 added in C2X (such warnings are also enabled by use of -Wpedantic 5450 if not using -std=c2x or -std=gnu2x). 5451 * New warnings: 5452 + [51]-Wabsolute-value warns for calls to standard functions 5453 that compute the absolute value of an argument when a more 5454 appropriate standard function is available. For example, 5455 calling abs(3.14) triggers the warning because the appropriate 5456 function to call to compute the absolute value of a double 5457 argument is fabs. The option also triggers warnings when the 5458 argument in a call to such a function has an unsigned type. 5459 This warning can be suppressed with an explicit type cast and 5460 it is also enabled by -Wextra. 5461 5462 [52]C++ 5463 5464 * New warnings: 5465 + [53]-Wdeprecated-copy, implied by -Wextra, warns about the 5466 C++11 deprecation of implicitly declared copy constructor and 5467 assignment operator if one of them is user-provided. 5468 -Wdeprecated-copy-dtor also warns if the destructor is 5469 user-provided, as specified in C++11. 5470 + [54]-Winit-list-lifetime, on by default, warns about uses of 5471 std::initializer_list that are likely to result in a dangling 5472 pointer, such as returning or assigning from a temporary list. 5473 + [55]-Wredundant-move, implied by -Wextra, warns about 5474 redundant calls to std::move. 5475 + [56]-Wpessimizing-move, implied by -Wall, warns when a call to 5476 std::move prevents copy elision. 5477 + [57]-Wclass-conversion, on by default, warns when a conversion 5478 function will never be called due to the type it converts to. 5479 * The C++ front end has experimental support for some of the upcoming 5480 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 5481 including range-based for statements with initializer, default 5482 constructible and assignable stateless lambdas, lambdas in 5483 unevaluated contexts, language support for empty data members, 5484 allowing pack expansion in lambda init-capture, likely and unlikely 5485 attributes, class types in non-type template parameters, allowing 5486 virtual function calls in constant expressions, explicit(bool), 5487 std::is_constant_evaluated, nested inline namespaces, etc. For a 5488 full list of new features, see [58]the C++ status page. 5489 * The C++ front end now preserves source locations for literals, 5490 id-expression, and mem-initializer for longer. For example it is 5491 now able to pin-point the pertinent locations for bad 5492 initializations such as these 5493 $ g++ -c bad-inits.cc 5494 bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization 5495 10 | { 3, json::object }, 5496 | ~~~~~~^~~~~~ 5497 | | 5498 | json 5499 bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f 5500 permissive] 5501 14 | char buffers[3][5] = { "red", "green", "blue" }; 5502 | ^~~~~~~ 5503 bad-inits.cc: In constructor 'X::X()': 5504 bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv 5505 e] 5506 17 | X() : one(42), two(42), three(42) 5507 | ^~ 5508 | | 5509 | int 5510 5511 rather than emitting the error at the final closing parenthesis or 5512 brace. 5513 * Error-reporting of overload resolution has been special-cased to 5514 make the case of a single failed candidate easier to read. For 5515 example: 5516 $ g++ param-type-mismatch.cc 5517 param-type-mismatch.cc: In function 'int test(int, const char*, float)': 5518 param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char* 5519 *' 5520 8 | return foo::member_1 (first, second, third); 5521 | ^~~~~~ 5522 | | 5523 | const char* 5524 param-type-mismatch.cc:3:46: note: initializing argument 2 of 'static int foo: 5525 :member_1(int, const char**, float)' 5526 3 | static int member_1 (int one, const char **two, float three); 5527 | ~~~~~~~~~~~~~^~~ 5528 5529 highlights both the problematic argument, and the parameter that it 5530 can't be converted to. 5531 * Diagnostics involving binary operators now use color to distinguish 5532 the two operands, and label them separately (as per the example of 5533 source labelling above). 5534 * Diagnostics involving function calls now highlight the pertinent 5535 parameter of the declaration in more places. 5536 $ g++ bad-conversion.cc 5537 bad-conversion.cc: In function 'void caller()': 5538 bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*' 5539 9 | callee (0, false, 2); 5540 | ^~~~~ 5541 | | 5542 | bool 5543 bad-conversion.cc:3:19: note: initializing argument 2 of 'void callee(int, voi 5544 d*, int)' 5545 3 | void callee (int, void *, int) 5546 | ^~~~~~ 5547 5548 * The C++ front end's implementation of [59]-Wformat now shows 5549 precise locations within string literals, and underlines the 5550 pertinent arguments at bogus call sites (the C front end has been 5551 doing this since GCC 7). For example: 5552 $ g++ -c bad-printf.cc -Wall 5553 bad-printf.cc: In function 'void print_field(const char*, float, long int, long 5554 int)': 5555 bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type 5556 'int', but argument 3 has type 'long int' [-Wformat=] 5557 6 | printf ("%s: %*ld ", fieldname, column - width, value); 5558 | ~^~~ ~~~~~~~~~~~~~~ 5559 | | | 5560 | int long int 5561 bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b 5562 ut argument 4 has type 'double' [-Wformat=] 5563 6 | printf ("%s: %*ld ", fieldname, column - width, value); 5564 | ~~~^ ~~~~~ 5565 | | | 5566 | long int double 5567 | %*f 5568 5569 * The C++ front end has gained new fix-it hints for forgetting the 5570 return *this; needed by various C++ operators: 5571 $ g++ -c operator.cc 5572 operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr 5573 &)': 5574 operator.cc:7:3: warning: no return statement in function returning non-void [-W 5575 return-type] 5576 6 | m_ptr = other.m_ptr; 5577 +++ |+ return *this; 5578 7 | } 5579 | ^ 5580 5581 for when the compiler needs a typename: 5582 $ g++ -c template.cc 5583 template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i 5584 s a dependent scope 5585 3 | Traits::type type; 5586 | ^~~~~~ 5587 | typename 5588 5589 when trying to use an accessor member as if it were a data member: 5590 $ g++ -c fncall.cc 5591 fncall.cc: In function 'void hangman(const mystring&)': 5592 fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length 5593 () const' (did you forget the '()' ?) 5594 12 | if (str.get_length > 0) 5595 | ~~~~^~~~~~~~~~ 5596 | () 5597 5598 for C++11's scoped enums: 5599 $ g++ -c enums.cc 5600 enums.cc: In function 'void json::test(const json::value&)': 5601 enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js 5602 on::kind::STRING'? 5603 12 | if (v.get_kind () == STRING) 5604 | ^~~~~~ 5605 | json::kind::STRING 5606 enums.cc:3:44: note: 'json::kind::STRING' declared here 5607 3 | enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ }; 5608 | ^~~~~~ 5609 5610 and a tweak to integrate the suggestions about misspelled members 5611 with that for accessors: 5612 $ g++ -c accessor-fixit.cc 5613 accessor-fixit.cc: In function 'int test(t*)': 5614 accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m 5615 ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const') 5616 17 | return ptr->ratio; 5617 | ^~~~~ 5618 | get_ratio() 5619 5620 In addition, various diagnostics in the C++ front-end have been 5621 streamlined by consolidating the suggestion into the initial error, 5622 rather than emitting a follow-up note: 5623 $ g++ typo.cc 5624 typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF 5625 _SIZE'? 5626 5 | uint8_t buf[BUFSIZE]; 5627 | ^~~~~~~ 5628 | BUF_SIZE 5629 5630 [60]Runtime Library (libstdc++) 5631 5632 * Improved support for C++17, including: 5633 + The C++17 implementation is no longer experimental. 5634 + Parallel algorithms and <execution> (requires [61]Thread 5635 Building Blocks 2018 or newer). 5636 + <memory_resource>. 5637 + Using the types and functions in <filesystem> does not require 5638 linking with -lstdc++fs now. 5639 * Improved experimental support for C++2a, including: 5640 + Type traits std::remove_cvref, std::unwrap_reference, 5641 std::unwrap_decay_ref, std::is_nothrow_convertible, and 5642 std::type_identity. 5643 + Headers <bit> and <version>. 5644 + Uniform container erasure (std::erase_if). 5645 + contains member of maps and sets. 5646 + String prefix and suffix checking (starts_with, ends_with). 5647 + Functions std::midpoint and std::lerp for interpolation. 5648 + std::bind_front. 5649 + std::visit<R>. 5650 + std::assume_aligned. 5651 + Uses-allocator construction utilities. 5652 + std::pmr::polymorphic_allocator<std::byte>. 5653 + Library support for char8_t type. 5654 + Destroying delete. 5655 + std::is_constant_evaluated() function. 5656 * Support for opening file streams with wide character paths on 5657 Windows 5658 * Incomplete support for the C++17 Filesystem library and the 5659 Filesystem TS on Windows. 5660 * Incomplete, experimental support for the Networking TS. 5661 5662 [62]D 5663 5664 * Support for the D programming language has been added to GCC, 5665 implementing version 2.076 of the language and run-time library. 5666 5667 [63]Fortran 5668 5669 * Asynchronous I/O is now fully supported. The program needs to be 5670 linked against the pthreads library to use it, otherwise the I/O is 5671 done synchronously. For systems which do not support POSIX 5672 condition variables, such as AIX, all I/O is still done 5673 synchronously. 5674 * The BACK argument for MINLOC and MAXLOC has been implemented. 5675 * The FINDLOC intrinsic function has been implemented. 5676 * The IS_CONTIGUOUS intrinsic function has been implemented. 5677 * Direct access to the real and imaginary parts of a complex variable 5678 via c%re and c%im has been implemented. 5679 * Type parameter inquiry via str%len and a%kind has been implemented. 5680 * C descriptors and the ISO_Fortran_binding.h source file have been 5681 implemented. 5682 * The MAX and MIN intrinsics are no longer guaranteed to return any 5683 particular value in case one of the arguments is a NaN. Note that 5684 this conforms to the Fortran standard and to what other Fortran 5685 compilers do. If there is a need to handle that case in some 5686 specific way, one needs to explicitly check for NaN's before 5687 calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the 5688 intrinsic module IEEE_ARITHMETIC. 5689 * A new command-line option [64]-fdec-include, set also by the 5690 [65]-fdec option, has been added to increase compatibility with 5691 legacy code. With this option, an INCLUDE directive is also parsed 5692 as a statement, which allows the directive to be spread across 5693 multiple source lines with line continuations. 5694 * A new [66]BUILTIN directive, has been added. The purpose of the 5695 directive is to provide an API between the GCC compiler and the GNU 5696 C Library which would define vector implementations of math 5697 routines. 5698 5699 [67]Go 5700 5701 * GCC 9 provides a complete implementation of the Go 1.12.2 user 5702 packages. 5703 5704 [68]libgccjit 5705 5706 * The libgccjit API gained a new entry point: 5707 [69]gcc_jit_context_add_driver_option. 5708 5709 [70]New Targets and Target Specific Improvements 5710 5711 [71]AArch64 & Arm 5712 5713 * Support has been added for the following processors (GCC 5714 identifiers in parentheses): 5715 + Arm Cortex-A76 (cortex-a76). 5716 + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE 5717 (cortex-a76.cortex-a55). 5718 + Arm Neoverse N1 (neoverse-n1). 5719 The GCC identifiers can be used as arguments to the -mcpu or -mtune 5720 options, for example: -mcpu=cortex-a76 or 5721 -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent 5722 target attributes and pragmas. 5723 * The Armv8.3-A complex number instructions are now supported via 5724 intrinsics when the option -march=armv8.3-a or equivalent is 5725 specified. For the half-precision floating-point variants of these 5726 instructions use the architecture extension flag +fp16, e.g. 5727 -march=armv8.3-a+fp16. 5728 The intrinsics are defined by the ACLE specification. 5729 * The Armv8.5-A architecture is now supported through the 5730 -march=armv8.5-a option. 5731 * The Armv8.5-A architecture also adds some security features that 5732 are optional to all older architecture versions. These are now 5733 supported and only affect the assembler. 5734 + Speculation Barrier instruction through the -march=armv8-a+sb 5735 option. 5736 + Execution and Data Prediction Restriction instructions through 5737 the -march=armv8-a+predres option. 5738 + Speculative Store Bypass Safe instruction through the 5739 -march=armv8-a+ssbs option. This does not require a compiler 5740 option for Arm and thus -march=armv8-a+ssbs is an 5741 AArch64-specific option. 5742 5743 AArch64 specific 5744 5745 * Support has been added for the Arm Neoverse E1 processor 5746 (-mcpu=neoverse-e1). 5747 * The AArch64 port now has support for stack clash protection using 5748 the [72]-fstack-clash-protection option. The probing interval/guard 5749 size can be set by using --param 5750 stack-clash-protection-guard-size=12|16. The value of this 5751 parameter must be in bytes represented as a power of two. The two 5752 supported values for this parameter are 12 (for a 4KiB size, 2^12) 5753 and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and 5754 can be changed at configure time using the flag 5755 --with-stack-clash-protection-guard-size=12|16. 5756 * The option -msign-return-address= has been deprecated. This has 5757 been replaced by the new -mbranch-protection= option. This new 5758 option can now be used to enable the return address signing as well 5759 as the new Branch Target Identification feature of Armv8.5-A 5760 architecture. For more information on the arguments accepted by 5761 this option, please refer to [73]AArch64-Options. 5762 * The following optional extensions to Armv8.5-A architecture are now 5763 supported and only affect the assembler. 5764 + Random Number Generation instructions through the 5765 -march=armv8.5-a+rng option. 5766 + Memory Tagging Extension through the -march=armv8.5-a+memtag 5767 option. 5768 5769 Arm specific 5770 5771 * Support for the deprecated Armv2 and Armv3 architectures and their 5772 variants has been removed. Their corresponding -march values and 5773 the -mcpu options that used these architectures have been removed. 5774 * Support for the Armv5 and Armv5E architectures (which have no known 5775 implementations) has been removed. Note that Armv5T, Armv5TE and 5776 Armv5TEJ architectures remain supported. 5777 * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using 5778 their respective -mcpu options. 5779 5780 [74]AMD GCN 5781 5782 * A new back end targeting AMD GCN GPUs has been contributed to GCC. 5783 The implementation is currently limited to compiling 5784 single-threaded, stand-alone programs. Future versions will add 5785 support for offloading multi-threaded kernels via OpenMP and 5786 OpenACC. The following devices are supported (GCC identifiers in 5787 parentheses): 5788 + Fiji (fiji). 5789 + Vega 10 (gfx900). 5790 5791 [75]ARC 5792 5793 * LRA is now on by default for the ARC target. This can be controlled 5794 by -mlra. 5795 * Add support for frame code-density and branch-and-index 5796 instructions. 5797 5798 [76]C-SKY 5799 5800 * A new back end targeting C-SKY V2 processors has been contributed 5801 to GCC. 5802 5803 [77]IA-32/x86-64 5804 5805 * Support of Intel MPX (Memory Protection Extensions) has been 5806 removed. 5807 * New ISA extension support for Intel PTWRITE was added to GCC. 5808 PTWRITE intrinsics are available via the -mptwrite compiler switch. 5809 * GCC now supports the Intel CPU named Cascade Lake with AVX512 5810 extensions through -march=cascadelake. The switch enables the 5811 following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW, 5812 AVX512DQ, AVX512VNNI. 5813 5814 [78]MIPS 5815 5816 * The Loongson loongson-mmi and loongson-ext extensions have been 5817 split from loongson3a: 5818 + loongson-mmi contains the Loongson MMI (MultiMedia extensions 5819 Instructions). 5820 + loongson-ext contains the Loongson EXT (EXTensions 5821 instructions). 5822 * The Loongson EXT2 (EXTensions R2 instructions) are now supported. 5823 + loongson-ext2 contains the Loongson EXT2 instructions. 5824 Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and 5825 -m[no-]loongson-ext2 enable or disable those extensions. 5826 * Support has been added for the following processors (GCC 5827 identifiers in parentheses): 5828 + Loongson 3A1000 (gs464) which enables loongson-mmi, 5829 loongson-ext by default. 5830 + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi, 5831 loongson-ext, loongson-ext2 by default. 5832 + Loongson 2K1000 (gs264e) which enables loongson-ext, 5833 loongson-ext2, msa by default. 5834 The GCC identifiers can be used as arguments to the -mcpu and 5835 -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments 5836 to the equivalent target attributes and pragmas. 5837 5838 [79]OpenRISC 5839 5840 * A new back end targeting OpenRISC processors has been contributed 5841 to GCC. 5842 5843 [80]S/390, System z, IBM z Systems 5844 5845 * Support for the arch13 architecture has been added. When using the 5846 -march=arch13 option, the compiler will generate code making use of 5847 the new instructions introduced with the vector enhancement 5848 facility 2 and the miscellaneous instruction extension facility 2. 5849 The -mtune=arch13 option enables arch13 specific instruction 5850 scheduling without making use of new instructions. 5851 * Builtins for the new vector instructions have been added and can be 5852 enabled using the -mzvector option. 5853 * Support for ESA architecture machines g5 and g6 is deprecated since 5854 GCC 6.1.0 and has been removed now. 5855 * When compiling with -march=z14 or higher GCC emits alignments hints 5856 on the vector load/store instructions (8 or 16 byte). 5857 * Functions now have a default alignment of 16 bytes. This helps with 5858 branch prediction effects. 5859 * -mfentry is now supported. As well as the mcount mechanism the 5860 __fentry__ is called before the function prologue. However, since 5861 just a single instruction is required to call __fentry__ the call 5862 sequence imposes a smaller overhead than mcount (4 instructions). 5863 The produced code is compatible only with newer glibc versions, 5864 which provide the __fentry__ symbol and do not clobber r0 when 5865 resolving lazily bound functions. -mfentry is only supported when 5866 generating 64-bit code and does not work with nested C functions. 5867 * The -mnop-mcount option can be used to emit NOP instructions 5868 instead of an mcount or fentry call stub. 5869 * With the -mrecord-mcount option a __mcount_loc section is generated 5870 containing pointers to each profiling call stub. This is useful for 5871 automatically patching in and out calls. 5872 5873 [81]Operating Systems 5874 5875 [82]Solaris 5876 5877 * g++ now unconditionally enables large file support when compiling 5878 32-bit code. 5879 * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has 5880 been merged from LLVM. For the moment, this only works for 32-bit 5881 code on both SPARC and x86. 5882 * An initial port of the D runtime library has been completed on 5883 Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC 5884 support is still work-in-progress. 5885 5886 [83]Windows 5887 5888 * A C++ Microsoft ABI bitfield layout bug, [84]PR87137 has been 5889 fixed. A non-field declaration could cause the current bitfield 5890 allocation unit to be completed, incorrectly placing a following 5891 bitfield into a new allocation unit. The Microsoft ABI is selected 5892 for: 5893 + Mingw targets 5894 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 5895 option is specified, or __attribute__((ms_struct)) is used 5896 + SuperH targets when the -mhitachi option is specified, or 5897 __attribute__((renesas)) is used 5898 5899 [85]Improvements for plugin authors 5900 5901 * GCC's diagnostic subsystem now has a way to logically group 5902 together related diagnostics, auto_diagnostic_group. Such 5903 diagnostics will be nested by the output of 5904 [86]-fdiagnostics-format=json. 5905 * GCC now has a set of [87]user experience guidelines for GCC, with 5906 information and advice on implementing new diagnostics. 5907 5908 Other significant improvements 5909 5910 * GCC's internal "selftest" suite now runs for C++ as well as C (in 5911 debug builds of the compiler). 5912 5913 [88]GCC 9.1 5914 5915 This is the [89]list of problem reports (PRs) from GCC's bug tracking 5916 system that are known to be fixed in the 9.1 release. This list might 5917 not be complete (that is, it is possible that some PRs that have been 5918 fixed are not listed here). 5919 5920 [90]GCC 9.2 5921 5922 This is the [91]list of problem reports (PRs) from GCC's bug tracking 5923 system that are known to be fixed in the 9.2 release. This list might 5924 not be complete (that is, it is possible that some PRs that have been 5925 fixed are not listed here). 5926 5927 [92]GCC 9.3 5928 5929 This is the [93]list of problem reports (PRs) from GCC's bug tracking 5930 system that are known to be fixed in the 9.3 release. This list might 5931 not be complete (that is, it is possible that some PRs that have been 5932 fixed are not listed here). 5933 5934 [94]GCC 9.4 5935 5936 This is the [95]list of problem reports (PRs) from GCC's bug tracking 5937 system that are known to be fixed in the 9.4 release. This list might 5938 not be complete (that is, it is possible that some PRs that have been 5939 fixed are not listed here). 5940 5941 Target Specific Changes 5942 5943 AArch64 5944 5945 * The option -moutline-atomics has been added to aid deployment of 5946 the Large System Extensions (LSE) on GNU/Linux systems built with a 5947 baseline architecture targeting Armv8-A. When the option is 5948 specified code is emitted to detect the presence of LSE 5949 instructions at run time and use them for standard atomic 5950 operations. For more information please refer to the documentation. 5951 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 5952 options are -mcpu=a64fx and -mtune=a64fx respectively. In 5953 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 5954 tunes the code for the A64FX. This includes tuning the SVE code, 5955 although by default the code is still length-agnostic and so works 5956 for all SVE implementations. Adding -msve-vector-bits=512 makes the 5957 code specific to 512-bit SVE. 5958 5959 [96]GCC 9.5 5960 5961 This is the [97]list of problem reports (PRs) from GCC's bug tracking 5962 system that are known to be fixed in the 9.5 release. This list might 5963 not be complete (that is, it is possible that some PRs that have been 5964 fixed are not listed here). 5965 5966 5967 For questions related to the use of GCC, please consult these web 5968 pages and the [98]GCC manuals. If that fails, the 5969 [99]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5970 web pages and the development of GCC are welcome on our developer 5971 list at [100]gcc (a] gcc.gnu.org. All of [101]our lists have public 5972 archives. 5973 5974 Copyright (C) [102]Free Software Foundation, Inc. Verbatim copying and 5975 distribution of this entire article is permitted in any medium, 5976 provided this notice is preserved. 5977 5978 These pages are [103]maintained by the GCC team. Last modified 5979 2025-01-31. 5980 5981 References 5982 5983 1. https://gcc.gnu.org/gcc-9/porting_to.html 5984 2. https://gcc.gnu.org/onlinedocs/index.html#current 5985 3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469 5986 4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html 5987 5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html 5988 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920 5989 7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo 5990 8. https://gcc.gnu.org/PR90361 5991 9. https://gcc.gnu.org/gcc-9/changes.html#general 5992 10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC 5993 11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching 5994 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers 5995 13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels 5996 14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 5997 15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions 5998 16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels 5999 17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops 6000 18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps 6001 19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files 6002 20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files 6003 21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info 6004 22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record 6005 23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment 6006 24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable 6007 25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability 6008 26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1 6009 27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1 6010 28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute 6011 29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion 6012 30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use 6013 31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides 6014 32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange 6015 33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam 6016 34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution 6017 35. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov 6018 36. https://gcc.gnu.org/gcc-9/changes.html#languages 6019 37. https://www.openacc.org/ 6020 38. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-9 6021 39. https://gcc.gnu.org/gcc-9/changes.html#c-family 6022 40. https://www.openmp.org/specifications/ 6023 41. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html 6024 42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector 6025 43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member 6026 44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds 6027 45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias 6028 46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow 6029 47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation 6030 48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 6031 49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 6032 50. https://gcc.gnu.org/gcc-9/changes.html#c 6033 51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value 6034 52. https://gcc.gnu.org/gcc-9/changes.html#cxx 6035 53. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy 6036 54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime 6037 55. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 6038 56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 6039 57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion 6040 58. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 6041 59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat 6042 60. https://gcc.gnu.org/gcc-9/changes.html#libstdcxx 6043 61. https://github.com/uxlfoundation/oneTBB 6044 62. https://gcc.gnu.org/gcc-9/changes.html#d 6045 63. https://gcc.gnu.org/gcc-9/changes.html#fortran 6046 64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include 6047 65. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec 6048 66. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive 6049 67. https://gcc.gnu.org/gcc-9/changes.html#go 6050 68. https://gcc.gnu.org/gcc-9/changes.html#jit 6051 69. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option 6052 70. https://gcc.gnu.org/gcc-9/changes.html#targets 6053 71. https://gcc.gnu.org/gcc-9/changes.html#arm-targets 6054 72. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector 6055 73. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options 6056 74. https://gcc.gnu.org/gcc-9/changes.html#amdgcn 6057 75. https://gcc.gnu.org/gcc-9/changes.html#arc 6058 76. https://gcc.gnu.org/gcc-9/changes.html#csky 6059 77. https://gcc.gnu.org/gcc-9/changes.html#x86 6060 78. https://gcc.gnu.org/gcc-9/changes.html#mips 6061 79. https://gcc.gnu.org/gcc-9/changes.html#or1k 6062 80. https://gcc.gnu.org/gcc-9/changes.html#s390 6063 81. https://gcc.gnu.org/gcc-9/changes.html#os 6064 82. https://gcc.gnu.org/gcc-9/changes.html#solaris 6065 83. https://gcc.gnu.org/gcc-9/changes.html#windows 6066 84. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 6067 85. https://gcc.gnu.org/gcc-9/changes.html#plugins 6068 86. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 6069 87. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html 6070 88. https://gcc.gnu.org/gcc-9/changes.html#GCC9.1 6071 89. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0 6072 90. https://gcc.gnu.org/gcc-9/changes.html#GCC9.2 6073 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2 6074 92. https://gcc.gnu.org/gcc-9/changes.html#GCC9.3 6075 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3 6076 94. https://gcc.gnu.org/gcc-9/changes.html#GCC9.4 6077 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4 6078 96. https://gcc.gnu.org/gcc-9/changes.html#GCC9.5 6079 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5 6080 98. https://gcc.gnu.org/onlinedocs/ 6081 99. mailto:gcc-help (a] gcc.gnu.org 6082 100. mailto:gcc (a] gcc.gnu.org 6083 101. https://gcc.gnu.org/lists.html 6084 102. https://www.fsf.org/ 6085 103. https://gcc.gnu.org/about.html 6086 ====================================================================== 6087 http://gcc.gnu.org/gcc-8/index.html 6088 6089 GCC 8 Release Series 6090 6091 (This release series is no longer supported.) 6092 6093 May 14, 2021 6094 6095 The [1]GNU project and the GCC developers are pleased to announce the 6096 release of GCC 8.5. 6097 6098 This release is a bug-fix release, containing fixes for regressions in 6099 GCC 8.4 relative to previous releases of GCC. 6100 6101 Release History 6102 6103 GCC 8.5 6104 May 14, 2021 ([2]changes, [3]documentation) 6105 6106 GCC 8.4 6107 Mar 4, 2020 ([4]changes, [5]documentation) 6108 6109 GCC 8.3 6110 Feb 22, 2019 ([6]changes, [7]documentation) 6111 6112 GCC 8.2 6113 Jul 26, 2018 ([8]changes, [9]documentation) 6114 6115 GCC 8.1 6116 May 2, 2018 ([10]changes, [11]documentation) 6117 6118 References and Acknowledgements 6119 6120 GCC used to stand for the GNU C Compiler, but since the compiler 6121 supports several other languages aside from C, it now stands for the 6122 GNU Compiler Collection. 6123 6124 A list of [12]successful builds is updated as new information becomes 6125 available. 6126 6127 The GCC developers would like to thank the numerous people that have 6128 contributed new features, improvements, bug fixes, and other changes as 6129 well as test results to GCC. This [13]amazing group of volunteers is 6130 what makes GCC successful. 6131 6132 For additional information about GCC please refer to the [14]GCC 6133 project web site or contact the [15]GCC development mailing list. 6134 6135 To obtain GCC please use [16]our mirror sites or [17]our version 6136 control system. 6137 6138 6139 For questions related to the use of GCC, please consult these web 6140 pages and the [18]GCC manuals. If that fails, the 6141 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6142 web pages and the development of GCC are welcome on our developer 6143 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6144 archives. 6145 6146 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6147 distribution of this entire article is permitted in any medium, 6148 provided this notice is preserved. 6149 6150 These pages are [23]maintained by the GCC team. Last modified 6151 2024-05-30. 6152 6153 References 6154 6155 1. http://www.gnu.org/ 6156 2. https://gcc.gnu.org/gcc-8/changes.html 6157 3. https://gcc.gnu.org/onlinedocs/8.5.0/ 6158 4. https://gcc.gnu.org/gcc-8/changes.html 6159 5. https://gcc.gnu.org/onlinedocs/8.4.0/ 6160 6. https://gcc.gnu.org/gcc-8/changes.html 6161 7. https://gcc.gnu.org/onlinedocs/8.3.0/ 6162 8. https://gcc.gnu.org/gcc-8/changes.html 6163 9. https://gcc.gnu.org/onlinedocs/8.2.0/ 6164 10. https://gcc.gnu.org/gcc-8/changes.html 6165 11. https://gcc.gnu.org/onlinedocs/8.1.0/ 6166 12. https://gcc.gnu.org/gcc-8/buildstat.html 6167 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 6168 14. https://gcc.gnu.org/index.html 6169 15. mailto:gcc (a] gcc.gnu.org 6170 16. https://gcc.gnu.org/mirrors.html 6171 17. https://gcc.gnu.org/git.html 6172 18. https://gcc.gnu.org/onlinedocs/ 6173 19. mailto:gcc-help (a] gcc.gnu.org 6174 20. mailto:gcc (a] gcc.gnu.org 6175 21. https://gcc.gnu.org/lists.html 6176 22. https://www.fsf.org/ 6177 23. https://gcc.gnu.org/about.html 6178 ====================================================================== 6179 http://gcc.gnu.org/gcc-8/changes.html 6180 6181 GCC 8 Release Series 6182 Changes, New Features, and Fixes 6183 6184 This page is a "brief" summary of some of the huge number of 6185 improvements in GCC 8. You may also want to check out our [1]Porting to 6186 GCC 8 page and the [2]full GCC documentation. 6187 6188 Caveats 6189 6190 * The default mode for C is now -std=gnu17 instead of -std=gnu11. 6191 * Support for the obsolete SDB/coff debug info format has been 6192 removed. The option -gcoff no longer does anything. 6193 * The Cilk+ extensions to the C and C++ languages have been removed. 6194 * The MPX extensions to the C and C++ languages have been deprecated 6195 and will be removed in a future release. 6196 * The extension allowing arithmetic on std::atomic<void*> and types 6197 like std::atomic<R(*)()> has been deprecated. 6198 * The non-standard C++0x std::copy_exception function was removed. 6199 std::make_exception_ptr should be used instead. 6200 * Support for the powerpc*-*-*spe* target ports which have been 6201 recently unmaintained and untested in GCC has been declared 6202 obsolete in GCC 8 as [3]announced. Unless there is activity to 6203 revive them, the next release of GCC will have their sources 6204 permanently removed. 6205 6206 [4]General Improvements 6207 6208 * Inter-procedural optimization improvements: 6209 + Reworked run-time estimation metrics leading to more realistic 6210 guesses driving inliner and cloning heuristics. 6211 + The ipa-pure-const pass is extended to propagate the malloc 6212 attribute, and the corresponding warning option 6213 -Wsuggest-attribute=malloc emits a diagnostic for functions 6214 which can be annotated with the malloc attribute. 6215 * Profile driven optimization improvements: 6216 + New infrastructure for representing profiles (both statically 6217 guessed and profile feedback) which allows propagation of 6218 additional information about the reliability of the profile. 6219 + A number of improvements in the profile updating code solving 6220 problems found by new verification code. 6221 + Static detection of code which is not executed in a valid run 6222 of the program. This includes paths which trigger undefined 6223 behavior as well as calls to functions declared with the cold 6224 attribute. Newly the noreturn attribute does not imply all 6225 effects of cold to differentiate between exit (which is 6226 noreturn) and abort (which is in addition not executed in 6227 valid runs). 6228 + -freorder-blocks-and-partition, a pass splitting function 6229 bodies into hot and cold regions, is now enabled by default at 6230 -O2 and higher for x86 and x86-64. 6231 * Link-time optimization improvements: 6232 + We have significantly improved debug information on ELF 6233 targets using DWARF by properly preserving language-specific 6234 information. This allows for example the libstdc++ 6235 pretty-printers to work with LTO optimized executables. 6236 * A new option -fcf-protection=[full|branch|return|none] is 6237 introduced to perform code instrumentation to increase program 6238 security by checking that target addresses of control-flow transfer 6239 instructions (such as indirect function call, function return, 6240 indirect jump) are valid. Currently the instrumentation is 6241 supported on x86 GNU/Linux targets only. See the user guide for 6242 further information about the option syntax and section "New 6243 Targets and Target Specific Improvements" for IA-32/x86-64 for more 6244 details. 6245 * The -gcolumn-info option is now enabled by default. It includes 6246 column information in addition to just filenames and line numbers 6247 in DWARF debugging information. 6248 * The polyhedral-based loop nest optimization pass 6249 -floop-nest-optimize has been overhauled. It's still considered 6250 experimental and may not result in any runtime improvements. 6251 * Two new classical loop nest optimization passes have been added. 6252 -floop-unroll-and-jam performs outer loop unrolling and fusing of 6253 the inner loop copies. -floop-interchange exchanges loops in a loop 6254 nest to improve data locality. Both passes are enabled by default 6255 at -O3 and above. 6256 * The classic loop nest optimization pass -ftree-loop-distribution 6257 has been improved and enabled by default at -O3 and above. It 6258 supports loop nest distribution in some restricted scenarios; it 6259 also supports cancellable innermost loop distribution with loop 6260 versioning under run-time alias checks. 6261 * The new option -fstack-clash-protection causes the compiler to 6262 insert probes whenever stack space is allocated statically or 6263 dynamically to reliably detect stack overflows and thus mitigate 6264 the attack vector that relies on jumping over a stack guard page as 6265 provided by the operating system. 6266 * A new pragma GCC unroll has been implemented in the C family of 6267 languages, as well as Fortran and Ada, so as to make it possible 6268 for the user to have a finer-grained control over the loop 6269 unrolling optimization. 6270 * GCC has been enhanced to detect more instances of meaningless or 6271 mutually exclusive attribute specifications and handle such 6272 conflicts more consistently. Mutually exclusive attribute 6273 specifications are ignored with a warning regardless of whether 6274 they appear on the same declaration or on distinct declarations of 6275 the same entity. For example, because the noreturn attribute on the 6276 second declaration below is mutually exclusive with the malloc 6277 attribute on the first, it is ignored and a warning is issued. 6278 > 6279 void* __attribute__ ((malloc)) f (unsigned); 6280 void* __attribute__ ((noreturn)) f (unsigned); 6281 6282 warning: ignoring attribute 'noreturn' because it conflicts with attribute 6283 'malloc' [-Wattributes] 6284 * The gcov tool can distinguish functions that begin on a same line 6285 in a source file. This can be a different template instantiation or 6286 a class constructor: 6287 6288 File 'ins.C' 6289 Lines executed:100.00% of 8 6290 Creating 'ins.C.gcov' 6291 6292 -: 0:Source:ins.C 6293 -: 0:Graph:ins.gcno 6294 -: 0:Data:ins.gcda 6295 -: 0:Runs:1 6296 -: 0:Programs:1 6297 -: 1:template<class T> 6298 -: 2:class Foo 6299 -: 3:{ 6300 -: 4: public: 6301 2: 5: Foo(): b (1000) {} 6302 ------------------ 6303 Foo<char>::Foo(): 6304 1: 5: Foo(): b (1000) {} 6305 ------------------ 6306 Foo<int>::Foo(): 6307 1: 5: Foo(): b (1000) {} 6308 ------------------ 6309 2: 6: void inc () { b++; } 6310 ------------------ 6311 Foo<char>::inc(): 6312 1: 6: void inc () { b++; } 6313 ------------------ 6314 Foo<int>::inc(): 6315 1: 6: void inc () { b++; } 6316 ------------------ 6317 -: 7: 6318 -: 8: private: 6319 -: 9: int b; 6320 -: 10:}; 6321 -: 11: 6322 1: 12:int main(int argc, char **argv) 6323 -: 13:{ 6324 1: 14: Foo<int> a; 6325 1: 15: Foo<char> b; 6326 -: 16: 6327 1: 17: a.inc (); 6328 1: 18: b.inc (); 6329 1: 19:} 6330 6331 * The gcov tool has more accurate numbers for execution of lines in a 6332 source file. 6333 * The gcov tool can use TERM colors to provide more readable output. 6334 * AddressSanitizer gained a new pair of sanitization options, 6335 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 6336 warn about subtraction (or comparison) of pointers that point to a 6337 different memory object: 6338 6339 int 6340 main () 6341 { 6342 /* Heap allocated memory. */ 6343 char *heap1 = (char *)__builtin_malloc (42); 6344 char *heap2 = (char *)__builtin_malloc (42); 6345 if (heap1 > heap2) 6346 return 1; 6347 6348 return 0; 6349 } 6350 6351 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 6352 00050 6353 #0 0x40070f in main /tmp/pointer-compare.c:7 6354 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6355 #2 0x400629 in _start (/tmp/a.out+0x400629) 6356 6357 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 6358 00000003a) 6359 allocated by thread T0 here: 6360 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 6361 _malloc_linux.cc:86 6362 #1 0x4006ea in main /tmp/pointer-compare.c:5 6363 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6364 6365 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 6366 00000007a) 6367 allocated by thread T0 here: 6368 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 6369 _malloc_linux.cc:86 6370 #1 0x4006f8 in main /tmp/pointer-compare.c:6 6371 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 6372 6373 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 6374 6375 * The store merging pass has been enhanced to handle bit-fields and 6376 not just constant stores, but also data copying from adjacent 6377 memory locations into other adjacent memory locations, including 6378 bitwise logical operations on the data. The pass can also handle 6379 byte swapping into memory locations. 6380 * The undefined behavior sanitizer gained two new options included in 6381 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 6382 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 6383 builtins, and -fsanitize=pointer-overflow which performs cheap run 6384 time tests for pointer wrapping. 6385 * A new attribute no_sanitize can be applied to functions to instruct 6386 the compiler not to do sanitization of the options provided as 6387 arguments to the attribute. Acceptable values for no_sanitize match 6388 those acceptable by the -fsanitize command-line option. 6389 6390 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 6391 f () { /* Do something. */; } 6392 6393 [5]New Languages and Language specific improvements 6394 6395 [6]Ada 6396 6397 * For its internal exception handling used on the host for error 6398 recovery in the front-end, the compiler now relies on the native 6399 exception handling mechanism of the host platform, which should be 6400 more efficient than the former mechanism. 6401 6402 [7]BRIG (HSAIL) 6403 6404 In this release cycle, the focus for the BRIGFE was on stabilization 6405 and performance improvements. Also a couple of completely new features 6406 were added. 6407 * Improved support for function and module scope group segment 6408 variables. PRM specs define function and module scope group segment 6409 variables as an experimental feature. However, PRM test suite uses 6410 them. Now group segment is handled by separate book keeping of 6411 module scope and function (kernel) offsets. Each function has a 6412 "frame" in the group segment offset to which is given as an 6413 argument, similar to traditional call stack frame handling. 6414 * Reduce the number of type conversions due to the untyped HSAIL 6415 registers. Instead of always representing the HSAIL's untyped 6416 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 6417 code and builds the register variables as a type used the most when 6418 storing or reading data to/from each register. This reduces the 6419 number of total casts which cannot be always optimized away. 6420 * Support for BRIG_KIND_NONE directives. 6421 * Made -O3 the default optimization level for BRIGFE. 6422 * Fixed illegal addresses generated from address expressions which 6423 refer only to offset 0. 6424 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 6425 mode, the offset is treated as 32-bit unless it's in global, 6426 read-only or kernarg address space. 6427 * Fixed a crash caused sometimes by calls with more than 4 arguments. 6428 * Fixed a mis-execution issue with kernels that have both unexpanded 6429 ID functions and calls to subfunctions. 6430 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 6431 avoid illegal optimizations. 6432 * Ensure per WI copies of private variables are aligned correctly. 6433 * libhsail-rt: Assume the host runtime allocates the work group 6434 memory. 6435 6436 [8]C family 6437 6438 * New command-line options have been added for the C and C++ 6439 compilers: 6440 + [9]-Wmultistatement-macros warns about unsafe macros expanding 6441 to multiple statements used as a body of a statement such as 6442 if, else, while, switch, or for. 6443 + [10]-Wstringop-truncation warns for calls to bounded string 6444 manipulation functions such as strncat, strncpy, and stpncpy 6445 that might either truncate the copied string or leave the 6446 destination unchanged. For example, the following call to 6447 strncat is diagnosed because it appends just three of the four 6448 characters from the source string. 6449 void append (char *buf, size_t bufsize) 6450 { 6451 strncat (buf, ".txt", 3); 6452 } 6453 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 6454 Wstringop-truncation] 6455 Similarly, in the following example, the call to strncpy 6456 specifies the size of the destination buffer as the bound. If 6457 the length of the source string is equal to or greater than 6458 this size the result of the copy will not be NUL-terminated. 6459 Therefore, the call is also diagnosed. To avoid the warning, 6460 specify sizeof buf - 1 as the bound and set the last element 6461 of the buffer to NUL. 6462 void copy (const char *s) 6463 { 6464 char buf[80]; 6465 strncpy (buf, s, sizeof buf); 6466 ... 6467 } 6468 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 6469 tion] 6470 The -Wstringop-truncation option is included in -Wall. 6471 Note that due to GCC bug [11]82944, defining strncat, strncpy, 6472 or stpncpy as a macro in a system header as some 6473 implementations do, suppresses the warning. 6474 + [12]-Wif-not-aligned controls warnings issued in response to 6475 invalid uses of objects declared with attribute 6476 [13]warn_if_not_aligned. 6477 The -Wif-not-aligned option is included in -Wall. 6478 + [14]-Wmissing-attributes warns when a declaration of a 6479 function is missing one or more attributes that a related 6480 function is declared with and whose absence may adversely 6481 affect the correctness or efficiency of generated code. For 6482 example, in C++, the warning is issued when an explicit 6483 specialization of a primary template declared with attribute 6484 alloc_align, alloc_size, assume_aligned, format, format_arg, 6485 malloc, or nonnull is declared without it. Attributes 6486 deprecated, error, and warning suppress the warning. 6487 The -Wmissing-attributes option is included in -Wall. 6488 + [15]-Wpacked-not-aligned warns when a struct or union declared 6489 with attribute packed defines a member with an explicitly 6490 specified alignment greater than 1. Such a member will wind up 6491 under-aligned. For example, a warning will be issued for the 6492 definition of struct A in the following: 6493 struct __attribute__ ((aligned (8))) 6494 S8 { char a[8]; }; 6495 6496 struct __attribute__ ((packed)) A 6497 { 6498 struct S8 s8; 6499 }; 6500 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 6501 The -Wpacked-not-aligned option is included in -Wall. 6502 + -Wcast-function-type warns when a function pointer is cast to 6503 an incompatible function pointer. This warning is enabled by 6504 -Wextra. 6505 + -Wsizeof-pointer-div warns for suspicious divisions of the 6506 size of a pointer by the size of the elements it points to, 6507 which looks like the usual way to compute the array size but 6508 won't work out correctly with pointers. This warning is 6509 enabled by -Wall. 6510 + -Wcast-align=strict warns whenever a pointer is cast such that 6511 the required alignment of the target is increased. For 6512 example, warn if a char * is cast to an int * regardless of 6513 the target machine. 6514 + -fprofile-abs-path creates absolute path names in the .gcno 6515 files. This allows gcov to find the correct sources in 6516 projects where compilations occur with different working 6517 directories. 6518 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 6519 signed integer overflow is now undefined by default at all 6520 optimization levels. Using -fsanitize=signed-integer-overflow is 6521 now the preferred way to audit code, -Wstrict-overflow is 6522 deprecated. 6523 * The [16]-Warray-bounds option has been improved to detect more 6524 instances of out-of-bounds array indices and pointer offsets. For 6525 example, negative or excessive indices into flexible array members 6526 and string literals are detected. 6527 * The [17]-Wrestrict option introduced in GCC 7 has been enhanced to 6528 detect many more instances of overlapping accesses to objects via 6529 restrict-qualified arguments to standard memory and string 6530 manipulation functions such as memcpy and strcpy. For example, the 6531 strcpy call in the function below attempts to truncate the string 6532 by replacing its initial characters with the last four. However, 6533 because the function writes the terminating NUL into a[4], the 6534 copies overlap and the call is diagnosed. 6535 void f (void) 6536 { 6537 char a[] = "abcd1234"; 6538 strcpy (a, a + 4); 6539 ... 6540 } 6541 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 6542 4 [-Wrestrict] 6543 The -Wrestrict option is included in -Wall. 6544 * Several optimizer enhancements have enabled improvements to the 6545 [18]-Wformat-overflow and [19]-Wformat-truncation options. The 6546 warnings detect more instances of buffer overflow and truncation 6547 than in GCC 7 and are better at avoiding certain kinds of false 6548 positives. 6549 * When reporting mismatching argument types at a function call, the C 6550 and C++ compilers now underline both the argument and the pertinent 6551 parameter in the declaration. 6552 $ gcc arg-type-mismatch.cc 6553 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 6554 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 6555 [-fpermissive] 6556 return callee(first, second, third); 6557 ^~~~~~ 6558 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 6559 onst char*, float)' 6560 extern int callee(int one, const char *two, float three); 6561 ~~~~~~~~~~~~^~~ 6562 6563 * When reporting on unrecognized identifiers, the C and C++ compilers 6564 will now emit fix-it hints suggesting #include directives for 6565 various headers in the C and C++ standard libraries. 6566 $ gcc incomplete.c 6567 incomplete.c: In function 'test': 6568 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 6569 return NULL; 6570 ^~~~ 6571 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 6572 t to '#include <stddef.h>'? 6573 incomplete.c:1:1: 6574 +#include <stddef.h> 6575 const char *test(void) 6576 incomplete.c:3:10: 6577 return NULL; 6578 ^~~~ 6579 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 6580 ch function it appears in 6581 6582 $ gcc incomplete.cc 6583 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 6584 std::string s("hello world"); 6585 ^~~~~~ 6586 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 6587 forget to '#include <string>'? 6588 +#include <string> 6589 std::string s("hello world"); 6590 ^~~ 6591 6592 * The C and C++ compilers now use more intuitive locations when 6593 reporting on missing semicolons, and offer fix-it hints: 6594 $ gcc t.c 6595 t.c: In function 'test': 6596 t.c:3:12: error: expected ';' before '}' token 6597 return 42 6598 ^ 6599 ; 6600 } 6601 ~ 6602 6603 * When reporting on missing '}' and ')' tokens, the C and C++ 6604 compilers will now highlight the corresponding '{' and '(' token, 6605 issuing a 'note' if it's on a separate line: 6606 $ gcc unclosed.c 6607 unclosed.c: In function 'log_when_out_of_range': 6608 unclosed.c:12:50: error: expected ')' before '{' token 6609 && (temperature < MIN || temperature > MAX) { 6610 ^~ 6611 ) 6612 unclosed.c:11:6: note: to match this '(' 6613 if (logging_enabled && check_range () 6614 ^ 6615 6616 or highlighting it directly if it's on the same line: 6617 $ gcc unclosed-2.c 6618 unclosed-2.c: In function 'test': 6619 unclosed-2.c:8:45: error: expected ')' before '{' token 6620 if (temperature < MIN || temperature > MAX { 6621 ~ ^~ 6622 ) 6623 6624 They will also emit fix-it hints. 6625 6626 [20]C 6627 6628 * New options -std=c17, to select support for the 2018 edition of the 6629 ISO C standard (__STDC_VERSION__ == 201710L), and -std=gnu17, for 6630 C17 with GNU extensions. 6631 * The default mode has been changed to -std=gnu17. 6632 6633 [21]C++ 6634 6635 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 6636 convention, which changes the ABI for some uncommon code: 6637 + Passing an empty class as an argument now takes up no space on 6638 x86_64, as required by the psABI. 6639 + Passing or returning a class with only deleted copy and move 6640 constructors now uses the same calling convention as a class 6641 with a non-trivial copy or move constructor. This only affects 6642 C++17 mode, as in earlier standards passing or returning such 6643 a class was impossible. 6644 + WARNING: In GCC 8.1 the second change mistakenly also affects 6645 classes with a deleted copy constructor and defaulted trivial 6646 move constructor (bug [22]c++/86094). This issue is fixed in 6647 GCC 8.2 (-fabi-version=13). 6648 You can test whether these changes affect your code with -Wabi=11 6649 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 6650 problematic for your project, the GCC 7 ABI can be selected with 6651 -fabi-version=11. 6652 * The value of the C++11 alignof operator has been corrected to match 6653 C _Alignof (minimum alignment) rather than GNU __alignof__ 6654 (preferred alignment); on ia32 targets this means that 6655 alignof(double) is now 4 rather than 8. Code that wants the 6656 preferred alignment should use __alignof__ instead. 6657 * New command-line options have been added for the C++ compiler to 6658 control warnings: 6659 + [23]-Wclass-memaccess warns when objects of non-trivial class 6660 types are manipulated in potentially unsafe ways by raw memory 6661 functions such as memcpy, or realloc. The warning helps detect 6662 calls that bypass user-defined constructors or copy-assignment 6663 operators, corrupt virtual table pointers, data members of 6664 const-qualified types or references, or member pointers. The 6665 warning also detects calls that would bypass access controls 6666 to data members. For example, a call such as: 6667 memcpy (&std::cout, &std::cerr, sizeof std::cout); 6668 results in 6669 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 6670 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 6671 o trivial copy-assignment [-Wclass-memaccess] 6672 The -Wclass-memaccess option is included in -Wall. 6673 * The C++ front end has experimental support for some of the upcoming 6674 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 6675 including designated initializers, default member initializers for 6676 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 6677 lambda [=, this] captures, etc. For a full list of new features, 6678 see [24]the C++ status page. 6679 * When reporting on attempts to access private fields of a class or 6680 struct, the C++ compiler will now offer fix-it hints showing how to 6681 use an accessor function to get at the field in question, if one 6682 exists. 6683 $ gcc accessor.cc 6684 accessor.cc: In function 'void test(foo*)': 6685 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 6686 if (ptr->m_ratio >= 0.5) 6687 ^~~~~~~ 6688 accessor.cc:7:10: note: declared private here 6689 double m_ratio; 6690 ^~~~~~~ 6691 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 6692 foo::get_ratio() const' 6693 if (ptr->m_ratio >= 0.5) 6694 ^~~~~~~ 6695 get_ratio() 6696 6697 * The C++ compiler can now give you a hint if you use a macro before 6698 it was defined (e.g. if you mess up the order of your #include 6699 directives): 6700 $ gcc ordering.cc 6701 ordering.cc:2:24: error: expected ';' at end of member declaration 6702 virtual void clone() const OVERRIDE { } 6703 ^~~~~ 6704 ; 6705 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 6706 virtual void clone() const OVERRIDE { } 6707 ^~~~~~~~ 6708 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 6709 In file included from ordering.cc:5: 6710 c++11-compat.h:2: note: it was later defined here 6711 #define OVERRIDE override 6712 6713 6714 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 6715 you when you can use a static_cast, const_cast, or 6716 reinterpret_cast. 6717 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 6718 old-style-cast-fixits.cc: In function 'void test(void*)': 6719 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 6720 -Wold-style-cast] 6721 foo *f = (foo *)ptr; 6722 ^~~ 6723 ---------- 6724 static_cast<foo *> (ptr) 6725 6726 * When reporting on problems within extern "C" linkage 6727 specifications, the C++ compiler will now display the location of 6728 the start of the extern "C". 6729 $ gcc -c extern-c.cc 6730 extern-c.cc:3:1: error: template with C linkage 6731 template <typename T> void test (void); 6732 ^~~~~~~~ 6733 In file included from extern-c.cc:1: 6734 unclosed.h:1:1: note: 'extern "C"' linkage started here 6735 extern "C" { 6736 ^~~~~~~~~~ 6737 extern-c.cc:3:39: error: expected '}' at end of input 6738 template <typename T> void test (void); 6739 ^ 6740 In file included from extern-c.cc:1: 6741 unclosed.h:1:12: note: to match this '{' 6742 extern "C" { 6743 ^ 6744 6745 * When reporting on mismatching template types, the C++ compiler will 6746 now use color to highlight the mismatching parts of the template, 6747 and will elide the parameters that are common between two 6748 mismatching templates, printing [...] instead: 6749 $ gcc templates.cc 6750 templates.cc: In function 'void test()': 6751 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 6752 e>' to 'vector<int>' 6753 fn_1(vector<double> ()); 6754 ^~~~~~~~~~~~~~~~~ 6755 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 6756 ,double>' to 'map<[...],int>' 6757 fn_2(map<int, double>()); 6758 ^~~~~~~~~~~~~~~~~~ 6759 6760 Those [...] elided parameters can be seen using -fno-elide-type: 6761 $ gcc templates.cc -fno-elide-type 6762 templates.cc: In function 'void test()': 6763 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 6764 e>' to 'vector<int>' 6765 fn_1(vector<double> ()); 6766 ^~~~~~~~~~~~~~~~~ 6767 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 6768 ouble>' to 'map<int,int>' 6769 fn_2(map<int, double>()); 6770 ^~~~~~~~~~~~~~~~~~ 6771 6772 The C++ compiler has also gained an option 6773 -fdiagnostics-show-template-tree which visualizes such mismatching 6774 templates in a hierarchical form: 6775 $ gcc templates-2.cc -fdiagnostics-show-template-tree 6776 templates-2.cc: In function 'void test()': 6777 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 6778 ble>' to 'vector<int>' 6779 vector< 6780 [double != int]> 6781 fn_1(vector<double> ()); 6782 ^~~~~~~~~~~~~~~~~ 6783 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 6784 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 6785 ap<[...],vector<float>>,vector<float>>' 6786 map< 6787 map< 6788 [...], 6789 vector< 6790 [double != float]>>, 6791 vector< 6792 [double != float]>> 6793 fn_2(map<map<int, vector<double>>, vector<double>> ()); 6794 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 6795 6796 which again works with -fno-elide-type: 6797 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 6798 templates-2.cc: In function 'void test()': 6799 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 6800 ble>' to 'vector<int>' 6801 vector< 6802 [double != int]> 6803 fn_1(vector<double> ()); 6804 ^~~~~~~~~~~~~~~~~ 6805 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 6806 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 6807 <int,vector<float>>,vector<float>>' 6808 map< 6809 map< 6810 int, 6811 vector< 6812 [double != float]>>, 6813 vector< 6814 [double != float]>> 6815 fn_2(map<map<int, vector<double>>, vector<double>> ()); 6816 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 6817 6818 * Flowing off the end of a non-void function is considered 6819 unreachable and may be subject to optimization on that basis. As a 6820 result of this change, -Wreturn-type warnings are enabled by 6821 default for C++. 6822 6823 [25]Runtime Library (libstdc++) 6824 6825 * Improved experimental support for C++17, including the following 6826 features: 6827 + Deduction guides to support class template argument deduction. 6828 + std::filesystem implementation. 6829 + std::char_traits<char> and std::char_traits<wchar_t> are 6830 usable in constant expressions. 6831 + std::to_chars and std::from_chars (for integers only, not for 6832 floating point types). 6833 * Experimental support for C++2a: std::to_address (thanks to Glen 6834 Fernandes) and std::endian. 6835 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 6836 entropy count for the random device, if known (thanks to Xi 6837 Ruoyao). 6838 * Support for std::experimental::source_location. 6839 * AddressSanitizer integration for std::vector, detecting 6840 out-of-range accesses to the unused capacity of a vector. 6841 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 6842 Mathematical Special Functions. 6843 6844 [26]Fortran 6845 6846 * The main version of libfortran has been changed to 5. 6847 * Parameterized derived types, a major feature of Fortran 2003, have 6848 been implemented. 6849 * Partial support is provided for Fortran 2018 teams, which are 6850 hierarchical subsets of images that execute independently of other 6851 image subsets. 6852 * The maximum rank for arrays has been increased to 15, conforming to 6853 the Fortran 2008 standard. 6854 * Transformational intrinsics are now fully supported in 6855 initialization expressions. 6856 * New flag -fc-prototypes to write C prototypes for BIND(C) 6857 procedures and variables. 6858 * If -fmax-stack-var-size is honored if given together with -Ofast, 6859 -fstack-arrays is no longer set in that case. 6860 * New options -fdefault-real-16 and -fdefault-real-10 to control the 6861 default kind of REAL variables. 6862 * A warning is now issued if an array subscript inside a DO loop 6863 could lead to an out-of-bounds-access. The new option 6864 -Wdo-subscript, enabled by -Wextra, warns about this even if the 6865 compiler can not prove that the code will be executed. 6866 * The Fortran front end now attempts to interchange loops if it is 6867 deemed profitable. So far, this is restricted to FORALL and DO 6868 CONCURRENT statements with multiple indices. This behavior be 6869 controlled with the new flag -ffrontend-loop-interchange, which is 6870 enabled with optimization by default. The 6871 -Wfrontend-loop-interchange option warns about such occurrences. 6872 * When an actual argument contains too few elements for a dummy 6873 argument, an error is now issued. The -std=legacy option can be 6874 used to still compile such code. 6875 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 6876 integers, making records larger than 2GiB possible. 6877 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 6878 effect. The record length for preconnected units is now larger than 6879 any practical limit, same as for sequential access units opened 6880 without an explicit RECL= specifier. 6881 * Character variables longer than HUGE(0) elements are now possible 6882 on 64-bit targets. Note that this changes the procedure call ABI 6883 for all procedures with character arguments on 64-bit targets, as 6884 the type of the hidden character length argument has changed. The 6885 hidden character length argument is now of type INTEGER(C_SIZE_T). 6886 * Partial support is provided for Fortran 2018 teams, which are 6887 hierarchical subsets of images that execute independently of other 6888 image subsets. 6889 6890 [27]Go 6891 6892 * GCC 8 provides a complete implementation of the Go 1.10.1 user 6893 packages. 6894 * The garbage collector is now fully concurrent. As before, values 6895 stored on the stack are scanned conservatively, but value stored in 6896 the heap are scanned precisely. 6897 * Escape analysis is fully implemented and enabled by default in the 6898 Go front end. This significantly reduces the number of heap 6899 allocations by allocating values on the stack instead. 6900 6901 [28]libgccjit 6902 6903 The libgccjit API gained four new entry points: 6904 * [29]gcc_jit_type_get_vector and 6905 * [30]gcc_jit_context_new_rvalue_from_vector for working with 6906 vectors, 6907 * [31]gcc_jit_type_get_aligned 6908 * [32]gcc_jit_function_get_address 6909 6910 The C code generated by [33]gcc_jit_context_dump_reproducer_to_file is 6911 now easier-to-read. 6912 6913 [34]New Targets and Target Specific Improvements 6914 6915 [35]AArch64 6916 6917 * The Armv8.4-A architecture is now supported. It can be used by 6918 specifying the -march=armv8.4-a option. 6919 * The Dot Product instructions are now supported as an optional 6920 extension to the Armv8.2-A architecture and newer and are mandatory 6921 on Armv8.4-A. The extension can be used by specifying the +dotprod 6922 architecture extension. E.g. -march=armv8.2-a+dotprod. 6923 * The Armv8-A +crypto extension has now been split into two 6924 extensions for finer grained control: 6925 + +aes which contains the Armv8-A AES crytographic instructions. 6926 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 6927 instructions. 6928 Using +crypto will now enable these two extensions. 6929 * New Armv8.4-A FP16 Floating Point Multiplication Variant 6930 instructions have been added. These instructions are mandatory in 6931 Armv8.4-A but available as an optional extension to Armv8.2-A and 6932 Armv8.3-A. The new extension can be used by specifying the +fp16fml 6933 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 6934 the instructions can be enabled by specifying +fp16. 6935 * New cryptographic instructions have been added as optional 6936 extensions to Armv8.2-A and newer. These instructions can be 6937 enabled with: 6938 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 6939 implies +sha2. 6940 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 6941 * The Scalable Vector Extension (SVE) is now supported as an optional 6942 extension to the Armv8.2-A architecture and newer. This support 6943 includes automatic vectorization with SVE instructions, but it does 6944 not yet include the SVE Arm C Language Extensions (ACLE). It can be 6945 enabled by specifying the +sve architecture extension (for example, 6946 -march=armv8.2-a+sve). By default, the generated code works with 6947 all vector lengths, but it can be made specific to N-bit vectors 6948 using -msve-vector-bits=N. 6949 * Support has been added for the following processors (GCC 6950 identifiers in parentheses): 6951 + Arm Cortex-A75 (cortex-a75). 6952 + Arm Cortex-A55 (cortex-a55). 6953 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 6954 (cortex-a75.cortex-a55). 6955 The GCC identifiers can be used as arguments to the -mcpu or -mtune 6956 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 6957 arguments to the equivalent target attributes and pragmas. 6958 6959 [36]ARC 6960 6961 * Added support for: 6962 + Fast interrupts. 6963 + Naked functions. 6964 + aux variable attributes. 6965 + uncached type qualifier. 6966 + Secure functions via sjli instruction. 6967 * New exception handling implementation. 6968 * Revamped trampoline implementation. 6969 * Refactored small data feature implementation, controlled via the -G 6970 command-line option. 6971 * New support for reduced register set ARC architecture 6972 configurations, controlled via the -mrf16 command-line option. 6973 * Refurbished and improved support for zero overhead loops. 6974 Introduced -mlpc-width command-line option to control the width of 6975 the lp_count register. 6976 6977 [37]ARM 6978 6979 * The -mfpu option now takes a new option setting of -mfpu=auto. When 6980 set to this the floating-point and SIMD settings are derived from 6981 the settings of the -mcpu or -march options. The internal CPU 6982 configurations have been updated with information about the 6983 permitted floating-point configurations supported. See the user 6984 guide for further information about the extended option syntax for 6985 controlling architectural extensions via the -march option. 6986 -mfpu=auto is now the default setting unless the compiler has been 6987 configured with an explicit --with-fpu option. 6988 * The -march and -mcpu options now accept optional extensions to the 6989 architecture or CPU option, allowing the user to enable or disable 6990 any such extensions supported by that architecture or CPU such as 6991 (but not limited to) floating-point and AdvancedSIMD. For example: 6992 the option -mcpu=cortex-a53+nofp will generate code for the 6993 Cortex-A53 processor with no floating-point support. This, in 6994 combination with the new -mfpu=auto option, provides a 6995 straightforward way of specifying a valid build target through a 6996 single -mcpu or -march option. The -mtune option accepts the same 6997 arguments as -mcpu but only the CPU name has an effect on tuning. 6998 The architecture extensions do not have any effect. For details of 6999 what extensions a particular architecture or CPU option supports 7000 please refer to the [38]documentation. 7001 * The -mstructure-size-boundary option has been deprecated and will 7002 be removed in a future release. 7003 * The default link behavior for Armv6 and Armv7-R targets has been 7004 changed to produce BE8 format when generating big-endian images. A 7005 new flag -mbe32 can be used to force the linker to produce legacy 7006 BE32 format images. There is no change of behavior for Armv6-M and 7007 other Armv7 or later targets: these already defaulted to BE8 7008 format. This change brings GCC into alignment with other compilers 7009 for the ARM architecture. 7010 * The Armv8-R architecture is now supported. It can be used by 7011 specifying the -march=armv8-r option. 7012 * The Armv8.3-A architecture is now supported. It can be used by 7013 specifying the -march=armv8.3-a option. 7014 * The Armv8.4-A architecture is now supported. It can be used by 7015 specifying the -march=armv8.4-a option. 7016 * The Dot Product instructions are now supported as an optional 7017 extension to the Armv8.2-A architecture and newer and are mandatory 7018 on Armv8.4-A. The extension can be used by specifying the +dotprod 7019 architecture extension. E.g. -march=armv8.2-a+dotprod. 7020 * Support for setting extensions and architectures using the GCC 7021 target pragma and attribute has been added. It can be used by 7022 specifying #pragma GCC target ("arch=..."), #pragma GCC target 7023 ("+extension"), __attribute__((target("arch=..."))) or 7024 __attribute__((target("+extension"))). 7025 * New Armv8.4-A FP16 Floating Point Multiplication Variant 7026 instructions have been added. These instructions are mandatory in 7027 Armv8.4-A but available as an optional extension to Armv8.2-A and 7028 Armv8.3-A. The new extension can be used by specifying the +fp16fml 7029 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 7030 the instructions can be enabled by specifying +fp16. 7031 * Support has been added for the following processors (GCC 7032 identifiers in parentheses): 7033 + Arm Cortex-A75 (cortex-a75). 7034 + Arm Cortex-A55 (cortex-a55). 7035 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 7036 (cortex-a75.cortex-a55). 7037 + Arm Cortex-R52 for Armv8-R (cortex-r52). 7038 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7039 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 7040 arguments to the equivalent target attributes and pragmas. 7041 7042 [39]AVR 7043 7044 * The AVR port now supports the following XMEGA-like devices: 7045 7046 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 7047 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 7048 ATtiny3214, ATtiny3216, ATtiny3217 7049 The new devices are listed under [40]-mmcu=avrxmega3. 7050 + These devices see flash memory in the RAM address space, so 7051 that features like PROGMEM and __flash are not needed any more 7052 (as opposed to other AVR families for which read-only data 7053 will be located in RAM except special, non-standard features 7054 are used to locate and access such data). This requires that 7055 the compiler is used with Binutils 2.29 or newer so that 7056 [41]read-only data will be located in flash memory. 7057 + A new command-line option -mshort-calls is supported. This 7058 option is used internally for multilib selection of the 7059 avrxmega3 variants. It is not an optimization option. Do not 7060 set it by hand. 7061 * The compiler now generates [42]efficient interrupt service routine 7062 (ISR) prologues and epilogues. This is achieved by using the new 7063 [43]AVR pseudo instruction __gcc_isr which is supported and 7064 resolved by the GNU assembler. 7065 + As the __gcc_isr pseudo-instruction will be resolved by the 7066 assembler, inline assembly is transparent to the process. This 7067 means that when inline assembly uses an instruction like INC 7068 that clobbers the condition code, then the assembler will 7069 detect this and generate an appropriate ISR prologue / 7070 epilogue chunk to save / restore SREG as needed. 7071 + A new command-line option -mno-gas-isr-prologues disables the 7072 generation of the __gcc_isr pseudo instruction. Any non-naked 7073 ISR will save and restore SREG, tmp_reg and zero_reg, no 7074 matter whether the respective register is clobbered or used. 7075 + The feature is turned on per default for all optimization 7076 levels except for -O0 and -Og. It is explicitly enabled by 7077 means of option -mgas-isr-prologues. 7078 + Support has been added for a new [44]AVR function attribute 7079 no_gccisr. It can be used to disable __gcc_isr pseudo 7080 instruction generation for individual ISRs. 7081 + This optimization is only available if GCC is configured with 7082 GNU Binutils 2.29 or newer; or at least with a version of 7083 Binutils that implements feature [45]PR21683. 7084 * The compiler no more saves / restores registers in main; the effect 7085 is the same as if attribute OS_task was specified for main. This 7086 optimization can be switched off by the new command-line option 7087 -mno-main-is-OS_task. 7088 7089 [46]IA-32/x86-64 7090 7091 * The x86 port now supports the naked function attribute. 7092 * Better tuning for znver1 and Intel Core based CPUs. 7093 * Vectorization cost metrics has been reworked leading to significant 7094 improvements on some benchmarks. 7095 * GCC now supports the Intel CPU named Cannonlake through 7096 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 7097 and SHA ISA extensions. 7098 * GCC now supports the Intel CPU named Icelake through 7099 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 7100 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 7101 ISA extensions. 7102 * GCC now supports the Intel Control-flow Enforcement Technology 7103 (CET) extension through -fcf-protection option. 7104 7105 [47]NDS32 7106 7107 * New command-line options -mext-perf, -mext-perf2, and -mext-string 7108 have been added for performance extension instructions. 7109 7110 [48]Nios II 7111 7112 * The Nios II back end has been improved to generate better-optimized 7113 code. Changes include switching to LRA, more accurate cost models, 7114 and more compact code for addressing static variables. 7115 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 7116 added. 7117 * The stack-smashing protection options are now enabled on Nios II. 7118 7119 [49]PA-RISC 7120 7121 * The default call ABI on 32-bit linux has been changed from callee 7122 copies to caller copies. This affects objects larger than eight 7123 bytes passed by value. The goal is to improve compatibility with 7124 x86 and resolve issues with OpenMP. 7125 * Other PA-RISC targets are unchanged. 7126 7127 [50]PowerPC / PowerPC64 / RS6000 7128 7129 * The PowerPC SPE support is split off to a separate powerpcspe port. 7130 The separate port is deprecated and might be removed in a future 7131 release. 7132 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 7133 powerpc*-*-linux*paired*) is deprecated and will be removed in a 7134 future release. 7135 * The Xilinx floating point support (-mxilinx-fpu, 7136 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 7137 release. 7138 * Support for using big-endian AltiVec intrinsics on a little-endian 7139 target (-maltivec=be) is deprecated and will be removed in a future 7140 release. 7141 7142 [51]Tile 7143 7144 * The TILE-Gx port is deprecated and will be removed in a future 7145 release. 7146 7147 [52]Operating Systems 7148 7149 [53]Windows 7150 7151 * GCC on Microsoft Windows can now be configured via 7152 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 7153 specific behavior for GCC itself with regards to supporting the 7154 wildcard character. Prior versions of GCC would follow the 7155 configuration of the MinGW runtime. This behavior can still be 7156 obtained by not using the above options or by using 7157 --enable-mingw-wildcard=platform. 7158 7159 [54]Improvements for plugin authors 7160 7161 * Plugins can now register a callback hook for when comments are 7162 encountered by the C and C++ compilers, e.g. allowing for plugins 7163 to handle documentation markup in code comments. 7164 * The gdbinit support script for debugging GCC now has a 7165 break-on-diagnostic command, providing an easy way to trigger a 7166 breakpoint whenever a diagnostic is emitted. 7167 * The API for creating fix-it hints now supports newlines, and for 7168 emitting mutually incompatible fix-it hints for one diagnostic. 7169 7170 [55]GCC 8.1 7171 7172 This is the [56]list of problem reports (PRs) from GCC's bug tracking 7173 system that are known to be fixed in the 8.1 release. This list might 7174 not be complete (that is, it is possible that some PRs that have been 7175 fixed are not listed here). 7176 7177 [57]GCC 8.2 7178 7179 This is the [58]list of problem reports (PRs) from GCC's bug tracking 7180 system that are known to be fixed in the 8.2 release. This list might 7181 not be complete (that is, it is possible that some PRs that have been 7182 fixed are not listed here). 7183 7184 General Improvements 7185 7186 * Fixed LTO link-time performance problems caused by an overflow in 7187 the partitioning algorithm while building large binaries. 7188 7189 Language Specific Changes 7190 7191 C++ 7192 7193 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 7194 returning of classes with a deleted copy constructor and defaulted 7195 trivial move constructor (bug [59]c++/86094). GCC 8.2 introduces 7196 -fabi-version=13 and makes it the default, ABI incompatibilities 7197 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [60]C++ 7198 changes for more details. 7199 7200 Target Specific Changes 7201 7202 IA-32/x86-64 7203 7204 * -mtune=native performance regression [61]PR84413 on Intel Skylake 7205 processors has been fixed. 7206 7207 [62]GCC 8.3 7208 7209 This is the [63]list of problem reports (PRs) from GCC's bug tracking 7210 system that are known to be fixed in the 8.3 release. This list might 7211 not be complete (that is, it is possible that some PRs that have been 7212 fixed are not listed here). 7213 7214 Windows 7215 7216 * A C++ Microsoft ABI bitfield layout bug, [64]PR87137 has been 7217 fixed. A non-field declaration could cause the current bitfield 7218 allocation unit to be completed, incorrectly placing a following 7219 bitfield into a new allocation unit. The Microsoft ABI is selected 7220 for: 7221 + Mingw targets 7222 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 7223 option is specified, or __attribute__((ms_struct)) is used 7224 + SuperH targets when the -mhitachi option is specified, or 7225 __attribute__((renesas)) is used 7226 GCC 8 introduced additional cases of this defect, but rather than 7227 resolve only those regressions, we decided to resolve all the cases 7228 of this defect in single change. 7229 7230 [65]GCC 8.4 7231 7232 This is the [66]list of problem reports (PRs) from GCC's bug tracking 7233 system that are known to be fixed in the 8.4 release. This list might 7234 not be complete (that is, it is possible that some PRs that have been 7235 fixed are not listed here). 7236 7237 [67]GCC 8.5 7238 7239 This is the [68]list of problem reports (PRs) from GCC's bug tracking 7240 system that are known to be fixed in the 8.5 release. This list might 7241 not be complete (that is, it is possible that some PRs that have been 7242 fixed are not listed here). 7243 7244 Target Specific Changes 7245 7246 AArch64 7247 7248 * The option -moutline-atomics has been added to aid deployment of 7249 the Large System Extensions (LSE) on GNU/Linux systems built with a 7250 baseline architecture targeting Armv8-A. When the option is 7251 specified code is emitted to detect the presence of LSE 7252 instructions at run time and use them for standard atomic 7253 operations. For more information please refer to the documentation. 7254 7255 7256 For questions related to the use of GCC, please consult these web 7257 pages and the [69]GCC manuals. If that fails, the 7258 [70]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7259 web pages and the development of GCC are welcome on our developer 7260 list at [71]gcc (a] gcc.gnu.org. All of [72]our lists have public 7261 archives. 7262 7263 Copyright (C) [73]Free Software Foundation, Inc. Verbatim copying and 7264 distribution of this entire article is permitted in any medium, 7265 provided this notice is preserved. 7266 7267 These pages are [74]maintained by the GCC team. Last modified 7268 2025-01-31. 7269 7270 References 7271 7272 1. https://gcc.gnu.org/gcc-8/porting_to.html 7273 2. https://gcc.gnu.org/onlinedocs/index.html#current 7274 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 7275 4. https://gcc.gnu.org/gcc-8/changes.html#general 7276 5. https://gcc.gnu.org/gcc-8/changes.html#languages 7277 6. https://gcc.gnu.org/gcc-8/changes.html#ada 7278 7. https://gcc.gnu.org/gcc-8/changes.html#brig 7279 8. https://gcc.gnu.org/gcc-8/changes.html#c-family 7280 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 7281 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 7282 11. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 7283 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 7284 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 7285 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 7286 15. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 7287 16. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 7288 17. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 7289 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 7290 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 7291 20. https://gcc.gnu.org/gcc-8/changes.html#c 7292 21. https://gcc.gnu.org/gcc-8/changes.html#cxx 7293 22. https://gcc.gnu.org/PR86094 7294 23. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 7295 24. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 7296 25. https://gcc.gnu.org/gcc-8/changes.html#libstdcxx 7297 26. https://gcc.gnu.org/gcc-8/changes.html#fortran 7298 27. https://gcc.gnu.org/gcc-8/changes.html#go 7299 28. https://gcc.gnu.org/gcc-8/changes.html#jit 7300 29. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 7301 30. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 7302 31. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 7303 32. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 7304 33. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 7305 34. https://gcc.gnu.org/gcc-8/changes.html#targets 7306 35. https://gcc.gnu.org/gcc-8/changes.html#aarch64 7307 36. https://gcc.gnu.org/gcc-8/changes.html#arc 7308 37. https://gcc.gnu.org/gcc-8/changes.html#arm 7309 38. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 7310 39. https://gcc.gnu.org/gcc-8/changes.html#avr 7311 40. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 7312 41. https://sourceware.org/PR21472 7313 42. https://gcc.gnu.org/PR20296 7314 43. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 7315 44. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 7316 45. https://sourceware.org/PR21683 7317 46. https://gcc.gnu.org/gcc-8/changes.html#x86 7318 47. https://gcc.gnu.org/gcc-8/changes.html#nds32 7319 48. https://gcc.gnu.org/gcc-8/changes.html#nios2 7320 49. https://gcc.gnu.org/gcc-8/changes.html#hppa 7321 50. https://gcc.gnu.org/gcc-8/changes.html#powerpc 7322 51. https://gcc.gnu.org/gcc-8/changes.html#Tile 7323 52. https://gcc.gnu.org/gcc-8/changes.html#os 7324 53. https://gcc.gnu.org/gcc-8/changes.html#windows 7325 54. https://gcc.gnu.org/gcc-8/changes.html#plugins 7326 55. https://gcc.gnu.org/gcc-8/changes.html#GCC8.1 7327 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 7328 57. https://gcc.gnu.org/gcc-8/changes.html#GCC8.2 7329 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 7330 59. https://gcc.gnu.org/PR86094 7331 60. https://gcc.gnu.org/gcc-8/changes.html#cxx 7332 61. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 7333 62. https://gcc.gnu.org/gcc-8/changes.html#GCC8.3 7334 63. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 7335 64. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 7336 65. https://gcc.gnu.org/gcc-8/changes.html#GCC8.4 7337 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 7338 67. https://gcc.gnu.org/gcc-8/changes.html#GCC8.5 7339 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5 7340 69. https://gcc.gnu.org/onlinedocs/ 7341 70. mailto:gcc-help (a] gcc.gnu.org 7342 71. mailto:gcc (a] gcc.gnu.org 7343 72. https://gcc.gnu.org/lists.html 7344 73. https://www.fsf.org/ 7345 74. https://gcc.gnu.org/about.html 7346 ====================================================================== 7347 http://gcc.gnu.org/gcc-7/index.html 7348 7349 GCC 7 Release Series 7350 7351 (This release series is no longer supported.) 7352 7353 Nov 14, 2019 7354 7355 The [1]GNU project and the GCC developers are pleased to announce the 7356 release of GCC 7.5. 7357 7358 This release is a bug-fix release, containing fixes for regressions in 7359 GCC 7.4 relative to previous releases of GCC. 7360 7361 Release History 7362 7363 GCC 7.5 7364 Nov 14, 2019 ([2]changes, [3]documentation) 7365 7366 GCC 7.4 7367 Dec 6, 2018 ([4]changes, [5]documentation) 7368 7369 GCC 7.3 7370 Jan 25, 2018 ([6]changes, [7]documentation) 7371 7372 GCC 7.2 7373 Aug 14, 2017 ([8]changes, [9]documentation) 7374 7375 GCC 7.1 7376 May 2, 2017 ([10]changes, [11]documentation) 7377 7378 References and Acknowledgements 7379 7380 GCC used to stand for the GNU C Compiler, but since the compiler 7381 supports several other languages aside from C, it now stands for the 7382 GNU Compiler Collection. 7383 7384 A list of [12]successful builds is updated as new information becomes 7385 available. 7386 7387 The GCC developers would like to thank the numerous people that have 7388 contributed new features, improvements, bug fixes, and other changes as 7389 well as test results to GCC. This [13]amazing group of volunteers is 7390 what makes GCC successful. 7391 7392 For additional information about GCC please refer to the [14]GCC 7393 project web site or contact the [15]GCC development mailing list. 7394 7395 To obtain GCC please use [16]our mirror sites or [17]our version 7396 control system. 7397 7398 7399 For questions related to the use of GCC, please consult these web 7400 pages and the [18]GCC manuals. If that fails, the 7401 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7402 web pages and the development of GCC are welcome on our developer 7403 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 7404 archives. 7405 7406 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 7407 distribution of this entire article is permitted in any medium, 7408 provided this notice is preserved. 7409 7410 These pages are [23]maintained by the GCC team. Last modified 7411 2024-05-30. 7412 7413 References 7414 7415 1. http://www.gnu.org/ 7416 2. https://gcc.gnu.org/gcc-7/changes.html 7417 3. https://gcc.gnu.org/onlinedocs/7.5.0/ 7418 4. https://gcc.gnu.org/gcc-7/changes.html 7419 5. https://gcc.gnu.org/onlinedocs/7.4.0/ 7420 6. https://gcc.gnu.org/gcc-7/changes.html 7421 7. https://gcc.gnu.org/onlinedocs/7.3.0/ 7422 8. https://gcc.gnu.org/gcc-7/changes.html 7423 9. https://gcc.gnu.org/onlinedocs/7.2.0/ 7424 10. https://gcc.gnu.org/gcc-7/changes.html 7425 11. https://gcc.gnu.org/onlinedocs/7.1.0/ 7426 12. https://gcc.gnu.org/gcc-7/buildstat.html 7427 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7428 14. https://gcc.gnu.org/index.html 7429 15. mailto:gcc (a] gcc.gnu.org 7430 16. https://gcc.gnu.org/mirrors.html 7431 17. https://gcc.gnu.org/git.html 7432 18. https://gcc.gnu.org/onlinedocs/ 7433 19. mailto:gcc-help (a] gcc.gnu.org 7434 20. mailto:gcc (a] gcc.gnu.org 7435 21. https://gcc.gnu.org/lists.html 7436 22. https://www.fsf.org/ 7437 23. https://gcc.gnu.org/about.html 7438 ====================================================================== 7439 http://gcc.gnu.org/gcc-7/changes.html 7440 7441 GCC 7 Release Series 7442 Changes, New Features, and Fixes 7443 7444 This page is a brief summary of some of the huge number of improvements 7445 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 7446 the [2]full GCC documentation. 7447 7448 Caveats 7449 7450 * GCC now uses [3]LRA (a new local register allocator) by default for 7451 new targets. 7452 * The non-standard C++0x type traits has_trivial_default_constructor, 7453 has_trivial_copy_constructor and has_trivial_copy_assign have been 7454 removed. 7455 * The libstdc++ [4]Profile Mode has been deprecated and will be 7456 removed in a future version. 7457 * The Cilk+ extensions to the C and C++ languages have been 7458 deprecated. 7459 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 7460 affects conformance to the procedure call standard (AAPCS) has been 7461 fixed. The bug affects some C++ code where class objects are passed 7462 by value to functions and could result in incorrect or inconsistent 7463 code being generated. This is an ABI change. If the option -Wpsabi 7464 is enabled (on by default) the compiler will emit a diagnostic note 7465 for code that might be affected. 7466 7467 [6]General Optimizer Improvements 7468 7469 * GCC 7 can determine the return value or range of return values of 7470 some calls to the sprintf family of functions and make it available 7471 to other optimization passes. Some calls to the snprintf function 7472 with a zero size argument can be folded into constants. This 7473 optimization is included in -O1 and can be selectively controlled 7474 by the -fprintf-return-value option. 7475 * A new store merging pass has been added. It merges constant stores 7476 to adjacent memory locations into fewer, wider, stores. It is 7477 enabled by the -fstore-merging option and at the -O2 optimization 7478 level or higher (and -Os). 7479 * A new code hoisting optimization has been added to the partial 7480 redundancy elimination pass. It attempts to move evaluation of 7481 expressions executed on all paths to the function exit as early as 7482 possible. This primarily helps improve code size, but can improve 7483 the speed of the generated code as well. It is enabled by the 7484 -fcode-hoisting option and at the -O2 optimization level or higher 7485 (and -Os). 7486 * A new interprocedural bitwise constant propagation optimization has 7487 been added, which propagates knowledge about which bits of 7488 variables are known to be zero (including pointer alignment 7489 information) across the call graph. It is enabled by the 7490 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 7491 at the -O2 optimization level and higher (and -Os). This 7492 optimization supersedes interprocedural alignment propagation of 7493 GCC 6, and therefore the option -fipa-cp-alignment is now 7494 deprecated and ignored. 7495 * A new interprocedural value range propagation optimization has been 7496 added, which propagates integral range information across the call 7497 graph when variable values can be proven to be within those ranges. 7498 It is enabled by the -fipa-vrp option and at the -O2 optimization 7499 level and higher (and -Os). 7500 * A new loop splitting optimization pass has been added. Certain 7501 loops which contain a condition that is always true on one side of 7502 the iteration space and always false on the other are split into 7503 two loops, such that each of the two new loops iterates on just one 7504 side of the iteration space and the condition does not need to be 7505 checked inside of the loop. It is enabled by the -fsplit-loops 7506 option and at the -O3 optimization level or higher. 7507 * The shrink-wrapping optimization can now separate portions of 7508 prologues and epilogues to improve performance if some of the work 7509 done traditionally by prologues and epilogues is not needed on 7510 certain paths. This is controlled by the -fshrink-wrap-separate 7511 option, enabled by default. It requires target support, which is 7512 currently only implemented in the PowerPC and AArch64 ports. 7513 * AddressSanitizer gained a new sanitization option, 7514 -fsanitize-address-use-after-scope, which enables sanitization of 7515 variables whose address is taken and used after a scope where the 7516 variable is defined: 7517 7518 int 7519 main (int argc, char **argv) 7520 { 7521 char *ptr; 7522 { 7523 char my_char; 7524 ptr = &my_char; 7525 } 7526 7527 *ptr = 123; 7528 return *ptr; 7529 } 7530 7531 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 7532 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 7533 WRITE of size 1 at 0x7fffb8dba990 thread T0 7534 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 7535 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 7536 #2 0x400739 in _start (/tmp/a.out+0x400739) 7537 7538 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 7539 #0 0x40067f in main /tmp/use-after-scope-1.c:3 7540 7541 This frame has 1 object(s): 7542 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 7543 7544 The option is enabled by default with -fsanitize=address and 7545 disabled by default with -fsanitize=kernel-address. Compared to the 7546 LLVM compiler, where the option already exists, the implementation 7547 in the GCC compiler has some improvements and advantages: 7548 + Complex uses of gotos and case labels are properly handled and 7549 should not report any false positive or false negatives. 7550 + C++ temporaries are sanitized. 7551 + Sanitization can handle invalid memory stores that are 7552 optimized out by the LLVM compiler when optimization is 7553 enabled. 7554 * The -fsanitize=signed-integer-overflow suboption of the 7555 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 7556 on arithmetic operations with generic vectors. 7557 * Version 5 of the DWARF debugging information standard is supported 7558 through the -gdwarf-5 option. The DWARF version 4 debugging 7559 information remains the default until consumers of debugging 7560 information are adjusted. 7561 7562 [7]New Languages and Language specific improvements 7563 7564 OpenACC support in C, C++, and Fortran continues to be maintained and 7565 improved. See the [8]OpenACC and [9]Offloading wiki pages for further 7566 information. 7567 7568 [10]Ada 7569 7570 * On mainstream native platforms, Ada programs no longer require the 7571 stack to be made executable in order to run properly. 7572 7573 [11]BRIG (HSAIL) 7574 7575 Support for processing BRIG 1.0 files was added in this release. BRIG 7576 is a binary format for HSAIL (Heterogeneous System Architecture 7577 Intermediate Language). The BRIG front end can be used for implementing 7578 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 7579 GCC-supported targets. An implementation of an HSAIL runtime library, 7580 libhsail-rt is also included. 7581 7582 [12]C family 7583 7584 * New command-line options have been added for the C and C++ 7585 compilers: 7586 + -Wimplicit-fallthrough warns when a switch case falls through. 7587 This warning has five different levels. The compiler is able 7588 to parse a wide range of fallthrough comments, depending on 7589 the level. It also handles control-flow statements, such as 7590 ifs. It's possible to suppress the warning by either adding a 7591 fallthrough comment, or by using a null statement: 7592 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 7593 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 7594 is enabled by -Wextra. 7595 + -Wpointer-compare warns when a pointer is compared with a zero 7596 character constant. Such code is now invalid in C++11 and GCC 7597 rejects it. This warning is enabled by default. 7598 + -Wduplicated-branches warns when an if-else has identical 7599 branches. 7600 + -Wrestrict warns when an argument passed to a 7601 restrict-qualified parameter aliases with another argument. 7602 + -Wmemset-elt-size warns for memset calls, when the first 7603 argument references an array, and the third argument is a 7604 number equal to the number of elements of the array, but not 7605 the size of the array. This warning is enabled by -Wall. 7606 + -Wint-in-bool-context warns about suspicious uses of integer 7607 values where boolean values are expected. This warning is 7608 enabled by -Wall. 7609 + -Wswitch-unreachable warns when a switch statement has 7610 statements between the controlling expression and the first 7611 case label which will never be executed. This warning is 7612 enabled by default. 7613 + -Wexpansion-to-defined warns when defined is used outside #if. 7614 This warning is enabled by -Wextra or -Wpedantic. 7615 + -Wregister warns about uses of the register storage specifier. 7616 In C++17 this keyword has been removed and for C++17 this is a 7617 pedantic warning enabled by default. The warning is not 7618 emitted for the GNU Explicit Register Variables extension. 7619 + -Wvla-larger-than=N warns about unbounded uses of 7620 variable-length arrays, and about bounded uses of 7621 variable-length arrays whose bound can be larger than N bytes. 7622 + -Wduplicate-decl-specifier warns when a declaration has 7623 duplicate const, volatile, restrict or _Atomic specifier. This 7624 warning is enabled by -Wall. 7625 * GCC 6's C and C++ front ends were able to offer suggestions for 7626 misspelled field names: 7627 7628 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 7629 you mean 'color'? 7630 return ptr->colour; 7631 ^~~~~~ 7632 7633 GCC 7 greatly expands the scope of these suggestions. Firstly, it 7634 adds fix-it hints to such suggestions: 7635 7636 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 7637 you mean 'color'? 7638 return ptr->colour; 7639 ^~~~~~ 7640 color 7641 7642 The suggestions now cover many other things, such as misspelled 7643 function names: 7644 7645 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 7646 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 7647 ion] 7648 gtk_widget_showall (w); 7649 ^~~~~~~~~~~~~~~~~~ 7650 gtk_widget_show_all 7651 7652 misspelled macro names and enum values: 7653 7654 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 7655 ction); did you mean 'MAX_ITEMS'? 7656 int array[MAX_ITEM]; 7657 ^~~~~~~~ 7658 MAX_ITEMS 7659 7660 misspelled type names: 7661 7662 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 7663 gned'? 7664 void test (singed char e); 7665 ^~~~~~ 7666 signed 7667 7668 and, in the C front end, named initializers: 7669 7670 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 7671 '? 7672 struct s test = { .colour = 3 }; 7673 ^~~~~~ 7674 color 7675 7676 * The preprocessor can now offer suggestions for misspelled 7677 directives, e.g.: 7678 7679 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 7680 #endfi 7681 ^~~~~ 7682 endif 7683 7684 * Warnings about format strings now underline the pertinent part of 7685 the string, and can offer suggested fixes. In some cases, the 7686 pertinent argument is underlined. 7687 7688 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 7689 nt 3 has type 'int' [-Wformat=] 7690 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 7691 ~^ ~~~~~ 7692 %d 7693 7694 * The new -Wdangling-else command-line option has been split out of 7695 -Wparentheses and warns about dangling else. 7696 * The -Wshadow warning has been split into three variants. 7697 -Wshadow=global warns for any shadowing. This is the default when 7698 using -Wshadow without any argument. -Wshadow=local only warns for 7699 a local variable shadowing another local variable or parameter. 7700 -Wshadow=compatible-local only warns for a local variable shadowing 7701 another local variable or parameter whose type is compatible (in 7702 C++ compatible means that the type of the shadowing variable can be 7703 converted to that of the shadowed variable). 7704 The following example shows the different kinds of shadow warnings: 7705 7706 enum operation { add, count }; 7707 struct container { int nr; }; 7708 7709 int 7710 container_count (struct container c, int count) 7711 { 7712 int r = 0; 7713 for (int count = 0; count > 0; count--) 7714 { 7715 struct container count = c; 7716 r += count.nr; 7717 } 7718 return r; 7719 } 7720 7721 -Wshadow=compatible-local will warn for the parameter being 7722 shadowed with the same type: 7723 7724 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 7725 compatible-local] 7726 for (int count = 0; count > 0; count--) 7727 ^~~~~ 7728 warn-test.c:5:42: note: shadowed declaration is here 7729 container_count (struct container c, int count) 7730 ^~~~~ 7731 7732 -Wshadow=local will warn for the above and for the shadowed 7733 declaration with incompatible type: 7734 7735 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 7736 hadow=local] 7737 struct container count = c; 7738 ^~~~~ 7739 warn-test.c:8:12: note: shadowed declaration is here 7740 for (int count = 0; count > 0; count--) 7741 ^~~~~ 7742 7743 -Wshadow=global will warn for all of the above and the shadowing of 7744 the global declaration: 7745 7746 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 7747 -Wshadow] 7748 container_count (struct container c, int count) 7749 ^~~~~ 7750 warn-test.c:1:23: note: shadowed declaration is here 7751 enum operation { add, count }; 7752 ^~~~~ 7753 7754 * GCC 7 contains a number of enhancements that help detect buffer 7755 overflow and other forms of invalid memory accesses. 7756 + The -Walloc-size-larger-than=size option detects calls to 7757 standard and user-defined memory allocation functions 7758 decorated with attribute alloc_size whose argument exceeds the 7759 specified size (PTRDIFF_MAX by default). The option also 7760 detects arithmetic overflow in the computation of the size in 7761 two-argument allocation functions like calloc where the total 7762 size is the product of the two arguments. Since calls with an 7763 excessive size cannot succeed they are typically the result of 7764 programming errors. Such bugs have been known to be the source 7765 of security vulnerabilities and a target of exploits. 7766 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 7767 For example, the following call to malloc incorrectly tries to 7768 avoid passing a negative argument to the function and instead 7769 ends up unconditionally invoking it with an argument less than 7770 or equal to zero. Since after conversion to the type of the 7771 argument of the function (size_t) a negative argument results 7772 in a value in excess of the maximum PTRDIFF_MAX the call is 7773 diagnosed. 7774 7775 void* f (int n) 7776 { 7777 return malloc (n > 0 ? 0 : n); 7778 } 7779 7780 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 7781 147483647 [-Walloc-size-larger-than=] 7782 7783 + The -Walloc-zero option detects calls to standard and 7784 user-defined memory allocation functions decorated with 7785 attribute alloc_size with a zero argument. -Walloc-zero is not 7786 included in either -Wall or -Wextra and must be explicitly 7787 enabled. 7788 + The -Walloca option detects all calls to the alloca function 7789 in the program. -Walloca is not included in either -Wall or 7790 -Wextra and must be explicitly enabled. 7791 + The -Walloca-larger-than=size option detects calls to the 7792 alloca function whose argument either may exceed the specified 7793 size, or that is not known to be sufficiently constrained to 7794 avoid exceeding it. -Walloca-larger-than is not included in 7795 either -Wall or -Wextra and must be explicitly enabled. 7796 For example, compiling the following snippet with 7797 -Walloca-larger-than=1024 results in a warning because even 7798 though the code appears to call alloca only with sizes of 1kb 7799 and less, since n is signed, a negative value would result in 7800 a call to the function well in excess of the limit. 7801 7802 void f (int n) 7803 { 7804 char *d; 7805 if (n < 1025) 7806 d = alloca (n); 7807 else 7808 d = malloc (n); 7809 ... 7810 } 7811 7812 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 7813 ong unsigned int' [-Walloca-larger-than=] 7814 7815 In contrast, a call to alloca that isn't bounded at all such 7816 as in the following function will elicit the warning below 7817 regardless of the size argument to the option. 7818 7819 void f (size_t n) 7820 { 7821 char *d = alloca (n); 7822 ... 7823 } 7824 7825 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 7826 7827 + The -Wformat-overflow=level option detects certain and likely 7828 buffer overflow in calls to the sprintf family of formatted 7829 output functions. Although the option is enabled even without 7830 optimization it works best with -O2 and higher. 7831 For example, in the following snippet the call to sprintf is 7832 diagnosed because even though its output has been constrained 7833 using the modulo operation it could result in as many as three 7834 bytes if mday were negative. The solution is to either 7835 allocate a larger buffer or make sure the argument is not 7836 negative, for example by changing mday's type to unsigned or 7837 by making the type of the second operand of the modulo 7838 expression unsigned: 100U. 7839 7840 void* f (int mday) 7841 { 7842 char *buf = malloc (3); 7843 sprintf (buf, "%02i", mday % 100); 7844 return buf; 7845 } 7846 7847 warning: 'sprintf may write a terminating nul past the end of the destination [- 7848 Wformat-overflow=] 7849 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 7850 7851 + The -Wformat-truncation=level option detects certain and 7852 likely output truncation in calls to the snprintf family of 7853 formatted output functions. -Wformat-truncation=1 is included 7854 in -Wall and enabled without optimization but works best with 7855 -O2 and higher. 7856 For example, the following function attempts to format an 7857 integer between 0 and 255 in hexadecimal, including the 0x 7858 prefix, into a buffer of four characters. But since the 7859 function must always terminate output by the null character 7860 ('\0') such a buffer is only big enough to fit just one digit 7861 plus the prefix. Therefore the snprintf call is diagnosed. To 7862 avoid the warning either use a bigger buffer or handle the 7863 function's return value which indicates whether or not its 7864 output has been truncated. 7865 7866 void f (unsigned x) 7867 { 7868 char d[4]; 7869 snprintf (d, sizeof d, "%#02x", x & 0xff); 7870 ... 7871 } 7872 7873 warning: 'snprintf' output may be truncated before the last format character [-W 7874 format-truncation=] 7875 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 7876 7877 + The -Wnonnull option has been enhanced to detect a broader set 7878 of cases of passing null pointers to functions that expect a 7879 non-null argument (those decorated with attribute nonnull). By 7880 taking advantage of optimizations the option can detect many 7881 more cases of the problem than in prior GCC versions. 7882 + The -Wstringop-overflow=type option detects buffer overflow in 7883 calls to string handling functions like memcpy and strcpy. The 7884 option relies on [13]Object Size Checking and has an effect 7885 similar to defining the _FORTIFY_SOURCE macro. 7886 -Wstringop-overflow=2 is enabled by default. 7887 For example, in the following snippet, because the call to 7888 strncat specifies a maximum that allows the function to write 7889 past the end of the destination, it is diagnosed. To correct 7890 the problem and avoid the overflow the function should be 7891 called with a size of at most sizeof d - strlen(d) - 1. 7892 7893 void f (const char *fname) 7894 { 7895 char d[8]; 7896 strncpy (d, "/tmp/", sizeof d); 7897 strncat (d, fname, sizeof d); 7898 ... 7899 } 7900 7901 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 7902 ow=] 7903 7904 * The <limits.h> header provided by GCC defines macros such as 7905 INT_WIDTH for the width in bits of integer types, if 7906 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 7907 included. The <stdint.h> header defines such macros as SIZE_WIDTH 7908 and INTMAX_WIDTH for the width of some standard typedef names for 7909 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 7910 before the header is included; note that GCC's implementation of 7911 this header is only used for freestanding compilations, not hosted 7912 compilations, on most systems. These macros come from ISO/IEC TS 7913 18661-1:2014. 7914 * The <float.h> header provided by GCC defines the macro 7915 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 7916 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 7917 included. This represents the number of decimal digits for which 7918 conversions between decimal character strings and binary formats, 7919 in both directions, are correctly rounded, and currently has the 7920 value of UINTMAX_MAX on all systems, reflecting that GCC's 7921 compile-time conversions are correctly rounded for any number of 7922 digits. 7923 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 7924 __builtin_mul_overflow_p built-in functions have been added. These 7925 work similarly to their siblings without the _p suffix, but do not 7926 actually store the result of the arithmetics anywhere, just return 7927 whether the operation would overflow. Calls to these built-ins with 7928 integer constant arguments evaluate to integer constants 7929 expressions. 7930 For example, in the following, c is assigned the result of a * b 7931 only if the multiplication does not overflow, otherwise it is 7932 assigned the value zero. The multiplication is performed at 7933 compile-time and without triggering a -Woverflow warning. 7934 7935 enum { 7936 a = 12345678, 7937 b = 87654321, 7938 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 7939 }; 7940 7941 [14]C 7942 7943 * The C front end now supports type names _FloatN for floating-point 7944 types with IEEE interchange formats and _FloatNx for floating-point 7945 types with IEEE extended formats. These type names come from 7946 ISO/IEC TS 18661-3:2015. 7947 The set of types supported depends on the target for which GCC is 7948 configured. Most targets support _Float32, _Float32x and _Float64. 7949 _Float128 is supported on targets where IEEE binary128 encoding was 7950 already supported as long double or __float128. _Float64x is 7951 supported on targets where a type with either binary128 or Intel 7952 extended precision format is available. 7953 Constants with these types are supported using suffixes fN, FN, fNx 7954 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 7955 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 7956 defined before it is included. 7957 These new types are always distinct from each other and from float, 7958 double and long double, even if they have the same encoding. 7959 Complex types such as _Complex _Float128 are also supported. 7960 Type-generic built-in functions such as __builtin_isinf support the 7961 new types, and the following type-specific built-in functions have 7962 versions (suffixed fN or fNx) for the new types: 7963 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 7964 __builtin_inf, __builtin_nan, __builtin_nans. 7965 * Compilation with -fopenmp is now compatible with the C11 _Atomic 7966 keyword. 7967 7968 [15]C++ 7969 7970 * The C++ front end has experimental support for all of the current 7971 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 7972 constexpr, class template argument deduction, auto template 7973 parameters, and structured bindings. For a full list of new 7974 features, see [16]the C++ status page. 7975 * C++17 support for new of over-aligned types can be enabled in other 7976 modes with the -faligned-new flag. 7977 * The C++17 evaluation order requirements can be selected in other 7978 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 7979 with -fno-strong-eval-order. 7980 * The default semantics of inherited constructors has changed in all 7981 modes, following [17]P0136. Essentially, overload resolution 7982 happens as if calling the inherited constructor directly, and the 7983 compiler fills in construction of the other bases and members as 7984 needed. Most uses should not need any changes. The old behavior can 7985 be restored with -fno-new-inheriting-ctors, or -fabi-version less 7986 than 11. 7987 * The resolution of DR 150 on matching of template template 7988 parameters, allowing default template arguments to make a template 7989 match a parameter, is currently enabled by default in C++17 mode 7990 only. The default can be overridden with -f{no-,}new-ttp-matching. 7991 * The C++ front end will now provide fix-it hints for some missing 7992 semicolons, allowing for automatic fixes by IDEs: 7993 7994 test.cc:4:11: error: expected ';' after class definition 7995 class a {} 7996 ^ 7997 ; 7998 7999 * -Waligned-new has been added to the C++ front end. It warns about 8000 new of type with extended alignment without -faligned-new. 8001 8002 [18]Runtime Library (libstdc++) 8003 8004 * The type of exception thrown by iostreams, std::ios_base::failure, 8005 now uses the [19]cxx11 ABI. 8006 * Experimental support for C++17, including the following new 8007 features: 8008 + std::string_view; 8009 + std::any, std::optional, and std::variant; 8010 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 8011 invoke_result; 8012 + std::is_swappable, and std::is_nothrow_swappable; 8013 + std::apply, and std::make_from_tuple; 8014 + std::void_t, std::bool_constant, std::conjunction, 8015 std::disjunction, and std::negation; 8016 + Variable templates for type traits; 8017 + Mathematical Special Functions; 8018 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 8019 std::chrono::abs; 8020 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 8021 + std::scoped_lock, std::shared_mutex, 8022 std::atomic<T>::is_always_lock_free; 8023 + std::sample, std::default_searcher, std::boyer_moore_searcher 8024 and std::boyer_moore_horspool_searcher; 8025 + Extraction and re-insertion of map and set nodes, try_emplace 8026 members for maps, and functions for accessing containers 8027 std::size, std::empty, and std::data; 8028 + std::shared_ptr support for arrays, 8029 std::shared_ptr<T>::weak_type, 8030 std::enable_shared_from_this<T>::weak_from_this(), and 8031 std::owner_less<void>; 8032 + std::byte; 8033 + std::as_const, std::not_fn, 8034 std::has_unique_object_representations, constexpr 8035 std::addressof. 8036 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 8037 Voutilainen for work on the C++17 support. 8038 * A new power-of-two rehashing policy for use with the _Hashtable 8039 internals, thanks to Franois Dumont. 8040 8041 [20]Fortran 8042 8043 * Support for a number of extensions for compatibility with legacy 8044 code with new flags: 8045 + -fdec-structure Support for DEC STRUCTURE and UNION 8046 + -fdec-intrinsic-ints Support for new integer intrinsics with 8047 B/I/J/K prefixes such as BABS, JIAND... 8048 + -fdec-math Support for additional math intrinsics, including 8049 COTAN and degree-valued trigonometric functions such as TAND, 8050 ASIND... 8051 + -fdec Enable the -fdec-* family of extensions. 8052 * New flag -finit-derived to allow default initialization of 8053 derived-type variables. 8054 * Improved DO loops with step equal to 1 or -1, generates faster code 8055 without a loop preheader. A new warning, -Wundefined-do-loop, warns 8056 when a loop iterates either to HUGE(i) (with step equal to 1), or 8057 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 8058 at run time with -fcheck=do enabled: 8059 8060 program test 8061 implicit none 8062 integer(1) :: i 8063 do i = -HUGE(i)+10, -HUGE(i)-1, -1 8064 print *, i 8065 end do 8066 end program test 8067 8068 At line 8 of file do_check_12.f90 8069 Fortran runtime error: Loop iterates infinitely 8070 8071 * Version 4.5 of the [21]OpenMP specification is now partially 8072 supported in the Fortran compiler; the largest missing item is 8073 structure element mapping. 8074 * User-defined derived-type input/output (UDTIO) is added. 8075 * Derived type coarrays with allocatable and pointer components are 8076 partially supported. 8077 * Non-constant stop codes and error stop codes (Fortran 2015 8078 feature). 8079 * Derived types with allocatable components of recursive type. 8080 * Intrinsic assignment to polymorphic variables. 8081 * Improved submodule support. 8082 * Improved diagnostics (polymorphic results in pure functions). 8083 * Coarray: Support for failed images (Fortan 2015 feature). 8084 8085 [22]Go 8086 8087 * GCC 7 provides a complete implementation of the Go 1.8.1 user 8088 packages. 8089 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 8090 conservative and less concurrent. 8091 * Escape analysis is available for experimental use via the 8092 -fgo-optimize-allocs option. The -fgo-debug-escape prints 8093 information useful for debugging escape analysis choices. 8094 8095 [23]Java (GCJ) 8096 8097 The GCC Java front end and associated libjava runtime library have been 8098 removed from GCC. 8099 8100 [24]libgccjit 8101 8102 The libgccjit API gained support for marking calls as requiring 8103 tail-call optimization via a new entry point: 8104 [25]gcc_jit_rvalue_set_bool_require_tail_call. 8105 8106 libgccjit performs numerous checks at the API boundary, but if these 8107 succeed, it previously ignored errors and other diagnostics emitted 8108 within the core of GCC, and treated the compile of a gcc_jit_context as 8109 having succeeded. As of GCC 7 it now ensures that if any diagnostics 8110 are emitted, they are visible from the libgccjit API, and that the the 8111 context is flagged as having failed. 8112 8113 [26]New Targets and Target Specific Improvements 8114 8115 [27]AArch64 8116 8117 * GCC has been updated to the latest revision of the procedure call 8118 standard (AAPCS64) to provide support for parameter passing when 8119 data types have been over-aligned. 8120 * The ARMv8.3-A architecture is now supported. It can be used by 8121 specifying the -march=armv8.3-a option. 8122 * The option -msign-return-address= is supported to enable return 8123 address protection using ARMv8.3-A Pointer Authentication 8124 Extensions. For more information on the arguments accepted by this 8125 option, please refer to [28]AArch64-Options. 8126 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 8127 Extensions are now supported. They can be used by specifying the 8128 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 8129 Floating-Point Extensions introduce new half-precision data 8130 processing floating-point instructions. 8131 * Support has been added for the following processors (GCC 8132 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 8133 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 8134 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 8135 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 8136 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 8137 The GCC identifiers can be used as arguments to the -mcpu or -mtune 8138 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 8139 arguments to the equivalent target attributes and pragmas. 8140 8141 [29]ARC 8142 8143 * Added support for ARC HS and ARC EM processors. 8144 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 8145 * Added support for NPS400 ARC700 based CPUs. 8146 * Thread Local Storage is now supported by ARC CPUs. 8147 * Fixed errors for ARC600 when using 32x16 multiplier option. 8148 * Fixed PIE for ARC CPUs. 8149 * New CPU templates are supported via multilib. 8150 8151 [30]ARM 8152 8153 * Support for the ARMv5 and ARMv5E architectures has been deprecated 8154 (which have no known implementations) and will be removed in a 8155 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 8156 architectures remain supported. The values armv5 and armv5e of 8157 -march are thus deprecated. 8158 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 8159 Extensions are now supported. They can be used by specifying the 8160 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 8161 Floating-Point Extensions introduce new half-precision data 8162 processing floating-point instructions. 8163 * The ARMv8-M architecture is now supported in its two architecture 8164 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 8165 Floating-Point Extensions. They can be used by specifying the 8166 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 8167 * Support has been added for the following processors (GCC 8168 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 8169 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 8170 identifiers can be used as arguments to the -mcpu or -mtune 8171 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 8172 * A new command-line option -mpure-code has been added. It does not 8173 allow constant data to be placed in code sections. This option is 8174 only available when generating non-PIC code for ARMv7-M targets. 8175 * Support for the ACLE Coprocessor Intrinsics has been added. This 8176 enables the generation of coprocessor instructions through the use 8177 of intrinsics such as cdp, ldc, and others. 8178 * The configure option --with-multilib-list now accepts the value 8179 rmprofile to build multilib libraries for a range of embedded 8180 targets. See our [31]installation instructions for details. 8181 8182 [32]AVR 8183 8184 * On the reduced Tiny cores, the progmem [33]variable attribute is 8185 now properly supported. Respective read-only variables are located 8186 in flash memory in section .progmem.data. No special code is needed 8187 to access such variables; the compiler automatically adds an offset 8188 of 0x4000 to all addresses, which is needed to access variables in 8189 flash memory. As opposed to ordinary cores where it is sufficient 8190 to specify the progmem attribute with definitions, on the reduced 8191 Tiny cores the attribute also has to be specified with (external) 8192 declarations: 8193 8194 extern const int array[] __attribute__((__progmem__)); 8195 8196 int get_value2 (void) 8197 { 8198 /* Access via addresses array + 0x4004 and array + 0x4005. */ 8199 return array[2]; 8200 } 8201 8202 const int* get_address (unsigned idx) 8203 { 8204 /* Returns array + 0x4000 + 2 * idx. */ 8205 return &array[idx]; 8206 } 8207 8208 * A new command-line option -Wmisspelled-isr has been added. It turns 8209 off -- or turns into errors -- warnings that are reported for 8210 interrupt service routines (ISRs) which don't follow AVR-LibC's 8211 naming convention of prefixing ISR names with __vector. 8212 * __builtin_avr_nops(n) is a new [34]built-in function that inserts n 8213 NOP instructions into the instruction stream. n must be a value 8214 known at compile time. 8215 8216 [35]IA-32/x86-64 8217 8218 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 8219 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 8220 variable precision (4VNNIW), AVX-512 Vector Population Count 8221 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 8222 been added. 8223 8224 [36]NVPTX 8225 8226 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 8227 See the [37]Offloading Wiki on how to configure it. 8228 8229 [38]PowerPC / PowerPC64 / RS6000 8230 8231 * The PowerPC port now uses LRA by default. 8232 * GCC now diagnoses inline assembly that clobbers register r2. This 8233 has always been invalid code, and is no longer quietly tolerated. 8234 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 8235 enhanced to generate more of the new instructions by default, and 8236 to provide more built-in functions to generate code for other new 8237 instructions. 8238 * The configuration option --enable-gnu-indirect-function is now 8239 enabled by default on PowerPC GNU/Linux builds. 8240 * The PowerPC port will now allow 64-bit and 32-bit integer types to 8241 be allocated to the VSX vector registers (ISA 2.06 and above). In 8242 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 8243 allocated in the vector registers. Previously, only 64-bit integer 8244 types were allowed in the traditional floating point registers. 8245 * New options -mstack-protector-guard=global, 8246 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 8247 -mstack-protector-guard-offset= change how the stack protector gets 8248 the value to use as canary. 8249 8250 [39]S/390, System z, IBM z Systems, IBM Z 8251 8252 * Support for the IBM z14 processor has been added. When using the 8253 -march=z14 option, the compiler will generate code making use of 8254 the new instructions introduced with the vector enhancement 8255 facility and the miscellaneous instruction extension facility 2. 8256 The -mtune=z14 option enables z14 specific instruction scheduling 8257 without making use of new instructions. 8258 * Builtins for the new vector instructions have been added and can be 8259 enabled using the -mzvector option. 8260 8261 [40]RISC-V 8262 8263 * Support for the RISC-V instruction set has been added. 8264 8265 [41]RX 8266 8267 Basic support for atomic built-in function has been added. It is 8268 currently implemented by flipping interrupts off and on as needed. 8269 8270 [42]SH 8271 8272 * Support for SH5/SH64 has been removed. 8273 * Improved utilization of delay slots on SH2A. 8274 * Improved utilization of zero-displacement conditional branches. 8275 * The following deprecated options have been removed 8276 + -mcbranchdi 8277 + -mcmpeqdi 8278 + -minvalid-symbols 8279 + -msoft-atomic 8280 + -mspace 8281 + -madjust-unroll 8282 * Support for the following SH2A instructions has been added 8283 + mov.b @-Rm,R0 8284 + mov.w @-Rm,R0 8285 + mov.l @-Rm,R0 8286 + mov.b R0,@Rn+ 8287 + mov.w R0,@Rn+ 8288 + mov.l R0,@Rn+ 8289 8290 [43]SPARC 8291 8292 * The SPARC port now uses LRA by default. 8293 * Support for the new Subtract-Extended-with-Carry instruction 8294 available in SPARC M7 (Niagara 7) has been added. 8295 8296 [44]Operating Systems 8297 8298 [45]AIX 8299 8300 * Visibility support has been enabled for AIX 7.1 and above. 8301 8302 [46]Fuchsia 8303 8304 * Support has been added for the [47]Fuchsia OS. 8305 8306 [48]RTEMS 8307 8308 * The ABI changes on ARM so that no short enums are used by default. 8309 8310 Other significant improvements 8311 8312 * -fverbose-asm previously emitted information on the meanings of 8313 assembly expressions. This has been extended so that it now also 8314 prints comments showing the source lines that correspond to the 8315 assembly, making it easier to read the generated assembly 8316 (especially with larger functions). For example, given this C 8317 source file: 8318 8319 int test (int n) 8320 { 8321 int i; 8322 int total = 0; 8323 8324 for (i = 0; i < n; i++) 8325 total += i * i; 8326 return total; 8327 } 8328 8329 -fverbose-asm now gives output similar to this for the function 8330 body (when compiling for x86_64, with -Os): 8331 8332 .text 8333 .globl test 8334 .type test, @@function 8335 test: 8336 .LFB0: 8337 .cfi_startproc 8338 # example.c:4: int total = 0; 8339 xorl %eax, %eax # <retval> 8340 # example.c:6: for (i = 0; i < n; i++) 8341 xorl %edx, %edx # i 8342 .L2: 8343 # example.c:6: for (i = 0; i < n; i++) 8344 cmpl %edi, %edx # n, i 8345 jge .L5 #, 8346 # example.c:7: total += i * i; 8347 movl %edx, %ecx # i, tmp92 8348 imull %edx, %ecx # i, tmp92 8349 # example.c:6: for (i = 0; i < n; i++) 8350 incl %edx # i 8351 # example.c:7: total += i * i; 8352 addl %ecx, %eax # tmp92, <retval> 8353 jmp .L2 # 8354 .L5: 8355 # example.c:10: } 8356 ret 8357 .cfi_endproc 8358 8359 * Two new options have been added for printing fix-it hints: 8360 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 8361 emitted in a machine-readable form, suitable for consumption 8362 by IDEs. For example, given: 8363 8364 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 8365 you mean 'color'? 8366 return ptr->colour; 8367 ^~~~~~ 8368 color 8369 8370 it will emit: 8371 8372 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 8373 8374 + -fdiagnostics-generate-patch will print a patch in "unified" 8375 format after any diagnostics are printed, showing the result 8376 of applying all fix-it hints. For the above example it would 8377 emit: 8378 8379 --- spellcheck-fields.cc 8380 +++ spellcheck-fields.cc 8381 @@ -49,5 +49,5 @@ 8382 8383 color get_color(struct s *ptr) 8384 { 8385 - return ptr->colour; 8386 + return ptr->color; 8387 } 8388 8389 * The gcc and g++ driver programs will now provide suggestions for 8390 misspelled arguments to command-line options. 8391 8392 $ gcc -c test.c -ftls-model=global-dinamic 8393 gcc: error: unknown TLS model 'global-dinamic' 8394 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 8395 cal-dynamic local-exec; did you mean 'global-dynamic'? 8396 8397 * The compiler will now provide suggestions for misspelled 8398 parameters. 8399 8400 $ gcc -c test.c --param max-early-inliner-iteration=3 8401 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 8402 x-early-inliner-iterations'? 8403 8404 * Profile-guided optimization (PGO) instrumentation, as well as test 8405 coverage (GCOV), can newly instrument constructors (functions marks 8406 with __attribute__((constructor))), destructors and C++ 8407 constructors (and destructors) of classes that are used as the type 8408 of a global variable. 8409 * A new option -fprofile-update=atomic prevents creation of corrupted 8410 profiles created during an instrumentation run (-fprofile=generate) 8411 of an application. The downside of the option is a speed penalty. 8412 Providing -pthread on the command line selects atomic profile 8413 updating (when supported by the target). 8414 * GCC's already extensive testsuite has gained some new capabilities, 8415 to further improve the reliability of the compiler: 8416 + GCC now has an internal unit-testing API and a suite of tests 8417 for programmatic self-testing of subsystems. 8418 + GCC's C front end has been extended so that it can parse dumps 8419 of GCC's internal representations, allowing for DejaGnu tests 8420 that more directly exercise specific optimization passes. This 8421 covers both the [49]GIMPLE representation (for testing 8422 higher-level optimizations) and the [50]RTL representation, 8423 allowing for more direct testing of lower-level details, such 8424 as register allocation and instruction selection. 8425 8426 [51]GCC 7.1 8427 8428 This is the [52]list of problem reports (PRs) from GCC's bug tracking 8429 system that are known to be fixed in the 7.1 release. This list might 8430 not be complete (that is, it is possible that some PRs that have been 8431 fixed are not listed here). 8432 8433 [53]GCC 7.2 8434 8435 This is the [54]list of problem reports (PRs) from GCC's bug tracking 8436 system that are known to be fixed in the 7.2 release. This list might 8437 not be complete (that is, it is possible that some PRs that have been 8438 fixed are not listed here). 8439 8440 Target Specific Changes 8441 8442 SPARC 8443 8444 * Support for the SPARC M8 processor has been added. 8445 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 8446 around an erratum in LEON3FT processors. 8447 * Use of the Floating-point Multiply Single to Double (FsMULd) 8448 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 8449 options. 8450 8451 Operating Systems 8452 8453 RTEMS 8454 8455 * The Ada run-time support uses now thread-local storage (TLS). 8456 * Support for RISC-V has been added. 8457 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 8458 double has been added. 8459 8460 [55]GCC 7.3 8461 8462 This is the [56]list of problem reports (PRs) from GCC's bug tracking 8463 system that are known to be fixed in the 7.3 release. This list might 8464 not be complete (that is, it is possible that some PRs that have been 8465 fixed are not listed here). 8466 8467 Target Specific Changes 8468 8469 SPARC 8470 8471 * Workarounds for the four [57]LEON3FT errata GRLIB-TN-0010..0013 8472 have been added. Relevant errata are activated by the target 8473 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 8474 8475 Operating Systems 8476 8477 RTEMS 8478 8479 * Support has been added for Epiphany target. 8480 8481 [58]GCC 7.4 8482 8483 This is the [59]list of problem reports (PRs) from GCC's bug tracking 8484 system that are known to be fixed in the 7.4 release. This list might 8485 not be complete (that is, it is possible that some PRs that have been 8486 fixed are not listed here). 8487 8488 [60]GCC 7.5 8489 8490 This is the [61]list of problem reports (PRs) from GCC's bug tracking 8491 system that are known to be fixed in the 7.5 release. This list might 8492 not be complete (that is, it is possible that some PRs that have been 8493 fixed are not listed here). 8494 8495 8496 For questions related to the use of GCC, please consult these web 8497 pages and the [62]GCC manuals. If that fails, the 8498 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8499 web pages and the development of GCC are welcome on our developer 8500 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 8501 archives. 8502 8503 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 8504 distribution of this entire article is permitted in any medium, 8505 provided this notice is preserved. 8506 8507 These pages are [67]maintained by the GCC team. Last modified 8508 2025-01-31. 8509 8510 References 8511 8512 1. https://gcc.gnu.org/gcc-7/porting_to.html 8513 2. https://gcc.gnu.org/onlinedocs/index.html#current 8514 3. https://gcc.gnu.org/wiki/LRAIsDefault 8515 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 8516 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 8517 6. https://gcc.gnu.org/gcc-7/changes.html#general 8518 7. https://gcc.gnu.org/gcc-7/changes.html#languages 8519 8. https://gcc.gnu.org/wiki/OpenACC 8520 9. https://gcc.gnu.org/wiki/Offloading 8521 10. https://gcc.gnu.org/gcc-7/changes.html#ada 8522 11. https://gcc.gnu.org/gcc-7/changes.html#brig 8523 12. https://gcc.gnu.org/gcc-7/changes.html#c-family 8524 13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 8525 14. https://gcc.gnu.org/gcc-7/changes.html#c 8526 15. https://gcc.gnu.org/gcc-7/changes.html#cxx 8527 16. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 8528 17. https://wg21.link/p0136 8529 18. https://gcc.gnu.org/gcc-7/changes.html#libstdcxx 8530 19. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 8531 20. https://gcc.gnu.org/gcc-7/changes.html#fortran 8532 21. https://www.openmp.org/specifications/ 8533 22. https://gcc.gnu.org/gcc-7/changes.html#go 8534 23. https://gcc.gnu.org/gcc-7/changes.html#java 8535 24. https://gcc.gnu.org/gcc-7/changes.html#jit 8536 25. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 8537 26. https://gcc.gnu.org/gcc-7/changes.html#targets 8538 27. https://gcc.gnu.org/gcc-7/changes.html#aarch64 8539 28. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 8540 29. https://gcc.gnu.org/gcc-7/changes.html#arc 8541 30. https://gcc.gnu.org/gcc-7/changes.html#arm 8542 31. https://gcc.gnu.org/install/configure.html 8543 32. https://gcc.gnu.org/gcc-7/changes.html#avr 8544 33. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 8545 34. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 8546 35. https://gcc.gnu.org/gcc-7/changes.html#x86 8547 36. https://gcc.gnu.org/gcc-7/changes.html#nvptx 8548 37. https://gcc.gnu.org/wiki/Offloading 8549 38. https://gcc.gnu.org/gcc-7/changes.html#powerpc 8550 39. https://gcc.gnu.org/gcc-7/changes.html#s390 8551 40. https://gcc.gnu.org/gcc-7/changes.html#riscv 8552 41. https://gcc.gnu.org/gcc-7/changes.html#rx 8553 42. https://gcc.gnu.org/gcc-7/changes.html#sh 8554 43. https://gcc.gnu.org/gcc-7/changes.html#sparc 8555 44. https://gcc.gnu.org/gcc-7/changes.html#os 8556 45. https://gcc.gnu.org/gcc-7/changes.html#aix 8557 46. https://gcc.gnu.org/gcc-7/changes.html#fuchsia 8558 47. https://fuchsia.googlesource.com/ 8559 48. https://gcc.gnu.org/gcc-7/changes.html#rtems 8560 49. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 8561 50. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 8562 51. https://gcc.gnu.org/gcc-7/changes.html#GCC7.1 8563 52. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 8564 53. https://gcc.gnu.org/gcc-7/changes.html#GCC7.2 8565 54. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 8566 55. https://gcc.gnu.org/gcc-7/changes.html#GCC7.3 8567 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 8568 57. https://www.gaisler.com/app-notes-tech-notes-and-white-papers 8569 58. https://gcc.gnu.org/gcc-7/changes.html#GCC7.4 8570 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 8571 60. https://gcc.gnu.org/gcc-7/changes.html#GCC7.5 8572 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 8573 62. https://gcc.gnu.org/onlinedocs/ 8574 63. mailto:gcc-help (a] gcc.gnu.org 8575 64. mailto:gcc (a] gcc.gnu.org 8576 65. https://gcc.gnu.org/lists.html 8577 66. https://www.fsf.org/ 8578 67. https://gcc.gnu.org/about.html 8579 ====================================================================== 8580 http://gcc.gnu.org/gcc-6/index.html 8581 8582 GCC 6 Release Series 8583 8584 (This release series is no longer supported.) 8585 8586 October 26, 2018 8587 8588 The [1]GNU project and the GCC developers are pleased to announce the 8589 release of GCC 6.5. 8590 8591 This release is a bug-fix release, containing fixes for regressions in 8592 GCC 6.4 relative to previous releases of GCC. 8593 8594 Release History 8595 8596 GCC 6.5 8597 October 26, 2018 ([2]changes, [3]documentation) 8598 8599 GCC 6.4 8600 July 4, 2017 ([4]changes, [5]documentation) 8601 8602 GCC 6.3 8603 December 21, 2016 ([6]changes, [7]documentation) 8604 8605 GCC 6.2 8606 August 22, 2016 ([8]changes, [9]documentation) 8607 8608 GCC 6.1 8609 April 27, 2016 ([10]changes, [11]documentation) 8610 8611 References and Acknowledgements 8612 8613 GCC used to stand for the GNU C Compiler, but since the compiler 8614 supports several other languages aside from C, it now stands for the 8615 GNU Compiler Collection. 8616 8617 A list of [12]successful builds is updated as new information becomes 8618 available. 8619 8620 The GCC developers would like to thank the numerous people that have 8621 contributed new features, improvements, bug fixes, and other changes as 8622 well as test results to GCC. This [13]amazing group of volunteers is 8623 what makes GCC successful. 8624 8625 For additional information about GCC please refer to the [14]GCC 8626 project web site or contact the [15]GCC development mailing list. 8627 8628 To obtain GCC please use [16]our mirror sites or [17]our version 8629 control system. 8630 8631 8632 For questions related to the use of GCC, please consult these web 8633 pages and the [18]GCC manuals. If that fails, the 8634 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8635 web pages and the development of GCC are welcome on our developer 8636 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 8637 archives. 8638 8639 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8640 distribution of this entire article is permitted in any medium, 8641 provided this notice is preserved. 8642 8643 These pages are [23]maintained by the GCC team. Last modified 8644 2024-05-30. 8645 8646 References 8647 8648 1. http://www.gnu.org/ 8649 2. https://gcc.gnu.org/gcc-6/changes.html 8650 3. https://gcc.gnu.org/onlinedocs/6.5.0/ 8651 4. https://gcc.gnu.org/gcc-6/changes.html 8652 5. https://gcc.gnu.org/onlinedocs/6.4.0/ 8653 6. https://gcc.gnu.org/gcc-6/changes.html 8654 7. https://gcc.gnu.org/onlinedocs/6.3.0/ 8655 8. https://gcc.gnu.org/gcc-6/changes.html 8656 9. https://gcc.gnu.org/onlinedocs/6.2.0/ 8657 10. https://gcc.gnu.org/gcc-6/changes.html 8658 11. https://gcc.gnu.org/onlinedocs/6.1.0/ 8659 12. https://gcc.gnu.org/gcc-6/buildstat.html 8660 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8661 14. https://gcc.gnu.org/index.html 8662 15. mailto:gcc (a] gcc.gnu.org 8663 16. https://gcc.gnu.org/mirrors.html 8664 17. https://gcc.gnu.org/git.html 8665 18. https://gcc.gnu.org/onlinedocs/ 8666 19. mailto:gcc-help (a] gcc.gnu.org 8667 20. mailto:gcc (a] gcc.gnu.org 8668 21. https://gcc.gnu.org/lists.html 8669 22. https://www.fsf.org/ 8670 23. https://gcc.gnu.org/about.html 8671 ====================================================================== 8672 http://gcc.gnu.org/gcc-6/changes.html 8673 8674 GCC 6 Release Series 8675 Changes, New Features, and Fixes 8676 8677 This page is a brief summary of some of the huge number of improvements 8678 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 8679 the [2]full GCC documentation. 8680 8681 Caveats 8682 8683 * The default mode for C++ is now -std=gnu++14 instead of 8684 -std=gnu++98. 8685 * Support for a number of older systems and recently unmaintained or 8686 untested target ports of GCC has been declared obsolete in GCC 6. 8687 Unless there is activity to revive them, the next release of GCC 8688 will have their sources permanently removed. 8689 The following ports for individual systems on particular 8690 architectures have been obsoleted: 8691 + SH5 / SH64 (sh64-*-*) as announced [3]here. 8692 * The AVR port requires binutils version 2.26.1 or later for the fix 8693 for [4]PR71151 to work. 8694 * The GCC 6.5 release has an accidental ABI incompatibility for 8695 nested std::pair objects, for more details see [5]PR 87822. The bug 8696 causes a layout change for pairs where the first member is also a 8697 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 8698 is closed so the bug in GCC 6.5 will not be fixed upstream, but 8699 there is a patch in the bug report to allow it to be fixed by 8700 anybody packaging GCC 6.5 or installing it themselves. 8701 8702 [6]General Optimizer Improvements 8703 8704 * UndefinedBehaviorSanitizer gained a new sanitization option, 8705 -fsanitize=bounds-strict, which enables strict checking of array 8706 bounds. In particular, it enables -fsanitize=bounds as well as 8707 instrumentation of flexible array member-like arrays. 8708 * Type-based alias analysis now disambiguates accesses to different 8709 pointers. This improves precision of the alias oracle by about 8710 20-30% on higher-level C++ programs. Programs doing invalid type 8711 punning of pointer types may now need -fno-strict-aliasing to work 8712 correctly. 8713 * Alias analysis now correctly supports the weakref and alias 8714 attributes. This allows accessing both a variable and its alias in 8715 one translation unit which is common with link-time optimization. 8716 * Value range propagation now assumes that the this pointer in C++ 8717 member functions is non-null. This eliminates common null pointer 8718 checks but also breaks some non-conforming code-bases (such as 8719 Qt-5, Chromium, KDevelop). As a temporary work-around 8720 -fno-delete-null-pointer-checks can be used. Wrong code can be 8721 identified by using -fsanitize=undefined. 8722 * Link-time optimization improvements: 8723 + warning and error attributes are now correctly preserved by 8724 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 8725 supported with -flto. 8726 + Type merging was fixed to handle C and Fortran 8727 interoperability rules as defined by the Fortran 2008 language 8728 standard. 8729 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 8730 with char in all cases because it is an array while char is 8731 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 8732 general, this inter-operability cannot be implemented, for 8733 example on targets where the argument passing convention for 8734 arrays differs from scalars. 8735 + More type information is now preserved at link time, reducing 8736 the loss of accuracy of the type-based alias analysis compared 8737 to builds without link-time optimization. 8738 + Invalid type punning on global variables and declarations is 8739 now reported with -Wodr-type-mismatch. 8740 + The size of LTO object files was reduced by about 11% 8741 (measured by compiling Firefox 46.0). 8742 + Link-time parallelization (enabled using -flto=n) was 8743 significantly improved by decreasing the size of streamed data 8744 when partitioning programs. The size of streamed IL while 8745 compiling Firefox 46.0 was reduced by 66%. 8746 + The linker plugin was extended to pass information about the 8747 type of binary produced to the GCC back end. (That can also be 8748 controlled manually by -flinker-output.) This makes it 8749 possible to properly configure the code generator and support 8750 incremental linking. Incremental linking of LTO objects by gcc 8751 -r is now supported for plugin-enabled setups. 8752 There are two ways to perform incremental linking: 8753 1. Linking by ld -r will result in an object file with all 8754 sections from individual object files mechanically 8755 merged. This delays the actual link-time optimization to 8756 the final linking step and thus permits whole program 8757 optimization. Linking the final binary with such object 8758 files is however slower. 8759 2. Linking by gcc -r will lead to link-time optimization and 8760 emit the final binary into the object file. Linking such 8761 an object file is fast but avoids any benefits from whole 8762 program optimization. 8763 GCC 7 will support incremental link-time optimization with gcc 8764 -r. 8765 * Inter-procedural optimization improvements: 8766 + Basic jump threading is now performed before profile 8767 construction and inline analysis, resulting in more realistic 8768 size and time estimates that drive the heuristics of the 8769 inliner and function cloning passes. 8770 + Function cloning now more aggressively eliminates unused 8771 function parameters. 8772 8773 [7]New Languages and Language specific improvements 8774 8775 Compared to GCC 5, the GCC 6 release series includes a much improved 8776 implementation of the [8]OpenACC 2.0a specification. Highlights are: 8777 * In addition to single-threaded host-fallback execution, offloading 8778 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 8779 little-endian GNU/Linux host systems. For nvptx offloading, with 8780 the OpenACC parallel construct, the execution model allows for an 8781 arbitrary number of gangs, up to 32 workers, and 32 vectors. 8782 * Initial support for parallelized execution of OpenACC kernels 8783 constructs: 8784 + Parallelization of a kernels region is switched on by 8785 -fopenacc combined with -O2 or higher. 8786 + Code is offloaded onto multiple gangs, but executes with just 8787 one worker, and a vector length of 1. 8788 + Directives inside a kernels region are not supported. 8789 + Loops with reductions can be parallelized. 8790 + Only kernels regions with one loop nest are parallelized. 8791 + Only the outer-most loop of a loop nest can be parallelized. 8792 + Loop nests containing sibling loops are not parallelized. 8793 Typically, using the OpenACC parallel construct gives much better 8794 performance, compared to the initial support of the OpenACC kernels 8795 construct. 8796 * The device_type clause is not supported. The bind and nohost 8797 clauses are not supported. The host_data directive is not supported 8798 in Fortran. 8799 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 8800 * Usage of OpenACC constructs inside multithreaded contexts (such as 8801 created by OpenMP, or pthread programming) is not supported. 8802 * If a call to the acc_on_device function has a compile-time constant 8803 argument, the function call evaluates to a compile-time constant 8804 value only for C and C++ but not for Fortran. 8805 8806 See the [9]OpenACC and [10]Offloading wiki pages for further 8807 information. 8808 8809 [11]C family 8810 8811 * Version 4.5 of the [12]OpenMP specification is now supported in the 8812 C and C++ compilers. 8813 * The C and C++ compilers now support attributes on enumerators. For 8814 instance, it is now possible to mark enumerators as deprecated: 8815 8816 enum { 8817 newval, 8818 oldval __attribute__ ((deprecated ("too old"))) 8819 }; 8820 8821 * Source locations for the C and C++ compilers are now tracked as 8822 ranges, rather than just points, making it easier to identify the 8823 subexpression of interest within a complicated expression. For 8824 example: 8825 8826 test.cc: In function 'int test(int, int, foo, int, int)': 8827 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 8828 ) 8829 return p + q * r * s + t; 8830 ~~^~~ 8831 8832 In addition, there is now initial support for precise diagnostic 8833 locations within strings: 8834 8835 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 8836 t' argument [-Wformat=] 8837 printf("%*d"); 8838 ^ 8839 8840 * Diagnostics can now contain "fix-it hints", which are displayed in 8841 context underneath the relevant source code. For example: 8842 8843 fixits.c: In function 'bad_deref': 8844 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 8845 return ptr.x; 8846 ^ 8847 -> 8848 8849 * The C and C++ compilers now offer suggestions for misspelled field 8850 names: 8851 8852 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 8853 you mean 'color'? 8854 return ptr->colour; 8855 ^~~~~~ 8856 8857 * New command-line options have been added for the C and C++ 8858 compilers: 8859 + -Wshift-negative-value warns about left shifting a negative 8860 value. 8861 + -Wshift-overflow warns about left shift overflows. This 8862 warning is enabled by default. -Wshift-overflow=2 also warns 8863 about left-shifting 1 into the sign bit. 8864 + -Wtautological-compare warns if a self-comparison always 8865 evaluates to true or false. This warning is enabled by -Wall. 8866 + -Wnull-dereference warns if the compiler detects paths that 8867 trigger erroneous or undefined behavior due to dereferencing a 8868 null pointer. This option is only active when 8869 -fdelete-null-pointer-checks is active, which is enabled by 8870 optimizations in most targets. The precision of the warnings 8871 depends on the optimization options used. 8872 + -Wduplicated-cond warns about duplicated conditions in an 8873 if-else-if chain. 8874 + -Wmisleading-indentation warns about places where the 8875 indentation of the code gives a misleading idea of the block 8876 structure of the code to a human reader. For example, given 8877 [13]CVE-2014-1266: 8878 8879 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 8880 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 8881 g-indentation] 8882 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 8883 ^~ 8884 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 8885 indented as if it is guarded by the 'if' 8886 goto fail; 8887 ^~~~ 8888 8889 This warning is enabled by -Wall. 8890 * The C and C++ compilers now emit saner error messages if 8891 merge-conflict markers are present in a source file. 8892 8893 test.c:3:1: error: version control conflict marker in file 8894 <<<<<<< HEAD 8895 ^~~~~~~ 8896 8897 [14]C 8898 8899 * It is possible to disable warnings when an initialized field of a 8900 structure or a union with side effects is being overridden when 8901 using designated initializers via a new warning option 8902 -Woverride-init-side-effects. 8903 * A new type attribute scalar_storage_order applying to structures 8904 and unions has been introduced. It specifies the storage order (aka 8905 endianness) in memory of scalar fields in structures or unions. 8906 8907 [15]C++ 8908 8909 * The default mode has been changed to -std=gnu++14. 8910 * [16]C++ Concepts are now supported when compiling with -fconcepts. 8911 * -flifetime-dse is more aggressive in dead-store elimination in 8912 situations where a memory store to a location precedes a 8913 constructor to that memory location. 8914 * G++ now supports [17]C++17 fold expressions, u8 character literals, 8915 extended static_assert, and nested namespace definitions. 8916 * G++ now allows constant evaluation for all non-type template 8917 arguments. 8918 * G++ now supports C++ Transactional Memory when compiling with 8919 -fgnu-tm. 8920 8921 [18]Runtime Library (libstdc++) 8922 8923 * Extensions to the C++ Library to support mathematical special 8924 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 8925 * Experimental support for C++17, including the following new 8926 features: 8927 + std::uncaught_exceptions function (this is also available for 8928 -std=gnu++NN modes); 8929 + new member functions try_emplace and insert_or_assign for 8930 unique_key maps; 8931 + non-member functions std::size, std::empty, and std::data for 8932 accessing containers and arrays; 8933 + std::invoke; 8934 + std::shared_mutex; 8935 + std::void_t and std::bool_constant metaprogramming utilities. 8936 Thanks to Ville Voutilainen for contributing many of the C++17 8937 features. 8938 * An experimental implementation of the File System TS. 8939 * Experimental support for most features of the second version of the 8940 Library Fundamentals TS. This includes polymorphic memory resources 8941 and array support in shared_ptr, thanks to Fan You. 8942 * Some assertions checked by Debug Mode can now also be enabled by 8943 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 8944 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 8945 don't affect the library ABI, so can be enabled per-translation 8946 unit. 8947 * Timed mutex types are supported on more targets, including Darwin. 8948 * Improved std::locale support for DragonFly and FreeBSD, thanks to 8949 John Marino and Andreas Tobler. 8950 8951 [19]Fortran 8952 8953 * Fortran 2008 SUBMODULE support. 8954 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 8955 support. 8956 * Improved support for Fortran 2003 deferred-length character 8957 variables. 8958 * Improved support for OpenMP and OpenACC. 8959 * The MATMUL intrinsic is now inlined for straightforward cases if 8960 front-end optimization is active. The maximum size for inlining can 8961 be set to n with the -finline-matmul-limit=n option and turned off 8962 with -finline-matmul-limit=0. 8963 * The -Wconversion-extra option will warn about REAL constants which 8964 have excess precision for their kind. 8965 * The -Winteger-division option has been added, which warns about 8966 divisions of integer constants which are truncated. This option is 8967 included in -Wall by default. 8968 8969 [20]libgccjit 8970 8971 * The driver code is now run in-process within libgccjit, providing a 8972 small speed-up of the compilation process. 8973 * The API has gained entrypoints for 8974 + [21]timing how long was spent in different parts of code, 8975 + [22]creating switch statements, 8976 + [23]allowing unreachable basic blocks in a function, and 8977 + [24]adding arbitrary command-line options to a compilation. 8978 8979 [25]New Targets and Target Specific Improvements 8980 8981 [26]AArch64 8982 8983 * A number of AArch64-specific options have been added. The most 8984 important ones are summarised in this section; for more detailed 8985 information please refer to the documentation. 8986 * The command-line options -march=native, -mcpu=native and 8987 -mtune=native are now available on native AArch64 GNU/Linux 8988 systems. Specifying these options causes GCC to auto-detect the 8989 host CPU and choose the optimal setting for that system. 8990 * -fpic is now supported when generating code for the small code 8991 model (-mcmodel=small). The size of the global offset table (GOT) 8992 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 8993 ILP32 SysV ABI. 8994 * The AArch64 port now supports target attributes and pragmas. Please 8995 refer to the [27]documentation for details of available attributes 8996 and pragmas as well as usage instructions. 8997 * Link-time optimization across translation units with different 8998 target-specific options is now supported. 8999 * The option -mtls-size= is now supported. It can be used to specify 9000 the bit size of TLS offsets, allowing GCC to generate better TLS 9001 instruction sequences. 9002 * The option -fno-plt is now fully functional. 9003 * The ARMv8.1-A architecture and the Large System Extensions are now 9004 supported. They can be used by specifying the -march=armv8.1-a 9005 option. Additionally, the +lse option extension can be used in a 9006 similar fashion to other option extensions. The Large System 9007 Extensions introduce new instructions that are used in the 9008 implementation of atomic operations. 9009 * The ACLE half-precision floating-point type __fp16 is now supported 9010 in the C and C++ languages. 9011 * The ARM Cortex-A35 processor is now supported via the 9012 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 9013 equivalent target attributes and pragmas. 9014 * The Qualcomm QDF24xx processor is now supported via the 9015 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 9016 target attributes and pragmas. 9017 * Code generation for the ARM Cortex-A57 processor is improved. Among 9018 general code generation improvements, a better algorithm is added 9019 for allocating registers to floating-point multiply-accumulate 9020 instructions offering increased performance when compiling with 9021 -mcpu=cortex-a57 or -mtune=cortex-a57. 9022 * Code generation for the ARM Cortex-A53 processor is improved. A 9023 more accurate instruction scheduling model for the processor is now 9024 used, and a number of compiler tuning parameters have been set to 9025 offer increased performance when compiling with -mcpu=cortex-a53 or 9026 -mtune=cortex-a53. 9027 * Code generation for the Samsung Exynos M1 processor is improved. A 9028 more accurate instruction scheduling model for the processor is now 9029 used, and a number of compiler tuning parameters have been set to 9030 offer increased performance when compiling with -mcpu=exynos-m1 or 9031 -mtune=exynos-m1. 9032 * Improvements in the generation of conditional branches and literal 9033 pools allow the compiler to compile functions of a large size. 9034 Constant pools are now placed into separate rodata sections. The 9035 new option -mpc-relative-literal-loads generates per-function 9036 literal pools, limiting the maximum size of functions to 1MiB. 9037 * Several correctness issues generating Advanced SIMD instructions 9038 for big-endian targets have been fixed resulting in improved code 9039 generation for ACLE intrinsics with -mbig-endian. 9040 9041 [28]ARM 9042 9043 * Support for revisions of the ARM architecture prior to ARMv4t has 9044 been deprecated and will be removed in a future GCC release. The 9045 -mcpu and -mtune values that are deprecated are: arm2, arm250, 9046 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 9047 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 9048 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 9049 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 9050 arm7tdmi is still supported. The values of -march that are 9051 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 9052 * The ARM port now supports target attributes and pragmas. Please 9053 refer to the [29]documentation for details of available attributes 9054 and pragmas as well as usage instructions. 9055 * Support has been added for the following processors (GCC 9056 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 9057 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 9058 identifiers can be used as arguments to the -mcpu or -mtune 9059 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 9060 9061 [30]Heterogeneous Systems Architecture 9062 9063 * GCC can now generate HSAIL (Heterogeneous System Architecture 9064 Intermediate Language) for simple OpenMP device constructs if 9065 configured with --enable-offload-targets=hsa. A new libgomp plugin 9066 then runs the HSA GPU kernels implementing these constructs on HSA 9067 capable GPUs via a standard HSA run time. 9068 If the HSA compilation back end determines it cannot output HSAIL 9069 for a particular input, it gives a warning by default. These 9070 warnings can be suppressed with -Wno-hsa. To give a few examples, 9071 the HSA back end does not implement compilation of code using 9072 function pointers, automatic allocation of variable sized arrays, 9073 functions with variadic arguments as well as a number of other less 9074 common programming constructs. 9075 When compilation for HSA is enabled, the compiler attempts to 9076 compile composite OpenMP constructs 9077 9078 #pragma omp target teams distribute parallel for 9079 9080 into parallel HSA GPU kernels. 9081 9082 [31]IA-32/x86-64 9083 9084 * GCC now supports the Intel CPU named Skylake with AVX-512 9085 extensions through -march=skylake-avx512. The switch enables the 9086 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 9087 AVX-512DQ. 9088 * Support for new AMD instructions monitorx and mwaitx has been 9089 added. This includes new intrinsic and built-in support. It is 9090 enabled through option -mmwaitx. The instructions monitorx and 9091 mwaitx implement the same functionality as the old monitor and 9092 mwait instructions. In addition, mwaitx adds a configurable timer. 9093 The timer value is received as third argument and stored in 9094 register %ebx. 9095 * x86-64 targets now allow stack realignment from a word-aligned 9096 stack pointer using the command-line option -mstackrealign or 9097 __attribute__ ((force_align_arg_pointer)). This allows functions 9098 compiled with a vector-aligned stack to be invoked from objects 9099 that keep only word-alignment. 9100 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 9101 can be used to access data via the %fs and %gs segments without 9102 having to resort to inline assembly. Please refer to the 9103 [32]documentation for usage instructions. 9104 * Support for AMD Zen (family 17h) processors is now available 9105 through the -march=znver1 and -mtune=znver1 options. 9106 9107 [33]MeP 9108 9109 * Support for the MeP (mep-elf) architecture has been deprecated and 9110 will be removed in a future GCC release. 9111 9112 [34]MSP430 9113 9114 * The MSP430 compiler now has the ability to automatically distribute 9115 code and data between low memory (addresses below 64K) and high 9116 memory. This only applies to parts that actually have both memory 9117 regions and only if the linker script for the part has been 9118 specifically set up to support this feature. 9119 A new attribute of either can be applied to both functions and 9120 data, and this tells the compiler to place the object into low 9121 memory if there is room and into high memory otherwise. Two other 9122 new attributes - lower and upper - can be used to explicitly state 9123 that an object should be placed in the specified memory region. If 9124 there is not enough left in that region the compilation will fail. 9125 Two new command-line options - -mcode-region=[lower|upper|either] 9126 and -mdata-region=[lower|upper|either] - can be used to tell the 9127 compiler what to do with objects that do not have one of these new 9128 attributes. 9129 9130 [35]PowerPC / PowerPC64 / RS6000 9131 9132 * PowerPC64 now supports IEEE 128-bit floating-point using the 9133 __float128 data type. In GCC 6, this is not enabled by default, but 9134 you can enable it with -mfloat128. The IEEE 128-bit floating-point 9135 support requires the use of the VSX instruction set. IEEE 128-bit 9136 floating-point values are passed and returned as a single vector 9137 value. The software emulator for IEEE 128-bit floating-point 9138 support is only built on PowerPC GNU/Linux systems where the 9139 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 9140 and later), you will be able to use the -mfloat128-hardware option 9141 to use the ISA 3.0 instructions that support IEEE 128-bit 9142 floating-point. An additional type (__ibm128) has been added to 9143 refer to the IBM extended double type that normally implements long 9144 double. This will allow for a future transition to implementing 9145 long double with IEEE 128-bit floating-point. 9146 * Basic support has been added for POWER9 hardware that will use the 9147 recently published OpenPOWER ISA 3.0 instructions. The following 9148 new switches are available: 9149 + -mcpu=power9: Implement all of the ISA 3.0 instructions 9150 supported by the compiler. 9151 + -mtune=power9: In the future, apply tuning for POWER9 systems. 9152 Currently, POWER8 tunings are used. 9153 + -mmodulo: Generate code using the ISA 3.0 integer instructions 9154 (modulus, count trailing zeros, array index support, integer 9155 multiply/add). 9156 + -mpower9-fusion: Generate code to suitably fuse instruction 9157 sequences for a POWER9 system. 9158 + -mpower9-dform: Generate code to use the new D-form 9159 (register+offset) memory instructions for the vector 9160 registers. 9161 + -mpower9-vector: Generate code using the new ISA 3.0 vector 9162 (VSX or Altivec) instructions. 9163 + -mpower9-minmax: Reserved for future development. 9164 + -mtoc-fusion: Keep TOC entries together to provide more fusion 9165 opportunities. 9166 * New constraints have been added to support IEEE 128-bit 9167 floating-point and ISA 3.0 instructions: 9168 + wb: Altivec register if -mpower9-dform is enabled. 9169 + we: VSX register if -mpower9-vector is enabled for 64-bit code 9170 generation. 9171 + wo: VSX register if -mpower9-vector is enabled. 9172 + wp: Reserved for future use if long double is implemented with 9173 IEEE 128-bit floating-point instead of IBM extended double. 9174 + wq: VSX register if -mfloat128 is enabled. 9175 + wF: Memory operand suitable for POWER9 fusion load/store. 9176 + wG: Memory operand suitable for TOC fusion memory references. 9177 + wL: Integer constant identifying the element number mfvsrld 9178 accesses within a vector. 9179 * Support has been added for __builtin_cpu_is() and 9180 __builtin_cpu_supports(), allowing for very fast access to 9181 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 9182 glibc 2.23 or later. 9183 * All hardware transactional memory builtins now correctly behave as 9184 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 9185 determine whether their "old" compiler treats the builtins as 9186 barriers. 9187 * Split-stack support has been added for gccgo on PowerPC64 for both 9188 big- and little-endian (but not for 32-bit). The gold linker from 9189 at least binutils 2.25.1 must be available in the PATH when 9190 configuring and building gccgo to enable split stack. (The 9191 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 9192 split-stack feature allows a small initial stack size to be 9193 allocated for each goroutine, which increases as needed. 9194 * GCC on PowerPC now supports the standard lround function. 9195 * A new configuration option ---with-advance-toolchain=at was added 9196 for PowerPC 64-bit GNU/Linux systems to use the header files, 9197 library files, and the dynamic linker from a specific Advance 9198 Toolchain release instead of the default versions that are provided 9199 by the GNU/Linux distribution. In general, this option is intended 9200 for the developers of GCC, and it is not intended for general use. 9201 * The "q", "S", "T", and "t" asm-constraints have been removed. 9202 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 9203 9204 [36]S/390, System z, IBM z Systems 9205 9206 * Support for the IBM z13 processor has been added. When using the 9207 -march=z13 option, the compiler will generate code making use of 9208 the new instructions and registers introduced with the vector 9209 extension facility. The -mtune=z13 option enables z13 specific 9210 instruction scheduling without making use of new instructions. 9211 Compiling code with -march=z13 reduces the default alignment of 9212 vector types bigger than 8 bytes to 8. This is an ABI change and 9213 care must be taken when linking modules compiled with different 9214 arch levels which interchange variables containing vector type 9215 values. For newly compiled code the GNU linker will emit a warning. 9216 * The -mzvector option enables a C/C++ language extension. This 9217 extension provides a new keyword vector which can be used to define 9218 vector type variables. (Note: This is not available when enforcing 9219 strict standard compliance e.g. with -std=c99. Either enable GNU 9220 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 9221 Additionally a set of overloaded builtins is provided which is 9222 partially compatible to the PowerPC Altivec builtins. In order to 9223 make use of these builtins the vecintrin.h header file needs to be 9224 included. 9225 * The new command-line options -march=native, and -mtune=native are 9226 now available on native IBM z Systems. Specifying these options 9227 causes GCC to auto-detect the host CPU and choose the optimal 9228 setting for that system. If GCC is unable to detect the host CPU 9229 these options have no effect. 9230 * The IBM z Systems port now supports target attributes and pragmas. 9231 Please refer to the [37]documentation for details of available 9232 attributes and pragmas as well as usage instructions. 9233 * -fsplit-stack is now supported as part of the IBM z Systems port. 9234 This feature requires a recent gold linker to be used. 9235 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 9236 been deprecated and will be removed in a future GCC release. -m31 9237 from now on defaults to -march=z900 if not specified otherwise. 9238 -march=native on a g5/g6 machine will default to -march=z900. 9239 9240 [38]SH 9241 9242 * Support for SH5 / SH64 has been declared obsolete and will be 9243 removed in future releases. 9244 * Support for the FDPIC ABI has been added. It can be enabled using 9245 the new -mfdpic target option and --enable-fdpic configure option. 9246 9247 [39]SPARC 9248 9249 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 9250 change will break binary compatibility with earlier releases for 9251 code it affects, but this should be pretty rare in practice. The 9252 conditions are: a 16-byte structure containing a double or a 8-byte 9253 vector in the second half is passed to a subprogram in slot #15, 9254 for example as 16th parameter if the first 15 ones have at most 8 9255 bytes. The double or vector was wrongly passed in floating-point 9256 register %d32 in lieu of on the stack as per the SPARC calling 9257 conventions. 9258 9259 [40]Operating Systems 9260 9261 [41]AIX 9262 9263 * DWARF debugging support for AIX 7.1 has been enabled as an optional 9264 debugging format. A more recent Technology Level (TL) and GCC built 9265 with that level are required for full exploitation of DWARF 9266 debugging capabilities. 9267 9268 [42]Linux 9269 9270 * Support for the [43]musl C library was added for the AArch64, ARM, 9271 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 9272 x86_64 targets. It can be selected using the new -mmusl option in 9273 case musl is not the default libc. GCC defaults to musl libc if it 9274 is built with a target triplet matching the *-linux-musl* pattern. 9275 9276 [44]RTEMS 9277 9278 * The RTEMS thread model implementation changed. Mutexes now use 9279 self-contained objects defined in newlib <sys/lock.h> instead of 9280 Classic API semaphores. The keys for thread specific data and the 9281 once function are directly defined via <pthread.h>. Self-contained 9282 condition variables are provided via newlib <sys/lock.h>. The RTEMS 9283 thread model also supports C++11 threads. 9284 * OpenMP support now uses self-contained objects provided by newlib 9285 <sys/lock.h> and offers a significantly better performance compared 9286 to the POSIX configuration of libgomp. It is possible to configure 9287 thread pools for each scheduler instance via the environment 9288 variable GOMP_RTEMS_THREAD_POOLS. 9289 9290 [45]Solaris 9291 9292 * Solaris 12 is now fully supported. Minimal support had already been 9293 present in GCC 5.3. 9294 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 9295 crtn.o), which GCC now prefers over its own ones. 9296 * Position independent executables (PIE) are now supported on Solaris 9297 12. 9298 * Constructor priority is now supported on Solaris 12 with the system 9299 linker. 9300 * libvtv has been ported to Solaris 11 and up. 9301 9302 [46]Windows 9303 9304 * The option -mstackrealign is now automatically activated in 32-bit 9305 mode whenever the use of SSE instructions is requested. 9306 9307 Other significant improvements 9308 9309 * The gcc and g++ driver programs will now provide suggestions for 9310 misspelled command-line options. 9311 9312 $ gcc -static-libfortran test.f95 9313 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 9314 '-static-libgfortran'? 9315 9316 * The --enable-default-pie configure option enables generation of PIE 9317 by default. 9318 9319 [47]GCC 6.2 9320 9321 This is the [48]list of problem reports (PRs) from GCC's bug tracking 9322 system that are known to be fixed in the 6.2 release. This list might 9323 not be complete (that is, it is possible that some PRs that have been 9324 fixed are not listed here). 9325 9326 Target Specific Changes 9327 9328 SPARC 9329 9330 * Support for --with-cpu-32 and --with-cpu-64 configure options has 9331 been added on bi-architecture platforms. 9332 * Support for the SPARC M7 (Niagara 7) processor has been added. 9333 * Support for the VIS 4.0 instruction set has been added. 9334 9335 [49]GCC 6.3 9336 9337 This is the [50]list of problem reports (PRs) from GCC's bug tracking 9338 system that are known to be fixed in the 6.3 release. This list might 9339 not be complete (that is, it is possible that some PRs that have been 9340 fixed are not listed here). 9341 9342 Target Specific Changes 9343 9344 IA-32/x86-64 9345 9346 * Support for the [51]deprecated pcommit instruction has been 9347 removed. 9348 9349 [52]GCC 6.4 9350 9351 This is the [53]list of problem reports (PRs) from GCC's bug tracking 9352 system that are known to be fixed in the 6.4 release. This list might 9353 not be complete (that is, it is possible that some PRs that have been 9354 fixed are not listed here). 9355 9356 [54]Operating Systems 9357 9358 [55]RTEMS 9359 9360 * The ABI changes on ARM so that no short enums are used by default. 9361 9362 [56]GCC 6.5 9363 9364 This is the [57]list of problem reports (PRs) from GCC's bug tracking 9365 system that are known to be fixed in the 6.5 release. This list might 9366 not be complete (that is, it is possible that some PRs that have been 9367 fixed are not listed here). 9368 9369 9370 For questions related to the use of GCC, please consult these web 9371 pages and the [58]GCC manuals. If that fails, the 9372 [59]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9373 web pages and the development of GCC are welcome on our developer 9374 list at [60]gcc (a] gcc.gnu.org. All of [61]our lists have public 9375 archives. 9376 9377 Copyright (C) [62]Free Software Foundation, Inc. Verbatim copying and 9378 distribution of this entire article is permitted in any medium, 9379 provided this notice is preserved. 9380 9381 These pages are [63]maintained by the GCC team. Last modified 9382 2025-01-31. 9383 9384 References 9385 9386 1. https://gcc.gnu.org/gcc-6/porting_to.html 9387 2. https://gcc.gnu.org/onlinedocs/index.html#current 9388 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 9389 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 9390 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 9391 6. https://gcc.gnu.org/gcc-6/changes.html#general 9392 7. https://gcc.gnu.org/gcc-6/changes.html#languages 9393 8. https://www.openacc.org/ 9394 9. https://gcc.gnu.org/wiki/OpenACC 9395 10. https://gcc.gnu.org/wiki/Offloading 9396 11. https://gcc.gnu.org/gcc-6/changes.html#c-family 9397 12. https://www.openmp.org/specifications/ 9398 13. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 9399 14. https://gcc.gnu.org/gcc-6/changes.html#c 9400 15. https://gcc.gnu.org/gcc-6/changes.html#cxx 9401 16. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 9402 17. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 9403 18. https://gcc.gnu.org/gcc-6/changes.html#libstdcxx 9404 19. https://gcc.gnu.org/gcc-6/changes.html#fortran 9405 20. https://gcc.gnu.org/gcc-6/changes.html#jit 9406 21. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 9407 22. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 9408 23. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 9409 24. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 9410 25. https://gcc.gnu.org/gcc-6/changes.html#targets 9411 26. https://gcc.gnu.org/gcc-6/changes.html#aarch64 9412 27. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 9413 28. https://gcc.gnu.org/gcc-6/changes.html#arm 9414 29. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 9415 30. https://gcc.gnu.org/gcc-6/changes.html#hsa 9416 31. https://gcc.gnu.org/gcc-6/changes.html#x86 9417 32. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 9418 33. https://gcc.gnu.org/gcc-6/changes.html#mep 9419 34. https://gcc.gnu.org/gcc-6/changes.html#msp430 9420 35. https://gcc.gnu.org/gcc-6/changes.html#powerpc 9421 36. https://gcc.gnu.org/gcc-6/changes.html#s390 9422 37. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 9423 38. https://gcc.gnu.org/gcc-6/changes.html#sh 9424 39. https://gcc.gnu.org/gcc-6/changes.html#sparc 9425 40. https://gcc.gnu.org/gcc-6/changes.html#os 9426 41. https://gcc.gnu.org/gcc-6/changes.html#aix 9427 42. https://gcc.gnu.org/gcc-6/changes.html#linux 9428 43. http://www.musl-libc.org/ 9429 44. https://gcc.gnu.org/gcc-6/changes.html#rtems 9430 45. https://gcc.gnu.org/gcc-6/changes.html#solaris 9431 46. https://gcc.gnu.org/gcc-6/changes.html#windows 9432 47. https://gcc.gnu.org/gcc-6/changes.html#GCC6.2 9433 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 9434 49. https://gcc.gnu.org/gcc-6/changes.html#GCC6.3 9435 50. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 9436 51. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 9437 52. https://gcc.gnu.org/gcc-6/changes.html#GCC6.4 9438 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 9439 54. https://gcc.gnu.org/gcc-6/changes.html#os64 9440 55. https://gcc.gnu.org/gcc-6/changes.html#rtems64 9441 56. https://gcc.gnu.org/gcc-6/changes.html#GCC6.5 9442 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 9443 58. https://gcc.gnu.org/onlinedocs/ 9444 59. mailto:gcc-help (a] gcc.gnu.org 9445 60. mailto:gcc (a] gcc.gnu.org 9446 61. https://gcc.gnu.org/lists.html 9447 62. https://www.fsf.org/ 9448 63. https://gcc.gnu.org/about.html 9449 ====================================================================== 9450 http://gcc.gnu.org/gcc-5/index.html 9451 9452 GCC 5 Release Series 9453 9454 (This release series is no longer supported.) 9455 9456 October 10, 2017 9457 9458 The [1]GNU project and the GCC developers are pleased to announce the 9459 release of GCC 5.5. 9460 9461 This release is a bug-fix release, containing fixes for regressions in 9462 GCC 5.4 relative to previous releases of GCC. 9463 9464 Release History 9465 9466 GCC 5.5 9467 October 10, 2017 ([2]changes, [3]documentation) 9468 9469 GCC 5.4 9470 June 3, 2016 ([4]changes, [5]documentation) 9471 9472 GCC 5.3 9473 December 4, 2015 ([6]changes, [7]documentation) 9474 9475 GCC 5.2 9476 July 16, 2015 ([8]changes, [9]documentation) 9477 9478 GCC 5.1 9479 April 22, 2015 ([10]changes, [11]documentation) 9480 9481 References and Acknowledgements 9482 9483 GCC used to stand for the GNU C Compiler, but since the compiler 9484 supports several other languages aside from C, it now stands for the 9485 GNU Compiler Collection. 9486 9487 A list of [12]successful builds is updated as new information becomes 9488 available. 9489 9490 The GCC developers would like to thank the numerous people that have 9491 contributed new features, improvements, bug fixes, and other changes as 9492 well as test results to GCC. This [13]amazing group of volunteers is 9493 what makes GCC successful. 9494 9495 For additional information about GCC please refer to the [14]GCC 9496 project web site or contact the [15]GCC development mailing list. 9497 9498 To obtain GCC please use [16]our mirror sites or [17]our version 9499 control system. 9500 9501 9502 For questions related to the use of GCC, please consult these web 9503 pages and the [18]GCC manuals. If that fails, the 9504 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9505 web pages and the development of GCC are welcome on our developer 9506 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 9507 archives. 9508 9509 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9510 distribution of this entire article is permitted in any medium, 9511 provided this notice is preserved. 9512 9513 These pages are [23]maintained by the GCC team. Last modified 9514 2024-05-30. 9515 9516 References 9517 9518 1. http://www.gnu.org/ 9519 2. https://gcc.gnu.org/gcc-5/changes.html 9520 3. https://gcc.gnu.org/onlinedocs/5.5.0/ 9521 4. https://gcc.gnu.org/gcc-5/changes.html 9522 5. https://gcc.gnu.org/onlinedocs/5.4.0/ 9523 6. https://gcc.gnu.org/gcc-5/changes.html 9524 7. https://gcc.gnu.org/onlinedocs/5.3.0/ 9525 8. https://gcc.gnu.org/gcc-5/changes.html 9526 9. https://gcc.gnu.org/onlinedocs/5.2.0/ 9527 10. https://gcc.gnu.org/gcc-5/changes.html 9528 11. https://gcc.gnu.org/onlinedocs/5.1.0/ 9529 12. https://gcc.gnu.org/gcc-5/buildstat.html 9530 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9531 14. https://gcc.gnu.org/index.html 9532 15. mailto:gcc (a] gcc.gnu.org 9533 16. https://gcc.gnu.org/mirrors.html 9534 17. https://gcc.gnu.org/git.html 9535 18. https://gcc.gnu.org/onlinedocs/ 9536 19. mailto:gcc-help (a] gcc.gnu.org 9537 20. mailto:gcc (a] gcc.gnu.org 9538 21. https://gcc.gnu.org/lists.html 9539 22. https://www.fsf.org/ 9540 23. https://gcc.gnu.org/about.html 9541 ====================================================================== 9542 http://gcc.gnu.org/gcc-5/changes.html 9543 9544 GCC 5 Release Series 9545 Changes, New Features, and Fixes 9546 9547 Caveats 9548 9549 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 9550 * The C++ runtime library (libstdc++) uses a new ABI by default (see 9551 [1]below). 9552 * The Graphite framework for loop optimizations no longer requires 9553 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 9554 The installation manual contains more information about 9555 requirements to build GCC. 9556 * The non-standard C++0x type traits has_trivial_default_constructor, 9557 has_trivial_copy_constructor and has_trivial_copy_assign have been 9558 deprecated and will be removed in a future version. The standard 9559 C++11 traits is_trivially_default_constructible, 9560 is_trivially_copy_constructible and is_trivially_copy_assignable 9561 should be used instead. 9562 9563 [2]General Optimizer Improvements 9564 9565 * Inter-procedural optimization improvements: 9566 + An Identical Code Folding (ICF) pass (controlled via 9567 -fipa-icf) has been added. Compared to the identical code 9568 folding performed by the Gold linker this pass does not 9569 require function sections. It also performs merging before 9570 inlining, so inter-procedural optimizations are aware of the 9571 code re-use. On the other hand not all unifications performed 9572 by a linker are doable by GCC which must honor aliasing 9573 information. During link-time optimization of Firefox, this 9574 pass unifies about 31000 functions, that is 14% overall. 9575 + The devirtualization pass was significantly improved by adding 9576 better support for speculative devirtualization and dynamic 9577 type detection. About 50% of virtual calls in Firefox are now 9578 speculatively devirtualized during link-time optimization. 9579 + A new comdat localization pass allows the linker to eliminate 9580 more dead code in presence of C++ inline functions. 9581 + Virtual tables are now optimized. Local aliases are used to 9582 reduce dynamic linking time of C++ virtual tables on ELF 9583 targets and data alignment has been reduced to limit data 9584 segment bloat. 9585 + A new -fno-semantic-interposition option can be used to 9586 improve code quality of shared libraries where interposition 9587 of exported symbols is not allowed. 9588 + Write-only variables are now detected and optimized out. 9589 + With profile feedback the function inliner can now bypass 9590 --param inline-insns-auto and --param inline-insns-single 9591 limits for hot calls. 9592 + The IPA reference pass was significantly sped up making it 9593 feasible to enable -fipa-reference with -fprofile-generate. 9594 This also solves a bottleneck seen when building Chromium with 9595 link-time optimization. 9596 + The symbol table and call-graph API was reworked to C++ and 9597 simplified. 9598 + The interprocedural propagation of constants now also 9599 propagates alignments of pointer parameters. This for example 9600 means that the vectorizer often does not need to generate loop 9601 prologues and epilogues to make up for potential 9602 misalignments. 9603 * Link-time optimization improvements: 9604 + One Definition Rule based merging of C++ types has been 9605 implemented. Type merging enables better devirtualization and 9606 alias analysis. Streaming extra information needed to merge 9607 types adds about 2-6% of memory size and object size increase. 9608 This can be controlled by -flto-odr-type-merging. 9609 + Command-line optimization and target options are now streamed 9610 on a per-function basis and honored by the link-time 9611 optimizer. This change makes link-time optimization a more 9612 transparent replacement of per-file optimizations. It is now 9613 possible to build projects that require different optimization 9614 settings for different translation units (such as -ffast-math, 9615 -mavx, or -finline). Contrary to earlier GCC releases, the 9616 optimization and target options passed on the link command 9617 line are ignored. 9618 Note that this applies only to those command-line options that 9619 can be passed to optimize and target attributes. Command-line 9620 options affecting global code generation (such as -fpic), 9621 warnings (such as -Wodr), optimizations affecting the way 9622 static variables are optimized (such as -fcommon), debug 9623 output (such as -g), and --param parameters can be applied 9624 only to the whole link-time optimization unit. In these cases, 9625 it is recommended to consistently use the same options at both 9626 compile time and link time. 9627 + GCC bootstrap now uses slim LTO object files. 9628 + Memory usage and link times were improved. Tree merging was 9629 sped up, memory usage of GIMPLE declarations and types was 9630 reduced, and, support for on-demand streaming of variable 9631 constructors was added. 9632 * Feedback directed optimization improvements: 9633 + A new auto-FDO mode uses profiles collected by low overhead 9634 profiling tools (perf) instead of more expensive program 9635 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 9636 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 9637 traditional feedback directed optimization. 9638 + Profile precision was improved in presence of C++ inline and 9639 extern inline functions. 9640 + The new gcov-tool utility allows manipulating profiles. 9641 + Profiles are now more tolerant to source file changes (this 9642 can be controlled by --param profile-func-internal-id). 9643 * Register allocation improvements: 9644 + A new local register allocator (LRA) sub-pass, controlled by 9645 -flra-remat, implements control-flow sensitive global register 9646 rematerialization. Instead of spilling and restoring a 9647 register value, it is recalculated if it is profitable. The 9648 sub-pass improved SPEC2000 generated code by 1% and 0.5% 9649 correspondingly on ARM and x86-64. 9650 + Reuse of the PIC hard register, instead of using a fixed 9651 register, was implemented on x86/x86-64 targets. This improves 9652 generated PIC code performance as more hard registers can be 9653 used. Shared libraries can significantly benefit from this 9654 optimization. Currently it is switched on only for x86/x86-64 9655 targets. As RA infrastructure is already implemented for PIC 9656 register reuse, other targets might follow this in the future. 9657 + A simple form of inter-procedural RA was implemented. When it 9658 is known that a called function does not use caller-saved 9659 registers, save/restore code is not generated around the call 9660 for such registers. This optimization can be controlled by 9661 -fipa-ra 9662 + LRA is now much more effective at generating spills of general 9663 registers into vector registers instead of memory on 9664 architectures (e.g., modern Intel processors) where this is 9665 profitable. 9666 * UndefinedBehaviorSanitizer gained a few new sanitization options: 9667 + -fsanitize=float-divide-by-zero: detect floating-point 9668 division by zero; 9669 + -fsanitize=float-cast-overflow: check that the result of 9670 floating-point type to integer conversions do not overflow; 9671 + -fsanitize=bounds: enable instrumentation of array bounds and 9672 detect out-of-bounds accesses; 9673 + -fsanitize=alignment: enable alignment checking, detect 9674 various misaligned objects; 9675 + -fsanitize=object-size: enable object size checking, detect 9676 various out-of-bounds accesses. 9677 + -fsanitize=vptr: enable checking of C++ member function calls, 9678 member accesses and some conversions between pointers to base 9679 and derived classes, detect if the referenced object does not 9680 have the correct dynamic type. 9681 * Pointer Bounds Checker, a bounds violation detector, has been added 9682 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 9683 instrumented with run-time checks of used pointers against their 9684 bounds to detect pointer bounds violations (overflows). The Pointer 9685 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 9686 new ISA extension Intel MPX support. See the Pointer Bounds Checker 9687 [3]Wiki page for more details. 9688 9689 [4]New Languages and Language specific improvements 9690 9691 * [5]OpenMP 4.0 specification offloading features are now supported 9692 by the C, C++, and Fortran compilers. Generic changes: 9693 + Infrastructure (suitable for any vendor). 9694 + Testsuite which covers offloading from the [6]OpenMP 4.0 9695 Examples document. 9696 Specific for upcoming Intel Xeon Phi products: 9697 + Run-time library. 9698 + Card emulator. 9699 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 9700 specification. OpenACC is intended for programming accelerator 9701 devices such as GPUs. See [7]the OpenACC wiki page for more 9702 information. 9703 9704 [8]C family 9705 9706 * The default setting of the -fdiagnostics-color= command-line option 9707 is now [9]configurable when building GCC using configuration option 9708 --with-diagnostics-color=. The possible values are: never, always, 9709 auto and auto-if-env. The new default auto uses color only when the 9710 standard error is a terminal. The default in GCC 4.9 was 9711 auto-if-env, which is equivalent to auto if there is a non-empty 9712 GCC_COLORS environment variable, and never otherwise. As in GCC 9713 4.9, an empty GCC_COLORS variable in the environment will always 9714 disable colors, no matter what the default is or what command-line 9715 options are used. 9716 * A new command-line option -Wswitch-bool has been added for the C 9717 and C++ compilers, which warns whenever a switch statement has an 9718 index of boolean type. 9719 * A new command-line option -Wlogical-not-parentheses has been added 9720 for the C and C++ compilers, which warns about "logical not" used 9721 on the left hand side operand of a comparison. 9722 * A new command-line option -Wsizeof-array-argument has been added 9723 for the C and C++ compilers, which warns when the sizeof operator 9724 is applied to a parameter that has been declared as an array in a 9725 function definition. 9726 * A new command-line option -Wbool-compare has been added for the C 9727 and C++ compilers, which warns about boolean expressions compared 9728 with an integer value different from true/false. 9729 * Full support for Cilk Plus has been added to the GCC compiler. Cilk 9730 Plus is an extension to the C and C++ languages to support data and 9731 task parallelism. 9732 * A new attribute no_reorder prevents reordering of selected symbols 9733 against other such symbols or inline assembler. This enables to 9734 link-time optimize the Linux kernel without having to resort to 9735 -fno-toplevel-reorder that disables several optimizations. 9736 * New preprocessor constructs, __has_include and __has_include_next, 9737 to test the availability of headers have been added. 9738 This demonstrates a way to include the header <optional> only if it 9739 is available: 9740 9741 #ifdef __has_include 9742 # if __has_include(<optional>) 9743 # include <optional> 9744 # define have_optional 1 9745 # elif __has_include(<experimental/optional>) 9746 # include <experimental/optional> 9747 # define have_optional 1 9748 # define experimental_optional 9749 # else 9750 # define have_optional 0 9751 # endif 9752 #endif 9753 9754 The header search paths for __has_include and __has_include_next 9755 are equivalent to those of the standard directive #include and the 9756 extension #include_next respectively. 9757 * A new built-in function-like macro to determine the existence of an 9758 attribute, __has_attribute, has been added. The equivalent built-in 9759 macro __has_cpp_attribute was added to C++ to support 9760 [10]Feature-testing recommendations for C++. The macro 9761 __has_attribute is added to all C-like languages as an extension: 9762 9763 int 9764 #ifdef __has_attribute 9765 # if __has_attribute(__noinline__) 9766 __attribute__((__noinline__)) 9767 # endif 9768 #endif 9769 foo(int x); 9770 9771 If an attribute exists, a nonzero constant integer is returned. For 9772 standardized C++ attributes a date is returned, otherwise the 9773 constant returned is 1. Both __has_attribute and 9774 __has_cpp_attribute will add underscores to an attribute name if 9775 necessary to resolve the name. For C++11 and onwards the attribute 9776 may be scoped. 9777 * A new set of built-in functions for arithmetics with overflow 9778 checking has been added: __builtin_add_overflow, 9779 __builtin_sub_overflow and __builtin_mul_overflow and for 9780 compatibility with clang also other variants. These builtins have 9781 two integral arguments (which don't need to have the same type), 9782 the arguments are extended to infinite precision signed type, +, - 9783 or * is performed on those, and the result is stored in an integer 9784 variable pointed to by the last argument. If the stored value is 9785 equal to the infinite precision result, the built-in functions 9786 return false, otherwise true. The type of the integer variable that 9787 will hold the result can be different from the types of the first 9788 two arguments. The following snippet demonstrates how this can be 9789 used in computing the size for the calloc function: 9790 9791 void * 9792 calloc (size_t x, size_t y) 9793 { 9794 size_t sz; 9795 if (__builtin_mul_overflow (x, y, &sz)) 9796 return NULL; 9797 void *ret = malloc (sz); 9798 if (ret) memset (res, 0, sz); 9799 return ret; 9800 } 9801 9802 On e.g. i?86 or x86-64 the above will result in a mul instruction 9803 followed by a jump on overflow. 9804 * The option -fextended-identifiers is now enabled by default for 9805 C++, and for C99 and later C versions. Various bugs in the 9806 implementation of extended identifiers have been fixed. 9807 9808 [11]C 9809 9810 * The default mode has been changed to -std=gnu11. 9811 * A new command-line option -Wc90-c99-compat has been added to warn 9812 about features not present in ISO C90, but present in ISO C99. 9813 * A new command-line option -Wc99-c11-compat has been added to warn 9814 about features not present in ISO C99, but present in ISO C11. 9815 * It is possible to disable warnings about conversions between 9816 pointers that have incompatible types via a new warning option 9817 -Wno-incompatible-pointer-types; warnings about implicit 9818 incompatible integer to pointer and pointer to integer conversions 9819 via a new warning option -Wno-int-conversion; and warnings about 9820 qualifiers on pointers being discarded via a new warning option 9821 -Wno-discarded-qualifiers. 9822 * To allow proper use of const qualifiers with multidimensional 9823 arrays, GCC will not warn about incompatible pointer types anymore 9824 for conversions between pointers to arrays with and without const 9825 qualifier (except when using -pedantic). Instead, a new warning is 9826 emitted only if the const qualifier is lost. This can be controlled 9827 with a new warning option -Wno-discarded-array-qualifiers. 9828 * The C front end now generates more precise caret diagnostics. 9829 * The -pg command-line option now only affects the current file in an 9830 LTO build. 9831 9832 [12]C++ 9833 9834 * G++ now supports [13]C++14 variable templates. 9835 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 9836 * Excessive template instantiation depth is now a fatal error. This 9837 prevents excessive diagnostics that usually do not help to identify 9838 the problem. 9839 * G++ and libstdc++ now implement the feature-testing macros from 9840 [14]Feature-testing recommendations for C++. 9841 * G++ now allows typename in a template template parameter. 9842 9843 template<template<typename> typename X> struct D; // OK 9844 9845 * G++ now supports [15]C++14 aggregates with non-static data member 9846 initializers. 9847 9848 struct A { int i, j = i; }; 9849 A a = { 42 }; // a.j is also 42 9850 9851 * G++ now supports [16]C++14 extended constexpr. 9852 9853 constexpr int f (int i) 9854 { 9855 int j = 0; 9856 for (; i > 0; --i) 9857 ++j; 9858 return j; 9859 } 9860 9861 constexpr int i = f(42); // i is 42 9862 9863 * G++ now supports the [17]C++14 sized deallocation functions. 9864 9865 void operator delete (void *, std::size_t) noexcept; 9866 void operator delete[] (void *, std::size_t) noexcept; 9867 9868 * A new One Definition Rule violation warning (controlled by -Wodr) 9869 detects mismatches in type definitions and virtual table contents 9870 during link-time optimization. 9871 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 9872 developers to annotate programs with final specifiers (or anonymous 9873 namespaces) to improve code generation. These warnings can be used 9874 at compile time, but they are more useful in combination with 9875 link-time optimization. 9876 * G++ no longer supports [18]N3639 variable length arrays, as they 9877 were removed from the C++14 working paper prior to ratification. 9878 GNU VLAs are still supported, so VLA support is now the same in 9879 C++14 mode as in C++98 and C++11 modes. 9880 * G++ now allows passing a non-trivially-copyable class via C 9881 varargs, which is conditionally-supported with 9882 implementation-defined semantics in the standard. This uses the 9883 same calling convention as a normal value parameter. 9884 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 9885 various mangling bugs are fixed, but G++ will still emit aliases 9886 with the old, wrong mangling where feasible. -Wabi=2 will warn 9887 about differences between ABI version 2 and the current setting. 9888 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 9889 to be unaffected, but -Wabi=8 will warn about a non-static data 9890 member with type std::nullptr_t which changes position due to this 9891 change. 9892 9893 [19]Runtime Library (libstdc++) 9894 9895 * A [20]Dual ABI is provided by the library. A new ABI is enabled by 9896 default. The old ABI is still supported and can be used by defining 9897 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 9898 standard library headers. 9899 * A new implementation of std::string is enabled by default, using 9900 the small string optimization instead of copy-on-write reference 9901 counting. 9902 * A new implementation of std::list is enabled by default, with an 9903 O(1) size() function; 9904 * [21]Full support for C++11, including the following new features: 9905 + std::deque and std::vector<bool> meet the allocator-aware 9906 container requirements; 9907 + movable and swappable iostream classes; 9908 + support for std::align and std::aligned_union; 9909 + type traits std::is_trivially_copyable, 9910 std::is_trivially_constructible, std::is_trivially_assignable 9911 etc.; 9912 + I/O manipulators std::put_time, std::get_time, std::hexfloat 9913 and std::defaultfloat; 9914 + generic locale-aware std::isblank; 9915 + locale facets for Unicode conversion; 9916 + atomic operations for std::shared_ptr; 9917 + std::notify_all_at_thread_exit() and functions for making 9918 futures ready at thread exit. 9919 * Support for the C++11 hexfloat manipulator changes how the num_put 9920 facet formats floating point types when 9921 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 9922 This change affects all language modes, even though the C++98 9923 standard gave no special meaning to that combination of flags. To 9924 prevent the use of hexadecimal notation for floating point types 9925 use str.unsetf(std::ios_base::floatfield) to clear the relevant 9926 bits in str.flags(). 9927 * [22]Full experimental support for C++14, including the following 9928 new features: 9929 + std::is_final type trait; 9930 + heterogeneous comparison lookup in associative containers. 9931 + global functions cbegin, cend, rbegin, rend, crbegin, and 9932 crend for range access to containers, arrays and initializer 9933 lists. 9934 * [23]Improved experimental support for the Library Fundamentals TS, 9935 including: 9936 + class std::experimental::any; 9937 + function template std::experimental::apply; 9938 + function template std::experimental::sample; 9939 + function template std::experimental::search and related 9940 searcher types; 9941 + variable templates for type traits; 9942 + function template std::experimental::not_fn. 9943 * New random number distributions logistic_distribution and 9944 uniform_on_sphere_distribution as extensions. 9945 * [24]GDB Xmethods for containers and std::unique_ptr. 9946 9947 [25]Fortran 9948 9949 * Compatibility notice: 9950 + The version of the module files (.mod) has been incremented. 9951 + For free-form source files [26]-Werror=line-truncation is now 9952 enabled by default. Note that comments exceeding the line 9953 length are not diagnosed. (For fixed-form source code, the 9954 same warning is available but turned off by default, such that 9955 excess characters are ignored. -ffree-line-length-n and 9956 -ffixed-line-length-n can be used to modify the default line 9957 lengths of 132 and 72 columns, respectively.) 9958 + The -Wtabs option is now more sensible: with -Wtabs the 9959 compiler warns if it encounters tabs and with -Wno-tabs this 9960 warning is turned off. Before, -Wno-tabs warned and -Wtabs 9961 disabled the warning. As before, this warning is also enabled 9962 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 9963 options of -std=. 9964 * Incomplete support for colorizing diagnostics emitted by gfortran 9965 has been added. The option [27]-fdiagnostics-color controls when 9966 color is used in diagnostics. The default value of this option can 9967 be [28]configured when building GCC. The GCC_COLORS environment 9968 variable can be used to customize the colors or disable coloring 9969 completely. Sample diagnostics output: 9970 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 9971 test.f90:6:1: 9972 9973 0 continue 9974 1 9975 Error: Zero is not a valid statement label at (1) 9976 test.f90:9:6: 9977 9978 USE foo 9979 1 9980 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 9981 9982 * The -Wuse-without-only option has been added to warn when a USE 9983 statement has no ONLY qualifier and thus implicitly imports all 9984 public entities of the used module. 9985 * Formatted READ and WRITE statements now work correctly in 9986 locale-aware programs. For more information and potential caveats, 9987 see [29]Section 5.3 Thread-safety of the runtime library in the 9988 manual. 9989 * [30]Fortran 2003: 9990 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 9991 IEEE_ARITHMETIC) are now supported. 9992 * [31]Fortran 2008: 9993 + [32]Coarrays: Full experimental support of Fortran 2008's 9994 coarrays with -fcoarray=lib except for allocatable/pointer 9995 components of derived-type coarrays. GCC currently only ships 9996 with a single-image library (libcaf_single), but multi-image 9997 support based on MPI and GASNet is provided by the libraries 9998 of the [33]OpenCoarrays project. 9999 * TS18508 Additional Parallel Features in Fortran: 10000 + Support for the collective intrinsic subroutines CO_MAX, 10001 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 10002 including -fcoarray=lib support. 10003 + Support for the new atomic intrinsics has been added, 10004 including -fcoarray=lib support. 10005 * Fortran 2015: 10006 + Support for IMPLICIT NONE (external, type). 10007 + ERROR STOP is now permitted in pure procedures. 10008 10009 [34]Go 10010 10011 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 10012 * Building GCC 5 with Go enabled will install two new programs: 10013 [35]go and [36]gofmt. 10014 10015 [37]libgccjit 10016 10017 New in GCC 5 is the ability to build GCC as a shared library for 10018 embedding in other processes (such as interpreters), suitable for 10019 Just-In-Time compilation to machine code. 10020 10021 The shared library has a [38]C API and a [39]C++ wrapper API providing 10022 some "syntactic sugar". There are also bindings available from 3rd 10023 parties for [40]Python and for [41]D. 10024 10025 For example, this library can be used by interpreters for [42]compiling 10026 functions from bytecode to machine code. 10027 10028 The library can also be used for ahead-of-time compilation, enabling 10029 GCC to be plugged into a pre-existing front end. An example of using 10030 this to build a compiler for an esoteric language we'll refer to as 10031 "brainf" can be seen [43]here. 10032 10033 libgccjit is licensed under the GPLv3 (or at your option, any later 10034 version) 10035 10036 It should be regarded as experimental at this time. 10037 10038 [44]New Targets and Target Specific Improvements 10039 10040 [45]Reporting stack usage 10041 10042 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 10043 -fstack-usage option. 10044 10045 [46]AArch64 10046 10047 * Code generation for the ARM Cortex-A57 processor has been improved. 10048 A more accurate instruction scheduling model for the processor is 10049 now used, and a number of compiler tuning parameters have been set 10050 to offer increased performance when compiling with -mcpu=cortex-a57 10051 or -mtune=cortex-a57. 10052 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 10053 and can be enabled by giving the -mfix-cortex-a53-835769 option. 10054 Alternatively it can be enabled by default by configuring GCC with 10055 the --enable-fix-cortex-a53-835769 option. 10056 * The optional cryptographic extensions to the ARMv8-A architecture 10057 are no longer enabled by default when specifying the 10058 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 10059 options. To enable these extensions add +crypto to the value of 10060 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 10061 * Support has been added for the following processors (GCC 10062 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 10063 initial support for its big.LITTLE combination with the ARM 10064 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 10065 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 10066 The GCC identifiers can be used as arguments to the -mcpu or -mtune 10067 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 10068 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 10069 support for the Cortex-A72. 10070 * The transitional options -mlra and -mno-lra have been removed. The 10071 AArch64 back end now uses the local register allocator (LRA) only. 10072 10073 [47]ARM 10074 10075 * Thumb-1 assembly code is now generated in unified syntax. The new 10076 option -masm-syntax-unified specifies whether inline assembly code 10077 is using unified syntax. By default the option is off which means 10078 non-unified syntax is used. However this is subject to change in 10079 future releases. Eventually the non-unified syntax will be 10080 deprecated. 10081 * It is now a configure-time error to use the --with-cpu configure 10082 option with either of --with-tune or --with-arch. 10083 * Code generation for the ARM Cortex-A57 processor has been improved. 10084 A more accurate instruction scheduling model for the processor is 10085 now used, and a number of compiler tuning parameters have been set 10086 to offer increased performance when compiling with -mcpu=cortex-a57 10087 or -mtune=cortex-a57. 10088 * Support has been added for the following processors (GCC 10089 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 10090 initial support for its big.LITTLE combination with the ARM 10091 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 10092 initial support for its big.LITTLE combination with the ARM 10093 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 10094 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 10095 The GCC identifiers can be used as arguments to the -mcpu or -mtune 10096 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 10097 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 10098 support for the Cortex-A72. 10099 * The deprecated option -mwords-little-endian has been removed. 10100 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 10101 -mtpcs-leaf-frame which are only applicable to the old ABI have 10102 been deprecated. 10103 * The transitional options -mlra and -mno-lra have been removed. The 10104 ARM back end now uses the local register allocator (LRA) only. 10105 10106 [48]AVR 10107 10108 * Support has been added for the devices ATtiny4/5/9/10/20/40. This 10109 requires Binutils 2.25 or newer. 10110 * The port uses a new scheme to describe supported devices: For each 10111 supported device, the compiler provides a device-specific [49]spec 10112 file. If the compiler is used together with AVR-LibC, this requires 10113 at least GCC 5.2 and a version of AVR-LibC which implements 10114 [50]feature #44574. 10115 As a consequence, the compiler no more supports individual devices 10116 like ATmega8. Specifying, say, -mmcu=atmega8 triggers the usage of 10117 the device-specific [51]spec file specs-atmega8 which is part of 10118 the installation and describes options for the sub-processes like 10119 compiler proper, assembler and linker. You can add support for a 10120 new device -mmcu=mydevice as follows: 10121 1. In an empty directory /someplace, create a new directory 10122 device-specs. 10123 2. Copy a device spec file from the installed device-specs 10124 folder, follow the comments in that file and then save it as 10125 /someplace/device-specs/specs-mydevice. 10126 3. Add -B /someplace -mmcu=mydevice to the compiler's 10127 command-line options. Notice that /someplace must specify an 10128 absolute path and that mydevice must not start with "avr". 10129 4. Provided you have a device-specific library libmydevice.a 10130 available, you can put it at /someplace, dito for a 10131 device-specific startup file crtmydevice.o. 10132 The contents of the device spec files depend on the compiler's 10133 configuration, in particular on --with-avrlibc=no and whether or 10134 not it is configured for RTEMS. 10135 * A new command-line option -nodevicelib has been added. It prevents 10136 the compiler from linking against AVR-LibC's device-specific 10137 library libdevice.a. 10138 * The following three command-line options have been added: 10139 10140 -mrmw 10141 Set if the device supports the read-modify-write 10142 instructions LAC, LAS, LAT and XCH. 10143 10144 -mn-flash=size 10145 Specify the flash size of the device in units of 64 KiB, 10146 rounded up to the next integer as needed. This option 10147 affects the availability of the [52]AVR address-spaces. 10148 10149 -mskip-bug 10150 Set if the device is affected by the respective silicon 10151 bug. 10152 10153 These options are used internally in order to communicate between 10154 the compiler and the device-spces file. They are set in the 10155 device-specs file as needed. Don't set them by hand. 10156 10157 [53]IA-32/x86-64 10158 10159 * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's 10160 CPU codenamed Skylake Server was added to GCC. That includes inline 10161 assembly support, new intrinsics, and basic autovectorization. 10162 These new AVX-512 extensions are available via the following GCC 10163 switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512 10164 Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword 10165 instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma 10166 and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi. 10167 * New ISA extensions support Intel MPX was added to GCC. This new 10168 extension is available via the -mmpx compiler switch. Intel MPX is 10169 a set of processor features which, with compiler, run-time library 10170 and OS support, brings increased robustness to software by run-time 10171 checking pointer references against their bounds. In GCC Intel MPX 10172 is supported by Pointer Bounds Checker and libmpx run-time 10173 libraries. 10174 * The new -mrecord-mcount option for -pg generates a Linux kernel 10175 style table of pointers to mcount or __fentry__ calls at the 10176 beginning of functions. The new -mnop-mcount option in addition 10177 also generates nops in place of the __fentry__ or mcount call, so 10178 that a call per function can be later patched in. This can be used 10179 for low overhead tracing or hot code patching. 10180 * The new -malign-data option controls how GCC aligns variables. 10181 -malign-data=compat uses increased alignment compatible with GCC 10182 4.8 and earlier, -malign-data=abi uses alignment as specified by 10183 the psABI, and -malign-data=cacheline uses increased alignment to 10184 match the cache line size. -malign-data=compat is the default. 10185 * The new -mskip-rax-setup option skips setting up the RAX register 10186 when SSE is disabled and there are no variable arguments passed in 10187 vector registers. This can be used to optimize the Linux kernel. 10188 10189 [54]MIPS 10190 10191 * MIPS Releases 3 and 5 are now directly supported. Use the 10192 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 10193 to enable code-generation for these processors. 10194 * The Imagination P5600 processor is now supported using the 10195 -march=p5600 command-line option. 10196 * The Cavium Octeon3 processor is now supported using the 10197 -march=octeon3 command-line option. 10198 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 10199 command-line options. 10200 * The o32 ABI has been modified and extended. The o32 64-bit 10201 floating-point register support is now obsolete and has been 10202 removed. It has been replaced by three ABI extensions FPXX, FP64A, 10203 and FP64. The meaning of the -mfp64 command-line option has 10204 changed. It is now used to enable the FP64A and FP64 ABI 10205 extensions. 10206 + The FPXX extension requires that code generated to access 10207 double-precision values use even-numbered registers. Code that 10208 adheres to this extension is link-compatible with all other 10209 o32 double-precision ABI variants and will execute correctly 10210 in all hardware FPU modes. The command-line options -mabi=32 10211 -mfpxx can be used to enable this extension. MIPS II is the 10212 minimum processor required. 10213 + The o32 FP64A extension requires that floating-point registers 10214 be 64-bit and odd-numbered single-precision registers are not 10215 allowed. Code that adheres to the o32 FP64A variant is 10216 link-compatible with all other o32 double-precision ABI 10217 variants. The command-line options -mabi=32 -mfp64 10218 -mno-odd-spreg can be used to enable this extension. MIPS32R2 10219 is the minimum processor required. 10220 + The o32 FP64 extension also requires that floating-point 10221 registers be 64-bit, but permits the use of single-precision 10222 registers. Code that adheres to the o32 FP64 variant is 10223 link-compatible with o32 FPXX and o32 FP64A variants only, 10224 i.e. it is not compatible with the original o32 10225 double-precision ABI. The command-line options -mabi=32 -mfp64 10226 -modd-spreg can be used to enable this extension. MIPS32R2 is 10227 the minimum processor required. 10228 The new ABI variants can be enabled by default using the configure 10229 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 10230 It is strongly recommended that all vendors begin to set o32 FPXX 10231 as the default ABI. This will be required to run the generated code 10232 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 10233 MIPSR6 cores. 10234 * GCC will now pass all floating-point options to the assembler if 10235 GNU binutils 2.25 is used. As a result, any inline assembly code 10236 that uses hard-float instructions should be amended to include a 10237 .set directive to override the global assembler options when 10238 compiling for soft-float targets. 10239 10240 [55]NDS32 10241 10242 * The variadic function ABI implementation is now compatible with 10243 past Andes toolchains where the caller uses registers to pass 10244 arguments and the callee is in charge of pushing them on stack. 10245 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 10246 been removed since they are not yet available in the nds32 port of 10247 GNU binutils. 10248 * A new option -mcmodel=[small|medium|large] supports varied code 10249 models on code generation. The -mgp-direct option became 10250 meaningless and can be discarded. 10251 10252 [56]RX 10253 10254 * A new command line option -mno-allow-string-insns can be used to 10255 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 10256 SWHILE and RMPA instructions. An erratum released by Renesas shows 10257 that it is unsafe to use these instructions on addresses within the 10258 I/O space of the processor. The new option can be used when the 10259 programmer is concerned that the I/O space might be accessed. The 10260 default is still to enable these instructions. 10261 10262 [57]SH 10263 10264 * The compiler will now pass the appropriate --isa= option to the 10265 assembler. 10266 * The default handling for the GBR has been changed from call 10267 clobbered to call preserved. The old behavior can be reinstated by 10268 specifying the option -fcall-used-gbr. 10269 * Support for the SH4A fpchg instruction has been added which will be 10270 utilized when switching between single and double precision FPU 10271 modes. 10272 * The compiler no longer uses the __fpscr_values array for switching 10273 between single and double FPU precision modes on non-SH4A targets. 10274 Instead mode switching will now be performed by storing, modifying 10275 and reloading the FPSCR, so that other FPSCR bits are preserved 10276 across mode switches. The __fpscr_values array that is defined in 10277 libgcc is still present for backwards compatibility, but it will 10278 not be referenced by compiler generated code anymore. 10279 * New builtin functions __builtin_sh_get_fpscr and 10280 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 10281 function will mask the specified bits in such a way that the SZ, PR 10282 and FR mode bits will be preserved, while changing the other bits. 10283 These new functions do not reference the __fpscr_values array. The 10284 old functions __set_fpscr and __get_fpscr in libgcc which access 10285 the __fpscr_values array are still present for backwards 10286 compatibility, but their usage is highly discouraged. 10287 * Some improvements to code generated for __atomic built-in 10288 functions. 10289 * When compiling for SH2E the compiler will no longer force the usage 10290 of delay slots for conditional branch instructions bt and bf. The 10291 old behavior can be reinstated (e.g. to work around a hardware bug 10292 in the original SH7055) by specifying the new option 10293 -mcbranch-force-delay-slot. 10294 10295 [58]Operating Systems 10296 10297 [59]AIX 10298 10299 * GCC now supports stabs debugging continuation lines to allow long 10300 stabs debug information without overflow that generates AIX linker 10301 errors. 10302 10303 [60]DragonFly BSD 10304 10305 * GCC now supports the DragonFly BSD operating system. 10306 10307 [61]FreeBSD 10308 10309 * GCC now supports the FreeBSD operating system for the arm port 10310 through the arm*-*-freebsd* target triplets. 10311 10312 [62]VxWorks MILS 10313 10314 * GCC now supports the MILS (Multiple Independent Levels of Security) 10315 variant of WindRiver's VxWorks operating system for PowerPC 10316 targets. 10317 10318 Other significant improvements 10319 10320 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 10321 to set the compiler to use. 10322 10323 * When the new command-line option -freport-bug is used, GCC 10324 automatically generates a developer-friendly reproducer whenever an 10325 internal compiler error is encountered. 10326 10327 [63]GCC 5.2 10328 10329 This is the [64]list of problem reports (PRs) from GCC's bug tracking 10330 system that are known to be fixed in the 5.2 release. This list might 10331 not be complete (that is, it is possible that some PRs that have been 10332 fixed are not listed here). 10333 10334 Target Specific Changes 10335 10336 IA-32/x86-64 10337 10338 * Support for new AMD instructions monitorx and mwaitx has been 10339 added. This includes new intrinsic and built-in support. It is 10340 enabled through option -mmwaitx. The instructions monitorx and 10341 mwaitx implement the same functionality as the old monitor and 10342 mwait instructions. In addition, mwaitx adds a configurable timer. 10343 The timer value is received as third argument and stored in 10344 register %ebx. 10345 10346 S/390, System z, IBM z Systems 10347 10348 * Support for the IBM z13 processor has been added. When using the 10349 -march=z13 option, the compiler will generate code making use of 10350 the new instructions and registers introduced with the vector 10351 extension facility. The -mtune=z13 option enables z13 specific 10352 instruction scheduling without making use of new instructions. 10353 Compiling code with -march=z13 reduces the default alignment of 10354 vector types bigger than 8 bytes to 8. This is an ABI change and 10355 care must be taken when linking modules compiled with different 10356 arch levels which interchange variables containing vector type 10357 values. For newly compiled code the GNU linker will emit a warning. 10358 * The -mzvector option enables a C/C++ language extension. This 10359 extension provides a new keyword vector which can be used to define 10360 vector type variables. (Note: This is not available when enforcing 10361 strict standard compliance e.g. with -std=c99. Either enable GNU 10362 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 10363 Additionally a set of overloaded builtins is provided which is 10364 partially compatible to the PowerPC Altivec builtins. In order to 10365 make use of these builtins the vecintrin.h header file needs to be 10366 included. 10367 10368 [65]GCC 5.3 10369 10370 This is the [66]list of problem reports (PRs) from GCC's bug tracking 10371 system that are known to be fixed in the 5.3 release. This list might 10372 not be complete (that is, it is possible that some PRs that have been 10373 fixed are not listed here). 10374 10375 Target Specific Changes 10376 10377 IA-32/x86-64 10378 10379 * GCC now supports the Intel CPU named Skylake with AVX-512 10380 extensions through -march=skylake-avx512. The switch enables the 10381 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 10382 AVX-512DQ. 10383 10384 S/390, System z, IBM z Systems 10385 10386 * With this version of GCC IBM z Systems support has been added to 10387 the GO runtime environment. GCC 5.3 has proven to be able to 10388 compile larger GO applications on IBM z Systems. 10389 10390 [67]GCC 5.4 10391 10392 This is the [68]list of problem reports (PRs) from GCC's bug tracking 10393 system that are known to be fixed in the 5.4 release. This list might 10394 not be complete (that is, it is possible that some PRs that have been 10395 fixed are not listed here). 10396 10397 [69]GCC 5.5 10398 10399 This is the [70]list of problem reports (PRs) from GCC's bug tracking 10400 system that are known to be fixed in the 5.5 release. This list might 10401 not be complete (that is, it is possible that some PRs that have been 10402 fixed are not listed here). 10403 10404 Target Specific Changes 10405 10406 IA-32/x86-64 10407 10408 * Support for the [71]deprecated pcommit instruction has been 10409 removed. 10410 10411 10412 For questions related to the use of GCC, please consult these web 10413 pages and the [72]GCC manuals. If that fails, the 10414 [73]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10415 web pages and the development of GCC are welcome on our developer 10416 list at [74]gcc (a] gcc.gnu.org. All of [75]our lists have public 10417 archives. 10418 10419 Copyright (C) [76]Free Software Foundation, Inc. Verbatim copying and 10420 distribution of this entire article is permitted in any medium, 10421 provided this notice is preserved. 10422 10423 These pages are [77]maintained by the GCC team. Last modified 10424 2025-02-28. 10425 10426 References 10427 10428 1. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 10429 2. https://gcc.gnu.org/gcc-5/changes.html#general 10430 3. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 10431 4. https://gcc.gnu.org/gcc-5/changes.html#languages 10432 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 10433 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 10434 7. https://gcc.gnu.org/wiki/OpenACC 10435 8. https://gcc.gnu.org/gcc-5/changes.html#c-family 10436 9. https://gcc.gnu.org/install/configure.html 10437 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 10438 11. https://gcc.gnu.org/gcc-5/changes.html#c 10439 12. https://gcc.gnu.org/gcc-5/changes.html#cxx 10440 13. https://gcc.gnu.org/projects/cxx1y.html 10441 14. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 10442 15. https://gcc.gnu.org/projects/cxx1y.html 10443 16. https://gcc.gnu.org/projects/cxx1y.html 10444 17. https://gcc.gnu.org/projects/cxx1y.html 10445 18. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 10446 19. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 10447 20. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 10448 21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 10449 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 10450 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 10451 24. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python 10452 25. https://gcc.gnu.org/gcc-5/changes.html#fortran 10453 26. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 10454 27. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 10455 28. https://gcc.gnu.org/install/configure.html 10456 29. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 10457 30. https://gcc.gnu.org/wiki/Fortran2003Status 10458 31. https://gcc.gnu.org/wiki/Fortran2008Status 10459 32. https://gcc.gnu.org/wiki/Coarray 10460 33. http://www.opencoarrays.org/ 10461 34. https://gcc.gnu.org/gcc-5/changes.html#go 10462 35. https://pkg.go.dev/cmd/go 10463 36. https://pkg.go.dev/cmd/gofmt 10464 37. https://gcc.gnu.org/gcc-5/changes.html#jit 10465 38. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 10466 39. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 10467 40. https://github.com/davidmalcolm/pygccjit 10468 41. https://github.com/ibuclaw/gccjitd 10469 42. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 10470 43. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 10471 44. https://gcc.gnu.org/gcc-5/changes.html#targets 10472 45. https://gcc.gnu.org/gcc-5/changes.html#stack-usage 10473 46. https://gcc.gnu.org/gcc-5/changes.html#aarch64 10474 47. https://gcc.gnu.org/gcc-5/changes.html#arm 10475 48. https://gcc.gnu.org/gcc-5/changes.html#avr 10476 49. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 10477 50. https://savannah.nongnu.org/bugs/?44574 10478 51. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 10479 52. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 10480 53. https://gcc.gnu.org/gcc-5/changes.html#x86 10481 54. https://gcc.gnu.org/gcc-5/changes.html#mips 10482 55. https://gcc.gnu.org/gcc-5/changes.html#nds32 10483 56. https://gcc.gnu.org/gcc-5/changes.html#rx 10484 57. https://gcc.gnu.org/gcc-5/changes.html#sh 10485 58. https://gcc.gnu.org/gcc-5/changes.html#os 10486 59. https://gcc.gnu.org/gcc-5/changes.html#aix 10487 60. https://gcc.gnu.org/gcc-5/changes.html#dragonfly 10488 61. https://gcc.gnu.org/gcc-5/changes.html#freebsd 10489 62. https://gcc.gnu.org/gcc-5/changes.html#vxmils 10490 63. https://gcc.gnu.org/gcc-5/changes.html#GCC5.2 10491 64. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 10492 65. https://gcc.gnu.org/gcc-5/changes.html#GCC5.3 10493 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 10494 67. https://gcc.gnu.org/gcc-5/changes.html#GCC5.4 10495 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 10496 69. https://gcc.gnu.org/gcc-5/changes.html#GCC5.5 10497 70. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 10498 71. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 10499 72. https://gcc.gnu.org/onlinedocs/ 10500 73. mailto:gcc-help (a] gcc.gnu.org 10501 74. mailto:gcc (a] gcc.gnu.org 10502 75. https://gcc.gnu.org/lists.html 10503 76. https://www.fsf.org/ 10504 77. https://gcc.gnu.org/about.html 10505 ====================================================================== 10506 http://gcc.gnu.org/gcc-4.9/index.html 10507 10508 GCC 4.9 Release Series 10509 10510 (This release series is no longer supported.) 10511 10512 Aug 3, 2016 10513 10514 The [1]GNU project and the GCC developers are pleased to announce the 10515 release of GCC 4.9.4. 10516 10517 This release is a bug-fix release, containing fixes for regressions in 10518 GCC 4.9.3 relative to previous releases of GCC. 10519 10520 Release History 10521 10522 GCC 4.9.4 10523 Aug 3, 2016 ([2]changes, [3]documentation) 10524 10525 GCC 4.9.3 10526 June 26, 2015 ([4]changes, [5]documentation) 10527 10528 GCC 4.9.2 10529 October 30, 2014 ([6]changes, [7]documentation) 10530 10531 GCC 4.9.1 10532 July 16, 2014 ([8]changes, [9]documentation) 10533 10534 GCC 4.9.0 10535 April 22, 2014 ([10]changes, [11]documentation) 10536 10537 References and Acknowledgements 10538 10539 GCC used to stand for the GNU C Compiler, but since the compiler 10540 supports several other languages aside from C, it now stands for the 10541 GNU Compiler Collection. 10542 10543 A list of [12]successful builds is updated as new information becomes 10544 available. 10545 10546 The GCC developers would like to thank the numerous people that have 10547 contributed new features, improvements, bug fixes, and other changes as 10548 well as test results to GCC. This [13]amazing group of volunteers is 10549 what makes GCC successful. 10550 10551 For additional information about GCC please refer to the [14]GCC 10552 project web site or contact the [15]GCC development mailing list. 10553 10554 To obtain GCC please use [16]our mirror sites or [17]our version 10555 control system. 10556 10557 10558 For questions related to the use of GCC, please consult these web 10559 pages and the [18]GCC manuals. If that fails, the 10560 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10561 web pages and the development of GCC are welcome on our developer 10562 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 10563 archives. 10564 10565 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 10566 distribution of this entire article is permitted in any medium, 10567 provided this notice is preserved. 10568 10569 These pages are [23]maintained by the GCC team. Last modified 10570 2024-05-30. 10571 10572 References 10573 10574 1. http://www.gnu.org/ 10575 2. https://gcc.gnu.org/gcc-4.9/changes.html 10576 3. https://gcc.gnu.org/onlinedocs/4.9.4/ 10577 4. https://gcc.gnu.org/gcc-4.9/changes.html 10578 5. https://gcc.gnu.org/onlinedocs/4.9.3/ 10579 6. https://gcc.gnu.org/gcc-4.9/changes.html 10580 7. https://gcc.gnu.org/onlinedocs/4.9.2/ 10581 8. https://gcc.gnu.org/gcc-4.9/changes.html 10582 9. https://gcc.gnu.org/onlinedocs/4.9.1/ 10583 10. https://gcc.gnu.org/gcc-4.9/changes.html 10584 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 10585 12. https://gcc.gnu.org/gcc-4.9/buildstat.html 10586 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10587 14. https://gcc.gnu.org/index.html 10588 15. mailto:gcc (a] gcc.gnu.org 10589 16. https://gcc.gnu.org/mirrors.html 10590 17. https://gcc.gnu.org/git.html 10591 18. https://gcc.gnu.org/onlinedocs/ 10592 19. mailto:gcc-help (a] gcc.gnu.org 10593 20. mailto:gcc (a] gcc.gnu.org 10594 21. https://gcc.gnu.org/lists.html 10595 22. https://www.fsf.org/ 10596 23. https://gcc.gnu.org/about.html 10597 ====================================================================== 10598 http://gcc.gnu.org/gcc-4.9/changes.html 10599 10600 GCC 4.9 Release Series 10601 Changes, New Features, and Fixes 10602 10603 Caveats 10604 10605 * The mudflap run time checker has been removed. The mudflap options 10606 remain, but do nothing. 10607 * Support for a number of older systems and recently unmaintained or 10608 untested target ports of GCC has been declared obsolete in GCC 4.9. 10609 Unless there is activity to revive them, the next release of GCC 10610 will have their sources permanently removed. 10611 The following ports for individual systems on particular 10612 architectures have been obsoleted: 10613 + Solaris 9 (*-*-solaris2.9). Details can be found in the 10614 [1]announcement. 10615 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 10616 float64x1_t exported by arm_neon.h are defined to be the same as 10617 their base types. This results in incorrect application of 10618 parameter passing rules to arguments of types int64x1_t and 10619 uint64x1_t, with respect to the AAPCS64 ABI specification. In 10620 addition, names of C++ functions with parameters of these types 10621 (including float64x1_t) are not mangled correctly. The current 10622 typedef declarations also unintentionally allow implicit casting 10623 between singleton vector types and their base types. These issues 10624 will be resolved in a near future release. See [2]PR60825 for more 10625 information. 10626 10627 More information on porting to GCC 4.9 from previous versions of GCC 10628 can be found in the [3]porting guide for this release. 10629 10630 General Optimizer Improvements 10631 10632 * AddressSanitizer, a fast memory error detector, is now available on 10633 ARM. 10634 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 10635 detector, has been added and can be enabled via 10636 -fsanitize=undefined. Various computations will be instrumented to 10637 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 10638 currently available for the C and C++ languages. 10639 * Link-time optimization (LTO) improvements: 10640 + Type merging was rewritten. The new implementation is 10641 significantly faster and uses less memory. 10642 + Better partitioning algorithm resulting in less streaming 10643 during link time. 10644 + Early removal of virtual methods reduces the size of object 10645 files and improves link-time memory usage and compile time. 10646 + Function bodies are now loaded on-demand and released early 10647 improving overall memory usage at link time. 10648 + C++ hidden keyed methods can now be optimized out. 10649 + When using a linker plugin, compiling with the -flto option 10650 now generates slim object files (.o) which only contain 10651 intermediate language representation for LTO. Use 10652 -ffat-lto-objects to create files which contain additionally 10653 the object code. To generate static libraries suitable for LTO 10654 processing, use gcc-ar and gcc-ranlib; to list symbols from a 10655 slim object file use gcc-nm. (This requires that ar, ranlib 10656 and nm have been compiled with plugin support.) 10657 Memory usage building Firefox with debug enabled was reduced from 10658 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 10659 * Inter-procedural optimization improvements: 10660 + New type inheritance analysis module improving 10661 devirtualization. Devirtualization now takes into account 10662 anonymous name-spaces and the C++11 final keyword. 10663 + New speculative devirtualization pass (controlled by 10664 -fdevirtualize-speculatively. 10665 + Calls that were speculatively made direct are turned back to 10666 indirect where direct call is not cheaper. 10667 + Local aliases are introduced for symbols that are known to be 10668 semantically equivalent across shared libraries improving 10669 dynamic linking times. 10670 * Feedback directed optimization improvements: 10671 + Profiling of programs using C++ inline functions is now more 10672 reliable. 10673 + New time profiling determines typical order in which functions 10674 are executed. 10675 + A new function reordering pass (controlled by 10676 -freorder-functions) significantly reduces startup time of 10677 large applications. Until binutils support is completed, it is 10678 effective only with link-time optimization. 10679 + Feedback driven indirect call removal and devirtualization now 10680 handle cross-module calls when link-time optimization is 10681 enabled. 10682 10683 [4]New Languages and Language specific improvements 10684 10685 * Version 4.0 of the [5]OpenMP specification is now supported in the 10686 C and C++ compilers and starting with the 4.9.1 release also in the 10687 Fortran compiler. The new -fopenmp-simd option can be used to 10688 enable OpenMP's SIMD directives while ignoring other OpenMP 10689 directives. The new [6]-fsimd-cost-model= option permits to tune 10690 the vectorization cost model for loops annotated with OpenMP and 10691 Cilk Plus simd directives. -Wopenmp-simd warns when the current 10692 cost model overrides simd directives set by the user. 10693 * The -Wdate-time option has been added for the C, C++ and Fortran 10694 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 10695 macros are used. Those macros might prevent bit-wise-identical 10696 reproducible compilations. 10697 10698 [7]Ada 10699 10700 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 10701 10702 [8]C family 10703 10704 * Support for colorizing diagnostics emitted by GCC has been added. 10705 The [9]-fdiagnostics-color=auto will enable it when outputting to 10706 terminals, -fdiagnostics-color=always unconditionally. The 10707 GCC_COLORS environment variable can be used to customize the colors 10708 or disable coloring. If GCC_COLORS variable is present in the 10709 environment, the default is -fdiagnostics-color=auto, otherwise 10710 -fdiagnostics-color=never. 10711 Sample diagnostics output: 10712 $ g++ -fdiagnostics-color=always -S -Wall test.C 10713 test.C: In function `int foo()': 10714 test.C:1:14: warning: no return statement in function returning non-void [-W 10715 return-type] 10716 int foo () { } 10717 ^ 10718 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 10719 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 10720 template <int N> struct X { static const int value = X<N-1>::value; }; temp 10721 late struct X<1000>; 10722 ^ 10723 test.C:2:46: recursively required from `const int X<999>::value' 10724 test.C:2:46: required from `const int X<1000>::value' 10725 test.C:2:88: required from here 10726 10727 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 10728 10729 * With the new [10]#pragma GCC ivdep, the user can assert that there 10730 are no loop-carried dependencies which would prevent concurrent 10731 execution of consecutive iterations using SIMD (single instruction 10732 multiple data) instructions. 10733 * Support for Cilk Plus has been added and can be enabled with the 10734 -fcilkplus option. Cilk Plus is an extension to the C and C++ 10735 languages to support data and task parallelism. The present 10736 implementation follows ABI version 1.2; all features but _Cilk_for 10737 have been implemented. 10738 10739 C 10740 10741 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 10742 <stdatomic.h> header) are now supported. 10743 * ISO C11 generic selections (_Generic keyword) are now supported. 10744 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 10745 __thread) is now supported. 10746 * ISO C11 support is now at a similar level of completeness to ISO 10747 C99 support: substantially complete modulo bugs, extended 10748 identifiers (supported except for corner cases when 10749 -fextended-identifiers is used), floating-point issues (mainly but 10750 not entirely relating to optional C99 features from Annexes F and 10751 G) and the optional Annexes K (Bounds-checking interfaces) and L 10752 (Analyzability). 10753 * A new C extension __auto_type provides a subset of the 10754 functionality of C++11 auto in GNU C. 10755 10756 [11]C++ 10757 10758 * The G++ implementation of [12]C++1y return type deduction for 10759 normal functions has been updated to conform to [13]N3638, the 10760 proposal accepted into the working paper. Most notably, it adds 10761 decltype(auto) for getting decltype semantics rather than the 10762 template argument deduction semantics of plain auto: 10763 10764 int& f(); 10765 auto i1 = f(); // int 10766 decltype(auto) i2 = f(); // int& 10767 10768 * G++ supports [14]C++1y lambda capture initializers: 10769 10770 [x = 42]{ ... }; 10771 10772 Actually, they have been accepted since GCC 4.5, but now the 10773 compiler doesn't warn about them with -std=c++1y, and supports 10774 parenthesized and brace-enclosed initializers as well. 10775 * G++ supports [15]C++1y variable length arrays. G++ has supported 10776 GNU/C99-style VLAs for a long time, but now additionally supports 10777 initializers and lambda capture by reference. In C++1y mode G++ 10778 will complain about VLA uses that are not permitted by the draft 10779 standard, such as forming a pointer to VLA type or applying sizeof 10780 to a VLA variable. Note that it now appears that VLAs will not be 10781 part of C++14, but will be part of a separate document and then 10782 perhaps C++17. 10783 10784 void f(int n) { 10785 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 10786 [&a]{ for (int i : a) { cout << i << endl; } }(); 10787 &a; // error, taking address of VLA 10788 } 10789 10790 * G++ supports the [16]C++1y [[deprecated]] attribute modulo bugs in 10791 the underlying [[gnu::deprecated]] attribute. Classes and functions 10792 can be marked deprecated and a diagnostic message added: 10793 10794 class A; 10795 int bar(int n); 10796 #if __cplusplus > 201103 10797 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 10798 [[deprecated("bar is unsafe; use foo() instead")]] 10799 int bar(int n); 10800 10801 int foo(int n); 10802 class B; 10803 #endif 10804 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 10805 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 10806 o() instead 10807 10808 * G++ supports [17]C++1y digit separators. Long numeric literals can 10809 be subdivided with a single quote ' to enhance readability: 10810 10811 int i = 1048576; 10812 int j = 1'048'576; 10813 int k = 0x10'0000; 10814 int m = 0'004'000'000; 10815 int n = 0b0001'0000'0000'0000'0000'0000; 10816 10817 double x = 1.602'176'565e-19; 10818 double y = 1.602'176'565e-1'9; 10819 10820 * G++ supports [18]C++1y generic (polymorphic) lambdas. 10821 10822 // a functional object that will increment any type 10823 auto incr = [](auto x) { return x++; }; 10824 10825 * As a GNU extension, G++ supports explicit template parameter syntax 10826 for generic lambdas. This can be combined in the expected way with 10827 the standard auto syntax. 10828 10829 // a functional object that will add two like-type objects 10830 auto add = [] <typename T> (T a, T b) { return a + b; }; 10831 10832 * G++ supports unconstrained generic functions as specified by 4.1.2 10833 and 5.1.1 of [19]N3889: Concepts Lite Specification. Briefly, auto 10834 may be used as a type-specifier in a parameter declaration of any 10835 function declarator in order to introduce an implicit function 10836 template parameter, akin to generic lambdas. 10837 10838 // the following two function declarations are equivalent 10839 auto incr(auto x) { return x++; } 10840 template <typename T> 10841 auto incr(T x) { return x++; } 10842 10843 Runtime Library (libstdc++) 10844 10845 * [20]Improved support for C++11, including: 10846 + support for <regex>; 10847 + The associative containers in <map> and <set> and the 10848 unordered associative containers in <unordered_map> and 10849 <unordered_set> meet the allocator-aware container 10850 requirements; 10851 * [21]Improved experimental support for the upcoming ISO C++ 10852 standard, C++14, including: 10853 + fixing constexpr member functions without const; 10854 + implementation of the std::exchange() utility function; 10855 + addressing tuples by type; 10856 + implemention of std::make_unique; 10857 + implemention of std::shared_lock; 10858 + making std::result_of SFINAE-friendly; 10859 + adding operator() to std::integral_constant; 10860 + adding user-defined literals for standard library types 10861 std::basic_string, std::chrono::duration, and std::complex; 10862 + adding two range overloads to non-modifying sequence oprations 10863 std::equal and std::mismatch; 10864 + adding IO manipulators for quoted strings; 10865 + adding constexpr members to <utility>, <complex>, <chrono>, 10866 and some containers; 10867 + adding compile-time std::integer_sequence; 10868 + adding cleaner transformation traits; 10869 + making <functional>s operator functors easier to use and more 10870 generic; 10871 * An implementation of std::experimental::optional. 10872 * An implementation of std::experimental::string_view. 10873 * The non-standard function std::copy_exception has been deprecated 10874 and will be removed in a future version. std::make_exception_ptr 10875 should be used instead. 10876 10877 [22]Fortran 10878 10879 * Compatibility notice: 10880 + Module files: The version of the module files (.mod) has been 10881 incremented; additionally, module files are now compressed. 10882 Fortran MODULEs compiled by earlier GCC versions have to be 10883 recompiled, when they are USEd by files compiled with GCC 4.9. 10884 GCC 4.9 is not able to read .mod files of earlier GCC 10885 versions; attempting to do so gives an error message. Note: 10886 The ABI of the produced assembler data itself has not changed: 10887 object files and libraries are fully compatible with older 10888 versions (except as stated below). 10889 + ABI changes: 10890 o The [23]argument passing ABI has changed for scalar dummy 10891 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 10892 which have both the VALUE and the OPTIONAL attributes. 10893 o To support finalization the virtual table associated with 10894 polymorphic variables has changed. Code containing CLASS 10895 should be recompiled, including all files which define 10896 derived types involved in the type definition used by 10897 polymorphic variables. (Note: Due to the incremented 10898 module version, trying to mix old code with new code will 10899 usually give an error message.) 10900 + GNU Fortran no longer deallocates allocatable variables or 10901 allocatable components of variables declared in the main 10902 program. Since Fortran 2008, the standard explicitly states 10903 that variables declared in the Fortran main program 10904 automatically have the SAVE attribute. 10905 + When opening files, the close-on-exec flag is set if the 10906 system supports such a feature. This is generally considered 10907 good practice these days, but if there is a need to pass file 10908 descriptors to child processes the parent process must now 10909 remember to clear the close-on-exec flag by calling fcntl(), 10910 e.g. via ISO_C_BINDING, before executing the child process. 10911 * The deprecated command-line option -fno-whole-file has been 10912 removed. (-fwhole-file is the default since GCC 4.6.) 10913 -fwhole-file/-fno-whole-file continue to be accepted but do not 10914 influence the code generation. 10915 * The compiler no longer unconditionally warns about DO loops with 10916 zero iterations. This warning is now controlled by the -Wzerotrip 10917 option, which is implied by -Wall. 10918 * The new NO_ARG_CHECK attribute of the [24]!GCC$ directive can be 10919 used to disable the type-kind-rank (TKR) argument check for a dummy 10920 argument. The feature is similar to ISO/IEC TS 29133:2012's 10921 TYPE(*), except that it additionally also disables the rank check. 10922 Variables with NO_ARG_CHECK have to be dummy arguments and may only 10923 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 10924 to another NO_ARG_CHECK dummy argument; also the other constraints 10925 of TYPE(*) apply. The dummy arguments should be declared as scalar 10926 or assumed-size variable of type type(*) (recommended) - or of type 10927 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 10928 the data without further type or shape information is passed, 10929 similar to C's void*. Note that also TS 29113's 10930 type(*),dimension(..) accepts arguments of any type and rank; 10931 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 10932 descriptor which contains the array shape and stride of the 10933 argument. 10934 * [25]Fortran 2003: 10935 + Finalization is now supported. It is currently only done for a 10936 subset of those situations in which it should occur. 10937 + Experimental support for scalar character components with 10938 deferred length (i.e. allocatable string length) in derived 10939 types has been added. (Deferred-length character variables are 10940 supported since GCC 4.6.) 10941 * [26]Fortran 2008: 10942 + When STOP or ERROR STOP are used to terminate the execution 10943 and any exception (but inexact) is signaling, a warning is 10944 printed to ERROR_UNIT, indicating which exceptions are 10945 signaling. The [27]-ffpe-summary= command-line option can be 10946 used to fine-tune for which exceptions the warning should be 10947 shown. 10948 + Rounding on input (READ) is now handled on systems where 10949 strtod honours the rounding mode. (For output, rounding is 10950 supported since GCC 4.5.) Note that for input, the compatible 10951 rounding mode is handled as nearest (i.e., rounding to an even 10952 least significant [cf. IEC 60559:1989] for a tie, while 10953 compatible rounds away from zero in that case). 10954 10955 [28]Go 10956 10957 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 10958 10959 [29]New Targets and Target Specific Improvements 10960 10961 [30]AArch64 10962 10963 * The ARMv8-A crypto and CRC instructions are now supported through 10964 intrinsics. These are enabled when the architecture supports these 10965 and are available through the -march=armv8-a+crc and 10966 -march=armv8-a+crypto options. 10967 * Initial support for ILP32 has now been added to the compiler. This 10968 is now available through the command-line option -mabi=ilp32. 10969 Support for ILP32 is considered experimental as the ABI 10970 specification is still beta. 10971 * Coverage of more of the ISA including the SIMD extensions has been 10972 added. The Advanced SIMD intrinsics have also been improved. 10973 * The new local register allocator (LRA) is now on by default for the 10974 AArch64 back end. 10975 * The REE (Redundant extension elimination) pass has now been enabled 10976 by default for the AArch64 back end. 10977 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 10978 * Initial big.LITTLE tuning support for the combination of Cortex-A57 10979 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 10980 option. 10981 * A number of structural changes have been made to both the ARM and 10982 AArch64 back ends to facilitate improved code-generation. 10983 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 10984 has been added and can be enabled by giving the 10985 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 10986 default by configuring GCC with the --enable-fix-cortex-a53-835769 10987 option. 10988 10989 [31]ARC 10990 10991 * A port for Synopsys Designware ARC has been contributed by Embecosm 10992 and Synopsys Inc. 10993 10994 [32]ARM 10995 10996 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 10997 disabled by default. This was found to generate better code in only 10998 a small number of cases. It can be turned back on with the 10999 -mneon-for-64bits option. 11000 * Further support for the ARMv8-A architecture, notably implementing 11001 the restriction around IT blocks in the Thumb32 instruction set has 11002 been added. The -mrestrict-it option can be used with 11003 -march=armv7-a or the -march=armv7ve options to make code 11004 generation fully compatible with the deprecated instructions in 11005 ARMv8-A. 11006 * Support has now been added for the ARMv7ve variant of the 11007 architecture. This can be used by the -march=armv7ve option. 11008 * The ARMv8-A crypto and CRC instructions are now supported through 11009 intrinsics and are available through the -march=armv8-a+crc and 11010 mfpu=crypto-neon-fp-armv8 options. 11011 * LRA is now on by default for the ARM target. This can be turned off 11012 using the -mno-lra option. This option is a purely transitionary 11013 command-line option and will be removed in a future release. We are 11014 interested in any bug reports regarding functional and performance 11015 regressions with LRA. 11016 * A new option -mslow-flash-data to improve performance of programs 11017 fetching data on slow flash memory has now been introduced for the 11018 ARMv7-M profile cores. 11019 * A new option -mpic-data-is-text-relative for targets that allows 11020 data segments to be relative to text segments has been added. This 11021 is on by default for all targets except VxWorks RTP. 11022 * A number of infrastructural changes have been made to both the ARM 11023 and AArch64 back ends to facilitate improved code-generation. 11024 * GCC now supports Cortex-A12 and the Cortex-R7 through the 11025 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 11026 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 11027 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 11028 * Initial big.LITTLE tuning support for the combination of Cortex-A57 11029 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 11030 option. Similar support was added for the combination of Cortex-A15 11031 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 11032 * Further performance optimizations for the Cortex-A15 and the 11033 Cortex-M4 have been added. 11034 * A number of code generation improvements for Thumb2 to reduce code 11035 size when compiling for the M-profile processors. 11036 11037 [33]AVR 11038 11039 * A new command-line option -mfract-convert-truncate has been added. 11040 It allows compiler to use truncation instead of rounding towards 11041 zero for fractional fixed-point types. 11042 11043 [34]IA-32/x86-64 11044 11045 * -mfpmath=sse is now implied by -ffast-math on all targets where 11046 SSE2 is supported. 11047 * Intel AVX-512 support was added to GCC. That includes inline 11048 assembly support, new registers and extending existing ones, new 11049 intrinsics (covered by corresponding testsuite), and basic 11050 autovectorization. AVX-512 instructions are available via the 11051 following GCC switches: AVX-512 foundation instructions: -mavx512f, 11052 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 11053 reciprocal instructions: -mavx512er, AVX-512 conflict detection 11054 instructions: -mavx512cd. 11055 * It is now possible to call x86 intrinsics from select functions in 11056 a file that are tagged with the corresponding target attribute 11057 without having to compile the entire file with the -mxxx option. 11058 This improves the usability of x86 intrinsics and is particularly 11059 useful when doing [35]Function Multiversioning. 11060 * GCC now supports the new Intel microarchitecture named Silvermont 11061 through -march=silvermont. 11062 * GCC now supports the new Intel microarchitecture named Broadwell 11063 through -march=broadwell. 11064 * Optimizing for other Intel microarchitectures have been renamed to 11065 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 11066 * -march=generic has been retuned for better support of Intel core 11067 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 11068 Pentium-M, and Pentium4 based CPUs is no longer considered 11069 important for generic. 11070 * -mtune=intel can now be used to generate code running well on the 11071 most current Intel processors, which are Haswell and Silvermont for 11072 GCC 4.9. 11073 * Support to encode 32-bit assembly instructions in 16-bit format is 11074 now available through the -m16 command-line option. 11075 * Better inlining of memcpy and memset that is aware of value ranges 11076 and produces shorter alignment prologues. 11077 * -mno-accumulate-outgoing-args is now honored when unwind 11078 information is output. Argument accumulation is also now turned off 11079 for portions of programs optimized for size. 11080 * Support for new AMD family 15h processors (Excavator core) is now 11081 available through the -march=bdver4 and -mtune=bdver4 options. 11082 11083 [36]MSP430 11084 11085 * A new command-line option -mcpu= has been added to the MSP430 back 11086 end. This option is used to specify the ISA to be used. Accepted 11087 values are msp430 (the default), msp430x and msp430xv2. The ISA is 11088 no longer deduced from the -mmcu= option as there are far too many 11089 different MCU names. The -mmcu= option is still supported, and this 11090 is still used to select linker scripts and generate a C 11091 preprocessor symbol that will be recognised by the msp430.h header 11092 file. 11093 11094 [37]NDS32 11095 11096 * A new nds32 port supports the 32-bit architecture from Andes 11097 Technology Corporation. 11098 * The port provides initial support for the V2, V3, V3m instruction 11099 set architectures. 11100 11101 [38]Nios II 11102 11103 * A port for the Altera Nios II has been contributed by Mentor 11104 Graphics. 11105 11106 [39]PowerPC / PowerPC64 / RS6000 11107 11108 * GCC now supports Power ISA 2.07, which includes support for 11109 Hardware Transactional Memory (HTM), Quadword atomics and several 11110 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 11111 integer and decimal integer operations. 11112 * Support for the POWER8 processor is now available through the 11113 -mcpu=power8 and -mtune=power8 options. 11114 * The libitm library has been modified to add a HTM fastpath that 11115 automatically uses POWER's HTM hardware instructions when it is 11116 executing on a HTM enabled processor. 11117 * Support for the new powerpc64le-linux platform has been added. It 11118 defaults to generating code that conforms to the ELFV2 ABI. 11119 11120 S/390, System z 11121 11122 * Support for the Transactional Execution Facility included with the 11123 IBM zEnterprise zEC12 processor has been added. A set of GCC style 11124 builtins as well as XLC style builtins are provided. The builtins 11125 are enabled by default when using the -march=zEC12 option but can 11126 explicitly be disabled with -mno-htm. Using the GCC builtins also 11127 libitm supports hardware transactions on S/390. 11128 * The hotpatch features allows to prepare functions for hotpatching. 11129 A certain amount of bytes is reserved before the function entry 11130 label plus a NOP is inserted at its very beginning to implement a 11131 backward jump when applying a patch. The feature can either be 11132 enabled per compilation unit via the command-line option -mhotpatch 11133 or per function using the hotpatch attribute. 11134 * The shrink wrap optimization is now supported on S/390 and enabled 11135 by default. 11136 * A major rework of the routines to determine which registers need to 11137 be saved and restored in function prologue/epilogue now allow to 11138 use floating point registers as save slots. This will happen for 11139 certain leaf function with -march=z10 or higher. 11140 * The LRA rtl pass replaces reload by default on S/390. 11141 11142 [40]RX 11143 11144 * The port now allows to specify the RX100, RX200, and RX600 11145 processors with the command-line options -mcpu=rx100, -mcpu=rx200 11146 and -mcpu=rx600. 11147 11148 [41]SH 11149 11150 * Minor improvements to code generated for integer arithmetic and 11151 code that involves the T bit. 11152 * Added support for the SH2A clips and clipu instructions. The 11153 compiler will now try to utilize them for min/max expressions such 11154 as max (-128, min (127, x)). 11155 * Added support for the cmp/str instruction through built-in 11156 functions such as __builtin_strlen. When not optimizing for size, 11157 the compiler will now expand calls to e.g. strlen as an inlined 11158 sequences which utilize the cmp/str instruction. 11159 * Improved code generated around volatile memory loads and stores. 11160 * The option -mcbranchdi has been deprecated. Specifying it will 11161 result in a warning and will not influence code generation. 11162 * The option -mcmpeqdi has been deprecated. Specifying it will result 11163 in a warning and will not influence code generation. 11164 11165 [42]GCC 4.9.1 11166 11167 This is the [43]list of problem reports (PRs) from GCC's bug tracking 11168 system that are known to be fixed in the 4.9.1 release. This list might 11169 not be complete (that is, it is possible that some PRs that have been 11170 fixed are not listed here). 11171 11172 Version 4.0 of the OpenMP specification is supported even in Fortran, 11173 not just C and C++. 11174 11175 [44]GCC 4.9.2 11176 11177 This is the [45]list of problem reports (PRs) from GCC's bug tracking 11178 system that are known to be fixed in the 4.9.2 release. This list might 11179 not be complete (that is, it is possible that some PRs that have been 11180 fixed are not listed here). 11181 11182 [46]GCC 4.9.3 11183 11184 This is the [47]list of problem reports (PRs) from GCC's bug tracking 11185 system that are known to be fixed in the 4.9.3 release. This list might 11186 not be complete (that is, it is possible that some PRs that have been 11187 fixed are not listed here). 11188 11189 [48]GCC 4.9.4 11190 11191 This is the [49]list of problem reports (PRs) from GCC's bug tracking 11192 system that are known to be fixed in the 4.9.4 release. This list might 11193 not be complete (that is, it is possible that some PRs that have been 11194 fixed are not listed here). 11195 11196 11197 For questions related to the use of GCC, please consult these web 11198 pages and the [50]GCC manuals. If that fails, the 11199 [51]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11200 web pages and the development of GCC are welcome on our developer 11201 list at [52]gcc (a] gcc.gnu.org. All of [53]our lists have public 11202 archives. 11203 11204 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 11205 distribution of this entire article is permitted in any medium, 11206 provided this notice is preserved. 11207 11208 These pages are [55]maintained by the GCC team. Last modified 11209 2025-01-31. 11210 11211 References 11212 11213 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 11214 2. https://gcc.gnu.org/PR60825 11215 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 11216 4. https://gcc.gnu.org/gcc-4.9/changes.html#languages 11217 5. https://www.openmp.org/specifications/ 11218 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 11219 7. https://gcc.gnu.org/gcc-4.9/changes.html#ada 11220 8. https://gcc.gnu.org/gcc-4.9/changes.html#c-family 11221 9. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 11222 10. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 11223 11. https://gcc.gnu.org/gcc-4.9/changes.html#cxx 11224 12. https://gcc.gnu.org/projects/cxx1y.html 11225 13. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 11226 14. https://gcc.gnu.org/projects/cxx1y.html 11227 15. https://gcc.gnu.org/projects/cxx1y.html 11228 16. https://gcc.gnu.org/projects/cxx1y.html 11229 17. https://gcc.gnu.org/projects/cxx1y.html 11230 18. https://gcc.gnu.org/projects/cxx1y.html 11231 19. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 11232 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 11233 21. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 11234 22. https://gcc.gnu.org/gcc-4.9/changes.html#fortran 11235 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 11236 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 11237 25. https://gcc.gnu.org/wiki/Fortran2003Status 11238 26. https://gcc.gnu.org/wiki/Fortran2008Status 11239 27. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 11240 28. https://gcc.gnu.org/gcc-4.9/changes.html#go 11241 29. https://gcc.gnu.org/gcc-4.9/changes.html#targets 11242 30. https://gcc.gnu.org/gcc-4.9/changes.html#aarch64 11243 31. https://gcc.gnu.org/gcc-4.9/changes.html#arc 11244 32. https://gcc.gnu.org/gcc-4.9/changes.html#arm 11245 33. https://gcc.gnu.org/gcc-4.9/changes.html#avr 11246 34. https://gcc.gnu.org/gcc-4.9/changes.html#x86 11247 35. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 11248 36. https://gcc.gnu.org/gcc-4.9/changes.html#msp430 11249 37. https://gcc.gnu.org/gcc-4.9/changes.html#nds32 11250 38. https://gcc.gnu.org/gcc-4.9/changes.html#nios2 11251 39. https://gcc.gnu.org/gcc-4.9/changes.html#powerpc 11252 40. https://gcc.gnu.org/gcc-4.9/changes.html#rx 11253 41. https://gcc.gnu.org/gcc-4.9/changes.html#sh 11254 42. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.1 11255 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 11256 44. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.2 11257 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 11258 46. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.3 11259 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 11260 48. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.4 11261 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 11262 50. https://gcc.gnu.org/onlinedocs/ 11263 51. mailto:gcc-help (a] gcc.gnu.org 11264 52. mailto:gcc (a] gcc.gnu.org 11265 53. https://gcc.gnu.org/lists.html 11266 54. https://www.fsf.org/ 11267 55. https://gcc.gnu.org/about.html 11268 ====================================================================== 11269 http://gcc.gnu.org/gcc-4.8/index.html 11270 11271 GCC 4.8 Release Series 11272 11273 (This release series is no longer supported.) 11274 11275 June 23, 2015 11276 11277 The [1]GNU project and the GCC developers are pleased to announce the 11278 release of GCC 4.8.5. 11279 11280 This release is a bug-fix release, containing fixes for regressions in 11281 GCC 4.8.4 relative to previous releases of GCC. 11282 11283 Release History 11284 11285 GCC 4.8.5 11286 June 23, 2015 ([2]changes, [3]documentation) 11287 11288 GCC 4.8.4 11289 December 19, 2014 ([4]changes, [5]documentation) 11290 11291 GCC 4.8.3 11292 May 22, 2014 ([6]changes, [7]documentation) 11293 11294 GCC 4.8.2 11295 October 16, 2013 ([8]changes, [9]documentation) 11296 11297 GCC 4.8.1 11298 May 31, 2013 ([10]changes, [11]documentation) 11299 11300 GCC 4.8.0 11301 March 22, 2013 ([12]changes, [13]documentation) 11302 11303 References and Acknowledgements 11304 11305 GCC used to stand for the GNU C Compiler, but since the compiler 11306 supports several other languages aside from C, it now stands for the 11307 GNU Compiler Collection. 11308 11309 A list of [14]successful builds is updated as new information becomes 11310 available. 11311 11312 The GCC developers would like to thank the numerous people that have 11313 contributed new features, improvements, bug fixes, and other changes as 11314 well as test results to GCC. This [15]amazing group of volunteers is 11315 what makes GCC successful. 11316 11317 For additional information about GCC please refer to the [16]GCC 11318 project web site or contact the [17]GCC development mailing list. 11319 11320 To obtain GCC please use [18]our mirror sites or [19]our version 11321 control system. 11322 11323 11324 For questions related to the use of GCC, please consult these web 11325 pages and the [20]GCC manuals. If that fails, the 11326 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11327 web pages and the development of GCC are welcome on our developer 11328 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 11329 archives. 11330 11331 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 11332 distribution of this entire article is permitted in any medium, 11333 provided this notice is preserved. 11334 11335 These pages are [25]maintained by the GCC team. Last modified 11336 2022-10-26. 11337 11338 References 11339 11340 1. http://www.gnu.org/ 11341 2. https://gcc.gnu.org/gcc-4.8/changes.html 11342 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 11343 4. https://gcc.gnu.org/gcc-4.8/changes.html 11344 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 11345 6. https://gcc.gnu.org/gcc-4.8/changes.html 11346 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 11347 8. https://gcc.gnu.org/gcc-4.8/changes.html 11348 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 11349 10. https://gcc.gnu.org/gcc-4.8/changes.html 11350 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 11351 12. https://gcc.gnu.org/gcc-4.8/changes.html 11352 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 11353 14. https://gcc.gnu.org/gcc-4.8/buildstat.html 11354 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11355 16. https://gcc.gnu.org/index.html 11356 17. mailto:gcc (a] gcc.gnu.org 11357 18. https://gcc.gnu.org/mirrors.html 11358 19. https://gcc.gnu.org/git.html 11359 20. https://gcc.gnu.org/onlinedocs/ 11360 21. mailto:gcc-help (a] gcc.gnu.org 11361 22. mailto:gcc (a] gcc.gnu.org 11362 23. https://gcc.gnu.org/lists.html 11363 24. https://www.fsf.org/ 11364 25. https://gcc.gnu.org/about.html 11365 ====================================================================== 11366 http://gcc.gnu.org/gcc-4.8/changes.html 11367 11368 GCC 4.8 Release Series 11369 Changes, New Features, and Fixes 11370 11371 Caveats 11372 11373 GCC now uses C++ as its implementation language. This means that to 11374 build GCC from sources, you will need a C++ compiler that understands 11375 C++ 2003. For more details on the rationale and specific changes, 11376 please refer to the [1]C++ conversion page. 11377 11378 To enable the Graphite framework for loop optimizations you now need 11379 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 11380 the [2]GCC infrastructure directory. The installation manual contains 11381 more information about requirements to build GCC. 11382 11383 GCC now uses a more aggressive analysis to derive an upper bound for 11384 the number of iterations of loops using constraints imposed by language 11385 standards. This may cause non-conforming programs to no longer work as 11386 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 11387 option, -fno-aggressive-loop-optimizations, was added to disable this 11388 aggressive analysis. In some loops that have known constant number of 11389 iterations, but undefined behavior is known to occur in the loop before 11390 reaching or during the last iteration, GCC will warn about the 11391 undefined behavior in the loop instead of deriving lower upper bound of 11392 the number of iterations for the loop. The warning can be disabled with 11393 -Wno-aggressive-loop-optimizations. 11394 11395 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 11396 for the layout of vectors that could lead to wrong code being 11397 generated. Vectors larger than 8 bytes in size are now by default 11398 aligned to an 8-byte boundary. This is an ABI change: code that makes 11399 explicit use of vector types may be incompatible with binary objects 11400 built with older versions of GCC. Auto-vectorized code is not affected 11401 by this change. 11402 11403 More information on porting to GCC 4.8 from previous versions of GCC 11404 can be found in the [3]porting guide for this release. 11405 11406 General Optimizer Improvements (and Changes) 11407 11408 * DWARF4 is now the default when generating DWARF debug information. 11409 When -g is used on a platform that uses DWARF debugging 11410 information, GCC will now default to -gdwarf-4 11411 -fno-debug-types-section. 11412 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 11413 consumers support DWARF4 by default. Before GCC 4.8 the default 11414 version used was DWARF2. To make GCC 4.8 generate an older DWARF 11415 version use -g together with -gdwarf-2 or -gdwarf-3. The default 11416 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 11417 * A new general optimization level, -Og, has been introduced. It 11418 addresses the need for fast compilation and a superior debugging 11419 experience while providing a reasonable level of run-time 11420 performance. Overall experience for development should be better 11421 than the default optimization level -O0. 11422 * A new option -ftree-partial-pre was added to control the partial 11423 redundancy elimination (PRE) optimization. This option is enabled 11424 by default at the -O3 optimization level, and it makes PRE more 11425 aggressive. 11426 * The option -fconserve-space has been removed; it was no longer 11427 useful on most targets since GCC supports putting variables into 11428 BSS without making them common. 11429 * The struct reorg and matrix reorg optimizations (command-line 11430 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 11431 removed. They did not always work correctly, nor did they work with 11432 link-time optimization (LTO), hence were only applicable to 11433 programs consisting of a single translation unit. 11434 * Several scalability bottle-necks have been removed from GCC's 11435 optimization passes. Compilation of extremely large functions, e.g. 11436 due to the use of the flatten attribute in the "Eigen" C++ linear 11437 algebra templates library, is significantly faster than previous 11438 releases of GCC. 11439 * Link-time optimization (LTO) improvements: 11440 + LTO partitioning has been rewritten for better reliability and 11441 maintanibility. Several important bugs leading to link 11442 failures have been fixed. 11443 * Interprocedural optimization improvements: 11444 + A new symbol table has been implemented. It builds on existing 11445 callgraph and varpool modules and provide a new API. Unusual 11446 symbol visibilities and aliases are handled more consistently 11447 leading to, for example, more aggressive unreachable code 11448 removal with LTO. 11449 + The inline heuristic can now bypass limits on the size of of 11450 inlined functions when the inlining is particularly 11451 profitable. This happens, for example, when loop bounds or 11452 array strides get propagated. 11453 + Values passed through aggregates (either by value or 11454 reference) are now propagated at the inter-procedural level 11455 leading to better inlining decisions (for example in the case 11456 of Fortran array descriptors) and devirtualization. 11457 * [4]AddressSanitizer , a fast memory error detector, has been added 11458 and can be enabled via -fsanitize=address. Memory access 11459 instructions will be instrumented to detect heap-, stack-, and 11460 global-buffer overflow as well as use-after-free bugs. To get nicer 11461 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 11462 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 11463 x86-64 Darwin. 11464 * [5]ThreadSanitizer has been added and can be enabled via 11465 -fsanitize=thread. Instructions will be instrumented to detect data 11466 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 11467 * A new local register allocator (LRA) has been implemented, which 11468 replaces the 26 year old reload pass and improves generated code 11469 quality. For now it is active on the IA-32 and x86-64 targets. 11470 * Support for transactional memory has been implemented on the 11471 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 11472 Alpha. 11473 11474 New Languages and Language specific improvements 11475 11476 C family 11477 11478 * Each diagnostic emitted now includes the original source line and a 11479 caret '^' indicating the column. The option 11480 -fno-diagnostics-show-caret suppresses this information. 11481 * The option -ftrack-macro-expansion=2 is now enabled by default. 11482 This allows the compiler to display the macro expansion stack in 11483 diagnostics. Combined with the caret information, an example 11484 diagnostic showing these two features is: 11485 11486 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 11487 ') 11488 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 11489 _b = (B); __a < __b ? __b : __a; }) 11490 11491 ^ 11492 t.c:7:7: note: in expansion of macro 'MYMAX' 11493 X = MYMAX(P, F); 11494 ^ 11495 11496 * A new -Wsizeof-pointer-memaccess warning has been added (also 11497 enabled by -Wall) to warn about suspicious length parameters to 11498 certain string and memory built-in functions if the argument uses 11499 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 11500 (ptr)); if ptr is not an array, but a pointer, and suggests a 11501 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 11502 * The new option -Wpedantic is an alias for -pedantic, which is now 11503 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 11504 -Wno-error=pedantic work in the same way as for any other -W 11505 option. One caveat is that -Werror=pedantic is not equivalent to 11506 -pedantic-errors, since the latter makes into errors some warnings 11507 that are not controlled by -Wpedantic, and the former only affects 11508 diagnostics that are disabled when using -Wno-pedantic. 11509 * The option -Wshadow no longer warns if a declaration shadows a 11510 function declaration, unless the former declares a function or 11511 pointer to function, because this is [6]a common and valid case in 11512 real-world code. 11513 11514 [7]C++ 11515 11516 * G++ now implements the [8]C++11 thread_local keyword; this differs 11517 from the GNU __thread keyword primarily in that it allows dynamic 11518 initialization and destruction semantics. Unfortunately, this 11519 support requires a run-time penalty for references to 11520 non-function-local thread_local variables defined in a different 11521 translation unit even if they don't need dynamic initialization, so 11522 users may want to continue to use __thread for TLS variables with 11523 static initialization semantics. 11524 If the programmer can be sure that no use of the variable in a 11525 non-defining TU needs to trigger dynamic initialization (either 11526 because the variable is statically initialized, or a use of the 11527 variable in the defining TU will be executed before any uses in 11528 another TU), they can avoid this overhead with the 11529 -fno-extern-tls-init option. 11530 OpenMP threadprivate variables now also support dynamic 11531 initialization and destruction by the same mechanism. 11532 * G++ now implements the [9]C++11 attribute syntax, e.g. 11533 11534 [[noreturn]] void f(); 11535 11536 and also the alignment specifier, e.g. 11537 11538 alignas(double) int i; 11539 11540 * G++ now implements [10]C++11 inheriting constructors, e.g. 11541 11542 struct A { A(int); }; 11543 struct B: A { using A::A; }; // defines B::B(int) 11544 B b(42); // OK 11545 11546 * As of GCC 4.8.1, G++ implements the change to decltype semantics 11547 from [11]N3276. 11548 11549 struct A f(); 11550 decltype(f()) g(); // OK, return type of f() is not required to be complete. 11551 11552 * As of GCC 4.8.1, G++ implements [12]C++11 ref-qualifiers, e.g. 11553 11554 struct A { int f() &; }; 11555 int i = A().f(); // error, f() requires an lvalue object 11556 11557 * G++ now supports a -std=c++1y option for experimentation with 11558 features proposed for the next revision of the standard, expected 11559 around 2014. Currently the only difference from -std=c++11 is 11560 support for return type deduction in normal functions, as proposed 11561 in [13]N3386. Status of C++1y features in GCC 4.8 can be found 11562 [14]here. 11563 * The G++ namespace association extension, __attribute ((strong)), 11564 has been deprecated. Inline namespaces should be used instead. 11565 * G++ now supports a -fext-numeric-literal option to control whether 11566 GNU numeric literal suffixes are accepted as extensions or 11567 processed as C++11 user-defined numeric literal suffixes. The flag 11568 is on (use suffixes for GNU literals) by default for -std=gnu++*, 11569 and -std=c++98. The flag is off (use suffixes for user-defined 11570 literals) by default for -std=c++11 and later. 11571 11572 Runtime Library (libstdc++) 11573 11574 * [15]Improved experimental support for the new ISO C++ standard, 11575 C++11, including: 11576 + forward_list meets the allocator-aware container requirements; 11577 + this_thread::sleep_for(), this_thread::sleep_until() and 11578 this_thread::yield() are defined without requiring the 11579 configure option --enable-libstdcxx-time; 11580 * Improvements to <random>: 11581 + SSE optimized normal_distribution. 11582 + Use of hardware RNG instruction for random_device on new x86 11583 processors (requires the assembler to support the 11584 instruction.) 11585 and <ext/random>: 11586 + New random number engine simd_fast_mersenne_twister_engine 11587 with an optimized SSE implementation. 11588 + New random number distributions beta_distribution, 11589 normal_mv_distribution, rice_distribution, 11590 nakagami_distribution, pareto_distribution, k_distribution, 11591 arcsine_distribution, hoyt_distribution. 11592 * Added --disable-libstdcxx-verbose configure option to disable 11593 diagnostic messages issued when a process terminates abnormally. 11594 This may be useful for embedded systems to reduce the size of 11595 executables that link statically to the library. 11596 11597 [16]Fortran 11598 11599 * Compatibility notice: 11600 + Module files: The version of module files (.mod) has been 11601 incremented. Fortran MODULEs compiled by earlier GCC versions 11602 have to be recompiled, when they are USEd by files compiled 11603 with GCC 4.8. GCC 4.8 is not able to read .mod files created 11604 by earlier versions; attempting to do so gives an error 11605 message. 11606 Note: The ABI of the produced assembler data itself has not 11607 changed; object files and libraries are fully compatible with 11608 older versions except as noted below. 11609 + ABI: Some internal names (used in the assembler/object file) 11610 have changed for symbols declared in the specification part of 11611 a module. If an affected module - or a file using it via use 11612 association - is recompiled, the module and all files which 11613 directly use such symbols have to be recompiled as well. This 11614 change only affects the following kind of module symbols: 11615 o Procedure pointers. Note: C-interoperable function 11616 pointers (type(c_funptr)) are not affected nor are 11617 procedure-pointer components. 11618 o Deferred-length character strings. 11619 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 11620 backtrace at an arbitrary place in user code; program execution 11621 continues normally afterwards. 11622 * The [18]-Wc-binding-type warning option has been added (disabled by 11623 default). It warns if the a variable might not be C interoperable; 11624 in particular, if the variable has been declared using an intrinsic 11625 type with default kind instead of using a kind parameter defined 11626 for C interoperability in the intrinsic ISO_C_Binding module. 11627 Before, this warning was always printed. The -Wc-binding-type 11628 option is enabled by -Wall. 11629 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 11630 options have been added, which diagnose when code is inserted for 11631 automatic (re)allocation of a variable during assignment. This 11632 option can be used to decide whether it is safe to use 11633 [20]-fno-realloc-lhs. Additionally, it can be used to find 11634 automatic (re)allocation in hot loops. (For arrays, replacing 11635 "var=" by "var(:)=" disables the automatic reallocation.) 11636 * The [21]-Wcompare-reals command-line option has been added. When 11637 this is set, warnings are issued when comparing REAL or COMPLEX 11638 types for equality and inequality; consider replacing a == b by 11639 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by 11640 -Wextra. 11641 * The [22]-Wtarget-lifetime command-line option has been added 11642 (enabled with -Wall), which warns if the pointer in a pointer 11643 assignment might outlive its target. 11644 * Reading floating point numbers which use "q" for the exponential 11645 (such as 4.0q0) is now supported as vendor extension for better 11646 compatibility with old data files. It is strongly recommended to 11647 use for I/O the equivalent but standard conforming "e" (such as 11648 4.0e0). 11649 (For Fortran source code, consider replacing the "q" in 11650 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 11651 suitable qp). Note that - in Fortran source code - replacing "q" by 11652 a simple "e" is not equivalent.) 11653 * The GFORTRAN_TMPDIR environment variable for specifying a 11654 non-default directory for files opened with STATUS="SCRATCH", is 11655 not used anymore. Instead gfortran checks the POSIX/GNU standard 11656 TMPDIR environment variable. If TMPDIR is not defined, gfortran 11657 falls back to other methods to determine the directory for 11658 temporary files as documented in the [23]user manual. 11659 * [24]Fortran 2003: 11660 + Support for unlimited polymorphic variables (CLASS(*)) has 11661 been added. Nonconstant character lengths are not yet 11662 supported. 11663 * [25]TS 29113: 11664 + Assumed types (TYPE(*)) are now supported. 11665 + Experimental support for assumed-rank arrays (dimension(..)) 11666 has been added. Note that currently gfortran's own array 11667 descriptor is used, which is different from the one defined in 11668 TS29113, see [26]gfortran's header file or use the [27]Chasm 11669 Language Interoperability Tools. 11670 11671 [28]Go 11672 11673 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 11674 release. 11675 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 11676 release. The library support is not quite complete. 11677 * Go has been tested on GNU/Linux and Solaris platforms for various 11678 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 11679 work on other platforms as well. 11680 11681 [29]New Targets and Target Specific Improvements 11682 11683 [30]AArch64 11684 11685 * A new port has been added to support AArch64, the new 64-bit 11686 architecture from ARM. Note that this is a separate port from the 11687 existing 32-bit ARM port. 11688 * The port provides initial support for the Cortex-A53 and the 11689 Cortex-A57 processors with the command line options 11690 -mcpu=cortex-a53 and -mcpu=cortex-a57. 11691 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 11692 has been added and can be enabled by giving the 11693 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 11694 default by configuring GCC with the --enable-fix-cortex-a53-835769 11695 option. 11696 11697 [31]ARM 11698 11699 * Initial support has been added for the AArch32 extensions defined 11700 in the ARMv8 architecture. 11701 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 11702 * A new option, -mcpu=marvell-pj4, has been added to generate code 11703 for the Marvell PJ4 processor. 11704 * The compiler can now automatically generate the VFMA, VFMS, REVSH 11705 and REV16 instructions. 11706 * A new vectorizer cost model for Advanced SIMD configurations to 11707 improve the auto-vectorization strategies used. 11708 * The scheduler now takes into account the number of live registers 11709 to reduce the amount of spilling that can occur. This should 11710 improve code performance in large functions. The limit can be 11711 removed by using the option -fno-sched-pressure. 11712 * Improvements have been made to the Marvell iWMMX code generation 11713 and support for the iWMMX2 SIMD unit has been added. The option 11714 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 11715 * A number of code generation improvements for Thumb2 to reduce code 11716 size when compiling for the M-profile processors. 11717 * The RTEMS (arm-rtems) port has been updated to use the EABI. 11718 * Code generation support for the old FPA and Maverick floating-point 11719 architectures has been removed. Ports that previously relied on 11720 these features have also been removed. This includes the targets: 11721 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 11722 + arm*-*-elf (use arm*-*-eabi) 11723 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 11724 + arm*-*-ecos-elf (no alternative) 11725 + arm*-*-freebsd (no alternative) 11726 + arm*-wince-pe* (no alternative). 11727 11728 [32]AVR 11729 11730 * Support for the command-line option -mshort-calls has been removed. 11731 It was deprecated in GCC 4.7. 11732 * The configure option --with-avrlibc supported since GCC 4.7.2 is 11733 turned on per default for all non-RTEMS configurations. This option 11734 arranges for a better integration of [33]AVR-LibC with avr-gcc. For 11735 technical details, see [34]PR54461. To turn off the option in 11736 non-AVR-LibC configurations, use --with-avrlibc=no. If the compiler 11737 is configured for RTEMS, the option is always turned off. 11738 * Support for the "Embedded C" fixed-point has been added. For 11739 details, see the [35]GCC wiki and the [36]user manual. The support 11740 is not complete. 11741 * A new print modifier %r for register operands in inline assembler 11742 is supported. It will print the raw register number without the 11743 register prefix 'r': 11744 /* Return the most significant byte of 'val', a 64-bit value. */ 11745 11746 unsigned char msb (long long val) 11747 { 11748 unsigned char c; 11749 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 11750 return c; 11751 } 11752 The inline assembler in this example will generate code like 11753 mov r24, 8+7 11754 provided c is allocated to R24 and val is allocated to R8...R15. 11755 This works because the GNU assembler accepts plain register numbers 11756 without register prefix. 11757 * Static initializers with 3-byte symbols are supported now: 11758 extern const __memx char foo; 11759 const __memx void *pfoo = &foo; 11760 This requires at least Binutils 2.23. 11761 11762 IA-32/x86-64 11763 11764 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 11765 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 11766 stack alignment, this is ABI incompatible and intended to be used 11767 in controlled environments where stack space is an important 11768 limitation. This option will lead to wrong code when functions 11769 compiled with 16 byte stack alignment (such as functions from a 11770 standard library) are called with misaligned stack. In this case, 11771 SSE instructions may lead to misaligned memory access traps. In 11772 addition, variable arguments will be handled incorrectly for 16 11773 byte aligned objects (including x87 long double and __int128), 11774 leading to wrong results. You must build all modules with 11775 -mpreferred-stack-boundary=3, including any libraries. This 11776 includes the system libraries and startup modules. 11777 * Support for the new Intel processor codename Broadwell with RDSEED, 11778 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 11779 -mrdseed command-line options. 11780 * Support for the Intel RTM and HLE intrinsics, built-in functions 11781 and code generation is available via -mrtm and -mhle. 11782 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 11783 Intrinsics and built-in functions are available via -mfxsr, -mxsave 11784 and -mxsaveopt respectively. 11785 * New -maddress-mode=[short|long] options for x32. 11786 -maddress-mode=short overrides default 64-bit addresses to 32-bit 11787 by emitting the 0x67 address-size override prefix. This is the 11788 default address mode for x32. 11789 * New built-in functions to detect run-time CPU type and ISA: 11790 + A built-in function __builtin_cpu_is has been added to detect 11791 if the run-time CPU is of a particular type. It returns a 11792 positive integer on a match and zero otherwise. It accepts one 11793 string literal argument, the CPU name. For example, 11794 __builtin_cpu_is("westmere") returns a positive integer if the 11795 run-time CPU is an Intel Core i7 Westmere processor. Please 11796 refer to the [37]user manual for the list of valid CPU names 11797 recognized. 11798 + A built-in function __builtin_cpu_supports has been added to 11799 detect if the run-time CPU supports a particular ISA feature. 11800 It returns a positive integer on a match and zero otherwise. 11801 It accepts one string literal argument, the ISA feature. For 11802 example, __builtin_cpu_supports("ssse3") returns a positive 11803 integer if the run-time CPU supports SSSE3 instructions. 11804 Please refer to the [38]user manual for the list of valid ISA 11805 names recognized. 11806 Caveat: If these built-in functions are called before any static 11807 constructors are invoked, like during IFUNC initialization, then 11808 the CPU detection initialization must be explicitly run using this 11809 newly provided built-in function, __builtin_cpu_init. The 11810 initialization needs to be done only once. For example, this is how 11811 the invocation would look like inside an IFUNC initializer: 11812 static void (*some_ifunc_resolver(void))(void) 11813 { 11814 __builtin_cpu_init(); 11815 if (__builtin_cpu_is("amdfam10h") ... 11816 if (__builtin_cpu_supports("popcnt") ... 11817 } 11818 11819 * Function Multiversioning Support with G++: 11820 It is now possible to create multiple function versions each 11821 targeting a specific processor and/or ISA. Function versions have 11822 the same signature but different target attributes. For example, 11823 here is a program with function versions: 11824 __attribute__ ((target ("default"))) 11825 int foo(void) 11826 { 11827 return 1; 11828 } 11829 11830 __attribute__ ((target ("sse4.2"))) 11831 int foo(void) 11832 { 11833 return 2; 11834 } 11835 11836 int main (void) 11837 { 11838 int (*p) = &foo; 11839 assert ((*p)() == foo()); 11840 return 0; 11841 } 11842 11843 Please refer to this [39]wiki for more information. 11844 * The x86 back end has been improved to allow option -fschedule-insns 11845 to work reliably. This option can be used to schedule instructions 11846 better and leads to improved performace in certain cases. 11847 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 11848 from the Mingw-w64 trunk. 11849 * Support for new AMD family 15h processors (Steamroller core) is now 11850 available through the -march=bdver3 and -mtune=bdver3 options. 11851 * Support for new AMD family 16h processors (Jaguar core) is now 11852 available through the -march=btver2 and -mtune=btver2 options. 11853 11854 [40]FRV 11855 11856 * This target now supports the -fstack-usage command-line option. 11857 11858 [41]MIPS 11859 11860 * GCC can now generate code specifically for the R4700, Broadcom XLP 11861 and MIPS 34kn processors. The associated -march options are 11862 -march=r4700, -march=xlp and -march=34kn respectively. 11863 * GCC now generates better DSP code for MIPS 74k cores thanks to 11864 further scheduling optimizations. 11865 * The MIPS port now supports the -fstack-check option. 11866 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 11867 * Previous versions of GCC would silently accept -fpic and -fPIC for 11868 -mno-abicalls targets like mips*-elf. This combination was not 11869 intended or supported, and did not generate position-independent 11870 code. GCC 4.8 now reports an error when this combination is used. 11871 11872 [42]PowerPC / PowerPC64 / RS6000 11873 11874 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 11875 restore or update the VRSAVE register by default. The respective 11876 operating systems manage the VRSAVE register directly. 11877 * Large TOC support has been added for AIX through the command line 11878 option -mcmodel=large. 11879 * Native Thread-Local Storage support has been added for AIX. 11880 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 11881 when targetting processors that support those hardware features on 11882 AIX 6.1 and above. 11883 11884 [43]RX 11885 11886 * This target will now issue a warning message whenever multiple fast 11887 interrupt handlers are found in the same compilation unit. This 11888 feature can be turned off by the new 11889 -mno-warn-multiple-fast-interrupts command-line option. 11890 11891 S/390, System z 11892 11893 * Support for the IBM zEnterprise zEC12 processor has been added. 11894 When using the -march=zEC12 option, the compiler will generate code 11895 making use of the following new instructions: 11896 + load and trap instructions 11897 + 2 new compare and trap instructions 11898 + rotate and insert selected bits - without CC clobber 11899 The -mtune=zEC12 option enables zEC12 specific instruction 11900 scheduling without making use of new instructions. 11901 * Register pressure sensitive instruction scheduling is enabled by 11902 default. 11903 * The ifunc function attribute is enabled by default. 11904 * memcpy and memcmp invokations on big memory chunks or with run time 11905 lengths are not generated inline anymore when tuning for z10 or 11906 higher. The purpose is to make use of the IFUNC optimized versions 11907 in Glibc. 11908 11909 [44]SH 11910 11911 * The default alignment settings have been reduced to be less 11912 aggressive. This results in more compact code for optimization 11913 levels other than -Os. 11914 * Improved support for the __atomic built-in functions: 11915 + A new option -matomic-model=model selects the model for the 11916 generated atomic sequences. The following models are 11917 supported: 11918 11919 soft-gusa 11920 Software gUSA sequences (SH3* and SH4* only). On 11921 SH4A targets this will now also partially utilize 11922 the movco.l and movli.l instructions. This is the 11923 default when the target is sh3*-*-linux* or 11924 sh4*-*-linux*. 11925 11926 hard-llcs 11927 Hardware movco.l / movli.l sequences (SH4A only). 11928 11929 soft-tcb 11930 Software thread control block sequences. 11931 11932 soft-imask 11933 Software interrupt flipping sequences (privileged 11934 mode only). This is the default when the target is 11935 sh1*-*-linux* or sh2*-*-linux*. 11936 11937 none 11938 Generates function calls to the respective __atomic 11939 built-in functions. This is the default for SH64 11940 targets or when the target is not sh*-*-linux*. 11941 11942 + The option -msoft-atomic has been deprecated. It is now an 11943 alias for -matomic-model=soft-gusa. 11944 + A new option -mtas makes the compiler generate the tas.b 11945 instruction for the __atomic_test_and_set built-in function 11946 regardless of the selected atomic model. 11947 + The __sync functions in libgcc now reflect the selected atomic 11948 model when building the toolchain. 11949 * Added support for the mov.b and mov.w instructions with 11950 displacement addressing. 11951 * Added support for the SH2A instructions movu.b and movu.w. 11952 * Various improvements to code generated for integer arithmetic. 11953 * Improvements to conditional branches and code that involves the T 11954 bit. A new option -mzdcbranch tells the compiler to favor 11955 zero-displacement branches. This is enabled by default for SH4* 11956 targets. 11957 * The pref instruction will now be emitted by the __builtin_prefetch 11958 built-in function for SH3* targets. 11959 * The fmac instruction will now be emitted by the fmaf standard 11960 function and the __builtin_fmaf built-in function. 11961 * The -mfused-madd option has been deprecated in favor of the 11962 machine-independent -ffp-contract option. Notice that the fmac 11963 instruction will now be generated by default for expressions like a 11964 * b + c. This is due to the compiler default setting 11965 -ffp-contract=fast. 11966 * Added new options -mfsrra and -mfsca to allow the compiler using 11967 the fsrra and fsca instructions on targets other than SH4A (where 11968 they are already enabled by default). 11969 * Added support for the __builtin_bswap32 built-in function. It is 11970 now expanded as a sequence of swap.b and swap.w instructions 11971 instead of a library function call. 11972 * The behavior of the -mieee option has been fixed and the negative 11973 form -mno-ieee has been added to control the IEEE conformance of 11974 floating point comparisons. By default -mieee is now enabled and 11975 the option -ffinite-math-only implicitly sets -mno-ieee. 11976 * Added support for the built-in functions __builtin_thread_pointer 11977 and __builtin_set_thread_pointer. This assumes that GBR is used to 11978 hold the thread pointer of the current thread. Memory loads and 11979 stores relative to the address returned by __builtin_thread_pointer 11980 will now also utilize GBR based displacement address modes. 11981 * The -mdiv= option for targets other than SHmedia has been fixed and 11982 documented. 11983 11984 [45]SPARC 11985 11986 * Added optimized instruction scheduling for Niagara4. 11987 11988 [46]TILE-Gx 11989 11990 * Added support for the -mcmodel=MODEL command-line option. The 11991 models supported are small and large. 11992 11993 [47]V850 11994 11995 * This target now supports the E3V5 architecture via the use of the 11996 new -mv850e3v5 command-line option. It also has experimental 11997 support for the e3v5 LOOP instruction which can be enabled via the 11998 new -mloop command-line option. 11999 12000 [48]XStormy16 12001 12002 * This target now supports the -fstack-usage command-line option. 12003 12004 [49]Operating Systems 12005 12006 [50]OpenBSD 12007 12008 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 12009 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 12010 12011 [51]Windows (Cygwin) 12012 12013 * Executables are now linked against shared libgcc by default. The 12014 previous default was to link statically, which can still be done by 12015 explicitly specifying -static or static-libgcc on the command line. 12016 However it is strongly advised against, as it will cause problems 12017 for any application that makes use of DLLs compiled by GCC. It 12018 should be alright for a monolithic stand-alone application that 12019 only links against the Windows DLLs, but offers little or no 12020 benefit. 12021 12022 [52]GCC 4.8.1 12023 12024 This is the [53]list of problem reports (PRs) from GCC's bug tracking 12025 system that are known to be fixed in the 4.8.1 release. This list might 12026 not be complete (that is, it is possible that some PRs that have been 12027 fixed are not listed here). 12028 12029 The C++11 <chrono> std::chrono::system_clock and 12030 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 12031 both are now separate (never typedefs of each other), both use 12032 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 12033 std::chrono::steady_clock is now finally monotonic, and both classes 12034 are mangled differently than in the previous GCC releases. 12035 std::chrono::system_clock::now() with std::chrono::microseconds resp. 12036 std::chrono::seconds resolution is still exported for backwards 12037 compatibility with default configured libstdc++. Note that libstdc++ 12038 configured with --enable-libstdcxx-time= used to be ABI incompatible 12039 with default configured libstdc++ for those two classes and no ABI 12040 compatibility can be offered for those configurations, so any C++11 12041 code that uses those classes and has been compiled and linked against 12042 libstdc++ configured with the non-default --enable-libstdcxx-time= 12043 configuration option needs to be recompiled. 12044 12045 [54]GCC 4.8.2 12046 12047 This is the [55]list of problem reports (PRs) from GCC's bug tracking 12048 system that are known to be fixed in the 4.8.2 release. This list might 12049 not be complete (that is, it is possible that some PRs that have been 12050 fixed are not listed here). 12051 12052 [56]GCC 4.8.3 12053 12054 This is the [57]list of problem reports (PRs) from GCC's bug tracking 12055 system that are known to be fixed in the 4.8.3 release. This list might 12056 not be complete (that is, it is possible that some PRs that have been 12057 fixed are not listed here). 12058 12059 Support for the new powerpc64le-linux platform has been added. It 12060 defaults to generating code that conforms to the ELFV2 ABI. 12061 12062 [58]GCC 4.8.4 12063 12064 This is the [59]list of problem reports (PRs) from GCC's bug tracking 12065 system that are known to be fixed in the 4.8.4 release. This list might 12066 not be complete (that is, it is possible that some PRs that have been 12067 fixed are not listed here). 12068 12069 [60]GCC 4.8.5 12070 12071 This is the [61]list of problem reports (PRs) from GCC's bug tracking 12072 system that are known to be fixed in the 4.8.5 release. This list might 12073 not be complete (that is, it is possible that some PRs that have been 12074 fixed are not listed here). 12075 12076 12077 For questions related to the use of GCC, please consult these web 12078 pages and the [62]GCC manuals. If that fails, the 12079 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12080 web pages and the development of GCC are welcome on our developer 12081 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 12082 archives. 12083 12084 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 12085 distribution of this entire article is permitted in any medium, 12086 provided this notice is preserved. 12087 12088 These pages are [67]maintained by the GCC team. Last modified 12089 2025-02-28. 12090 12091 References 12092 12093 1. https://gcc.gnu.org/wiki/cxx-conversion 12094 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 12095 3. https://gcc.gnu.org/gcc-4.8/porting_to.html 12096 4. https://github.com/google/sanitizers 12097 5. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 12098 6. https://lkml.org/lkml/2006/11/28/239 12099 7. https://gcc.gnu.org/gcc-4.8/changes.html#cxx 12100 8. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12101 9. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12102 10. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12103 11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 12104 12. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 12105 13. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 12106 14. https://gcc.gnu.org/projects/cxx1y.html 12107 15. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 12108 16. https://gcc.gnu.org/gcc-4.8/changes.html#fortran 12109 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 12110 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12111 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12112 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 12113 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12114 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 12115 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 12116 24. https://gcc.gnu.org/wiki/Fortran2003Status 12117 25. https://gcc.gnu.org/wiki/TS29113Status 12118 26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h 12119 27. https://chasm-interop.sourceforge.net/ 12120 28. https://gcc.gnu.org/gcc-4.8/changes.html#go 12121 29. https://gcc.gnu.org/gcc-4.8/changes.html#targets 12122 30. https://gcc.gnu.org/gcc-4.8/changes.html#aarch64 12123 31. https://gcc.gnu.org/gcc-4.8/changes.html#arm 12124 32. https://gcc.gnu.org/gcc-4.8/changes.html#avr 12125 33. http://www.nongnu.org/avr-libc/ 12126 34. https://gcc.gnu.org/PR54461 12127 35. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 12128 36. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 12129 37. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 12130 38. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 12131 39. https://gcc.gnu.org/wiki/FunctionMultiVersioning 12132 40. https://gcc.gnu.org/gcc-4.8/changes.html#frv 12133 41. https://gcc.gnu.org/gcc-4.8/changes.html#mips 12134 42. https://gcc.gnu.org/gcc-4.8/changes.html#powerpc 12135 43. https://gcc.gnu.org/gcc-4.8/changes.html#rx 12136 44. https://gcc.gnu.org/gcc-4.8/changes.html#sh 12137 45. https://gcc.gnu.org/gcc-4.8/changes.html#sparc 12138 46. https://gcc.gnu.org/gcc-4.8/changes.html#tilegx 12139 47. https://gcc.gnu.org/gcc-4.8/changes.html#v850 12140 48. https://gcc.gnu.org/gcc-4.8/changes.html#xstormy16 12141 49. https://gcc.gnu.org/gcc-4.8/changes.html#os 12142 50. https://gcc.gnu.org/gcc-4.8/changes.html#openbsd 12143 51. https://gcc.gnu.org/gcc-4.8/changes.html#windows 12144 52. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.1 12145 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 12146 54. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.2 12147 55. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 12148 56. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.3 12149 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 12150 58. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.4 12151 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 12152 60. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.5 12153 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 12154 62. https://gcc.gnu.org/onlinedocs/ 12155 63. mailto:gcc-help (a] gcc.gnu.org 12156 64. mailto:gcc (a] gcc.gnu.org 12157 65. https://gcc.gnu.org/lists.html 12158 66. https://www.fsf.org/ 12159 67. https://gcc.gnu.org/about.html 12160 ====================================================================== 12161 http://gcc.gnu.org/gcc-4.7/index.html 12162 12163 GCC 4.7 Release Series 12164 12165 (This release series is no longer supported.) 12166 12167 June 12, 2014 12168 12169 The [1]GNU project and the GCC developers are pleased to announce the 12170 release of GCC 4.7.4. 12171 12172 This release is a bug-fix release, containing fixes for regressions in 12173 GCC 4.7.3 relative to previous releases of GCC. 12174 12175 Release History 12176 12177 GCC 4.7.4 12178 June 12, 2014 ([2]changes, [3]documentation) 12179 12180 GCC 4.7.3 12181 April 11, 2013 ([4]changes, [5]documentation) 12182 12183 GCC 4.7.2 12184 September 20, 2012 ([6]changes, [7]documentation) 12185 12186 GCC 4.7.1 12187 June 14, 2012 ([8]changes, [9]documentation) 12188 12189 GCC 4.7.0 12190 March 22, 2012 ([10]changes, [11]documentation) 12191 12192 References and Acknowledgements 12193 12194 GCC used to stand for the GNU C Compiler, but since the compiler 12195 supports several other languages aside from C, it now stands for the 12196 GNU Compiler Collection. 12197 12198 A list of [12]successful builds is updated as new information becomes 12199 available. 12200 12201 The GCC developers would like to thank the numerous people that have 12202 contributed new features, improvements, bug fixes, and other changes as 12203 well as test results to GCC. This [13]amazing group of volunteers is 12204 what makes GCC successful. 12205 12206 For additional information about GCC please refer to the [14]GCC 12207 project web site or contact the [15]GCC development mailing list. 12208 12209 To obtain GCC please use [16]our mirror sites or [17]our version 12210 control system. 12211 12212 12213 For questions related to the use of GCC, please consult these web 12214 pages and the [18]GCC manuals. If that fails, the 12215 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12216 web pages and the development of GCC are welcome on our developer 12217 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 12218 archives. 12219 12220 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 12221 distribution of this entire article is permitted in any medium, 12222 provided this notice is preserved. 12223 12224 These pages are [23]maintained by the GCC team. Last modified 12225 2022-10-26. 12226 12227 References 12228 12229 1. http://www.gnu.org/ 12230 2. https://gcc.gnu.org/gcc-4.7/changes.html 12231 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 12232 4. https://gcc.gnu.org/gcc-4.7/changes.html 12233 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 12234 6. https://gcc.gnu.org/gcc-4.7/changes.html 12235 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 12236 8. https://gcc.gnu.org/gcc-4.7/changes.html 12237 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 12238 10. https://gcc.gnu.org/gcc-4.7/changes.html 12239 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 12240 12. https://gcc.gnu.org/gcc-4.7/buildstat.html 12241 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12242 14. https://gcc.gnu.org/index.html 12243 15. mailto:gcc (a] gcc.gnu.org 12244 16. https://gcc.gnu.org/mirrors.html 12245 17. https://gcc.gnu.org/git.html 12246 18. https://gcc.gnu.org/onlinedocs/ 12247 19. mailto:gcc-help (a] gcc.gnu.org 12248 20. mailto:gcc (a] gcc.gnu.org 12249 21. https://gcc.gnu.org/lists.html 12250 22. https://www.fsf.org/ 12251 23. https://gcc.gnu.org/about.html 12252 ====================================================================== 12253 http://gcc.gnu.org/gcc-4.7/changes.html 12254 12255 GCC 4.7 Release Series 12256 Changes, New Features, and Fixes 12257 12258 Caveats 12259 12260 * The -fconserve-space flag has been deprecated. The flag had no 12261 effect for most targets: only targets without a global .bss section 12262 and without support for switchable sections. Furthermore, the flag 12263 only had an effect for G++, where it could result in wrong 12264 semantics (please refer to the GCC manual for further details). The 12265 flag will be removed in GCC 4.8 12266 * Support for a number of older systems and recently unmaintained or 12267 untested target ports of GCC has been declared obsolete in GCC 4.7. 12268 Unless there is activity to revive them, the next release of GCC 12269 will have their sources permanently removed. 12270 All GCC ports for the following processor architectures have been 12271 declared obsolete: 12272 + picoChip (picochip-*) 12273 The following ports for individual systems on particular 12274 architectures have been obsoleted: 12275 + IRIX 6.5 (mips-sgi-irix6.5) 12276 + MIPS OpenBSD (mips*-*-openbsd*) 12277 + Solaris 8 (*-*-solaris2.8). Details can be found in the 12278 [1]announcement. 12279 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 12280 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 12281 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 12282 default, which for some sources generates code that accesses memory 12283 on unaligned addresses. This requires the kernel of those systems 12284 to enable such accesses (controlled by CP15 register c1, refer to 12285 ARM documentation). Alternatively, or for compatibility with 12286 kernels where unaligned accesses are not supported, all code has to 12287 be compiled with -mno-unaligned-access. Upstream Linux kernel 12288 releases have automatically and unconditionally supported unaligned 12289 accesses as emitted by GCC due to this option being active since 12290 version 2.6.28. 12291 * Support on ARM for the legacy floating-point accelerator (FPA) and 12292 the mixed-endian floating-point format that it used has been 12293 obsoleted. The ports that still use this format have been obsoleted 12294 as well. Many legacy ARM ports already provide an alternative that 12295 uses the VFP floating-point format. The obsolete ports will be 12296 deleted in the next release. 12297 The obsolete ports with alternatives are: 12298 + arm*-*-rtems (use arm*-*-rtemseabi) 12299 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 12300 + arm*-*-elf (use arm*-*-eabi) 12301 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 12302 Note, however, that these alternatives are not binary compatible 12303 with their legacy counterparts (although some can support running 12304 legacy applications). 12305 The obsolete ports that currently lack a modern alternative are: 12306 + arm*-*-ecos-elf 12307 + arm*-*-freebsd 12308 + arm*-wince-pe* 12309 New ports that support more recent versions of the architecture are 12310 welcome. 12311 * Support for the Maverick co-processor on ARM has been obsoleted. 12312 Code to support it will be deleted in the next release. 12313 * Support has been removed for Unix International threads on Solaris 12314 2, so the --enable-threads=solaris configure option and the 12315 -threads compiler option don't work any longer. 12316 * Support has been removed for the Solaris BSD Compatibility Package, 12317 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 12318 from Solaris 11, and was only intended as a migration aid from 12319 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 12320 recognized any longer. 12321 * The ARM port's -mwords-little-endian option has been deprecated. It 12322 will be removed in a future release. 12323 * Support has been removed for the NetWare x86 configuration 12324 obsoleted in GCC 4.6. 12325 * It is no longer possible to use the "l" constraint in MIPS16 asm 12326 statements. 12327 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 12328 library which affected the ABI in C++11 mode: a data member was 12329 added to std::list changing its size and altering the definitions 12330 of some member functions, and std::pair's move constructor was 12331 non-trivial which altered the calling convention for functions with 12332 std::pair arguments or return types. The ABI incompatibilities have 12333 been fixed for GCC version 4.7.2 but as a result C++11 code 12334 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 12335 code compiled with different GCC versions and with C++98/C++03 code 12336 compiled with any version. 12337 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 12338 rules for the layout of vectors that could lead to wrong code being 12339 generated. Vectors larger than 8 bytes in size are now by default 12340 aligned to an 8-byte boundary. This is an ABI change: code that 12341 makes explicit use of vector types may be incompatible with binary 12342 objects built with older versions of GCC. Auto-vectorized code is 12343 not affected by this change. (This change affects GCC versions 12344 4.7.2 and later.) 12345 * More information on porting to GCC 4.7 from previous versions of 12346 GCC can be found in the [2]porting guide for this release. 12347 12348 General Optimizer Improvements 12349 12350 * Support for a new parameter --param case-values-threshold=n was 12351 added to allow users to control the cutoff between doing switch 12352 statements as a series of if statements and using a jump table. 12353 * Link-time optimization (LTO) improvements: 12354 + Improved scalability and reduced memory usage. Link time 12355 optimization of Firefox now requires 3GB of RAM on a 64-bit 12356 system, while over 8GB was needed previously. Linking time has 12357 been improved, too. The serial stage of linking Firefox has 12358 been sped up by about a factor of 10. 12359 + Reduced size of object files and temporary storage used during 12360 linking. 12361 + Streaming performance (both outbound and inbound) has been 12362 improved. 12363 + ld -r is now supported with LTO. 12364 + Several bug fixes, especially in symbol table handling and 12365 merging. 12366 * Interprocedural optimization improvements: 12367 + Heuristics now take into account that after inlining code will 12368 be optimized out because of known values (or properties) of 12369 function parameters. For example: 12370 void foo(int a) 12371 { 12372 if (a > 10) 12373 ... huge code ... 12374 } 12375 void bar (void) 12376 { 12377 foo (0); 12378 } 12379 12380 The call of foo will be inlined into bar even when optimizing 12381 for code size. Constructs based on __builtin_constant_p are 12382 now understood by the inliner and code size estimates are 12383 evaluated a lot more realistically. 12384 + The representation of C++ virtual thunks and aliases (both 12385 implicit and defined via the alias attribute) has been 12386 re-engineered. Aliases no longer pose optimization barriers 12387 and calls to an alias can be inlined and otherwise optimized. 12388 + The inter-procedural constant propagation pass has been 12389 rewritten. It now performs generic function specialization. 12390 For example when compiling the following: 12391 void foo(bool flag) 12392 { 12393 if (flag) 12394 ... do something ... 12395 else 12396 ... do something else ... 12397 } 12398 void bar (void) 12399 { 12400 foo (false); 12401 foo (true); 12402 foo (false); 12403 foo (true); 12404 foo (false); 12405 foo (true); 12406 } 12407 12408 GCC will now produce two copies of foo. One with flag being 12409 true, while other with flag being false. This leads to 12410 performance improvements previously possible only by inlining 12411 all calls. Cloning causes a lot less code size growth. 12412 * A string length optimization pass has been added. It attempts to 12413 track string lengths and optimize various standard C string 12414 functions like strlen, strchr, strcpy, strcat, stpcpy and their 12415 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 12416 enabled by default at -O2 or above, unless optimizing for size, and 12417 can be disabled by the -fno-optimize-strlen option. The pass can 12418 e.g. optimize 12419 char *bar (const char *a) 12420 { 12421 size_t l = strlen (a) + 2; 12422 char *p = malloc (l); if (p == NULL) return p; 12423 strcpy (p, a); strcat (p, "/"); return p; 12424 } 12425 12426 into: 12427 char *bar (const char *a) 12428 { 12429 size_t tmp = strlen (a); 12430 char *p = malloc (tmp + 2); if (p == NULL) return p; 12431 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 12432 } 12433 12434 or for hosted compilations where stpcpy is available in the runtime 12435 and headers provide its prototype, e.g. 12436 void foo (char *a, const char *b, const char *c, const char *d) 12437 { 12438 strcpy (a, b); strcat (a, c); strcat (a, d); 12439 } 12440 12441 can be optimized into: 12442 void foo (char *a, const char *b, const char *c, const char *d) 12443 { 12444 strcpy (stpcpy (stpcpy (a, b), c), d); 12445 } 12446 12447 New Languages and Language specific improvements 12448 12449 * Version 3.1 of the OpenMP specification is now supported for the C, 12450 C++, and Fortran compilers. 12451 12452 Ada 12453 12454 * The command-line option -feliminate-unused-debug-types has been 12455 re-enabled by default, as it is for the other languages, leading to 12456 a reduction in debug info size of 12.5% and more for relevant 12457 cases, as well as to a small compilation speedup. 12458 12459 C family 12460 12461 * A new built-in, __builtin_assume_aligned, has been added, through 12462 which the compiler can be hinted about pointer alignment and can 12463 use it to improve generated code. 12464 * A new warning option -Wunused-local-typedefs was added for C, C++, 12465 Objective-C and Objective-C++. This warning diagnoses typedefs 12466 locally defined in a function, and otherwise not used. 12467 * A new experimental command-line option -ftrack-macro-expansion was 12468 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 12469 the compiler to emit diagnostic about the current macro expansion 12470 stack when a compilation error occurs in a macro expansion. 12471 * Experimental support for transactional memory has been added. It 12472 includes support in the compiler, as well as a supporting runtime 12473 library called libitm. To compile code with transactional memory 12474 constructs, use the -fgnu-tm option. 12475 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 12476 and 32-bit/64-bit x86 platforms. 12477 For more details on transactional memory see [3]the GCC WiKi. 12478 * Support for atomic operations specifying the C++11/C11 memory model 12479 has been added. These new __atomic routines replace the existing 12480 __sync built-in routines. 12481 Atomic support is also available for memory blocks. Lock-free 12482 instructions will be used if a memory block is the same size and 12483 alignment as a supported integer type. Atomic operations which do 12484 not have lock-free support are left as function calls. A set of 12485 library functions is available on the GCC atomic wiki in the 12486 "External Atomics Library" section. 12487 For more details on the memory models and features, see the 12488 [4]atomic wiki. 12489 * When a binary operation is performed on vector types and one of the 12490 operands is a uniform vector, it is possible to replace the vector 12491 with the generating element. For example: 12492 typedef int v4si __attribute__ ((vector_size (16))); 12493 v4si res, a = {1,2,3,4}; 12494 int x; 12495 12496 res = 2 + a; /* means {2,2,2,2} + a */ 12497 res = a - x; /* means a - {x,x,x,x} */ 12498 12499 C 12500 12501 * There is support for some more features from the C11 revision of 12502 the ISO C standard. GCC now accepts the options -std=c11 and 12503 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 12504 + Unicode strings (previously supported only with options such 12505 as -std=gnu11, now supported with -std=c11), and the 12506 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 12507 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 12508 + Alignment support (_Alignas, _Alignof, max_align_t, 12509 <stdalign.h>). 12510 + A built-in function __builtin_complex is provided to support C 12511 library implementation of the CMPLX family of macros. 12512 12513 [5]C++ 12514 12515 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 12516 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 12517 -Wc++0x-compat, respectively. 12518 * G++ now implements [6]C++11 extended friend syntax: 12519 12520 template<class W> 12521 class Q 12522 { 12523 static const int I = 2; 12524 public: 12525 friend W; 12526 }; 12527 12528 struct B 12529 { 12530 int ar[Q<B>::I]; 12531 }; 12532 12533 * Thanks to Ville Voutilainen, G++ now implements [7]C++11 explicit 12534 override control. 12535 12536 struct B { 12537 virtual void f() const final; 12538 virtual void f(int); 12539 }; 12540 12541 struct D : B { 12542 void f() const; // error: D::f attempts to override final B::f 12543 void f(long) override; // error: doesn't override anything 12544 void f(int) override; // ok 12545 }; 12546 12547 struct E final { }; 12548 struct F: E { }; // error: deriving from final class 12549 12550 * G++ now implements [8]C++11 non-static data member initializers. 12551 12552 struct A { 12553 int i = 42; 12554 } a; // initializes a.i to 42 12555 12556 * Thanks to Ed Smith-Rowland, G++ now implements [9]C++11 12557 user-defined literals. 12558 12559 // Not actually a good approximation. :) 12560 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 12561 long double pi = 180.0_degrees; 12562 12563 * G++ now implements [10]C++11 alias-declarations. 12564 12565 template <class T> using Ptr = T*; 12566 Ptr<int> ip; // decltype(ip) is int* 12567 12568 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 12569 [11]C++11 delegating constructors. 12570 12571 struct A { 12572 A(int); 12573 A(): A(42) { } // delegate to the A(int) constructor 12574 }; 12575 12576 * G++ now fully implements C++11 atomic classes rather than just 12577 integer derived classes. 12578 12579 class POD { 12580 int a; 12581 int b; 12582 }; 12583 std::atomic<POD> my_atomic_POD; 12584 12585 * G++ now sets the predefined macro __cplusplus to the correct value, 12586 199711L for C++98/03, and 201103L for C++11. 12587 * G++ now correctly implements the two-phase lookup rules such that 12588 an unqualified name used in a template must have an appropriate 12589 declaration found either in scope at the point of definition of the 12590 template or by argument-dependent lookup at the point of 12591 instantiation. As a result, code that relies on a second 12592 unqualified lookup at the point of instantiation to find functions 12593 declared after the template or in dependent bases will be rejected. 12594 The compiler will suggest ways to fix affected code, and using the 12595 -fpermissive compiler flag will allow the code to compile with a 12596 warning. 12597 12598 template <class T> 12599 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 12600 void g(int) { } // fix by moving this declaration before the declaration of f 12601 12602 template <class T> 12603 struct A: T { 12604 // error, B::g(B) not found by argument-dependent lookup 12605 void f() { g(T()); } // fix by using this->g or A::g 12606 }; 12607 12608 struct B { void g(B); }; 12609 12610 int main() 12611 { 12612 f<int>(); 12613 A<B>().f(); 12614 } 12615 12616 * G++ now properly re-uses stack space allocated for temporary 12617 objects when their lifetime ends, which can significantly lower 12618 stack consumption for some C++ functions. As a result of this, some 12619 code with undefined behavior will now break: 12620 12621 const int &f(const int &i) { return i; } 12622 .... 12623 const int &x = f(1); 12624 const int &y = f(2); 12625 12626 Here, x refers to the temporary allocated to hold the 1 argument, 12627 which only lives until the end of the initialization; it 12628 immediately becomes a dangling reference. So the next statement 12629 re-uses the stack slot to hold the 2 argument, and users of x get 12630 that value instead. 12631 Note that this should not cause any change of behavior for 12632 temporaries of types with non-trivial destructors, as they are 12633 already destroyed at end of full-expression; the change is that now 12634 the storage is released as well. 12635 * A new command-line option -Wdelete-non-virtual-dtor has been added 12636 to warn when delete is used to destroy an instance of a class which 12637 has virtual functions and non-virtual destructor. It is unsafe to 12638 delete an instance of a derived class through a pointer to a base 12639 class if the base class does not have a virtual destructor. This 12640 warning is enabled by -Wall. 12641 * A new command-line option -Wzero-as-null-pointer-constant has been 12642 added to warn when a literal '0' is used as null pointer constant. 12643 It can be useful to facilitate the conversion to nullptr in C++11. 12644 * As per C++98, access-declarations are now deprecated by G++. 12645 Using-declarations are to be used instead. Furthermore, some 12646 efforts have been made to improve the support of class scope 12647 using-declarations. In particular, using-declarations referring to 12648 a dependent type now work as expected ([12]bug c++/14258). 12649 * The ELF symbol visibility of a template instantiation is now 12650 properly constrained by the visibility of its template arguments 12651 ([13]bug c++/35688). 12652 12653 Runtime Library (libstdc++) 12654 12655 * [14]Improved experimental support for the new ISO C++ standard, 12656 C++11, including: 12657 + using noexcept in most of the library; 12658 + implementations of pointer_traits, allocator_traits and 12659 scoped_allocator_adaptor; 12660 + uses-allocator construction for tuple; 12661 + vector meets the allocator-aware container requirements; 12662 + replacing monotonic_clock with steady_clock; 12663 + enabling the thread support library on most POSIX targets; 12664 + many small improvements to conform to the FDIS. 12665 * Added --enable-clocale=newlib configure option. 12666 * Debug Mode iterators for unordered associative containers. 12667 * Avoid polluting the global namespace and do not include <unistd.h>. 12668 12669 [15]Fortran 12670 12671 * The compile flag [16]-fstack-arrays has been added, which causes 12672 all local arrays to be put on stack memory. For some programs this 12673 will improve the performance significantly. If your program uses 12674 very large local arrays, it is possible that you will have to 12675 extend your runtime limits for stack memory. 12676 * The [17]-Ofast flag now also implies [18]-fno-protect-parens and 12677 [19]-fstack-arrays. 12678 * Front-end optimizations can now be selected by the 12679 [20]-ffrontend-optimize option and deselected by the 12680 -fno-frontend-optimize option. 12681 * When front-end optimization removes a function call, 12682 [21]-Wfunction-elimination warns about that. 12683 * When performing front-end-optimization, the 12684 [22]-faggressive-function-elimination option allows the removal of 12685 duplicate function calls even for impure functions. 12686 * The flag [23]-Wreal-q-constant has been added, which warns if 12687 floating-point literals have been specified using q (such as 12688 1.0q0); the q marker is now supported as a vendor extension to 12689 denote quad precision (REAL(16) or, if not available, REAL(10)). 12690 Consider using a kind parameter (such as in 1.0_qp) instead, which 12691 can be obtained via [24]SELECTED_REAL_KIND. 12692 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 12693 Fortran now always prints error messages to standard error. If you 12694 wish to redirect standard error, please consult the manual for your 12695 OS, shell, batch environment etc. as appropriate. 12696 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 12697 variable have been removed. When encountering a serious error, 12698 gfortran will now always abort the program. Whether a core dump is 12699 generated depends on the user environment settings; see the ulimit 12700 -c setting for POSIX shells, limit coredumpsize for C shells, and 12701 the [25]WER user-mode dumps settings on Windows. 12702 * The [26]-fbacktrace option is now enabled by default. When 12703 encountering a fatal error, gfortran will attempt to print a 12704 backtrace to standard error before aborting. It can be disabled 12705 with -fno-backtrace. Note: On POSIX targets with the addr2line 12706 utility from GNU binutils, GNU Fortran can print a backtrace with 12707 function name, file name, line number information in addition to 12708 the addresses; otherwise only the addresses are printed. 12709 * [27]Fortran 2003: 12710 + Generic interface names which have the same name as derived 12711 types are now supported, which allows to write constructor 12712 functions. Note that Fortran does not support static 12713 constructor functions; only default initialization or an 12714 explicit structure-constructor initialization are available. 12715 + [28]Polymorphic (class) arrays are now supported. 12716 * [29]Fortran 2008: 12717 + Support for the DO CONCURRENT construct has been added, which 12718 allows the user to specify that individual loop iterations 12719 have no interdependencies. 12720 + [30]Coarrays: Full single-image support except for polymorphic 12721 coarrays. Additionally, preliminary support for multiple 12722 images via an MPI-based [31]coarray communication library has 12723 been added. Note: The library version is not yet usable as 12724 remote coarray access is not yet possible. 12725 * [32]TS 29113: 12726 + New flag [33]-std=f2008ts permits programs that are expected 12727 to conform to the Fortran 2008 standard and the draft 12728 Technical Specification (TS) 29113 on Further Interoperability 12729 of Fortran with C. 12730 + The OPTIONAL attribute is now allowed for dummy arguments of 12731 BIND(C) procedures. 12732 + The RANK intrinsic has been added. 12733 + The implementation of the ASYNCHRONOUS attribute in GCC is 12734 compatible with the candidate draft of TS 29113 (since GCC 12735 4.6). 12736 12737 [34]Go 12738 12739 * GCC 4.7 implements the [35]Go 1 language standard. The library 12740 support in 4.7.0 is not quite complete, due to release timing. 12741 Release 4.7.1 includes complete support for Go 1. The Go library is 12742 from the Go 1.0.1 release. 12743 * Go has been tested on GNU/Linux and Solaris platforms. It may work 12744 on other platforms as well. 12745 12746 [36]New Targets and Target Specific Improvements 12747 12748 [37]ARM 12749 12750 * GCC now supports the Cortex-A7 processor implementing the v7-a 12751 version of the architecture using the option -mcpu=cortex-a7. 12752 * The default vector size in auto-vectorization for NEON is now 128 12753 bits. If vectorization fails thusly, the vectorizer tries again 12754 with 64-bit vectors. 12755 * A new option -mvectorize-with-neon-double was added to allow users 12756 to change the vector size to 64 bits. 12757 12758 [38]AVR 12759 12760 * The AVR port's libgcc has been improved and its multilib structure 12761 has been enhanced. As a result, all objects contributing to an 12762 application must either be compiled with GCC versions up to 4.6.x 12763 or with GCC versions 4.7.1 or later. If the compiler is used with 12764 AVR-LibC, you need a version that supports the new layout, i.e. 12765 implements [39]#35407. 12766 * The -mshort-calls command-line option has been deprecated. It will 12767 be removed in the GCC 4.8 release. See -mrelax for a replacement. 12768 * The compiled code only references startup code that clears .bss and 12769 the common section resp. initializes the .data and .rodata section 12770 provided respective sections (or subsections thereof) are not 12771 empty, see [40]PR18145. Applications that put all static storage 12772 objects into non-standard sections and / or define all static 12773 storage objects in assembler modules, must reference __do_clear_bss 12774 resp. __do_copy_data by hand or undefine the symbol(s) by means of 12775 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 12776 * GCC now supports the XMEGA architecture. This requires GNU Binutils 12777 2.22 or later. 12778 * Support for the [41]named address spaces __flash, __flash1, ..., 12779 __flash5 and __memx has been added. These address spaces locate 12780 read-only data in flash memory and allow reading from flash memory 12781 by means of ordinary C code, i.e. without the need of (inline) 12782 assembler code: 12783 12784 const __flash int values[] = { 42, 31 }; 12785 12786 int add_values (const __flash int *p, int i) 12787 { 12788 return values[i] + *p; 12789 } 12790 12791 * Support has been added for the AVR-specific configure option 12792 --with-avrlibc=yes in order to arrange for better integration of 12793 [42]AVR-Libc. This configure option is supported in avr-gcc v4.7.2 12794 and newer and will only take effect in non-RTEMS configurations. If 12795 avr-gcc is configured for RTEMS, the option will be ignored which 12796 is the same as specifying --with-avrlibc=no. See [43]PR54461 for 12797 more technical details. 12798 * Support for AVR-specific [44]built-in functions has been added. 12799 * Support has been added for the signed and unsigned 24-bit scalar 12800 integer types __int24 and __uint24. 12801 * New command-line options -maccumulate-args, -mbranch-cost=cost and 12802 -mstrict-X were added to allow better fine-tuning of code 12803 optimization. 12804 * The command-line option -fdata-sections now also takes affect on 12805 the section names of variables with the progmem attribute. 12806 * A new inline assembler print modifier %i to print a RAM address as 12807 I/O address has been added: 12808 12809 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 12810 12811 void set_portb (uint8_t value) 12812 { 12813 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 12814 } 12815 12816 The offset between an I/O address and the RAM address for that I/O 12817 location is device-specific. This offset is taken into account when 12818 printing a RAM address with the %i modifier so that the address is 12819 suitable to be used as operand in an I/O command. The address must 12820 be a constant integer known at compile time. 12821 * The inline assembler constraint "R" to represent integers in the 12822 range -6 ... 5 has been removed without replacement. 12823 * Many optimizations to: 12824 + 64-bit integer arithmetic 12825 + Widening multiplication 12826 + Integer division by a constant 12827 + Avoid constant reloading in multi-byte instructions. 12828 + Micro-optimizations for special instruction sequences. 12829 + Generic built-in functions like __builtin_ffs*, 12830 __builtin_clz*, etc. 12831 + If-else decision trees generated by switch instructions 12832 + Merging of data located in flash memory 12833 + New libgcc variants for devices with 8-bit wide stack pointer 12834 + ... 12835 * Better documentation: 12836 + Handling of EIND and indirect jumps on devices with more than 12837 128 KiB of program memory. 12838 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 12839 registers. 12840 + Function attributes OS_main and OS_task. 12841 + AVR-specific built-in macros. 12842 12843 C6X 12844 12845 * Support has been added for the Texas Instruments C6X family of 12846 processors. 12847 12848 CR16 12849 12850 * Support has been added for National Semiconductor's CR16 12851 architecture. 12852 12853 Epiphany 12854 12855 * Support has been added for Adapteva's Epiphany architecture. 12856 12857 IA-32/x86-64 12858 12859 * Support for Intel AVX2 intrinsics, built-in functions and code 12860 generation is available via -mavx2. 12861 * Support for Intel BMI2 intrinsics, built-in functions and code 12862 generation is available via -mbmi2. 12863 * Implementation and automatic generation of __builtin_clz* using the 12864 lzcnt instruction is available via -mlzcnt. 12865 * Support for Intel FMA3 intrinsics and code generation is available 12866 via -mfma. 12867 * A new -mfsgsbase command-line option is available that makes GCC 12868 generate new segment register read/write instructions through 12869 dedicated built-ins. 12870 * Support for the new Intel rdrnd instruction is available via 12871 -mrdrnd. 12872 * Two additional AVX vector conversion instructions are available via 12873 -mf16c. 12874 * Support for new Intel processor codename IvyBridge with RDRND, 12875 FSGSBASE and F16C is available through -march=core-avx-i. 12876 * Support for the new Intel processor codename Haswell with AVX2, 12877 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 12878 * Support for new AMD family 15h processors (Piledriver core) is now 12879 available through -march=bdver2 and -mtune=bdver2 options. 12880 * Support for [45]the x32 psABI is now available through the -mx32 12881 option. 12882 * Windows mingw targets are using the -mms-bitfields option by 12883 default. 12884 * Windows x86 targets are using the __thiscall calling convention for 12885 C++ class-member functions. 12886 * Support for the configure option --with-threads=posix for Windows 12887 mingw targets. 12888 12889 [46]MIPS 12890 12891 * GCC now supports thread-local storage (TLS) for MIPS16. This 12892 requires GNU binutils 2.22 or later. 12893 * GCC can now generate code specifically for the Cavium Octeon+ and 12894 Octeon2 processors. The associated command-line options are 12895 -march=octeon+ and -march=octeon2 respectively. Both options 12896 require GNU binutils 2.22 or later. 12897 * GCC can now work around certain 24k errata, under the control of 12898 the command-line option -mfix-24k. These workarounds require GNU 12899 binutils 2.20 or later. 12900 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 12901 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 12902 toolchain that generates 32-bit code by default. Use the 12903 configure-time option --enable-targets=all to select these extra 12904 multilibs. 12905 * Passing -fno-delayed-branch now also stops the assembler from 12906 automatically filling delay slots. 12907 12908 PowerPC/PowerPC64 12909 12910 * Vectors of type vector long long or vector long are passed and 12911 returned using the same method as other vectors with the VSX 12912 instruction set. Previously GCC did not adhere to the ABI for 12913 128-bit vectors with 64-bit integer base types (PR 48857). This 12914 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 12915 * A new option -mno-pointers-to-nested-functions was added to allow 12916 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 12917 that the compiler should not load up the chain register (r11) 12918 before calling a function through a pointer. If you use this 12919 option, you cannot call nested functions through a pointer, or call 12920 other languages that might use the static chain. 12921 * A new option msave-toc-indirect was added to allow AIX 12922 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 12923 save the TOC in the prologue for indirect calls or generate the 12924 save inline. This can speed up some programs that call through a 12925 function pointer a lot, but it can slow down other functions that 12926 only call through a function pointer in exceptional cases. 12927 * The PowerPC port will now enable machine-specific built-in 12928 functions when the user switches the target machine using the 12929 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 12930 sequences. In addition, the target macros are updated. However, due 12931 to the way the -save-temps switch is implemented, you won't see the 12932 effect of these additional macros being defined in preprocessor 12933 output. 12934 12935 SH 12936 12937 * A new option -msoft-atomic has been added. When it is specified, 12938 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 12939 the new __atomic routines. 12940 * Since it is neither supported by GAS nor officially documented, 12941 code generation for little endian SH2A has been disabled. 12942 Specifying -ml with -m2a* will now result in a compiler error. 12943 * The defunct -mbranch-cost option has been fixed. 12944 * Some improvements to the generated code of: 12945 + Utilization of the tst #imm,R0 instruction. 12946 + Dynamic shift instructions on SH2A. 12947 + Integer absolute value calculations. 12948 * The -mdiv= option for targets other than SHmedia has been fixed and 12949 documented. 12950 12951 SPARC 12952 12953 * The option -mflat has been reinstated. When it is specified, the 12954 compiler will generate code for a single register window model. 12955 This is essentially a new implementation and the corresponding 12956 debugger support has been added to GDB 7.4. 12957 * Support for the options -mtune=native and -mcpu=native has been 12958 added on selected native platforms (GNU/Linux and Solaris). 12959 * Support for the SPARC T3 (Niagara 3) processor has been added. 12960 * VIS: 12961 + An intrinsics header visintrin.h has been added. 12962 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 12963 compare instructions have been added. 12964 + The little-endian version of alignaddr is now supported. 12965 + When possible, VIS builtins are marked const, which should 12966 increase the compiler's ability to optimize VIS operations. 12967 + The compiler now properly tracks the %gsr register and how it 12968 behaves as an input for various VIS instructions. 12969 + Akin to fzero, the compiler can now generate fone instructions 12970 in order to set all of the bits of a floating-point register 12971 to 1. 12972 + The documentation for the VIS intrinsics in the GCC manual has 12973 been brought up to date and many inaccuracies were fixed. 12974 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 12975 non-condition-code setting edge instructions have been added. 12976 Their availability is controlled by the new -mvis2 and 12977 -mno-vis2 options. They are enabled by default on 12978 UltraSPARC-III and later CPUs. 12979 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 12980 has been added. These instructions are enabled by default on SPARC 12981 T3 (Niagara 3) and later CPUs. 12982 12983 TILE-Gx/TILEPro 12984 12985 * Support has been added for the Tilera TILE-Gx and TILEPro families 12986 of processors. 12987 12988 Other significant improvements 12989 12990 * A new option (-grecord-gcc-switches) was added that appends 12991 compiler command-line options that might affect code generation to 12992 the DW_AT_producer attribute string in the DWARF debugging 12993 information. 12994 * GCC now supports various new GNU extensions to the DWARF debugging 12995 information format, like [47]entry value and [48]call site 12996 information, [49]typed DWARF stack or [50]a more compact macro 12997 representation. Support for these extensions has been added to GDB 12998 7.4. They can be disabled through the -gstrict-dwarf command-line 12999 option. 13000 13001 [51]GCC 4.7.1 13002 13003 This is the [52]list of problem reports (PRs) from GCC's bug tracking 13004 system that are known to be fixed in the 4.7.1 release. This list might 13005 not be complete (that is, it is possible that some PRs that have been 13006 fixed are not listed here). 13007 13008 The Go front end in the 4.7.1 release fully supports the [53]Go 1 13009 language standard. 13010 13011 [54]GCC 4.7.2 13012 13013 This is the [55]list of problem reports (PRs) from GCC's bug tracking 13014 system that are known to be fixed in the 4.7.2 release. This list might 13015 not be complete (that is, it is possible that some PRs that have been 13016 fixed are not listed here). 13017 13018 [56]GCC 4.7.3 13019 13020 This is the [57]list of problem reports (PRs) from GCC's bug tracking 13021 system that are known to be fixed in the 4.7.3 release. This list might 13022 not be complete (that is, it is possible that some PRs that have been 13023 fixed are not listed here). 13024 13025 [58]GCC 4.7.4 13026 13027 This is the [59]list of problem reports (PRs) from GCC's bug tracking 13028 system that are known to be fixed in the 4.7.4 release. This list might 13029 not be complete (that is, it is possible that some PRs that have been 13030 fixed are not listed here). 13031 13032 13033 For questions related to the use of GCC, please consult these web 13034 pages and the [60]GCC manuals. If that fails, the 13035 [61]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13036 web pages and the development of GCC are welcome on our developer 13037 list at [62]gcc (a] gcc.gnu.org. All of [63]our lists have public 13038 archives. 13039 13040 Copyright (C) [64]Free Software Foundation, Inc. Verbatim copying and 13041 distribution of this entire article is permitted in any medium, 13042 provided this notice is preserved. 13043 13044 These pages are [65]maintained by the GCC team. Last modified 13045 2025-02-28. 13046 13047 References 13048 13049 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 13050 2. https://gcc.gnu.org/gcc-4.7/porting_to.html 13051 3. https://gcc.gnu.org/wiki/TransactionalMemory 13052 4. https://gcc.gnu.org/wiki/Atomic/GCCMM 13053 5. https://gcc.gnu.org/gcc-4.7/changes.html#cxx 13054 6. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13055 7. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13056 8. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13057 9. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13058 10. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13059 11. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 13060 12. https://gcc.gnu.org/PR14258 13061 13. https://gcc.gnu.org/PR35688 13062 14. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 13063 15. https://gcc.gnu.org/gcc-4.7/changes.html#fortran 13064 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 13065 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 13066 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 13067 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 13068 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 13069 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 13070 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 13071 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 13072 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 13073 25. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 13074 26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 13075 27. https://gcc.gnu.org/wiki/Fortran2003Status 13076 28. https://gcc.gnu.org/wiki/OOP 13077 29. https://gcc.gnu.org/wiki/Fortran2008Status 13078 30. https://gcc.gnu.org/wiki/Coarray 13079 31. https://gcc.gnu.org/wiki/CoarrayLib 13080 32. https://gcc.gnu.org/wiki/TS29113Status 13081 33. 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 13082 34. https://gcc.gnu.org/gcc-4.7/changes.html#go 13083 35. https://go.dev/doc/go1 13084 36. https://gcc.gnu.org/gcc-4.7/changes.html#targets 13085 37. https://gcc.gnu.org/gcc-4.7/changes.html#arm 13086 38. https://gcc.gnu.org/gcc-4.7/changes.html#avr 13087 39. http://savannah.nongnu.org/bugs/?35407 13088 40. https://gcc.gnu.org/PR18145 13089 41. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 13090 42. http://www.nongnu.org/avr-libc/ 13091 43. https://gcc.gnu.org/PR54461 13092 44. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 13093 45. https://sites.google.com/site/x32abi/ 13094 46. https://gcc.gnu.org/gcc-4.7/changes.html#mips 13095 47. https://dwarfstd.org/issues/100909.1.html 13096 48. https://dwarfstd.org/issues/100909.2.html 13097 49. https://dwarfstd.org/issues/140425.1.html 13098 50. https://dwarfstd.org/issues/110722.1.html 13099 51. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.1 13100 52. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 13101 53. https://go.dev/doc/go1 13102 54. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.2 13103 55. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 13104 56. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.3 13105 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 13106 58. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.4 13107 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 13108 60. https://gcc.gnu.org/onlinedocs/ 13109 61. mailto:gcc-help (a] gcc.gnu.org 13110 62. mailto:gcc (a] gcc.gnu.org 13111 63. https://gcc.gnu.org/lists.html 13112 64. https://www.fsf.org/ 13113 65. https://gcc.gnu.org/about.html 13114 ====================================================================== 13115 http://gcc.gnu.org/gcc-4.6/index.html 13116 13117 GCC 4.6 Release Series 13118 13119 (This release series is no longer supported.) 13120 13121 April 12, 2013 13122 13123 The [1]GNU project and the GCC developers are pleased to announce the 13124 release of GCC 4.6.4. 13125 13126 This release is a bug-fix release, containing fixes for regressions in 13127 GCC 4.6.3 relative to previous releases of GCC. 13128 13129 Release History 13130 13131 GCC 4.6.4 13132 April 12, 2013 ([2]changes, [3]documentation) 13133 13134 GCC 4.6.3 13135 March 1, 2012 ([4]changes, [5]documentation) 13136 13137 GCC 4.6.2 13138 October 26, 2011 ([6]changes, [7]documentation) 13139 13140 GCC 4.6.1 13141 June 27, 2011 ([8]changes, [9]documentation) 13142 13143 GCC 4.6.0 13144 March 25, 2011 ([10]changes, [11]documentation) 13145 13146 References and Acknowledgements 13147 13148 GCC used to stand for the GNU C Compiler, but since the compiler 13149 supports several other languages aside from C, it now stands for the 13150 GNU Compiler Collection. 13151 13152 A list of [12]successful builds is updated as new information becomes 13153 available. 13154 13155 The GCC developers would like to thank the numerous people that have 13156 contributed new features, improvements, bug fixes, and other changes as 13157 well as test results to GCC. This [13]amazing group of volunteers is 13158 what makes GCC successful. 13159 13160 For additional information about GCC please refer to the [14]GCC 13161 project web site or contact the [15]GCC development mailing list. 13162 13163 To obtain GCC please use [16]our mirror sites or [17]our version 13164 control system. 13165 13166 13167 For questions related to the use of GCC, please consult these web 13168 pages and the [18]GCC manuals. If that fails, the 13169 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13170 web pages and the development of GCC are welcome on our developer 13171 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 13172 archives. 13173 13174 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 13175 distribution of this entire article is permitted in any medium, 13176 provided this notice is preserved. 13177 13178 These pages are [23]maintained by the GCC team. Last modified 13179 2022-10-26. 13180 13181 References 13182 13183 1. http://www.gnu.org/ 13184 2. https://gcc.gnu.org/gcc-4.6/changes.html 13185 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 13186 4. https://gcc.gnu.org/gcc-4.6/changes.html 13187 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 13188 6. https://gcc.gnu.org/gcc-4.6/changes.html 13189 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 13190 8. https://gcc.gnu.org/gcc-4.6/changes.html 13191 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 13192 10. https://gcc.gnu.org/gcc-4.6/changes.html 13193 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 13194 12. https://gcc.gnu.org/gcc-4.6/buildstat.html 13195 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13196 14. https://gcc.gnu.org/index.html 13197 15. mailto:gcc (a] gcc.gnu.org 13198 16. https://gcc.gnu.org/mirrors.html 13199 17. https://gcc.gnu.org/git.html 13200 18. https://gcc.gnu.org/onlinedocs/ 13201 19. mailto:gcc-help (a] gcc.gnu.org 13202 20. mailto:gcc (a] gcc.gnu.org 13203 21. https://gcc.gnu.org/lists.html 13204 22. https://www.fsf.org/ 13205 23. https://gcc.gnu.org/about.html 13206 ====================================================================== 13207 http://gcc.gnu.org/gcc-4.6/changes.html 13208 13209 GCC 4.6 Release Series 13210 Changes, New Features, and Fixes 13211 13212 Caveats 13213 13214 * The options -b <machine> and -V <version> have been removed because 13215 they were unreliable. Instead, users should directly run 13216 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 13217 run a different version of gcc. 13218 * GCC now has stricter checks for invalid command-line options. In 13219 particular, when gcc was called to link object files rather than 13220 compile source code, it would previously accept and ignore all 13221 options starting with --, including linker options such as 13222 --as-needed and --export-dynamic, although such options would 13223 result in errors if any source code was compiled. Such options, if 13224 unknown to the compiler, are now rejected in all cases; if the 13225 intent was to pass them to the linker, options such as 13226 -Wl,--as-needed should be used. 13227 * Versions of the GNU C library up to and including 2.11.1 included 13228 an [1]incorrect implementation of the cproj function. GCC optimizes 13229 its builtin cproj according to the behavior specified and allowed 13230 by the ISO C99 standard. If you want to avoid discrepancies between 13231 the C library and GCC's builtin transformations when using cproj in 13232 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 13233 and actually rely on the incorrect behavior of cproj, then you can 13234 disable GCC's transformations using -fno-builtin-cproj. 13235 * The C-only intermodule optimization framework (IMA, enabled by 13236 -combine) has been removed in favor of the new generic link-time 13237 optimization framework (LTO) introduced in [2]GCC 4.5.0. 13238 * GCC now ships with the LGPL-licensed libquadmath library, which 13239 provides quad-precision mathematical functions for targets with a 13240 __float128 datatype. __float128 is available for targets on 32-bit 13241 x86, x86-64 and Itanium architectures. The libquadmath library is 13242 automatically built on such targets when building the Fortran 13243 compiler. 13244 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 13245 warnings were added for C, C++, Objective-C and Objective-C++. 13246 These warnings diagnose variables respective parameters which are 13247 only set in the code and never otherwise used. Usually such 13248 variables are useless and often even the value assigned to them is 13249 computed needlessly, sometimes expensively. The 13250 -Wunused-but-set-variable warning is enabled by default by -Wall 13251 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 13252 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 13253 rules for the layout of vectors that could lead to wrong code being 13254 generated. Vectors larger than 8 bytes in size are now by default 13255 aligned to an 8-byte boundary. This is an ABI change: code that 13256 makes explicit use of vector types may be incompatible with binary 13257 objects built with older versions of GCC. Auto-vectorized code is 13258 not affected by this change. (This change affects GCC versions 13259 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 13260 * On AVR, variables with the progmem attribute to locate data in 13261 flash memory must be qualified as const. 13262 * Support for a number of older systems and recently unmaintained or 13263 untested target ports of GCC has been declared obsolete in GCC 4.6. 13264 Unless there is activity to revive them, the next release of GCC 13265 will have their sources permanently removed. 13266 All GCC ports for the following processor architectures have been 13267 declared obsolete: 13268 + Argonaut ARC (arc-*) 13269 + National Semiconductor CRX (crx-*) 13270 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 13271 m68hc12-*-*, m6812-*-*) 13272 + Sunplus S+core (score-*) 13273 The following ports for individual systems on particular 13274 architectures have been obsoleted: 13275 + Interix (i[34567]86-*-interix3*) 13276 + NetWare x86 (i[3456x]86-*-netware*) 13277 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 13278 + MCore PE (mcore-*-pe*) 13279 + SH SymbianOS (sh*-*-symbianelf*) 13280 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 13281 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 13282 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 13283 vax-*-netbsd*, but not *-*-netbsdelf*) 13284 The i[34567]86-*-pe alias for Cygwin targets has also been 13285 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 13286 Certain configure options to control the set of libraries built 13287 with GCC on some targets have been obsoleted. On ARM targets, the 13288 options --disable-fpu, --disable-26bit, --disable-underscore, 13289 --disable-interwork, --disable-biendian and --disable-nofmult have 13290 been obsoleted. On MIPS targets, the options 13291 --disable-single-float, --disable-biendian and --disable-softfloat 13292 have been obsoleted. 13293 * Support has been removed for all the [3]configurations obsoleted in 13294 GCC 4.5. 13295 * More information on porting to GCC 4.6 from previous versions of 13296 GCC can be found in the [4]porting guide for this release. 13297 13298 General Optimizer Improvements 13299 13300 * A new general optimization level, -Ofast, has been introduced. It 13301 combines the existing optimization level -O3 with options that can 13302 affect standards compliance but result in better optimized code. 13303 For example, -Ofast enables -ffast-math. 13304 * Link-time optimization improvements: 13305 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 13306 stabilized to the point of being usable. It has become the 13307 default mode when using the LTO optimization model. Link time 13308 optimization can now split itself into multiple parallel 13309 compilations. Parallelism is controlled with -flto=n (where n 13310 specifies the number of compilations to execute in parallel). 13311 GCC can also cooperate with a GNU make job server by 13312 specifying the -flto=jobserver option and adding + to the 13313 beginning of the Makefile rule executing the linker. 13314 Classical LTO mode can be enforced by -flto-partition=none. 13315 This may result in small code quality improvements. 13316 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 13317 and other large applications can be built with LTO enabled. 13318 + The linker plugin support improvements 13319 o Linker plugin is now enabled by default when the linker 13320 is detected to have plugin support. This is the case for 13321 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 13322 the Gold linker on ELF targets. Plugin support of the 13323 Apple linker on Darwin is not compatible with GCC. The 13324 linker plugin can also be controlled by the 13325 -fuse-linker-plugin command-line option. 13326 o Resolution information from the linker plugin is used to 13327 drive whole program assumptions. Use of the linker plugin 13328 results in more aggressive optimization on binaries and 13329 on shared libraries that use the hidden visibility 13330 attribute. Consequently the use of -fwhole-program is not 13331 necessary in addition to LTO. 13332 + Hidden symbols used from non-LTO objects now have to be 13333 explicitly annotated with externally_visible when the linker 13334 plugin is not used. 13335 + C++ inline functions and virtual tables are now privatized 13336 more aggressively, leading to better inter-procedural 13337 optimization and faster dynamic linking. 13338 + Memory usage and intermediate language streaming performance 13339 have been improved. 13340 + Static constructors and destructors from individual units are 13341 inlined into a single function. This can significantly improve 13342 startup times of large C++ applications where static 13343 constructors are very common. For example, static constructors 13344 are used when including the iostream header. 13345 + Support for the Ada language has been added. 13346 * Interprocedural optimization improvements 13347 + The interprocedural framework was re-tuned for link time 13348 optimization. Several scalability issues were resolved. 13349 + Improved auto-detection of const and pure functions. Newly, 13350 noreturn functions are auto-detected. 13351 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 13352 available that informs users when adding attributes to headers 13353 might improve code generation. 13354 + A number of inlining heuristic improvements. In particular: 13355 o Partial inlining is now supported and enabled by default 13356 at -O2 and greater. The feature can be controlled via 13357 -fpartial-inlining. 13358 Partial inlining splits functions with short hot path to 13359 return. This allows more aggressive inlining of the hot 13360 path leading to better performance and often to code size 13361 reductions (because cold parts of functions are not 13362 duplicated). 13363 o Scalability for large compilation units was improved 13364 significantly. 13365 o Inlining of callbacks is now more aggressive. 13366 o Virtual methods are considered for inlining when the 13367 caller is inlined and devirtualization is then possible. 13368 o Inlining when optimizing for size (either in cold regions 13369 of a program or when compiling with -Os) was improved to 13370 better handle C++ programs with larger abstraction 13371 penalty, leading to smaller and faster code. 13372 + The IPA reference optimization pass detecting global variables 13373 used or modified by functions was strengthened and sped up. 13374 + Functions whose address was taken are now optimized out when 13375 all references to them are dead. 13376 + A new inter-procedural static profile estimation pass detects 13377 functions that are executed once or unlikely to be executed. 13378 Unlikely executed functions are optimized for size. Functions 13379 executed once are optimized for size except for the inner 13380 loops. 13381 + On most targets with named section support, functions used 13382 only at startup (static constructors and main), functions used 13383 only at exit and functions detected to be cold are placed into 13384 separate text segment subsections. This extends the 13385 -freorder-functions feature and is controlled by the same 13386 switch. The goal is to improve the startup time of large C++ 13387 programs. 13388 Proper function placement requires linker support. GNU ld 13389 2.21.51 on ELF targets was updated to place those functions 13390 together within the text section leading to better code 13391 locality and faster startup times of large C++ programs. The 13392 feature is also supported in the Apple linker. Support in the 13393 gold linker is planned. 13394 * A new switch -fstack-usage has been added. It makes the compiler 13395 output stack usage information for the program, on a per-function 13396 basis, in an auxiliary file. 13397 * A new switch -fcombine-stack-adjustments has been added. It can be 13398 used to enable or disable the compiler's stack-slot combining pass 13399 which before was enabled automatically at -O1 and above, but could 13400 not be controlled on its own. 13401 * A new switch -fstrict-volatile-bitfields has been added. Using it 13402 indicates that accesses to volatile bitfields should use a single 13403 access of the width of the field's type. This option can be useful 13404 for precisely defining and accessing memory-mapped peripheral 13405 registers from C or C++. 13406 13407 Compile time and memory usage improvements 13408 13409 * Datastructures used by the dataflow framework in GCC were 13410 reorganized for better memory usage and more cache locality. 13411 Compile time is improved especially on units with large functions 13412 (possibly resulting from a lot of inlining) not fitting into the 13413 processor cache. The compile time of the GCC C compiler binary with 13414 link-time optimization went down by over 10% (benchmarked on x86-64 13415 target). 13416 13417 New Languages and Language specific improvements 13418 13419 Ada 13420 13421 * Stack checking has been improved on selected architectures (Alpha, 13422 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 13423 overflows in all cases on these architectures. 13424 * Initial support for Ada 2012 has been added. 13425 13426 C family 13427 13428 * A new warning, enabled by -Wdouble-promotion, has been added that 13429 warns about cases where a value of type float is implicitly 13430 promoted to double. This is especially helpful for CPUs that handle 13431 the former in hardware, but emulate the latter in software. 13432 * A new function attribute leaf was introduced. This attribute allows 13433 better inter-procedural optimization across calls to functions that 13434 return to the current unit only via returning or exception 13435 handling. This is the case for most library functions that have no 13436 callbacks. 13437 * Support for a new data type __int128 for targets having wide enough 13438 machine-mode support. 13439 * The new function attribute callee_pop_aggregate allows to specify 13440 if the caller or callee is responsible for popping the aggregate 13441 return pointer value from the stack. 13442 * Support for selectively enabling and disabling warnings via #pragma 13443 GCC diagnostic has been added. For instance: 13444 #pragma GCC diagnostic error "-Wuninitialized" 13445 foo(a); /* error is given for this one */ 13446 #pragma GCC diagnostic push 13447 #pragma GCC diagnostic ignored "-Wuninitialized" 13448 foo(b); /* no diagnostic for this one */ 13449 #pragma GCC diagnostic pop 13450 foo(c); /* error is given for this one */ 13451 #pragma GCC diagnostic pop 13452 foo(d); /* depends on command-line options */ 13453 13454 * The -fmax-errors=N option is now supported. Using this option 13455 causes the compiler to exit after N errors have been issued. 13456 13457 [7]C 13458 13459 * There is now experimental support for some features from the 13460 upcoming C1X revision of the ISO C standard. This support may be 13461 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 13462 Note that this support is experimental and may change incompatibly 13463 in future releases for consistency with changes to the C1X standard 13464 draft. The following features are newly supported as described in 13465 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 13466 meeting); some other features were already supported with no 13467 compiler changes being needed, or have some support but not in full 13468 accord with N1539 (as amended). 13469 + Static assertions (_Static_assert keyword) 13470 + Typedef redefinition 13471 + New macros in <float.h> 13472 + Anonymous structures and unions 13473 * The new -fplan9-extensions option directs the compiler to support 13474 some extensions for anonymous struct fields which are implemented 13475 by the Plan 9 compiler. A pointer to a struct may be automatically 13476 converted to a pointer to an anonymous field when calling a 13477 function, in order to make the types match. An anonymous struct 13478 field whose type is a typedef name may be referred to using the 13479 typedef name. 13480 13481 [8]C++ 13482 13483 * Improved [9]experimental support for the upcoming C++0x ISO C++ 13484 standard, including support for constexpr (thanks to Gabriel Dos 13485 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 13486 noexcept, unrestricted unions, range-based for loops (thanks to 13487 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 13488 Rodrigo), implicitly deleted functions and implicit move 13489 constructors. 13490 * When an extern declaration within a function does not match a 13491 declaration in the enclosing context, G++ now properly declares the 13492 name within the namespace of the function rather than the namespace 13493 which was open just before the function definition ([10]c++/43145). 13494 * GCC now warns by default when casting integers to larger pointer 13495 types. These warnings can be disabled with the option 13496 -Wno-int-to-pointer-cast, which is now also available in C++. 13497 * G++ no longer optimizes using the assumption that a value of 13498 enumeration type will fall within the range specified by the 13499 standard, since that assumption is easily violated with a 13500 conversion from integer type ([11]c++/43680). The old behavior can 13501 be restored with -fstrict-enums. 13502 * The new -fnothrow-opt flag changes the semantics of a throw() 13503 exception specification to match the proposed semantics of the 13504 noexcept specification: just call terminate if an exception tries 13505 to propagate out of a function with such an exception 13506 specification. This dramatically reduces or eliminates the code 13507 size overhead from adding the exception specification. 13508 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 13509 a function that the compiler can tell doesn't throw if it would 13510 change the value of a noexcept expression. 13511 * The -Wshadow option now warns if a local variable or type 13512 declaration shadows another type in C++. Note that the compiler 13513 will not warn if a local variable shadows a struct/class/enum, but 13514 will warn if it shadows an explicit typedef. 13515 * When an identifier is not found in the current scope, G++ now 13516 offers suggestions about which identifier might have been intended. 13517 * G++ now issues clearer diagnostics for missing semicolons after 13518 class, struct, and union definitions. 13519 * G++ now issues clearer diagnostics for missing semicolons after 13520 class member declarations. 13521 * G++ now issues clearer diagnostics when a colon is used in a place 13522 where a double-colon was intended. 13523 * G++ no longer accepts mutable on reference members ([12]c++/33558). 13524 Use -fpermissive to allow the old, non-conforming behaviour. 13525 * A few mangling fixes have been made, to attribute const/volatile on 13526 function pointer types, decltype of a plain decl, and use of a 13527 function parameter in the declaration of another parameter. By 13528 default the compiler still uses the old mangling, but emits aliases 13529 with the new mangling on targets that support strong aliases. Users 13530 can switch over entirely to the new mangling with -fabi-version=5 13531 or -fabi-version=0. -Wabi will now warn about code that uses the 13532 old mangling. 13533 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 13534 type to be default initialized unless the type has a user-declared 13535 default constructor. In 4.6.2 G++ implements the proposed 13536 resolution of [13]DR 253, so default initialization is allowed if 13537 it initializes all subobjects. Code that fails to compile can be 13538 fixed by providing an initializer e.g. 13539 struct A { A(); }; 13540 struct B : A { int i; }; 13541 const B b = B(); 13542 Use -fpermissive to allow the old, non-conforming behaviour. 13543 13544 Runtime Library (libstdc++) 13545 13546 * [14]Improved experimental support for the upcoming ISO C++ 13547 standard, C++0x, including using constexpr and nullptr. 13548 * Performance improvements to the [15]Debug Mode, thanks to Franois 13549 Dumont. 13550 * Atomic operations used for reference-counting are annotated so that 13551 they can be understood by race detectors such as Helgrind, see 13552 [16]Data Race Hunting. 13553 * Most libstdc++ standard headers have been changed to no longer 13554 include the cstddef header as an implementation detail. Code that 13555 relied on that header being included as side-effect of including 13556 other standard headers will need to include cstddef explicitly. 13557 13558 [17]Fortran 13559 13560 * On systems supporting the libquadmath library, GNU Fortran now also 13561 supports a quad-precision, kind=16 floating-point data type 13562 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 13563 hardware, calculations might be one to two orders of magnitude 13564 slower than with the 4, 8 or 10 bytes floating-point data types. 13565 This change does not affect systems which support REAL(16) in 13566 hardware nor those which do not support libquadmath. 13567 * Much improved compile time for large array constructors. 13568 * In order to reduce execution time and memory consumption, use of 13569 temporary arrays in assignment expressions is avoided for many 13570 cases. The compiler now reverses loops in order to avoid generating 13571 a temporary array where possible. 13572 * Improved diagnostics, especially with -fwhole-file. 13573 * The -fwhole-file flag is now enabled by default. This improves code 13574 generation and diagnostics. It can be disabled using the deprecated 13575 -fno-whole-file flag. 13576 * Support the generation of Makefile dependencies via the [18]-M... 13577 flags of GCC; you may need to specify the -cpp option in addition. 13578 The dependencies take modules, Fortran's include, and CPP's 13579 #include into account. Note: Using -M for the module path is no 13580 longer supported, use -J instead. 13581 * The flag -Wconversion has been modified to only issue warnings 13582 where a conversion leads to information loss. This drastically 13583 reduces the number of warnings; -Wconversion is thus now enabled 13584 with -Wall. The flag -Wconversion-extra has been added and also 13585 warns about other conversions; -Wconversion-extra typically issues 13586 a huge number of warnings, most of which can be ignored. 13587 * A new command-line option -Wunused-dummy-argument warns about 13588 unused dummy arguments and is included in -Wall. Before, 13589 -Wunused-variable also warned about unused dummy arguments. 13590 * Fortran 2003 support has been extended: 13591 + Improved support for polymorphism between libraries and 13592 programs and for complicated inheritance patterns (cf. 13593 [19]object-oriented programming). 13594 + Experimental support of the ASSOCIATE construct. 13595 + In pointer assignments it is now possible to specify the lower 13596 bounds of the pointer and, for a rank-1 or a simply contiguous 13597 data-target, to remap the bounds. 13598 + Automatic (re)allocation: In intrinsic assignments to 13599 allocatable variables the left-hand side will be automatically 13600 allocated (if unallocated) or reallocated (if the shape or 13601 type parameter is different). To avoid the small performance 13602 penalty, you can use a(:) = ... instead of a = ... for arrays 13603 and character strings - or disable the feature using -std=f95 13604 or -fno-realloc-lhs. 13605 + Deferred type parameter: For scalar allocatable and pointer 13606 variables the character length can be deferred. 13607 + Namelist variables with allocatable and pointer attribute and 13608 nonconstant length type parameter are supported. 13609 * Fortran 2008 support has been extended: 13610 + Experimental [20]coarray support (for one image only, i.e. 13611 num_images() == 1); use the [21]-fcoarray=single flag to 13612 enable it. 13613 + The STOP and the new ERROR STOP statements now support all 13614 constant expressions. 13615 + Support for the CONTIGUOUS attribute. 13616 + Support for ALLOCATE with MOLD. 13617 + Support for the STORAGE_SIZE intrinsic inquiry function. 13618 + Support of the NORM2 and PARITY intrinsic functions. 13619 + The following bit intrinsics were added: POPCNT and POPPAR for 13620 counting the number of 1 bits and returning the parity; BGE, 13621 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 13622 for combined left and right shifts, MASKL and MASKR for simple 13623 left and right justified masks, MERGE_BITS for a bitwise merge 13624 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 13625 and the transformational bit intrinsics IALL, IANY and 13626 IPARITY. 13627 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 13628 + Support for the IMPURE attribute for procedures, which allows 13629 for ELEMENTAL procedures without the restrictions of PURE. 13630 + Null pointers (including NULL()) and not allocated variables 13631 can be used as actual argument to optional non-pointer, 13632 non-allocatable dummy arguments, denoting an absent argument. 13633 + Non-pointer variables with TARGET attribute can be used as 13634 actual argument to POINTER dummies with INTENT(IN) 13635 + Pointers including procedure pointers and those in a derived 13636 type (pointer components) can now be initialized by a target 13637 instead of only by NULL. 13638 + The EXIT statement (with construct-name) can now be used to 13639 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 13640 SELECT CASE and SELECT TYPE constructs. 13641 + Internal procedures can now be used as actual argument. 13642 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 13643 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 13644 have been added; these arrays contain the supported kind 13645 values for the respective types. 13646 + The module procedures C_SIZEOF of the intrinsic module 13647 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 13648 ISO_FORTRAN_ENV have been implemented. 13649 + Minor changes: obsolescence diagnostics for ENTRY was added 13650 for -std=f2008; a line may start with a semicolon; for 13651 internal and module procedures END can be used instead of END 13652 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 13653 a RADIX argument; intrinsic types are supported for 13654 TYPE(intrinsic-type-spec); multiple type-bound procedures can 13655 be declared in a single PROCEDURE statement; implied-shape 13656 arrays are supported for named constants (PARAMETER). The 13657 transformational, three argument versions of BESSEL_JN and 13658 BESSEL_YN were added - the elemental, two-argument version had 13659 been added in GCC 4.4; note that the transformational 13660 functions use a recurrence algorithm. 13661 13662 [22]Go 13663 13664 Support for the Go programming language has been added to GCC. It is 13665 not enabled by default when you build GCC; use the --enable-languages 13666 configure option to build it. The driver program for compiling Go code 13667 is gccgo. 13668 13669 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 13670 is in progress. It may or may not work on other platforms. 13671 13672 [23]Objective-C and Objective-C++ 13673 13674 * The -fobjc-exceptions flag is now required to enable Objective-C 13675 exception and synchronization syntax (introduced by the keywords 13676 @try, @catch, @finally and @synchronized). 13677 * A number of Objective-C 2.0 features and extensions are now 13678 supported by GCC. These features are enabled by default; you can 13679 disable them by using the new -fobjc-std=objc1 command-line option. 13680 * The Objective-C 2.0 dot-syntax is now supported. It is an 13681 alternative syntax for using getters and setters; object.count is 13682 automatically converted into [object count] or [object setCount: 13683 ...] depending on context; for example if (object.count > 0) is 13684 automatically compiled into the equivalent of if ([object count] > 13685 0) while object.count = 0; is automatically compiled into the 13686 equivalent ot [object setCount: 0];. The dot-syntax can be used 13687 with instance and class objects and with any setters or getters, no 13688 matter if they are part of a declared property or not. 13689 * Objective-C 2.0 declared properties are now supported. They are 13690 declared using the new @property keyword, and are most commonly 13691 used in conjunction with the new Objective-C 2.0 dot-syntax. The 13692 nonatomic, readonly, readwrite, assign, retain, copy, setter and 13693 getter attributes are all supported. Marking declared properties 13694 with __attribute__ ((deprecated)) is supported too. 13695 * The Objective-C 2.0 @synthesize and @dynamic keywords are 13696 supported. @synthesize causes the compiler to automatically 13697 synthesize a declared property, while @dynamic is used to disable 13698 all warnings for a declared property for which no implementation is 13699 provided at compile time. Synthesizing declared properties requires 13700 runtime support in most useful cases; to be able to use it with the 13701 GNU runtime, appropriate helper functions have been added to the 13702 GNU Objective-C runtime ABI, and are implemented by the GNU 13703 Objective-C runtime library shipped with GCC. 13704 * The Objective-C 2.0 fast enumeration syntax is supported in 13705 Objective-C. This is currently not yet available in Objective-C++. 13706 Fast enumeration requires support in the runtime, and such support 13707 has been added to the GNU Objective-C runtime library (shipped with 13708 GCC). 13709 * The Objective-C 2.0 @optional keyword is supported. It allows you 13710 to mark methods or properties in a protocol as optional as opposed 13711 to required. 13712 * The Objective-C 2.0 @package keyword is supported. It has currently 13713 the same effect as the @public keyword. 13714 * Objective-C 2.0 method attributes are supported. Currently the 13715 supported attributes are deprecated, sentinel, noreturn and format. 13716 * Objective-C 2.0 method argument attributes are supported. The most 13717 widely used attribute is unused, to mark an argument as unused in 13718 the implementation. 13719 * Objective-C 2.0 class and protocol attributes are supported. 13720 Currently the only supported attribute is deprecated. 13721 * Objective-C 2.0 class extensions are supported. A class extension 13722 has the same syntax as a category declaration with no category 13723 name, and the methods and properties declared in it are added 13724 directly to the main class. It is mostly used as an alternative to 13725 a category to add methods to a class without advertising them in 13726 the public headers, with the advantage that for class extensions 13727 the compiler checks that all the privately declared methods are 13728 actually implemented. 13729 * As a result of these enhancements, GCC can now be used to build 13730 Objective-C and Objective-C++ software that uses Foundation and 13731 other important system frameworks with the NeXT runtime on Darwin 9 13732 and Darwin 10 (OSX 10.5 and 10.6). 13733 * Many bugs in the compiler have been fixed in this release; in 13734 particular, LTO can now be used when compiling Objective-C and 13735 Objective-C++ and the parser is much more robust in dealing with 13736 invalid code. 13737 13738 Runtime Library (libobjc) 13739 13740 * The GNU Objective-C runtime library now defines the macro 13741 __GNU_LIBOBJC__ (with a value that is increased at every release 13742 where there is any change to the API) in objc/objc.h, making it 13743 easy to determine if the GNU Objective-C runtime library is being 13744 used, and if so, which version. Previous versions of the GNU 13745 Objective-C runtime library (and other Objective-C runtime 13746 libraries such as the Apple one) do not define this macro. 13747 * A new Objective-C 2.0 API, almost identical to the one implemented 13748 by the Apple Objective-C runtime, has been implemented in the GNU 13749 Objective-C runtime library. The new API hides the internals of 13750 most runtime structures but provides a more extensive set of 13751 functions to operate on them. It is much easier, for example, to 13752 create or modify classes at runtime. The new API also makes it 13753 easier to port software from Apple to GNU as almost no changes 13754 should be required. The old API is still supported for backwards 13755 compatibility; including the old objc/objc-api.h header file 13756 automatically selects the old API, while including the new 13757 objc/runtime.h header file automatically selects the new API. 13758 Support for the old API is being phased out and upgrading the 13759 software to use the new API is strongly recommended. To check for 13760 the availability of the new API, the __GNU_LIBOBJC__ macro can be 13761 used as older versions of the GNU Objective-C runtime library, 13762 which do not support the new API, do not define such a macro. 13763 * Runtime support for @synchronized has been added. 13764 * Runtime support for Objective-C 2.0 synthesized property accessors 13765 has been added. 13766 * Runtime support for Objective-C 2.0 fast enumeration has been 13767 added. 13768 13769 [24]New Targets and Target Specific Improvements 13770 13771 [25]ARM 13772 13773 * GCC now supports the Cortex-M4 processor implementing the v7-em 13774 version of the architecture using the option -mcpu=cortex-m4. 13775 * Scheduling descriptions for the Cortex-M4, the Neon and the 13776 floating point units of the Cortex-A9 and a pipeline description 13777 for the Cortex-A5 have been added. 13778 * Synchronization primitives such as __sync_fetch_and_add and friends 13779 are now inlined for supported architectures rather than calling 13780 into a kernel helper function. 13781 * SSA loop prefetching is enabled by default for the Cortex-A9 at 13782 -O3. 13783 * Several improvements were committed to improve code generation for 13784 the ARM architecture including a rewritten implementation for load 13785 and store multiples. 13786 * Several enhancements were committed to improve SIMD code generation 13787 for NEON by adding support for widening instructions, misaligned 13788 loads and stores, vector conditionals and support for 64-bit 13789 arithmetic. 13790 * Support was added for the Faraday cores fa526, fa606te, fa626te, 13791 fmp626te, fmp626 and fa726te and can be used with the respective 13792 names as parameters to the -mcpu= option. 13793 * Basic support was added for Cortex-A15 and is available through 13794 -mcpu=cortex-a15. 13795 * GCC for AAPCS configurations now more closely adheres to the AAPCS 13796 specification by enabling -fstrict-volatile-bitfields by default. 13797 13798 IA-32/x86-64 13799 13800 * The new -fsplit-stack option permits programs to use a 13801 discontiguous stack. This is useful for threaded programs, in that 13802 it is no longer necessary to specify the maximum stack size when 13803 creating a thread. This feature is currently only implemented for 13804 32-bit and 64-bit x86 GNU/Linux targets. 13805 * Support for emitting profiler counter calls before function 13806 prologues. This is enabled via a new command-line option -mfentry. 13807 * Optimization for the Intel Core 2 processors is now available 13808 through the -march=core2 and -mtune=core2 options. 13809 * Support for Intel Core i3/i5/i7 processors is now available through 13810 the -march=corei7 and -mtune=corei7 options. 13811 * Support for Intel Core i3/i5/i7 processors with AVX is now 13812 available through the -march=corei7-avx and -mtune=corei7-avx 13813 options. 13814 * Support for AMD Bobcat (family 14) processors is now available 13815 through the -march=btver1 and -mtune=btver1 options. 13816 * Support for AMD Bulldozer (family 15) processors is now available 13817 through the -march=bdver1 and -mtune=bdver1 options. 13818 * The default setting (when not optimizing for size) for 32-bit 13819 GNU/Linux and Darwin x86 targets has been changed to 13820 -fomit-frame-pointer. The default can be reverted to 13821 -fno-omit-frame-pointer by configuring GCC with the 13822 --enable-frame-pointer configure option. 13823 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 13824 __float128 on 32-bit and 64-bit x86 targets. 13825 * AVX floating-point arithmetic can now be enabled by default at 13826 configure time with the new --with-fpmath=avx option. 13827 * The SSA loop prefetching pass is enabled when using -O3 when 13828 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 13829 than K6). 13830 * Support for TBM (Trailing Bit Manipulation) built-in functions and 13831 code generation is available via -mtbm. 13832 * Support for AMD's BMI (Bit Manipulation) built-in functions and 13833 code generation is available via -mbmi. 13834 13835 [26]MicroBlaze 13836 13837 * Support has been added for the Xilinx MicroBlaze softcore processor 13838 (microblaze-elf) embedded target. This configurable processor is 13839 supported on several Xilinx Spartan and Virtex FPGAs. 13840 13841 MIPS 13842 13843 * GCC now supports the Loongson 3A processor. Its canonical -march= 13844 and -mtune= name is loongson3a. 13845 13846 [27]MN10300 / AM33 13847 13848 * The inline assembly register constraint "A" has been renamed "c". 13849 This constraint is used to select a floating-point register that 13850 can be used as the destination of a multiply-accumulate 13851 instruction. 13852 * New inline assembly register constraints "A" and "D" have been 13853 added. These constraint letters resolve to all general registers 13854 when compiling for AM33, and resolve to address registers only or 13855 data registers only when compiling for MN10300. 13856 * The MDR register is represented in the compiler. One can access the 13857 register via the "z" constraint in inline assembly. It can be 13858 marked as clobbered or used as a local register variable via the 13859 "mdr" name. The compiler uses the RETF instruction if the function 13860 does not modify the MDR register, so it is important that inline 13861 assembly properly annotate any usage of the register. 13862 13863 PowerPC/PowerPC64 13864 13865 * GCC now supports the Applied Micro Titan processor with 13866 -mcpu=titan. 13867 * The -mrecip option has been added, which indicates whether the 13868 reciprocal and reciprocal square root instructions should be used. 13869 * The -mveclibabi=mass option can be used to enable the compiler to 13870 autovectorize mathematical functions using the Mathematical 13871 Acceleration Subsystem library. 13872 * The -msingle-pic-base option has been added, which instructs the 13873 compiler to avoid loading the PIC base register in function 13874 prologues. The PIC base register must be initialized by the runtime 13875 system. 13876 * The -mblock-move-inline-limit option has been added, which enables 13877 the user to control the maximum size of inlined memcpy calls and 13878 similar. 13879 * PowerPC64 GNU/Linux support for applications requiring a large TOC 13880 section has been improved. A new command-line option, 13881 -mcmodel=MODEL, controls this feature; valid values for MODEL are 13882 small, medium, or large. 13883 * The Altivec builtin functions vec_ld and vec_st have been modified 13884 to generate the Altivec memory instructions LVX and STVX, even if 13885 the -mvsx option is used. In the initial GCC 4.5 release, these 13886 builtin functions were changed to generate VSX memory reference 13887 instructions instead of Altivec memory instructions, but there are 13888 differences between the two instructions. If the VSX instruction 13889 set is available, you can now use the new builtin functions 13890 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 13891 instructions. 13892 * The GCC compiler on AIX now defaults to a process layout with a 13893 larger data space allowing larger programs to be compiled. 13894 * The GCC long double type on AIX 6.1 and above has reverted to 13895 64-bit double precision, matching the AIX XL compiler default, 13896 because of missing C99 symbols required by the GCC runtime. 13897 * The default processor scheduling model and tuning for PowerPC64 13898 GNU/Linux and for AIX 6.1 and above now is POWER7. 13899 * Starting with GCC 4.6.1, vectors of type vector long long or vector 13900 long are passed and returned in the same method as other vectors 13901 with the VSX instruction set. Previously the GCC compiler did not 13902 adhere to the ABI for 128-bit vectors with 64-bit integer base 13903 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 13904 13905 S/390, zSeries and System z9/z10, IBM zEnterprise z196 13906 13907 * Support for the zEnterprise z196 processor has been added. When 13908 using the -march=z196 option, the compiler will generate code 13909 making use of the following instruction facilities: 13910 + Conditional load/store 13911 + Distinct-operands 13912 + Floating-point-extension 13913 + Interlocked-access 13914 + Population-count 13915 The -mtune=z196 option avoids the compare and branch instructions 13916 as well as the load address instruction with an index register as 13917 much as possible and performs instruction scheduling appropriate 13918 for the new out-of-order pipeline architecture. 13919 * When using the -m31 -mzarch options the generated code still 13920 conforms to the 32-bit ABI but uses the general purpose registers 13921 as 64-bit registers internally. This requires a Linux kernel saving 13922 the whole 64-bit registers when doing a context switch. Kernels 13923 providing that feature indicate that by the 'highgprs' string in 13924 /proc/cpuinfo. 13925 * The SSA loop prefetching pass is enabled when using -O3. 13926 13927 SPARC 13928 13929 * GCC now supports the LEON series of SPARC V8 processors. The code 13930 generated by the compiler can either be tuned to it by means of the 13931 --with-tune=leon configure option and -mtune=leon compilation 13932 option, or the compiler can be built for the sparc-leon-{elf,linux} 13933 and sparc-leon3-{elf,linux} targets directly. 13934 * GCC has stopped sign/zero-extending parameter registers in the 13935 callee for functions taking parameters with sub-word size in 32-bit 13936 mode, since this is redundant with the specification of the ABI. 13937 GCC has never done so in 64-bit mode since this is also redundant. 13938 * The command-line option -mfix-at697f has been added to enable the 13939 documented workaround for the single erratum of the Atmel AT697F 13940 processor. 13941 13942 [28]Operating Systems 13943 13944 [29]Android 13945 13946 * GCC now supports the Bionic C library and provides a convenient way 13947 of building native libraries and applications for the Android 13948 platform. Refer to the documentation of the -mandroid and -mbionic 13949 options for details on building native code. At the moment, Android 13950 support is enabled only for ARM. 13951 13952 [30]Darwin/Mac OS X 13953 13954 * General 13955 + Initial support for CFString types has been added. 13956 This allows GCC to build projects including the system Core 13957 Foundation frameworks. The GCC Objective-C family supports 13958 CFString "toll-free bridged" as per the Mac OS X system tools. 13959 CFString is also recognized in the context of format 13960 attributes and arguments (see the documentation for format 13961 attributes for limitations). At present, 8-bit character types 13962 are supported. 13963 + Object file size reduction. 13964 The Darwin zeroed memory allocators have been re-written to 13965 make more use of .zerofill sections. For non-debug code, this 13966 can reduce object file size significantly. 13967 + Objective-C family 64-bit support (NeXT ABI 2). 13968 Initial support has been added to support 64-bit Objective-C 13969 code using the Darwin/OS X native (NeXT) runtime. ABI version 13970 2 will be selected automatically when 64-bit code is built. 13971 + Objective-C family 32-bit ABI 1. 13972 For 32-bit code ABI 1 is also now also allowed. At present it 13973 must be selected manually using -fobjc-abi-version=1 where 13974 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 13975 * x86 Architecture 13976 + The -mdynamic-no-pic option has been enabled. 13977 Code supporting -mdynamic-no-pic optimization has been added 13978 and is applicable to -m32 builds. The compiler bootstrap uses 13979 the option where appropriate. 13980 + The default value for -mtune= has been changed. 13981 Since Darwin systems are primarily Xeon, Core-2 or similar the 13982 default tuning has been changed to -mtune=core2. 13983 + Enable 128-bit long double (__float128) support on Darwin. 13984 * PPC Architecture 13985 + Darwin64 ABI. 13986 Several significant bugs have been fixed, such that GCC now 13987 produces code compatible with the Darwin64 PowerPC ABI. 13988 + libffi and boehm-gc. 13989 The Darwin ports of the libffi and boehm-gc libraries have 13990 been upgraded to include a Darwin64 implementation. This means 13991 that powerpc*-*-darwin9 platforms may now, for example, build 13992 Java applications with -m64 enabled. 13993 + Plug-in support has been enabled. 13994 + The -fsection-anchors option is now available although, 13995 presently, not heavily tested. 13996 13997 [31]Solaris 2 13998 13999 New Features 14000 14001 * Support symbol versioning with the Sun linker. 14002 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 14003 * Support thread-local storage (TLS) with the Sun assembler on 14004 Solaris 2/x86. 14005 * Support TLS on Solaris 8/9 if prerequisites are met. 14006 * Support COMDAT group with the GNU assembler and recent Sun linker. 14007 * Support the Sun assembler visibility syntax. 14008 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 14009 -march=pentiumpro (Solaris 8/9). 14010 * Don't use SSE on Solaris 8/9 x86 by default. 14011 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 14012 14013 ABI Change 14014 14015 * Change the ABI for returning 8-byte vectors like __m64 in MMX 14016 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 14017 compilers. This is an incompatible change. If you use such types, 14018 you must either recompile all your code with the new compiler or 14019 use the new -mvect8-ret-in-mem option to remain compatible with 14020 previous versions of GCC and Sun Studio. 14021 14022 [32]Windows x86/x86_64 14023 14024 * Initial support for decimal floating point. 14025 * Support for the __thiscall calling-convention. 14026 * Support for hot-patchable function prologues via the 14027 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 14028 * Improvements of stack-probing and stack-allocation mechanisms. 14029 * Support of push/pop-macro pragma as preprocessor command. 14030 With #pragma push_macro("macro-name") the current definition of 14031 macro-name is saved and can be restored with #pragma 14032 pop_macro("macro-name") to its saved definition. 14033 * Enable 128-bit long double (__float128) support on MinGW and 14034 Cygwin. 14035 14036 Other significant improvements 14037 14038 Installation changes 14039 14040 * An install-strip make target is provided that installs stripped 14041 executables, and may install libraries with unneeded or debugging 14042 sections stripped. 14043 * On Power7 systems, there is a potential problem if you build the 14044 GCC compiler with a host compiler using options that enable the VSX 14045 instruction set generation. If the host compiler has been patched 14046 so that the vec_ld and vec_st builtin functions generate Altivec 14047 memory instructions instead of VSX memory instructions, then you 14048 should be able to build the compiler with VSX instruction 14049 generation. 14050 14051 Changes for GCC Developers 14052 14053 Note: these changes concern developers that develop GCC itself or 14054 software that integrates with GCC, such as plugins, and not the general 14055 GCC users. 14056 * The gengtype utility, which previously was internal to the GCC 14057 build process, has been enchanced to provide GC root information 14058 for plugins as necessary. 14059 * The old GC allocation interface of ggc_alloc and friends was 14060 replaced with a type-safe alternative. 14061 14062 [33]GCC 4.6.1 14063 14064 This is the [34]list of problem reports (PRs) from GCC's bug tracking 14065 system that are known to be fixed in the 4.6.1 release. This list might 14066 not be complete (that is, it is possible that some PRs that have been 14067 fixed are not listed here). 14068 14069 [35]GCC 4.6.2 14070 14071 This is the [36]list of problem reports (PRs) from GCC's bug tracking 14072 system that are known to be fixed in the 4.6.2 release. This list might 14073 not be complete (that is, it is possible that some PRs that have been 14074 fixed are not listed here). 14075 14076 [37]GCC 4.6.3 14077 14078 This is the [38]list of problem reports (PRs) from GCC's bug tracking 14079 system that are known to be fixed in the 4.6.3 release. This list might 14080 not be complete (that is, it is possible that some PRs that have been 14081 fixed are not listed here). 14082 14083 [39]GCC 4.6.4 14084 14085 This is the [40]list of problem reports (PRs) from GCC's bug tracking 14086 system that are known to be fixed in the 4.6.4 release. This list might 14087 not be complete (that is, it is possible that some PRs that have been 14088 fixed are not listed here). 14089 14090 14091 For questions related to the use of GCC, please consult these web 14092 pages and the [41]GCC manuals. If that fails, the 14093 [42]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14094 web pages and the development of GCC are welcome on our developer 14095 list at [43]gcc (a] gcc.gnu.org. All of [44]our lists have public 14096 archives. 14097 14098 Copyright (C) [45]Free Software Foundation, Inc. Verbatim copying and 14099 distribution of this entire article is permitted in any medium, 14100 provided this notice is preserved. 14101 14102 These pages are [46]maintained by the GCC team. Last modified 14103 2025-01-31. 14104 14105 References 14106 14107 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 14108 2. https://gcc.gnu.org/gcc-4.5/changes.html 14109 3. https://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 14110 4. https://gcc.gnu.org/gcc-4.6/porting_to.html 14111 5. https://gcc.gnu.org/projects/lto/whopr.pdf 14112 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 14113 7. https://gcc.gnu.org/gcc-4.6/changes.html#c 14114 8. https://gcc.gnu.org/gcc-4.6/changes.html#cplusplus 14115 9. https://gcc.gnu.org/gcc-4.6/cxx0x_status.html 14116 10. https://gcc.gnu.org/PR43145 14117 11. https://gcc.gnu.org/PR43680 14118 12. https://gcc.gnu.org/PR33558 14119 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 14120 14. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 14121 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 14122 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 14123 17. https://gcc.gnu.org/gcc-4.6/changes.html#fortran 14124 18. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 14125 19. https://gcc.gnu.org/wiki/OOP 14126 20. https://gcc.gnu.org/wiki/Coarray 14127 21. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 14128 22. https://gcc.gnu.org/gcc-4.6/changes.html#go 14129 23. https://gcc.gnu.org/gcc-4.6/changes.html#objective-c 14130 24. https://gcc.gnu.org/gcc-4.6/changes.html#targets 14131 25. https://gcc.gnu.org/gcc-4.6/changes.html#arm 14132 26. https://gcc.gnu.org/gcc-4.6/changes.html#microblaze 14133 27. https://gcc.gnu.org/gcc-4.6/changes.html#mn10300 14134 28. https://gcc.gnu.org/gcc-4.6/changes.html#os 14135 29. https://gcc.gnu.org/gcc-4.6/changes.html#android 14136 30. https://gcc.gnu.org/gcc-4.6/changes.html#darwin 14137 31. https://gcc.gnu.org/gcc-4.6/changes.html#solaris 14138 32. https://gcc.gnu.org/gcc-4.6/changes.html#windows 14139 33. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.1 14140 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 14141 35. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.2 14142 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 14143 37. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.3 14144 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 14145 39. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.4 14146 40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 14147 41. https://gcc.gnu.org/onlinedocs/ 14148 42. mailto:gcc-help (a] gcc.gnu.org 14149 43. mailto:gcc (a] gcc.gnu.org 14150 44. https://gcc.gnu.org/lists.html 14151 45. https://www.fsf.org/ 14152 46. https://gcc.gnu.org/about.html 14153 ====================================================================== 14154 http://gcc.gnu.org/gcc-4.5/index.html 14155 14156 GCC 4.5 Release Series 14157 14158 (This release series is no longer supported.) 14159 14160 Jul 2, 2012 14161 14162 The [1]GNU project and the GCC developers are pleased to announce the 14163 release of GCC 4.5.4. 14164 14165 This release is a bug-fix release, containing fixes for regressions in 14166 GCC 4.5.3 relative to previous releases of GCC. 14167 14168 Release History 14169 14170 GCC 4.5.4 14171 Jul 2, 2012 ([2]changes) 14172 14173 GCC 4.5.3 14174 Apr 28, 2011 ([3]changes) 14175 14176 GCC 4.5.2 14177 Dec 16, 2010 ([4]changes) 14178 14179 GCC 4.5.1 14180 Jul 31, 2010 ([5]changes) 14181 14182 GCC 4.5.0 14183 April 14, 2010 ([6]changes) 14184 14185 References and Acknowledgements 14186 14187 GCC used to stand for the GNU C Compiler, but since the compiler 14188 supports several other languages aside from C, it now stands for the 14189 GNU Compiler Collection. 14190 14191 A list of [7]successful builds is updated as new information becomes 14192 available. 14193 14194 The GCC developers would like to thank the numerous people that have 14195 contributed new features, improvements, bug fixes, and other changes as 14196 well as test results to GCC. This [8]amazing group of volunteers is 14197 what makes GCC successful. 14198 14199 For additional information about GCC please refer to the [9]GCC project 14200 web site or contact the [10]GCC development mailing list. 14201 14202 To obtain GCC please use [11]our mirror sites or [12]our version 14203 control system. 14204 14205 14206 For questions related to the use of GCC, please consult these web 14207 pages and the [13]GCC manuals. If that fails, the 14208 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14209 web pages and the development of GCC are welcome on our developer 14210 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 14211 archives. 14212 14213 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14214 distribution of this entire article is permitted in any medium, 14215 provided this notice is preserved. 14216 14217 These pages are [18]maintained by the GCC team. Last modified 14218 2022-10-26. 14219 14220 References 14221 14222 1. http://www.gnu.org/ 14223 2. https://gcc.gnu.org/gcc-4.5/changes.html 14224 3. https://gcc.gnu.org/gcc-4.5/changes.html 14225 4. https://gcc.gnu.org/gcc-4.5/changes.html 14226 5. https://gcc.gnu.org/gcc-4.5/changes.html 14227 6. https://gcc.gnu.org/gcc-4.5/changes.html 14228 7. https://gcc.gnu.org/gcc-4.5/buildstat.html 14229 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14230 9. https://gcc.gnu.org/index.html 14231 10. mailto:gcc (a] gcc.gnu.org 14232 11. https://gcc.gnu.org/mirrors.html 14233 12. https://gcc.gnu.org/git.html 14234 13. https://gcc.gnu.org/onlinedocs/ 14235 14. mailto:gcc-help (a] gcc.gnu.org 14236 15. mailto:gcc (a] gcc.gnu.org 14237 16. https://gcc.gnu.org/lists.html 14238 17. https://www.fsf.org/ 14239 18. https://gcc.gnu.org/about.html 14240 ====================================================================== 14241 http://gcc.gnu.org/gcc-4.5/changes.html 14242 14243 GCC 4.5 Release Series 14244 Changes, New Features, and Fixes 14245 14246 Caveats 14247 14248 * GCC now requires the [1]MPC library in order to build. See the 14249 [2]prerequisites page for version requirements. 14250 * Support for a number of older systems and recently unmaintained or 14251 untested target ports of GCC has been declared obsolete in GCC 4.5. 14252 Unless there is activity to revive them, the next release of GCC 14253 will have their sources permanently removed. 14254 The following ports for individual systems on particular 14255 architectures have been obsoleted: 14256 + IRIX releases before 6.5 (mips-sgi-irix5*, 14257 mips-sgi-irix6.[0-4]) 14258 + Solaris 7 (*-*-solaris2.7) 14259 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 14260 alpha-dec-osf5.0*) 14261 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 14262 can be found in the [3]announcement. 14263 Support for the classic POWER architecture implemented in the 14264 original RIOS and RIOS2 processors of the old IBM RS/6000 product 14265 line has been obsoleted in the rs6000 port. This does not affect 14266 the new generation Power and PowerPC architectures. 14267 * Support has been removed for all the [4]configurations obsoleted in 14268 GCC 4.4. 14269 * Support has been removed for the protoize and unprotoize utilities, 14270 obsoleted in GCC 4.4. 14271 * Support has been removed for tuning for Itanium1 (Merced) variants. 14272 Note that code tuned for Itanium2 should also run correctly on 14273 Itanium1. 14274 * GCC now generates unwind info also for epilogues. DWARF debuginfo 14275 generated by GCC now uses more features of DWARF3 than before, and 14276 also some DWARF4 features. GDB older than 7.0 is not able to handle 14277 either of these, so to debug GCC 4.5 generated binaries or 14278 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 14279 features with the -gdwarf-3 -gstrict-dwarf options, or use 14280 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 14281 epilogue unwind info is emitted unconditionally whenever unwind 14282 info is emitted. 14283 * On x86 targets, code containing floating-point calculations may run 14284 significantly more slowly when compiled with GCC 4.5 in strict C99 14285 conformance mode than they did with earlier GCC versions. This is 14286 due to stricter standard conformance of the compiler and can be 14287 avoided by using the option -fexcess-precision=fast; also see 14288 [5]below. 14289 * The function attribute noinline no longer prevents GCC from cloning 14290 the function. A new attribute noclone has been introduced for this 14291 purpose. Cloning a function means that it is duplicated and the new 14292 copy is specialized for certain contexts (for example when a 14293 parameter is a known constant). 14294 14295 General Optimizer Improvements 14296 14297 * The -save-temps now takes an optional argument. The -save-temps and 14298 -save-temps=cwd switches write the temporary files in the current 14299 working directory based on the original source file. The 14300 -save-temps=obj switch will write files into the directory 14301 specified with the -o option, and the intermediate filenames are 14302 based on the output file. This will allow the user to get the 14303 compiler intermediate files when doing parallel builds without two 14304 builds of the same filename located in different directories from 14305 interfering with each other. 14306 * Debugging dumps are now created in the same directory as the object 14307 file rather than in the current working directory. This allows the 14308 user to get debugging dumps when doing parallel builds without two 14309 builds of the same filename interfering with each other. 14310 * GCC has been integrated with the MPC library. This allows GCC to 14311 evaluate complex arithmetic at compile time [6]more accurately. It 14312 also allows GCC to evaluate calls to complex built-in math 14313 functions having constant arguments and replace them at compile 14314 time with their mathematically equivalent results. In doing so, GCC 14315 can generate correct results regardless of the math library 14316 implementation or floating point precision of the host platform. 14317 This also allows GCC to generate identical results regardless of 14318 whether one compiles in native or cross-compile configurations to a 14319 particular target. The following built-in functions take advantage 14320 of this new capability: cacos, cacosh, casin, casinh, catan, 14321 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 14322 and ctanh. The float and long double variants of these functions 14323 (e.g. csinf and csinl) are also handled. 14324 * A new link-time optimizer has been added ([7]-flto). When this 14325 option is used, GCC generates a bytecode representation of each 14326 input file and writes it to specially-named sections in each object 14327 file. When the object files are linked together, all the function 14328 bodies are read from these named sections and instantiated as if 14329 they had been part of the same translation unit. This enables 14330 interprocedural optimizations to work across different files (and 14331 even different languages), potentially improving the performance of 14332 the generated code. To use the link-timer optimizer, -flto needs to 14333 be specified at compile time and during the final link. If the 14334 program does not require any symbols to be exported, it is possible 14335 to combine -flto and the experimental [8]-fwhopr with 14336 [9]-fwhole-program to allow the interprocedural optimizers to use 14337 more aggressive assumptions. 14338 * The automatic parallelization pass was enhanced to support 14339 parallelization of outer loops. 14340 * Automatic parallelization can be enabled as part of Graphite. In 14341 addition to -ftree-parallelize-loops=, specify 14342 -floop-parallelize-all to enable the Graphite-based optimization. 14343 * The infrastructure for optimizing based on [10]restrict qualified 14344 pointers has been rewritten and should result in code generation 14345 improvements. Optimizations based on restrict qualified pointers 14346 are now also available when using -fno-strict-aliasing. 14347 * There is a new optimization pass that attempts to change prototype 14348 of functions to avoid unused parameters, pass only relevant parts 14349 of structures and turn arguments passed by reference to arguments 14350 passed by value when possible. It is enabled by -O2 and above as 14351 well as -Os and can be manually invoked using the new command-line 14352 switch -fipa-sra. 14353 * GCC now optimize exception handling code. In particular cleanup 14354 regions that are proved to not have any effect are optimized out. 14355 14356 New Languages and Language specific improvements 14357 14358 All languages 14359 14360 * The -fshow-column option is now on by default. This means error 14361 messages now have a column associated with them. 14362 14363 Ada 14364 14365 * Compilation of programs heavily using discriminated record types 14366 with variant parts has been sped up and generates more compact 14367 code. 14368 * Stack checking now works reasonably well on most plaforms. In some 14369 specific cases, stack overflows may still fail to be detected, but 14370 a compile-time warning will be issued for these cases. 14371 14372 C family 14373 14374 * If a header named in a #include directive is not found, the 14375 compiler exits immediately. This avoids a cascade of errors arising 14376 from declarations expected to be found in that header being 14377 missing. 14378 * A new built-in function __builtin_unreachable() has been added that 14379 tells the compiler that control will never reach that point. It may 14380 be used after asm statements that terminate by transferring control 14381 elsewhere, and in other places that are known to be unreachable. 14382 * The -Wlogical-op option now warns for logical expressions such as 14383 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 14384 mistakes. This option is disabled by default. 14385 * An asm goto feature has been added to allow asm statements that 14386 jump to C labels. 14387 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 14388 * The deprecated attribute now takes an optional string argument, for 14389 example, __attribute__((deprecated("text string"))), that will be 14390 printed together with the deprecation warning. 14391 14392 C 14393 14394 * The -Wenum-compare option, which warns when comparing values of 14395 different enum types, now works for C. It formerly only worked for 14396 C++. This warning is enabled by -Wall. It may be avoided by using a 14397 type cast. 14398 * The -Wcast-qual option now warns about casts which are unsafe in 14399 that they permit const-correctness to be violated without further 14400 warnings. Specifically, it warns about cases where a qualifier is 14401 added when all the lower types are not const. For example, it warns 14402 about a cast from char ** to const char **. 14403 * The -Wc++-compat option is significantly improved. It issues new 14404 warnings for: 14405 + Using C++ reserved operator names as identifiers. 14406 + Conversions to enum types without explicit casts. 14407 + Using va_arg with an enum type. 14408 + Using different enum types in the two branches of ?:. 14409 + Using ++ or -- on a variable of enum type. 14410 + Using the same name as both a struct, union or enum tag and a 14411 typedef, unless the typedef refers to the tagged type itself. 14412 + Using a struct, union, or enum which is defined within another 14413 struct or union. 14414 + A struct field defined using a typedef if there is a field in 14415 the struct, or an enclosing struct, whose name is the typedef 14416 name. 14417 + Duplicate definitions at file scope. 14418 + Uninitialized const variables. 14419 + A global variable with an anonymous struct, union, or enum 14420 type. 14421 + Using a string constant to initialize a char array whose size 14422 is the length of the string. 14423 * The new -Wjump-misses-init option warns about cases where a goto or 14424 switch skips the initialization of a variable. This sort of branch 14425 is an error in C++ but not in C. This warning is enabled by 14426 -Wc++-compat. 14427 * GCC now ensures that a C99-conforming <stdint.h> is present on most 14428 targets, and uses information about the types in this header to 14429 implement the Fortran bindings to those types. GCC does not ensure 14430 the presence of such a header, and does not implement the Fortran 14431 bindings, on the following targets: NetBSD, VxWorks, VMS, 14432 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 14433 * GCC now implements C90- and C99-conforming rules for constant 14434 expressions. This may cause warnings or errors for some code using 14435 expressions that can be folded to a constant but are not constant 14436 expressions as defined by ISO C. 14437 * All known target-independent C90 and C90 Amendment 1 conformance 14438 bugs, and all known target-independent C99 conformance bugs not 14439 related to floating point or extended identifiers, have been fixed. 14440 * The C decimal floating point support now includes support for the 14441 FLOAT_CONST_DECIMAL64 pragma. 14442 * The named address space feature from ISO/IEC TR 18037 is now 14443 supported. This is currently only implemented for the SPU 14444 processor. 14445 14446 [11]C++ 14447 14448 * Improved [12]experimental support for the upcoming C++0x ISO C++ 14449 standard, including support for raw strings, lambda expressions and 14450 explicit type conversion operators. 14451 * When printing the name of a class template specialization, G++ will 14452 now omit any template arguments which come from default template 14453 arguments. This behavior (and the pretty-printing of function 14454 template specializations as template signature and arguments) can 14455 be disabled with the -fno-pretty-templates option. 14456 * Access control is now applied to typedef names used in a template, 14457 which may cause G++ to reject some ill-formed code that was 14458 accepted by earlier releases. The -fno-access-control option can be 14459 used as a temporary workaround until the code is corrected. 14460 * Compilation time for code that uses templates should now scale 14461 linearly with the number of instantiations rather than 14462 quadratically, as template instantiations are now looked up using 14463 hash tables. 14464 * Declarations of functions that look like builtin declarations of 14465 library functions are only considered to be redeclarations if they 14466 are declared with extern "C". This may cause problems with code 14467 that omits extern "C" on hand-written declarations of C library 14468 functions such as abort or memcpy. Such code is ill-formed, but was 14469 accepted by earlier releases. 14470 * Diagnostics that used to complain about passing non-POD types to 14471 ... or jumping past the declaration of a non-POD variable now check 14472 for triviality rather than PODness, as per C++0x. 14473 * In C++0x mode local and anonymous classes are now allowed as 14474 template arguments, and in declarations of variables and functions 14475 with linkage, so long as any such declaration that is used is also 14476 defined ([13]DR 757). 14477 * Labels may now have attributes, as has been permitted for a while 14478 in C. This is only permitted when the label definition and the 14479 attribute specifier is followed by a semicolon--i.e., the label 14480 applies to an empty statement. The only useful attribute for a 14481 label is unused. 14482 * G++ now implements [14]DR 176. Previously G++ did not support using 14483 the injected-class-name of a template base class as a type name, 14484 and lookup of the name found the declaration of the template in the 14485 enclosing scope. Now lookup of the name finds the 14486 injected-class-name, which can be used either as a type or as a 14487 template, depending on whether or not the name is followed by a 14488 template argument list. As a result of this change, some code that 14489 was previously accepted may be ill-formed because 14490 1. The injected-class-name is not accessible because it's from a 14491 private base, or 14492 2. The injected-class-name cannot be used as an argument for a 14493 template template parameter. 14494 In either of these cases, the code can be fixed by adding a 14495 nested-name-specifier to explicitly name the template. The first 14496 can be worked around with -fno-access-control; the second is only 14497 rejected with -pedantic. 14498 * A new standard mangling for SIMD vector types has been added, to 14499 avoid name clashes on systems with vectors of varying length. By 14500 default the compiler still uses the old mangling, but emits aliases 14501 with the new mangling on targets that support strong aliases. Users 14502 can switch over entirely to the new mangling with -fabi-version=4 14503 or -fabi-version=0. -Wabi will now warn about code that uses the 14504 old mangling. 14505 * The command-line option -ftemplate-depth-N is now written as 14506 -ftemplate-depth=N and the old form is deprecated. 14507 * Conversions between NULL and non-pointer types are now warned by 14508 default. The new option -Wno-conversion-null disables these 14509 warnings. Previously these warnings were only available when using 14510 -Wconversion explicitly. 14511 14512 Runtime Library (libstdc++) 14513 14514 * Improved experimental support for the upcoming ISO C++ standard, 14515 C++0x, including: 14516 + Support for <future>, <functional>, and <random>. 14517 + Existing facilities now exploit explicit operators and the 14518 newly implemented core C++0x features. 14519 + The header <cstdatomic> has been renamed to <atomic>. 14520 * An experimental [15]profile mode has been added. This is an 14521 implementation of many C++ standard library constructs with an 14522 additional analysis layer that gives performance improvement advice 14523 based on recognition of suboptimal usage patterns. For example, 14524 #include <vector> 14525 int main() 14526 { 14527 std::vector<int> v; 14528 for (int k = 0; k < 1024; ++k) 14529 v.insert(v.begin(), k); 14530 } 14531 14532 When instrumented via the profile mode, can return suggestions 14533 about the initial size and choice of the container used as follows: 14534 vector-to-list: improvement = 5: call stack = 0x804842c ... 14535 : advice = change std::vector to std::list 14536 vector-size: improvement = 3: call stack = 0x804842c ... 14537 : advice = change initial container size from 0 to 1024 14538 14539 These constructs can be substituted for the normal libstdc++ 14540 constructs on a piecemeal basis, or all existing components can be 14541 transformed via the -D_GLIBCXX_PROFILE macro. 14542 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR 14543 24733) has been added. This support is in header file 14544 <decimal/decimal>, uses namespace std::decimal, and includes 14545 classes decimal32, decimal64, and decimal128. 14546 * Sources have been audited for application of function attributes 14547 nothrow, const, pure, and noreturn. 14548 * Python pretty-printers have been added for many standard library 14549 components that simplify the internal representation and present a 14550 more intuitive view of components when used with 14551 appropriately-advanced versions of GDB. For more information, 14552 please consult the more [17]detailed description. 14553 * The default behavior for comparing typeinfo names has changed, so 14554 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 14555 * The new -static-libstdc++ option directs g++ to link the C++ 14556 library statically, even if the default would normally be to link 14557 it dynamically. 14558 14559 [18]Fortran 14560 14561 * The COMMON default padding has been changed - instead of adding the 14562 padding before a variable it is now added afterwards, which 14563 increases the compatibility with other vendors and helps to obtain 14564 the correct output in some cases. Cf. also the -falign-commons 14565 option ([19]added in 4.4). 14566 * The -finit-real= option now also supports the value snan for 14567 signaling not-a-number; to be effective, one additionally needs to 14568 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 14569 optimizations can turn a signaling NaN into a quiet one. 14570 * The new option -fcheck= has been added with the options bounds, 14571 array-temps, do, pointer, and recursive. The bounds and array-temps 14572 options are equivalent to -fbounds-check and 14573 -fcheck-array-temporaries. The do option checks for invalid 14574 modification of loop iteration variables, and the recursive option 14575 tests for recursive calls to subroutines/functions which are not 14576 marked as recursive. With pointer pointer association checks in 14577 calls are performed; however, neither undefined pointers nor 14578 pointers in expressions are handled. Using -fcheck=all enables all 14579 these run-time checks. 14580 * The run-time checking -fcheck=bounds now warns about invalid string 14581 lengths of character dummy arguments. Additionally, more 14582 compile-time checks have been added. 14583 * The new option [20]-fno-protect-parens has been added; if set, the 14584 compiler may reorder REAL and COMPLEX expressions without regard to 14585 parentheses. 14586 * GNU Fortran no longer links against libgfortranbegin. As before, 14587 MAIN__ (assembler symbol name) is the actual Fortran main program, 14588 which is invoked by the main function. However, main is now 14589 generated and put in the same object file as MAIN__. For the time 14590 being, libgfortranbegin still exists for backward compatibility. 14591 For details see the new [21]Mixed-Language Programming chapter in 14592 the manual. 14593 * The I/O library was restructured for performance and cleaner code. 14594 * Array assignments and WHERE are now run in parallel when OpenMP's 14595 WORKSHARE is used. 14596 * The experimental option -fwhole-file was added. The option allows 14597 whole-file checking of procedure arguments and allows for better 14598 optimizations. It can also be used with -fwhole-program, which is 14599 now also supported in gfortran. 14600 * More Fortran 2003 and Fortran 2008 mathematical functions can now 14601 be used as initialization expressions. 14602 * Some extended attributes such as STDCALL are now supported via the 14603 [22]GCC$ compiler directive. 14604 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 14605 intrinsic behaves now as if zero were always positive. 14606 * For legacy compatibiliy: On Cygwin and MinGW, the special files 14607 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 14608 supported. 14609 * Fortran 2003 support has been extended: 14610 + Procedure-pointer function results and procedure-pointer 14611 components (including PASS), 14612 + allocatable scalars (experimental), 14613 + DEFERRED type-bound procedures, 14614 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 14615 have been implemented. 14616 + The ALLOCATE statement supports type-specs and the SOURCE= 14617 argument. 14618 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 14619 type-bound procedure (i.e. as type-bound operators). 14620 + Rounding (ROUND=, RZ, ...) for output is now supported. 14621 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 14622 intrinsic module ISO_C_BINDING are now supported, except for 14623 the targets listed above as ones where GCC does not have 14624 <stdint.h> type information. 14625 + Extensible derived types with type-bound procedure or 14626 procedure pointer with PASS attribute now have to use CLASS in 14627 line with the Fortran 2003 standard; the workaround to use 14628 TYPE is no longer supported. 14629 + [23]Experimental, incomplete support for polymorphism, 14630 including CLASS, SELECT TYPE and dynamic dispatch of 14631 type-bound procedure calls. Some features do not work yet such 14632 as unlimited polymorphism (CLASS(*)). 14633 * Fortran 2008 support has been extended: 14634 + The OPEN statement now supports the NEWUNIT= option, which 14635 returns a unique file unit, thus preventing inadvertent use of 14636 the same unit in different parts of the program. 14637 + Support for unlimited format items has been added. 14638 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 14639 the intrinsic module ISO_FORTRAN_ENV are now supported. 14640 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 14641 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 14642 and ATANH have been added (for real and complex arguments) and 14643 ATAN(Y,X) is now an alias for ATAN2(Y,X). 14644 + The BLOCK construct has been implemented. 14645 14646 [24]New Targets and Target Specific Improvements 14647 14648 AIX 14649 14650 * Full cross-toolchain support now available with GNU Binutils 14651 14652 ARM 14653 14654 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 14655 * GCC now supports the ARM v7E-M architecture. 14656 * GCC now supports VFPv4-based FPUs and FPUs with 14657 single-precision-only VFP. 14658 * GCC has many improvements to optimization for other ARM processors, 14659 including scheduling support for the integer pipeline on Cortex-A9. 14660 * GCC now supports the IEEE 754-2008 half-precision floating-point 14661 type, and a variant ARM-specific half-precision type. This type is 14662 specified using __fp16, with the layout determined by 14663 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 14664 VFPv4 half-precision instructions will be used. 14665 * GCC now supports the variant of AAPCS that uses VFP registers for 14666 parameter passing and return values. 14667 14668 AVR 14669 14670 * The -mno-tablejump option has been removed because it has the same 14671 effect as the -fno-jump-tables option. 14672 * Added support for these new AVR devices: 14673 + ATmega8U2 14674 + ATmega16U2 14675 + ATmega32U2 14676 14677 [25]IA-32/x86-64 14678 14679 * GCC now will set the default for -march= based on the configure 14680 target. 14681 * GCC now supports handling floating-point excess precision arising 14682 from use of the x87 floating-point unit in a way that conforms to 14683 ISO C99. This is enabled with -fexcess-precision=standard and with 14684 standards conformance options such as -std=c99, and may be disabled 14685 using -fexcess-precision=fast. 14686 * Support for the Intel Atom processor is now available through the 14687 -march=atom and -mtune=atom options. 14688 * A new -mcrc32 option is now available to enable crc32 intrinsics. 14689 * A new -mmovbe option is now available to enable GCC to use the 14690 movbe instruction to implement __builtin_bswap32 and 14691 __builtin_bswap64. 14692 * SSE math now can be enabled by default at configure time with the 14693 new --with-fpmath=sse option. 14694 * There is a new intrinsic header file, <x86intrin.h>. It should be 14695 included before using any IA-32/x86-64 intrinsics. 14696 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 14697 Orochi processors are now available with the -mxop, -mfma4, and 14698 -mlwp options. 14699 * The -mabm option enables GCC to use the popcnt and lzcnt 14700 instructions on AMD processors. 14701 * The -mpopcnt option enables GCC to use the popcnt instructions on 14702 both AMD and Intel processors. 14703 14704 M68K/ColdFire 14705 14706 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 14707 and 5441x devices. 14708 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 14709 processors. 14710 14711 [26]MeP 14712 14713 Support has been added for the Toshiba Media embedded Processor (MeP, 14714 or mep-elf) embedded target. 14715 14716 MIPS 14717 14718 * GCC now supports MIPS 1004K processors. 14719 * GCC can now be configured with options --with-arch-32, 14720 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 14721 default optimization separately for 32-bit and 64-bit modes. 14722 * MIPS targets now support an alternative _mcount interface, in which 14723 register $12 points to the function's save slot for register $31. 14724 This interface is selected by the -mcount-ra-address option; see 14725 the documentation for more details. 14726 * GNU/Linux targets can now generate read-only .eh_frame sections. 14727 This optimization requires GNU binutils 2.20 or above, and is only 14728 available if GCC is configured with a suitable version of binutils. 14729 * GNU/Linux targets can now attach special relocations to indirect 14730 calls, so that the linker can turn them into direct jumps or 14731 branches. This optimization requires GNU binutils 2.20 or later, 14732 and is automatically selected if GCC is configured with an 14733 appropriate version of binutils. It can be explicitly enabled or 14734 disabled using the -mrelax-pic-calls command-line option. 14735 * GCC now generates more heavily-optimized atomic operations on 14736 Octeon processors. 14737 * MIPS targets now support the -fstack-protector option. 14738 * GCC now supports an -msynci option, which specifies that synci is 14739 enough to flush the instruction cache, without help from the 14740 operating system. GCC uses this information to optimize 14741 automatically-generated cache flush operations, such as those used 14742 for nested functions in C. There is also a --with-synci 14743 configure-time option, which makes -msynci the default. 14744 * GCC supports four new function attributes for interrupt handlers: 14745 interrupt, use_shadow_register_set, keep_interrupts_masked and 14746 use_debug_exception_return. See the documentation for more details 14747 about these attributes. 14748 14749 [27]RS/6000 (POWER/PowerPC) 14750 14751 * GCC now supports the Power ISA 2.06, which includes the VSX 14752 instructions that add vector 64-bit floating point support, new 14753 population count instructions, and conversions between floating 14754 point and unsigned types. 14755 * Support for the power7 processor is now available through the 14756 -mcpu=power7 and -mtune=power7. 14757 * GCC will now vectorize loops that contain simple math functions 14758 like copysign when generating code for altivec or VSX targets. 14759 * Support for the A2 processor is now available through the -mcpu=a2 14760 and -mtune=a2 options. 14761 * Support for the 476 processor is now available through the 14762 -mcpu={476,476fp} and -mtune={476,476fp} options. 14763 * Support for the e500mc64 processor is now available through the 14764 -mcpu=e500mc64 and -mtune=e500mc64 options. 14765 * GCC can now be configured with options --with-cpu-32, 14766 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 14767 default optimization separately for 32-bit and 64-bit modes. 14768 * Starting with GCC 4.5.4, vectors of type vector long long or vector 14769 long are passed and returned in the same method as other vectors 14770 with the VSX instruction set. Previously the GCC compiler did not 14771 adhere to the ABI for 128-bit vectors with 64-bit integer base 14772 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 14773 14774 [28]RX 14775 14776 Support has been added for the Renesas RX Processor (rx-elf) target. 14777 14778 [29]Operating Systems 14779 14780 [30]Windows (Cygwin and MinGW) 14781 14782 * GCC now installs all the major language runtime libraries as DLLs 14783 when configured with the --enable-shared option. 14784 * GCC now makes use of the new support for aligned common variables 14785 in versions of binutils >= 2.20 to fix bugs in the support for SSE 14786 data types. 14787 * Improvements to the libffi support library increase the reliability 14788 of code generated by GCJ on all Windows platforms. Libgcj is 14789 enabled by default for the first time. 14790 * Libtool improvements simplify installation by placing the generated 14791 DLLs in the correct binaries directory. 14792 * Numerous other minor bugfixes and improvements, and substantial 14793 enhancements to the Fortran language support library. 14794 14795 > 14796 14797 Other significant improvements 14798 14799 Plugins 14800 14801 * It is now possible to extend the compiler without having to modify 14802 its source code. A new option -fplugin=file.so tells GCC to load 14803 the shared object file.so and execute it as part of the compiler. 14804 The internal documentation describes the details on how plugins can 14805 interact with the compiler. 14806 14807 Installation changes 14808 14809 * The move to newer autotools changed default installation 14810 directories and switches to control them: The --with-datarootdir, 14811 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 14812 used any more. Instead, you can now use --datarootdir, --docdir, 14813 --htmldir, and --pdfdir. The default installation directories have 14814 changed as follows according to the GNU Coding Standards: 14815 14816 datarootdir read-only architecture-independent data root [PREFIX/share] 14817 localedir locale-specific message catalogs [DATAROOTDIR/locale] 14818 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 14819 htmldir html documentation [DOCDIR] 14820 dvidir dvi documentation [DOCDIR] 14821 pdfdir pdf documentation [DOCDIR] 14822 psdir ps documentation [DOCDIR] 14823 The following variables have new default values: 14824 14825 datadir read-only architecture-independent data [DATAROOTDIR] 14826 infodir info documentation [DATAROOTDIR/info] 14827 mandir man documentation [DATAROOTDIR/man] 14828 14829 [31]GCC 4.5.1 14830 14831 This is the [32]list of problem reports (PRs) from GCC's bug tracking 14832 system that are known to be fixed in the 4.5.1 release. This list might 14833 not be complete (that is, it is possible that some PRs that have been 14834 fixed are not listed here). 14835 14836 All languages 14837 14838 * GCC's new link-time optimizer ([33]-flto) now also works on a few 14839 non-ELF targets: 14840 + Cygwin (*-cygwin*) 14841 + MinGW (*-mingw*) 14842 + Darwin on x86-64 (x86_64-apple-darwin*) 14843 LTO is not enabled by default for these targets. To enable LTO, you 14844 should configure with the --enable-lto option. 14845 14846 [34]GCC 4.5.2 14847 14848 This is the [35]list of problem reports (PRs) from GCC's bug tracking 14849 system that are known to be fixed in the 4.5.2 release. This list might 14850 not be complete (that is, it is possible that some PRs that have been 14851 fixed are not listed here). 14852 14853 [36]GCC 4.5.3 14854 14855 This is the [37]list of problem reports (PRs) from GCC's bug tracking 14856 system that are known to be fixed in the 4.5.3 release. This list might 14857 not be complete (that is, it is possible that some PRs that have been 14858 fixed are not listed here). 14859 14860 On the PowerPC compiler, the Altivec builtin functions vec_ld and 14861 vec_st have been modified to generate the Altivec memory instructions 14862 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 14863 release, these builtin functions were changed to generate VSX memory 14864 reference instructions instead of Altivec memory instructions, but 14865 there are differences between the two instructions. If the VSX 14866 instruction set is available, you can now use the new builtin functions 14867 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 14868 instructions. 14869 14870 [38]GCC 4.5.4 14871 14872 This is the [39]list of problem reports (PRs) from GCC's bug tracking 14873 system that are known to be fixed in the 4.5.4 release. This list might 14874 not be complete (that is, it is possible that some PRs that have been 14875 fixed are not listed here). 14876 14877 14878 For questions related to the use of GCC, please consult these web 14879 pages and the [40]GCC manuals. If that fails, the 14880 [41]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14881 web pages and the development of GCC are welcome on our developer 14882 list at [42]gcc (a] gcc.gnu.org. All of [43]our lists have public 14883 archives. 14884 14885 Copyright (C) [44]Free Software Foundation, Inc. Verbatim copying and 14886 distribution of this entire article is permitted in any medium, 14887 provided this notice is preserved. 14888 14889 These pages are [45]maintained by the GCC team. Last modified 14890 2025-01-31. 14891 14892 References 14893 14894 1. https://www.multiprecision.org/ 14895 2. https://gcc.gnu.org/install/prerequisites.html 14896 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 14897 4. https://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 14898 5. https://gcc.gnu.org/gcc-4.5/changes.html#x86 14899 6. https://gcc.gnu.org/PR30789 14900 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 14901 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 14902 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 14903 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 14904 11. https://gcc.gnu.org/gcc-4.5/changes.html#cplusplus 14905 12. https://gcc.gnu.org/gcc-4.5/cxx0x_status.html 14906 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 14907 14. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 14908 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 14909 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 14910 17. https://sourceware.org/gdb/wiki/STLSupport 14911 18. https://gcc.gnu.org/gcc-4.5/changes.html#Fortran 14912 19. https://gcc.gnu.org/gcc-4.4/changes.html 14913 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 14914 21. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 14915 22. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 14916 23. https://gcc.gnu.org/wiki/OOP 14917 24. https://gcc.gnu.org/gcc-4.5/changes.html#targets 14918 25. https://gcc.gnu.org/gcc-4.5/changes.html#x86 14919 26. https://gcc.gnu.org/gcc-4.5/changes.html#mep 14920 27. https://gcc.gnu.org/gcc-4.5/changes.html#rs6000 14921 28. https://gcc.gnu.org/gcc-4.5/changes.html#rx 14922 29. https://gcc.gnu.org/gcc-4.5/changes.html#os 14923 30. https://gcc.gnu.org/gcc-4.5/changes.html#windows 14924 31. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.1 14925 32. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 14926 33. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 14927 34. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.2 14928 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 14929 36. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.3 14930 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 14931 38. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.4 14932 39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 14933 40. https://gcc.gnu.org/onlinedocs/ 14934 41. mailto:gcc-help (a] gcc.gnu.org 14935 42. mailto:gcc (a] gcc.gnu.org 14936 43. https://gcc.gnu.org/lists.html 14937 44. https://www.fsf.org/ 14938 45. https://gcc.gnu.org/about.html 14939 ====================================================================== 14940 http://gcc.gnu.org/gcc-4.4/index.html 14941 14942 GCC 4.4 Release Series 14943 14944 This release series is no longer maintained. 14945 14946 March 13, 2012 14947 14948 The [1]GNU project and the GCC developers are pleased to announce the 14949 release of GCC 4.4.7. 14950 14951 This release is a bug-fix release, containing fixes for regressions in 14952 GCC 4.4.6 relative to previous releases of GCC. 14953 14954 Release History 14955 14956 GCC 4.4.7 14957 March 13, 2012 ([2]changes) 14958 14959 GCC 4.4.6 14960 April 16, 2011 ([3]changes) 14961 14962 GCC 4.4.5 14963 October 1, 2010 ([4]changes) 14964 14965 GCC 4.4.4 14966 April 29, 2010 ([5]changes) 14967 14968 GCC 4.4.3 14969 January 21, 2010 ([6]changes) 14970 14971 GCC 4.4.2 14972 October 15, 2009 ([7]changes) 14973 14974 GCC 4.4.1 14975 July 22, 2009 ([8]changes) 14976 14977 GCC 4.4.0 14978 April 21, 2009 ([9]changes) 14979 14980 References and Acknowledgements 14981 14982 GCC used to stand for the GNU C Compiler, but since the compiler 14983 supports several other languages aside from C, it now stands for the 14984 GNU Compiler Collection. 14985 14986 A list of [10]successful builds is updated as new information becomes 14987 available. 14988 14989 The GCC developers would like to thank the numerous people that have 14990 contributed new features, improvements, bug fixes, and other changes as 14991 well as test results to GCC. This [11]amazing group of volunteers is 14992 what makes GCC successful. 14993 14994 For additional information about GCC please refer to the [12]GCC 14995 project web site or contact the [13]GCC development mailing list. 14996 14997 To obtain GCC please use [14]our mirror sites or [15]our version 14998 control system. 14999 15000 15001 For questions related to the use of GCC, please consult these web 15002 pages and the [16]GCC manuals. If that fails, the 15003 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15004 web pages and the development of GCC are welcome on our developer 15005 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 15006 archives. 15007 15008 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 15009 distribution of this entire article is permitted in any medium, 15010 provided this notice is preserved. 15011 15012 These pages are [21]maintained by the GCC team. Last modified 15013 2022-10-26. 15014 15015 References 15016 15017 1. http://www.gnu.org/ 15018 2. https://gcc.gnu.org/gcc-4.4/changes.html 15019 3. https://gcc.gnu.org/gcc-4.4/changes.html 15020 4. https://gcc.gnu.org/gcc-4.4/changes.html 15021 5. https://gcc.gnu.org/gcc-4.4/changes.html 15022 6. https://gcc.gnu.org/gcc-4.4/changes.html 15023 7. https://gcc.gnu.org/gcc-4.4/changes.html 15024 8. https://gcc.gnu.org/gcc-4.4/changes.html 15025 9. https://gcc.gnu.org/gcc-4.4/changes.html 15026 10. https://gcc.gnu.org/gcc-4.4/buildstat.html 15027 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15028 12. https://gcc.gnu.org/index.html 15029 13. mailto:gcc (a] gcc.gnu.org 15030 14. https://gcc.gnu.org/mirrors.html 15031 15. https://gcc.gnu.org/git.html 15032 16. https://gcc.gnu.org/onlinedocs/ 15033 17. mailto:gcc-help (a] gcc.gnu.org 15034 18. mailto:gcc (a] gcc.gnu.org 15035 19. https://gcc.gnu.org/lists.html 15036 20. https://www.fsf.org/ 15037 21. https://gcc.gnu.org/about.html 15038 ====================================================================== 15039 http://gcc.gnu.org/gcc-4.4/changes.html 15040 15041 GCC 4.4 Release Series 15042 Changes, New Features, and Fixes 15043 15044 The latest release in the 4.4 release series is [1]GCC 4.4.7. 15045 15046 Caveats 15047 15048 * __builtin_stdarg_start has been completely removed from GCC. 15049 Support for <varargs.h> had been deprecated since GCC 4.0. Use 15050 __builtin_va_start as a replacement. 15051 * Some of the errors issued by the C++ front end that could be 15052 downgraded to warnings in previous releases by using -fpermissive 15053 are now warnings by default. They can be converted into errors by 15054 using -pedantic-errors. 15055 * Use of the cpp assertion extension will now emit a warning when 15056 -Wdeprecated or -pedantic is used. This extension has been 15057 deprecated for many years, but never warned about. 15058 * Packed bit-fields of type char were not properly bit-packed on many 15059 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 15060 causes an ABI change. For example there is no longer a 4-bit 15061 padding between field a and b in this structure: 15062 struct foo 15063 { 15064 char a:4; 15065 char b:8; 15066 } __attribute__ ((packed)); 15067 There is a new warning to help identify fields that are affected: 15068 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 15069 The warning can be disabled with -Wno-packed-bitfield-compat. 15070 * On ARM EABI targets, the C++ mangling of the va_list type has been 15071 changed to conform to the current revision of the EABI. This does 15072 not affect the libstdc++ library included with GCC. 15073 * The SCOUNT and POS bits of the MIPS DSP control register are now 15074 treated as global. Previous versions of GCC treated these fields as 15075 call-clobbered instead. 15076 * The MIPS port no longer recognizes the h asm constraint. It was 15077 necessary to remove this constraint in order to avoid generating 15078 unpredictable code sequences. 15079 One of the main uses of the h constraint was to extract the high 15080 part of a multiplication on 64-bit targets. For example: 15081 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 15082 You can now achieve the same effect using 128-bit types: 15083 typedef unsigned int uint128_t __attribute__((mode(TI))); 15084 result = ((uint128_t) x * y) >> 64; 15085 The second sequence is better in many ways. For example, if x and y 15086 are constants, the compiler can perform the multiplication at 15087 compile time. If x and y are not constants, the compiler can 15088 schedule the runtime multiplication better than it can schedule an 15089 asm statement. 15090 * Support for a number of older systems and recently unmaintained or 15091 untested target ports of GCC has been declared obsolete in GCC 4.4. 15092 Unless there is activity to revive them, the next release of GCC 15093 will have their sources permanently removed. 15094 The following ports for individual systems on particular 15095 architectures have been obsoleted: 15096 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 15097 m68k-*-aout*) 15098 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 15099 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 15100 sh-*-*). This does not affect other more specific targets 15101 using the COFF object format on those architectures, or the 15102 more specific H8300 and SH targets (h8300-*-rtems*, 15103 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 15104 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 15105 + 2BSD on PDP-11 (pdp11-*-bsd) 15106 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 15107 powerpc-ibm-aix4.[12]*) 15108 + Tuning support for Itanium1 (Merced) variants. Note that code 15109 tuned for Itanium2 should also run correctly on Itanium1. 15110 * The protoize and unprotoize utilities have been obsoleted and will 15111 be removed in GCC 4.5. These utilities have not been installed by 15112 default since GCC 3.0. 15113 * Support has been removed for all the [2]configurations obsoleted in 15114 GCC 4.3. 15115 * Unknown -Wno-* options are now silently ignored by GCC if no other 15116 diagnostics are issued. If other diagnostics are issued, then GCC 15117 warns about the unknown options. 15118 * More information on porting to GCC 4.4 from previous versions of 15119 GCC can be found in the [3]porting guide for this release. 15120 15121 General Optimizer Improvements 15122 15123 * A new command-line switch -findirect-inlining has been added. When 15124 turned on it allows the inliner to also inline indirect calls that 15125 are discovered to have known targets at compile time thanks to 15126 previous inlining. 15127 * A new command-line switch -ftree-switch-conversion has been added. 15128 This new pass turns simple initializations of scalar variables in 15129 switch statements into initializations from a static array, given 15130 that all the values are known at compile time and the ratio between 15131 the new array size and the original switch branches does not exceed 15132 the parameter --param switch-conversion-max-branch-ratio (default 15133 is eight). 15134 * A new command-line switch -ftree-builtin-call-dce has been added. 15135 This optimization eliminates unnecessary calls to certain builtin 15136 functions when the return value is not used, in cases where the 15137 calls can not be eliminated entirely because the function may set 15138 errno. This optimization is on by default at -O2 and above. 15139 * A new command-line switch -fconserve-stack directs the compiler to 15140 minimize stack usage even if it makes the generated code slower. 15141 This affects inlining decisions. 15142 * When the assembler supports it, the compiler will now emit unwind 15143 information using assembler .cfi directives. This makes it possible 15144 to use such directives in inline assembler code. The new option 15145 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 15146 directives. 15147 * The [4]Graphite branch has been merged. This merge has brought in a 15148 new framework for loop optimizations based on a polyhedral 15149 intermediate representation. These optimizations apply to all the 15150 languages supported by GCC. The following new code transformations 15151 are available in GCC 4.4: 15152 + -floop-interchange performs loop interchange transformations 15153 on loops. Interchanging two nested loops switches the inner 15154 and outer loops. For example, given a loop like: 15155 DO J = 1, M 15156 DO I = 1, N 15157 A(J, I) = A(J, I) * C 15158 ENDDO 15159 ENDDO 15160 15161 loop interchange will transform the loop as if the user had 15162 written: 15163 DO I = 1, N 15164 DO J = 1, M 15165 A(J, I) = A(J, I) * C 15166 ENDDO 15167 ENDDO 15168 15169 which can be beneficial when N is larger than the caches, 15170 because in Fortran, the elements of an array are stored in 15171 memory contiguously by column, and the original loop iterates 15172 over rows, potentially creating at each access a cache miss. 15173 + -floop-strip-mine performs loop strip mining transformations 15174 on loops. Strip mining splits a loop into two nested loops. 15175 The outer loop has strides equal to the strip size and the 15176 inner loop has strides of the original loop within a strip. 15177 For example, given a loop like: 15178 DO I = 1, N 15179 A(I) = A(I) + C 15180 ENDDO 15181 15182 loop strip mining will transform the loop as if the user had 15183 written: 15184 DO II = 1, N, 4 15185 DO I = II, min (II + 3, N) 15186 A(I) = A(I) + C 15187 ENDDO 15188 ENDDO 15189 15190 + -floop-block performs loop blocking transformations on loops. 15191 Blocking strip mines each loop in the loop nest such that the 15192 memory accesses of the element loops fit inside caches. For 15193 example, given a loop like: 15194 DO I = 1, N 15195 DO J = 1, M 15196 A(J, I) = B(I) + C(J) 15197 ENDDO 15198 ENDDO 15199 15200 loop blocking will transform the loop as if the user had 15201 written: 15202 DO II = 1, N, 64 15203 DO JJ = 1, M, 64 15204 DO I = II, min (II + 63, N) 15205 DO J = JJ, min (JJ + 63, M) 15206 A(J, I) = B(I) + C(J) 15207 ENDDO 15208 ENDDO 15209 ENDDO 15210 ENDDO 15211 15212 which can be beneficial when M is larger than the caches, 15213 because the innermost loop will iterate over a smaller amount 15214 of data that can be kept in the caches. 15215 * A new register allocator has replaced the old one. It is called 15216 integrated register allocator (IRA) because coalescing, register 15217 live range splitting, and hard register preferencing are done 15218 on-the-fly during coloring. It also has better integration with the 15219 reload pass. IRA is a regional register allocator which uses modern 15220 Chaitin-Briggs coloring instead of Chow's priority coloring used in 15221 the old register allocator. More info about IRA internals and 15222 options can be found in the GCC manuals. 15223 * A new instruction scheduler and software pipeliner, based on the 15224 selective scheduling approach, has been added. The new pass 15225 performs instruction unification, register renaming, substitution 15226 through register copies, and speculation during scheduling. The 15227 software pipeliner is able to pipeline non-countable loops. The new 15228 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 15229 it is available for the Intel Itanium platform working by default 15230 as the second scheduling pass (after register allocation) at the 15231 -O3 optimization level. 15232 * When using -fprofile-generate with a multi-threaded program, the 15233 profile counts may be slightly wrong due to race conditions. The 15234 new -fprofile-correction option directs the compiler to apply 15235 heuristics to smooth out the inconsistencies. By default the 15236 compiler will give an error message when it finds an inconsistent 15237 profile. 15238 * The new -fprofile-dir=PATH option permits setting the directory 15239 where profile data files are stored when using -fprofile-generate 15240 and friends, and the directory used when reading profile data files 15241 using -fprofile-use and friends. 15242 15243 New warning options 15244 15245 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 15246 warning if any stack frame is larger than NUMBER bytes. This may be 15247 used to help ensure that code fits within a limited amount of stack 15248 space. 15249 * The command-line option -Wlarger-than-N is now written as 15250 -Wlarger-than=N and the old form is deprecated. 15251 * The new -Wno-mudflap option disables warnings about constructs 15252 which can not be instrumented when using -fmudflap. 15253 15254 New Languages and Language specific improvements 15255 15256 * Version 3.0 of the OpenMP specification is now supported for the C, 15257 C++, and Fortran compilers. 15258 * New character data types, per [5]TR 19769: New character types in 15259 C, are now supported for the C compiler in -std=gnu99 mode, as 15260 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 15261 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 15262 15263 C family 15264 15265 * A new optimize attribute was added to allow programmers to change 15266 the optimization level and particular optimization options for an 15267 individual function. You can also change the optimization options 15268 via the GCC optimize pragma for functions defined after the pragma. 15269 The GCC push_options pragma and the GCC pop_options pragma allow 15270 you temporarily save and restore the options used. The GCC 15271 reset_options pragma restores the options to what was specified on 15272 the command line. 15273 * Uninitialized warnings do not require enabling optimization 15274 anymore, that is, -Wuninitialized can be used together with -O0. 15275 Nonetheless, the warnings given by -Wuninitialized will probably be 15276 more accurate if optimization is enabled. 15277 * -Wparentheses now warns about expressions such as (!x | y) and (!x 15278 & y). Using explicit parentheses, such as in ((!x) | y), silences 15279 this warning. 15280 * -Wsequence-point now warns within if, while,do while and for 15281 conditions, and within for begin/end expressions. 15282 * A new option -dU is available to dump definitions of preprocessor 15283 macros that are tested or expanded. 15284 15285 C++ 15286 15287 * [6]Improved experimental support for the upcoming ISO C++ standard, 15288 C++0x. Including support for auto, inline namespaces, generalized 15289 initializer lists, defaulted and deleted functions, new character 15290 types, and scoped enums. 15291 * Those errors that may be downgraded to warnings to build legacy 15292 code now mention -fpermissive when -fdiagnostics-show-option is 15293 enabled. 15294 * -Wconversion now warns if the result of a static_cast to enumeral 15295 type is unspecified because the value is outside the range of the 15296 enumeral type. 15297 * -Wuninitialized now warns if a non-static reference or non-static 15298 const member appears in a class without constructors. 15299 * G++ now properly implements value-initialization, so objects with 15300 an initializer of () and an implicitly defined default constructor 15301 will be zero-initialized before the default constructor is called. 15302 15303 Runtime Library (libstdc++) 15304 15305 * Improved experimental support for the upcoming ISO C++ standard, 15306 C++0x, including: 15307 + Support for <chrono>, <condition_variable>, <cstdatomic>, 15308 <forward_list>, <initializer_list>, <mutex>, <ratio>, 15309 <system_error>, and <thread>. 15310 + unique_ptr, <algorithm> additions, exception propagation, and 15311 support for the new character types in <string> and <limits>. 15312 + Existing facilities now exploit initializer lists, defaulted 15313 and deleted functions, and the newly implemented core C++0x 15314 features. 15315 + Some standard containers are more efficient together with 15316 stateful allocators, i.e., no allocator is constructed on the 15317 fly at element construction time. 15318 * Experimental support for non-standard pointer types in containers. 15319 * The long standing libstdc++/30928 has been fixed for targets 15320 running glibc 2.10 or later. 15321 * As usual, many small and larger bug fixes, in particular quite a 15322 few corner cases in <locale>. 15323 15324 Fortran 15325 15326 * GNU Fortran now employs libcpp directly instead of using cc1 as an 15327 external preprocessor. The [7]-cpp option was added to allow manual 15328 invocation of the preprocessor without relying on filename 15329 extensions. 15330 * The [8]-Warray-temporaries option warns about array temporaries 15331 generated by the compiler, as an aid to optimization. 15332 * The [9]-fcheck-array-temporaries option has been added, printing a 15333 notification at run time, when an array temporary had to be created 15334 for an function argument. Contrary to -Warray-temporaries the 15335 warning is only printed if the array is noncontiguous. 15336 * Improved generation of DWARF debugging symbols 15337 * If using an intrinsic not part of the selected standard (via -std= 15338 and -fall-intrinsics) gfortran will now treat it as if this 15339 procedure were declared EXTERNAL and try to link to a user-supplied 15340 procedure. -Wintrinsics-std will warn whenever this happens. The 15341 now-useless option -Wnonstd-intrinsic was removed. 15342 * The flag -falign-commons has been added to control the alignment of 15343 variables in COMMON blocks, which is enabled by default in line 15344 with previous GCC version. Using -fno-align-commons one can force 15345 commons to be contiguous in memory as required by the Fortran 15346 standard, however, this slows down the memory access. The option 15347 -Walign-commons, which is enabled by default, warns when padding 15348 bytes were added for alignment. The proper solution is to sort the 15349 common objects by decreasing storage size, which avoids the 15350 alignment problems. 15351 * Fortran 2003 support has been extended: 15352 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 15353 now supported (except internal reads from/writes to wide 15354 strings). [10]-fbackslash now supports also \unnnn and 15355 \Unnnnnnnn to enter Unicode characters. 15356 + Asynchronous I/O (implemented as synchronous I/O) and the 15357 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 15358 are now supported in I/O statements. 15359 + Support for Fortran 2003 structure constructors and for array 15360 constructor with typespec has been added. 15361 + Procedure Pointers (but not yet as component in derived types 15362 and as function results) are now supported. 15363 + Abstract types, type extension, and type-bound procedures 15364 (both PROCEDURE and GENERIC but not as operators). Note: As 15365 CLASS/polymorphyic types are not implemented, type-bound 15366 procedures with PASS accept as non-standard extension TYPE 15367 arguments. 15368 * Fortran 2008 support has been added: 15369 + The -std=f2008 option and support for the file extensions 15370 .f2008 and .F2008 has been added. 15371 + The g0 format descriptor is now supported. 15372 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 15373 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 15374 are now available (some of them existed as GNU extension 15375 before). Note: The hyperbolic functions are not yet supporting 15376 complex arguments and the three- argument version of BESSEL_*N 15377 is not available. 15378 + The bit intrinsics LEADZ and TRAILZ have been added. 15379 15380 Java (GCJ) 15381 15382 Ada 15383 15384 * The Ada runtime now supports multilibs on many platforms including 15385 x86_64, SPARC and PowerPC. Their build is enabled by default. 15386 15387 [11]New Targets and Target Specific Improvements 15388 15389 ARM 15390 15391 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 15392 Cortex-R4F processors and has many other improvements to 15393 optimization for ARM processors. 15394 * GCC now supports the VFPv3 variant with 16 double-precision 15395 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 15396 renamed to -mfpu=vfpv3. 15397 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 15398 erratum on Cortex-M3 processors. 15399 * GCC now supports the __sync_* atomic operations for ARM EABI 15400 GNU/Linux. 15401 * The section anchors optimization is now enabled by default when 15402 optimizing for ARM. 15403 * GCC now uses a new EABI-compatible profiling interface for EABI 15404 targets. This requires a function __gnu_mcount_nc, which is 15405 provided by GNU libc versions 2.8 and later. 15406 15407 AVR 15408 15409 * The -mno-tablejump option has been deprecated because it has the 15410 same effect as the -fno-jump-tables option. 15411 * Added support for these new AVR devices: 15412 + ATA6289 15413 + ATtiny13A 15414 + ATtiny87 15415 + ATtiny167 15416 + ATtiny327 15417 + ATmega8C1 15418 + ATmega16C1 15419 + ATmega32C1 15420 + ATmega8M1 15421 + ATmega16M1 15422 + ATmega32M1 15423 + ATmega32U4 15424 + ATmega16HVB 15425 + ATmega4HVD 15426 + ATmega8HVD 15427 + ATmega64C1 15428 + ATmega64M1 15429 + ATmega16U4 15430 + ATmega32U6 15431 + ATmega128RFA1 15432 + AT90PWM81 15433 + AT90SCR100 15434 + M3000F 15435 + M3000S 15436 + M3001B 15437 15438 IA-32/x86-64 15439 15440 * Support for Intel AES built-in functions and code generation is 15441 available via -maes. 15442 * Support for Intel PCLMUL built-in function and code generation is 15443 available via -mpclmul. 15444 * Support for Intel AVX built-in functions and code generation is 15445 available via -mavx. 15446 * Automatically align the stack for local variables with alignment 15447 requirement. 15448 * GCC can now utilize the SVML library for vectorizing calls to a set 15449 of C99 functions if -mveclibabi=svml is specified and you link to 15450 an SVML ABI compatible library. 15451 * On x86-64, the ABI has been changed in the following cases to 15452 conform to the x86-64 ABI: 15453 + Passing/returning structures with flexible array member: 15454 struct foo 15455 { 15456 int i; 15457 int flex[]; 15458 }; 15459 + Passing/returning structures with complex float member: 15460 struct foo 15461 { 15462 int i; 15463 __complex__ float f; 15464 }; 15465 + Passing/returning unions with long double member: 15466 union foo 15467 { 15468 int x; 15469 long double ld; 15470 }; 15471 Code built with previous versions of GCC that uses any of these is 15472 not compatible with code built with GCC 4.4.0 or later. 15473 * A new target attribute was added to allow programmers to change the 15474 target options like -msse2 or -march=k8 for an individual function. 15475 You can also change the target options via the GCC target pragma 15476 for functions defined after the pragma. 15477 * GCC can now be configured with options --with-arch-32, 15478 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 15479 --with-tune-64 to control the default optimization separately for 15480 32-bit and 64-bit modes. 15481 15482 IA-32/IA64 15483 15484 * Support for __float128 (TFmode) IEEE quad type and corresponding 15485 TCmode IEEE complex quad type is available via the soft-fp library 15486 on IA-32/IA64 targets. This includes basic arithmetic operations 15487 (addition, subtraction, negation, multiplication and division) on 15488 __float128 real and TCmode complex values, the full set of IEEE 15489 comparisons between __float128 values, conversions to and from 15490 float, double and long double floating point types, as well as 15491 conversions to and from signed or unsigned integer, signed or 15492 unsigned long integer and signed or unsigned quad (TImode, IA64 15493 only) integer types. Additionally, all operations generate the full 15494 set of IEEE exceptions and support the full set of IEEE rounding 15495 modes. 15496 15497 M68K/ColdFire 15498 15499 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 15500 processors. (Scheduling support for ColdFire V2 processors was 15501 added in GCC 4.3.) 15502 * GCC now supports the -mxgot option to support programs requiring 15503 many GOT entries on ColdFire. 15504 * The m68k-*-linux-gnu target now builds multilibs by default. 15505 15506 MIPS 15507 15508 * MIPS Technologies have extended the original MIPS SVR4 ABI to 15509 include support for procedure linkage tables (PLTs) and copy 15510 relocations. These extensions allow GNU/Linux executables to use a 15511 significantly more efficient code model than the one defined by the 15512 original ABI. 15513 GCC support for this code model is available via a new command-line 15514 option, -mplt. There is also a new configure-time option, 15515 --with-mips-plt, to make -mplt the default. 15516 The new code model requires support from the assembler, the linker, 15517 and the runtime C library. This support is available in binutils 15518 2.19 and GLIBC 2.9. 15519 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 15520 and 32-bit GNU/Linux shared libraries. This feature requires GNU 15521 binutils 2.19 or above. 15522 * Support for RMI's XLR processor is now available through the 15523 -march=xlr and -mtune=xlr options. 15524 * 64-bit targets can now perform 128-bit multiplications inline, 15525 instead of relying on a libgcc function. 15526 * Native GNU/Linux toolchains now support -march=native and 15527 -mtune=native, which select the host processor. 15528 * GCC now supports the R10K, R12K, R14K and R16K processors. The 15529 canonical -march= and -mtune= names for these processors are 15530 r10000, r12000, r14000 and r16000 respectively. 15531 * GCC can now work around the side effects of speculative execution 15532 on R10K processors. Please see the documentation of the 15533 -mr10k-cache-barrier option for details. 15534 * Support for the MIPS64 Release 2 instruction set has been added. 15535 The option -march=mips64r2 enables generation of these 15536 instructions. 15537 * GCC now supports Cavium Networks' Octeon processor. This support is 15538 available through the -march=octeon and -mtune=octeon options. 15539 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 15540 canonical -march= and -mtune= names for these processors are 15541 loongson2e and loongson2f. 15542 15543 [12]picochip 15544 15545 Picochip is a 16-bit processor. A typical picoChip contains over 250 15546 small cores, each with small amounts of memory. There are three 15547 processor variants (STAN, MEM and CTRL) with different instruction sets 15548 and memory configurations and they can be chosen using the -mae option. 15549 15550 This port is intended to be a "C" only port. 15551 15552 Power Architecture and PowerPC 15553 15554 * GCC now supports the e300c2, e300c3 and e500mc processors. 15555 * GCC now supports Xilinx processors with a single-precision FPU. 15556 * Decimal floating point is now supported for e500 processors. 15557 15558 S/390, zSeries and System z9/z10 15559 15560 * Support for the IBM System z10 EC/BC processor has been added. When 15561 using the -march=z10 option, the compiler will generate code making 15562 use of instructions provided by the General-Instruction-Extension 15563 Facility and the Execute-Extension Facility. 15564 15565 VxWorks 15566 15567 * GCC now supports the thread-local storage mechanism used on 15568 VxWorks. 15569 15570 Xtensa 15571 15572 * GCC now supports thread-local storage (TLS) for Xtensa processor 15573 configurations that include the Thread Pointer option. TLS also 15574 requires support from the assembler and linker; this support is 15575 provided in the GNU binutils beginning with version 2.19. 15576 15577 Documentation improvements 15578 15579 Other significant improvements 15580 15581 [13]GCC 4.4.1 15582 15583 This is the [14]list of problem reports (PRs) from GCC's bug tracking 15584 system that are known to be fixed in the 4.4.1 release. This list might 15585 not be complete (that is, it is possible that some PRs that have been 15586 fixed are not listed here). 15587 15588 [15]GCC 4.4.2 15589 15590 This is the [16]list of problem reports (PRs) from GCC's bug tracking 15591 system that are known to be fixed in the 4.4.2 release. This list might 15592 not be complete (that is, it is possible that some PRs that have been 15593 fixed are not listed here). 15594 15595 [17]GCC 4.4.3 15596 15597 This is the [18]list of problem reports (PRs) from GCC's bug tracking 15598 system that are known to be fixed in the 4.4.3 release. This list might 15599 not be complete (that is, it is possible that some PRs that have been 15600 fixed are not listed here). 15601 15602 [19]GCC 4.4.4 15603 15604 This is the [20]list of problem reports (PRs) from GCC's bug tracking 15605 system that are known to be fixed in the 4.4.4 release. This list might 15606 not be complete (that is, it is possible that some PRs that have been 15607 fixed are not listed here). 15608 15609 [21]GCC 4.4.5 15610 15611 This is the [22]list of problem reports (PRs) from GCC's bug tracking 15612 system that are known to be fixed in the 4.4.5 release. This list might 15613 not be complete (that is, it is possible that some PRs that have been 15614 fixed are not listed here). 15615 15616 [23]GCC 4.4.6 15617 15618 This is the [24]list of problem reports (PRs) from GCC's bug tracking 15619 system that are known to be fixed in the 4.4.6 release. This list might 15620 not be complete (that is, it is possible that some PRs that have been 15621 fixed are not listed here). 15622 15623 [25]GCC 4.4.7 15624 15625 This is the [26]list of problem reports (PRs) from GCC's bug tracking 15626 system that are known to be fixed in the 4.4.7 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 15631 For questions related to the use of GCC, please consult these web 15632 pages and the [27]GCC manuals. If that fails, the 15633 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15634 web pages and the development of GCC are welcome on our developer 15635 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 15636 archives. 15637 15638 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 15639 distribution of this entire article is permitted in any medium, 15640 provided this notice is preserved. 15641 15642 These pages are [32]maintained by the GCC team. Last modified 15643 2025-01-31. 15644 15645 References 15646 15647 1. https://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 15648 2. https://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 15649 3. https://gcc.gnu.org/gcc-4.4/porting_to.html 15650 4. https://gcc.gnu.org/wiki/Graphite 15651 5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 15652 6. https://gcc.gnu.org/gcc-4.4/cxx0x_status.html 15653 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 15654 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 15655 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 15656 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 15657 11. https://gcc.gnu.org/gcc-4.4/changes.html#targets 15658 12. https://gcc.gnu.org/gcc-4.4/changes.html#picochip 15659 13. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.1 15660 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 15661 15. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.2 15662 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 15663 17. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.3 15664 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 15665 19. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.4 15666 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 15667 21. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.5 15668 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 15669 23. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.6 15670 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 15671 25. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.7 15672 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 15673 27. https://gcc.gnu.org/onlinedocs/ 15674 28. mailto:gcc-help (a] gcc.gnu.org 15675 29. mailto:gcc (a] gcc.gnu.org 15676 30. https://gcc.gnu.org/lists.html 15677 31. https://www.fsf.org/ 15678 32. https://gcc.gnu.org/about.html 15679 ====================================================================== 15680 http://gcc.gnu.org/gcc-4.3/index.html 15681 15682 GCC 4.3 Release Series 15683 15684 (This release series is no longer supported.) 15685 15686 Jun 27, 2011 15687 15688 The [1]GNU project and the GCC developers are pleased to announce the 15689 release of GCC 4.3.6. 15690 15691 This release is a bug-fix release, containing fixes for regressions in 15692 GCC 4.3.5 relative to previous releases of GCC. 15693 15694 Release History 15695 15696 GCC 4.3.6 15697 Jun 27, 2011 ([2]changes) 15698 15699 GCC 4.3.5 15700 May 22, 2010 ([3]changes) 15701 15702 GCC 4.3.4 15703 August 4, 2009 ([4]changes) 15704 15705 GCC 4.3.3 15706 January 24, 2009 ([5]changes) 15707 15708 GCC 4.3.2 15709 August 27, 2008 ([6]changes) 15710 15711 GCC 4.3.1 15712 June 6, 2008 ([7]changes) 15713 15714 GCC 4.3.0 15715 March 5, 2008 ([8]changes) 15716 15717 References and Acknowledgements 15718 15719 GCC used to stand for the GNU C Compiler, but since the compiler 15720 supports several other languages aside from C, it now stands for the 15721 GNU Compiler Collection. 15722 15723 A list of [9]successful builds is updated as new information becomes 15724 available. 15725 15726 The GCC developers would like to thank the numerous people that have 15727 contributed new features, improvements, bug fixes, and other changes as 15728 well as test results to GCC. This [10]amazing group of volunteers is 15729 what makes GCC successful. 15730 15731 For additional information about GCC please refer to the [11]GCC 15732 project web site or contact the [12]GCC development mailing list. 15733 15734 To obtain GCC please use [13]our mirror sites or [14]our version 15735 control system. 15736 15737 15738 For questions related to the use of GCC, please consult these web 15739 pages and the [15]GCC manuals. If that fails, the 15740 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15741 web pages and the development of GCC are welcome on our developer 15742 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 15743 archives. 15744 15745 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 15746 distribution of this entire article is permitted in any medium, 15747 provided this notice is preserved. 15748 15749 These pages are [20]maintained by the GCC team. Last modified 15750 2022-10-26. 15751 15752 References 15753 15754 1. http://www.gnu.org/ 15755 2. https://gcc.gnu.org/gcc-4.3/changes.html 15756 3. https://gcc.gnu.org/gcc-4.3/changes.html 15757 4. https://gcc.gnu.org/gcc-4.3/changes.html 15758 5. https://gcc.gnu.org/gcc-4.3/changes.html 15759 6. https://gcc.gnu.org/gcc-4.3/changes.html 15760 7. https://gcc.gnu.org/gcc-4.3/changes.html 15761 8. https://gcc.gnu.org/gcc-4.3/changes.html 15762 9. https://gcc.gnu.org/gcc-4.3/buildstat.html 15763 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15764 11. https://gcc.gnu.org/index.html 15765 12. mailto:gcc (a] gcc.gnu.org 15766 13. https://gcc.gnu.org/mirrors.html 15767 14. https://gcc.gnu.org/git.html 15768 15. https://gcc.gnu.org/onlinedocs/ 15769 16. mailto:gcc-help (a] gcc.gnu.org 15770 17. mailto:gcc (a] gcc.gnu.org 15771 18. https://gcc.gnu.org/lists.html 15772 19. https://www.fsf.org/ 15773 20. https://gcc.gnu.org/about.html 15774 ====================================================================== 15775 http://gcc.gnu.org/gcc-4.3/changes.html 15776 15777 GCC 4.3 Release Series 15778 Changes, New Features, and Fixes 15779 15780 The latest release in the 4.3 release series is [1]GCC 4.3.5. 15781 15782 Caveats 15783 15784 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 15785 various front-end languages it supports. See the [4]prerequisites 15786 page for version requirements. 15787 * ColdFire targets now treat long double as having the same format as 15788 double. In earlier versions of GCC, they used the 68881 long double 15789 format instead. 15790 * The m68k-uclinux target now uses the same calling conventions as 15791 m68k-linux-gnu. You can select the original calling conventions by 15792 configuring for m68k-uclinuxoldabi instead. Note that 15793 m68k-uclinuxoldabi also retains the original 80-bit long double on 15794 ColdFire targets. 15795 * The -fforce-mem option has been removed because it has had no 15796 effect in the last few GCC releases. 15797 * The i386 -msvr3-shlib option has been removed since it is no longer 15798 used. 15799 * Fastcall for i386 has been changed not to pass aggregate arguments 15800 in registers, following Microsoft compilers. 15801 * Support for the AOF assembler has been removed from the ARM back 15802 end; this affects only the targets arm-semi-aof and armel-semi-aof, 15803 which are no longer recognized. We removed these targets without a 15804 deprecation period because we discovered that they have been 15805 unusable since GCC 4.0.0. 15806 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 15807 has been removed. This support had been deprecated since GCC 4.0.0. 15808 * Support for a number of older systems and recently unmaintained or 15809 untested target ports of GCC has been declared obsolete in GCC 4.3. 15810 Unless there is activity to revive them, the next release of GCC 15811 will have their sources permanently removed. 15812 All GCC ports for the following processor architectures have been 15813 declared obsolete: 15814 + Morpho MT (mt-*) 15815 The following aliases for processor architectures have been 15816 declared obsolete. Users should use the indicated generic target 15817 names instead, with compile-time options such as -mcpu or 15818 configure-time options such as --with-cpu to control the 15819 configuration more precisely. 15820 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 15821 instead). 15822 + parisc*-*-* (use hppa*-*-* instead). 15823 + m680[012]0-*-* (use m68k-*-* instead). 15824 All GCC ports for the following operating systems have been 15825 declared obsolete: 15826 + BeOS (*-*-beos*) 15827 + kaOS (*-*-kaos*) 15828 + GNU/Linux using the a.out object format (*-*-linux*aout*) 15829 + GNU/Linux using version 1 of the GNU C Library 15830 (*-*-linux*libc1*) 15831 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 15832 *-*-solaris2.[0-6].*) 15833 + Miscellaneous System V (*-*-sysv*) 15834 + WindISS (*-*-windiss*) 15835 Also, those for some individual systems on particular architectures 15836 have been obsoleted: 15837 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 15838 + CRIS with a.out object format (cris-*-aout) 15839 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 15840 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 15841 + PRO on PA-RISC (hppa1.1-*-pro*) 15842 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 15843 i[34567]86-sequent-sysv4*) 15844 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 15845 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 15846 was previously [5]removed in 2001, leaving only the support 15847 for UWIN as a target now being deprecated) 15848 + ChorusOS on PowerPC (powerpc-*-chorusos*) 15849 + All VAX configurations apart from NetBSD and OpenBSD 15850 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 15851 * The [6]-Wconversion option has been modified. Its purpose now is to 15852 warn for implicit conversions that may alter a value. This new 15853 behavior is available for both C and C++. Warnings about 15854 conversions between signed and unsigned integers can be disabled by 15855 using -Wno-sign-conversion. In C++, they are disabled by default 15856 unless -Wsign-conversion is explicitly requested. The old behavior 15857 of -Wconversion, that is, warn for prototypes causing a type 15858 conversion that is different from what would happen to the same 15859 argument in the absence of a prototype, has been moved to a new 15860 option -Wtraditional-conversion, which is only available for C. 15861 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 15862 been removed because they were deprecated for more than 3 GCC major 15863 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 15864 -mtune=pentiumpro as a replacement. 15865 * The -funsafe-math-optimizations option now automatically turns on 15866 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 15867 reassociation and thus may introduce or remove traps. 15868 * The -ftree-vectorize option is now on by default under -O3. In 15869 order to generate code for a SIMD extension, it has to be enabled 15870 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 15871 i?86 and x86_64. 15872 * More information on porting to GCC 4.3 from previous versions of 15873 GCC can be found in the [7]porting guide for this release. 15874 15875 General Optimizer Improvements 15876 15877 * The GCC middle-end has been integrated with the MPFR library. This 15878 allows GCC to evaluate and replace at compile-time calls to 15879 built-in math functions having constant arguments with their 15880 mathematically equivalent results. In making use of MPFR, GCC can 15881 generate correct results regardless of the math library 15882 implementation or floating point precision of the host platform. 15883 This also allows GCC to generate identical results regardless of 15884 whether one compiles in native or cross-compile configurations to a 15885 particular target. The following built-in functions take advantage 15886 of this new capability: acos, acosh, asin, asinh, atan2, atan, 15887 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 15888 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 15889 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 15890 tan, tanh, tgamma, y0, y1 and yn. The float and long double 15891 variants of these functions (e.g. sinf and sinl) are also handled. 15892 The sqrt and cabs functions with constant arguments were already 15893 optimized in prior GCC releases. Now they also use MPFR. 15894 * A new forward propagation pass on RTL was added. The new pass 15895 replaces several slower transformations, resulting in compile-time 15896 improvements as well as better code generation in some cases. 15897 * A new command-line switch -frecord-gcc-switches has been added to 15898 GCC, although it is only enabled for some targets. The switch 15899 causes the command line that was used to invoke the compiler to be 15900 recorded into the object file that is being created. The exact 15901 format of this recording is target and binary file format 15902 dependent, but it usually takes the form of a note section 15903 containing ASCII text. The switch is related to the -fverbose-asm 15904 switch, but that one only records the information in the assembler 15905 output file as comments, so the information never reaches the 15906 object file. 15907 * The inliner heuristic is now aware of stack frame consumption. New 15908 command-line parameters --param large-stack-frame and --param 15909 large-stack-frame-growth can be used to limit stack frame size 15910 growth caused by inlining. 15911 * During feedback directed optimizations, the expected block size the 15912 memcpy, memset and bzero functions operate on is discovered and for 15913 cases of commonly used small sizes, specialized inline code is 15914 generated. 15915 * __builtin_expect no longer requires its argument to be a compile 15916 time constant. 15917 * Interprocedural optimization was reorganized to work on functions 15918 in SSA form. This enables more precise and cheaper dataflow 15919 analysis and makes writing interprocedural optimizations easier. 15920 The following improvements have been implemented on top of this 15921 framework: 15922 + Pre-inline optimization: Selected local optimization passes 15923 are run before the inliner (and other interprocedural passes) 15924 are executed. This significantly improves the accuracy of code 15925 growth estimates used by the inliner and reduces the overall 15926 memory footprint for large compilation units. 15927 + Early inlining (a simple bottom-up inliner pass inlining only 15928 functions whose body is smaller than the expected call 15929 overhead) is now executed with the early optimization passes, 15930 thus inlining already optimized function bodies into an 15931 unoptimized function that is subsequently optimized by early 15932 optimizers. This enables the compiler to quickly eliminate 15933 abstraction penalty in C++ programs. 15934 + Interprocedural constant propagation now operate on SSA form 15935 increasing accuracy of the analysis. 15936 * A new internal representation for GIMPLE statements has been 15937 contributed, resulting in compile-time memory savings. 15938 * The vectorizer was enhanced to support vectorization of outer 15939 loops, intra-iteration parallelism (loop-aware SLP), vectorization 15940 of strided accesses and loops with multiple data-types. Run-time 15941 dependency testing using loop versioning was added. The cost model, 15942 turned on by -fvect-cost-model, was developed. 15943 15944 New Languages and Language specific improvements 15945 15946 * We have added new command-line options 15947 -finstrument-functions-exclude-function-list and 15948 -finstrument-functions-exclude-file-list. They provide more control 15949 over which functions are annotated by the -finstrument-functions 15950 option. 15951 15952 C family 15953 15954 * Implicit conversions between generic vector types are now only 15955 permitted when the two vectors in question have the same number of 15956 elements and compatible element types. (Note that the restriction 15957 involves compatible element types, not implicitly-convertible 15958 element types: thus, a vector type with element type int may not be 15959 implicitly converted to a vector type with element type unsigned 15960 int.) This restriction, which is in line with specifications for 15961 SIMD architectures such as AltiVec, may be relaxed using the flag 15962 -flax-vector-conversions. This flag is intended only as a 15963 compatibility measure and should not be used for new code. 15964 * -Warray-bounds has been added and is now enabled by default for 15965 -Wall . It produces warnings for array subscripts that can be 15966 determined at compile time to be always out of bounds. 15967 -Wno-array-bounds will disable the warning. 15968 * The constructor and destructor function attributes now accept 15969 optional priority arguments which control the order in which the 15970 constructor and destructor functions are run. 15971 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 15972 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 15973 -Wignored-qualifiers have been added for finer control of the 15974 diverse warnings enabled by -Wextra. 15975 * A new function attribute alloc_size has been added to mark up 15976 malloc style functions. For constant sized allocations this can be 15977 used to find out the size of the returned pointer using the 15978 __builtin_object_size() function for buffer overflow checking and 15979 similar. This supplements the already built-in malloc and calloc 15980 constant size handling. 15981 * Integer constants written in binary are now supported as a GCC 15982 extension. They consist of a prefix 0b or 0B, followed by a 15983 sequence of 0 and 1 digits. 15984 * A new predefined macro __COUNTER__ has been added. It expands to 15985 sequential integral values starting from 0. In conjunction with the 15986 ## operator, this provides a convenient means to generate unique 15987 identifiers. 15988 * A new command-line option -fdirectives-only has been added. It 15989 enables a special preprocessing mode which improves the performance 15990 of applications like distcc and ccache. 15991 * Fixed-point data types and operators have been added. They are 15992 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 15993 Currently, only MIPS targets are supported. 15994 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 15995 N1241, is now supported as a GCC extension to C for targets 15996 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 15997 and x86_64-*-linux-gnu. The feature introduces new data types 15998 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 15999 DD, and DL. 16000 16001 C++ 16002 16003 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 16004 * -Wc++0x-compat has been added and is now enabled by default for 16005 -Wall. It produces warnings for constructs whose meaning differs 16006 between ISO C++ 1998 and C++0x. 16007 * The -Wparentheses option now works for C++ as it does for C. It 16008 warns if parentheses are omitted when operators with confusing 16009 precedence are nested. It also warns about ambiguous else 16010 statements. Since -Wparentheses is enabled by -Wall, this may cause 16011 additional warnings with existing C++ code which uses -Wall. These 16012 new warnings may be disabled by using -Wall -Wno-parentheses. 16013 * The -Wmissing-declarations now works for C++ as it does for C. 16014 * The -fvisibility-ms-compat flag was added, to make it easier to 16015 port larger projects using shared libraries from Microsoft's Visual 16016 Studio to ELF and Mach-O systems. 16017 * C++ attribute handling has been overhauled for template arguments 16018 (ie dependent types). In particular, __attribute__((aligned(T))); 16019 works for C++ types. 16020 16021 Runtime Library (libstdc++) 16022 16023 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 16024 * Support for TR1 mathematical special functions and regular 16025 expressions. 16026 * Default what implementations give more elaborate exception strings 16027 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 16028 * Header dependencies have been streamlined, reducing unnecessary 16029 includes and pre-processed bloat. 16030 * Variadic template implementations of items in <tuple> and 16031 <functional>. 16032 * An experimental [11]parallel mode has been added. This is a 16033 parallel implementation of many C++ Standard library algorithms, 16034 like std::accumulate, std::for_each, std::transform, or std::sort, 16035 to give but four examples. These algorithms can be substituted for 16036 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 16037 or all existing algorithms can be transformed via the 16038 -D_GLIBCXX_PARALLEL macro. 16039 * Debug mode versions of classes in <unordered_set> and 16040 <unordered_map>. 16041 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 16042 now <backward/hash_set> and <backward/hash_map>. This code: 16043 #include <ext/hash_set> 16044 __gnu_cxx::hash_set<int> s; 16045 16046 Can be transformed (in order of preference) to: 16047 #include <tr1/unordered_set> 16048 std::tr1::unordered_set<int> s; 16049 16050 or 16051 #include <backward/hash_set> 16052 __gnu_cxx::hash_set<int> s; 16053 16054 Similar transformations apply to __gnu_cxx::hash_map, 16055 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 16056 __gnu_cxx::hash_multiset. 16057 16058 Fortran 16059 16060 * Due to the fact that the GMP and MPFR libraries are required for 16061 all languages, Fortran is no longer special in this regard and is 16062 available by default. 16063 * The [12]-fexternal-blas option has been added, which generates 16064 calls to BLAS routines for intrinsic matrix operations such as 16065 matmul rather than using the built-in algorithms. 16066 * Support to give a backtrace (compiler flag -fbacktrace or 16067 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 16068 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 16069 run-time error occured. 16070 * GNU Fortran now defines __GFORTRAN__ when it runs the C 16071 preprocessor (CPP). 16072 * The [13]-finit-local-zero, -finit-real, -finit-integer, 16073 -finit-character, and -finit-logical options have been added, which 16074 can be used to initialize local variables. 16075 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 16076 which calculate the Gamma function and its logarithm. Use EXTERNAL 16077 gamma if you want to use your own gamma function. 16078 * GNU Fortran now regards the backslash character as literal (as 16079 required by the Fortran 2003 standard); using [16]-fbackslash GNU 16080 Fortran interprets backslashes as C-style escape characters. 16081 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 16082 literal constants has been changed. Before they were always 16083 interpreted as integer; now they are bit-wise transferred as 16084 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 16085 2003 standard, and for real and complex variables in DATA 16086 statements or when directly assigned to real and complex variables. 16087 Everywhere else and especially in expressions they are still 16088 regarded as integer constants. 16089 * Fortran 2003 support has been extended: 16090 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 16091 + Pointer intent 16092 + Intrinsic module ISO_ENV_FORTRAN 16093 + Interoperability with C (ISO C Bindings) 16094 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 16095 attribute) 16096 + Fortran 2003 BOZ 16097 16098 Java (GCJ) 16099 16100 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 16101 This enables the use of all 1.5 language features, and fixes most 16102 existing front end bugs. 16103 * libgcj now supports all 1.5 language features which require runtime 16104 support: foreach, enum, annotations, generics, and auto-boxing. 16105 * We've made many changes to the tools shipped with gcj. 16106 + The old jv-scan tool has been removed. This tool never really 16107 worked properly. There is no replacement. 16108 + gcjh has been rewritten. Some of its more obscure options no 16109 longer work, but are still recognized in an attempt at 16110 compatibility. gjavah is a new program with similar 16111 functionality but different command-line options. 16112 + grmic and grmiregistry have been rewritten. grmid has been 16113 added. 16114 + gjar replaces the old fastjar. 16115 + gjarsigner (used for signing jars), gkeytool (used for key 16116 management), gorbd (for CORBA), gserialver (computes 16117 serialization UIDs), and gtnameserv (also for CORBA) are now 16118 installed. 16119 * The ability to dump the contents of the java run time heap to a 16120 file for off-line analysis has been added. The heap dumps may be 16121 analyzed with the new gc-analyze tool. They may be generated on 16122 out-of-memory conditions or on demand and are controlled by the new 16123 run time class gnu.gcj.util.GCInfo. 16124 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 16125 provide correct, updated, timezone information. This means that 16126 packagers no longer have to update libgcj when a time zone change 16127 is published. 16128 16129 [18]New Targets and Target Specific Improvements 16130 16131 IA-32/x86-64 16132 16133 * Tuning for Intel Core 2 processors is available via -mtune=core2 16134 and -march=core2. 16135 * Tuning for AMD Geode processors is available via -mtune=geode and 16136 -march=geode. 16137 * Code generation of block move (memcpy) and block set (memset) was 16138 rewritten. GCC can now pick the best algorithm (loop, unrolled 16139 loop, instruction with rep prefix or a library call) based on the 16140 size of the block being copied and the CPU being optimized for. A 16141 new option -minline-stringops-dynamically has been added. With this 16142 option string operations of unknown size are expanded such that 16143 small blocks are copied by in-line code, while for large blocks a 16144 library call is used. This results in faster code than 16145 -minline-all-stringops when the library implementation is capable 16146 of using cache hierarchy hints. The heuristic choosing the 16147 particular algorithm can be overwritten via -mstringop-strategy. 16148 Newly also memset of values different from 0 is inlined. 16149 * GCC no longer places the cld instruction before string operations. 16150 Both i386 and x86-64 ABI documents mandate the direction flag to be 16151 clear at the entry of a function. It is now invalid to set the flag 16152 in asm statement without reseting it afterward. 16153 * Support for SSSE3 built-in functions and code generation are 16154 available via -mssse3. 16155 * Support for SSE4.1 built-in functions and code generation are 16156 available via -msse4.1. 16157 * Support for SSE4.2 built-in functions and code generation are 16158 available via -msse4.2. 16159 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 16160 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 16161 allow explicit control of x87 floating point precision. 16162 * Support for __float128 (TFmode) IEEE quad type and corresponding 16163 TCmode IEEE complex quad type is available via the soft-fp library 16164 on x86_64 targets. This includes basic arithmetic operations 16165 (addition, subtraction, negation, multiplication and division) on 16166 __float128 real and TCmode complex values, the full set of IEEE 16167 comparisons between __float128 values, conversions to and from 16168 float, double and long double floating point types, as well as 16169 conversions to and from signed or unsigned integer, signed or 16170 unsigned long integer and signed or unsigned quad (TImode) integer 16171 types. Additionally, all operations generate the full set of IEEE 16172 exceptions and support the full set of IEEE rounding modes. 16173 * GCC can now utilize the ACML library for vectorizing calls to a set 16174 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 16175 link to an ACML ABI compatible library. 16176 16177 ARM 16178 16179 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 16180 has been added. 16181 16182 [19]CRIS 16183 16184 New features 16185 16186 * Compiler and Library support for the CRIS v32 architecture, as 16187 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 16188 added. 16189 16190 Configuration changes 16191 16192 * The cris-*-elf target now includes support for CRIS v32, including 16193 libraries, through the -march=v32 option. 16194 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 16195 * A new crisv32-*-linux* target defaults to generate code for CRIS 16196 v32. 16197 * The cris-*-aout target has been obsoleted. 16198 16199 Improved support for built-in functions 16200 16201 * GCC can now use the lz and swapwbr instructions to implement the 16202 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 16203 * __builtin_bswap32 is now implemented using the swapwb instruction, 16204 when available. 16205 16206 [20]m68k and ColdFire 16207 16208 New features 16209 16210 * Support for several new ColdFire processors has been added. You can 16211 generate code for them using the new -mcpu option. 16212 * All targets now support ColdFire processors. 16213 * m68k-uclinux targets have improved support for C++ constructors and 16214 destructors, and for shared libraries. 16215 * It is now possible to set breakpoints on the first or last line of 16216 a function, even if there are no statements on that line. 16217 16218 Optimizations 16219 16220 * Support for sibling calls has been added. 16221 * More use is now made of the ColdFire mov3q instruction. 16222 * __builtin_clz is now implemented using the ff1 ColdFire 16223 instruction, when available. 16224 * GCC now honors the -m68010 option. 68010 code now uses clr rather 16225 than move to zero volatile memory. 16226 * 68020 targets and above can now use symbol(index.size*scale) 16227 addresses for indexed array accesses. Earlier compilers would 16228 always load the symbol into a base register first. 16229 16230 Configuration changes 16231 16232 * All m68k and ColdFire targets now allow the default processor to be 16233 set at configure time using --with-cpu. 16234 * A --with-arch configuration option has been added. This option 16235 allows you to restrict a target to ColdFire or non-ColdFire 16236 processors. 16237 16238 Preprocessor macros 16239 16240 * An __mcfv*__ macro is now defined for all ColdFire targets. 16241 (Earlier versions of GCC only defined __mcfv4e__.) 16242 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 16243 * All targets now define __mc68010 and __mc68010__ when generating 16244 68010 code. 16245 16246 Command-line changes 16247 16248 * New command-line options -march, -mcpu, -mtune and -mhard-float 16249 have been added. These options apply to both m68k and ColdFire 16250 targets. 16251 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 16252 versions of -mshort, etc. 16253 * -fforce-addr has been removed. It is now ignored by the compiler. 16254 16255 Other improvements 16256 16257 * ColdFire targets now try to maintain a 4-byte-aligned stack where 16258 possible. 16259 * m68k-uclinux targets now try to avoid situations that lead to the 16260 load-time error: BINFMT_FLAT: reloc outside program. 16261 16262 MIPS 16263 16264 Changes to existing configurations 16265 16266 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 16267 and n64. Every GNU/Linux configuration now builds these libraries 16268 by default. 16269 * GNU/Linux configurations now generate -mno-shared code unless 16270 overridden by -fpic, -fPIC, -fpie or -fPIE. 16271 * mipsisa32*-linux-gnu configurations now generate hard-float code by 16272 default, just like other mipsisa32* and mips*-linux-gnu 16273 configurations. You can build a soft-float version of any 16274 mips*-linux-gnu configuration by passing --with-float=soft to 16275 configure. 16276 * mips-wrs-vxworks now supports run-time processes (RTPs). 16277 16278 Changes to existing command-line options 16279 16280 * The -march and -mtune options no longer accept 24k as a processor 16281 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 16282 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 16283 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 16284 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 16285 24kx, 24kex and 34kx. 16286 16287 New configurations 16288 16289 GCC now supports the following configurations: 16290 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 16291 default. Earlier releases also recognized this configuration, but 16292 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 16293 you can customize any mips*-linux-gnu* configuration to a 16294 particular ISA or processor by passing an appropriate --with-arch 16295 option to configure. 16296 * mipsisa*-sde-elf*, which provides compatibility with MIPS 16297 Technologies' SDE toolchains. The configuration uses the SDE 16298 libraries by default, but you can use it like other newlib-based 16299 ELF configurations by passing --with-newlib to configure. It is the 16300 only configuration besides mips64vr*-elf* to build MIPS16 as well 16301 as non-MIPS16 libraries. 16302 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 16303 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 16304 and 64-bit forms of the EABI. 16305 16306 New processors and application-specific extensions 16307 16308 * Support for the SmartMIPS ASE is available through the new 16309 -msmartmips option. 16310 * Support for revision 2 of the DSP ASE is available through the new 16311 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 16312 indicates the revision of the ASE in use. 16313 * Support for the 4KS and 74K families of processors is available 16314 through the -march and -mtune options. 16315 16316 Improved support for built-in functions 16317 16318 * GCC can now use load-linked, store-conditional and sync 16319 instructions to implement atomic built-in functions such as 16320 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 16321 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 16322 * GCC can now use the clz and dclz instructions to implement the 16323 __builtin_ctz and __builtin_ffs families of functions. 16324 * There is a new __builtin___clear_cache function for flushing the 16325 instruction cache. GCC expands this function inline on MIPS32 16326 revision 2 targets, otherwise it calls the function specified by 16327 -mcache-flush-func. 16328 16329 MIPS16 improvements 16330 16331 * GCC can now compile objects that contain a mixture of MIPS16 and 16332 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 16333 for specifying which mode a function should use. 16334 * A new option called -minterlink-mips16 makes non-MIPS16 code 16335 link-compatible with MIPS16 code. 16336 * After many bug fixes, the long-standing MIPS16 -mhard-float support 16337 should now work fairly reliably. 16338 * GCC can now use the MIPS16e save and restore instructions. 16339 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 16340 with -G0 -fsection-anchors is often smaller than code compiled with 16341 -G8. However, please note that you must usually compile all objects 16342 in your application with the same -G option; see the documentation 16343 of -G for details. 16344 * A new option called-mcode-readable specifies which instructions are 16345 allowed to load from the code segment. -mcode-readable=yes is the 16346 default and says that any instruction may load from the code 16347 segment. The other alternatives are -mcode-readable=pcrel, which 16348 says that only PC-relative MIPS16 instructions may load from the 16349 code segment, and -mcode-readable=no, which says that no 16350 instruction may do so. Please see the documentation for more 16351 details, including example uses. 16352 16353 Small-data improvements 16354 16355 There are three new options for controlling small data: 16356 * -mno-extern-sdata, which disables small-data accesses for 16357 externally-defined variables. Code compiled with -Gn 16358 -mno-extern-sdata will be link-compatible with any -G setting 16359 between -G0 and -Gn inclusive. 16360 * -mno-local-sdata, which disables the use of small-data sections for 16361 data that is not externally visible. This option can be a useful 16362 way of reducing small-data usage in less performance-critical parts 16363 of an application. 16364 * -mno-gpopt, which disables the use of the $gp register while still 16365 honoring the -G limit when placing externally-visible data. This 16366 option implies -mno-extern-sdata and -mno-local-sdata and it can be 16367 useful in situations where $gp does not necessarily hold the 16368 expected value. 16369 16370 Miscellaneous improvements 16371 16372 * There is a new option called -mbranch-cost for tweaking the 16373 perceived cost of branches. 16374 * If GCC is configured to use a version of GAS that supports the 16375 .gnu_attribute directive, it will use that directive to record 16376 certain properties of the output code. .gnu_attribute is new to GAS 16377 2.18. 16378 * There are two new function attributes, near and far, for overriding 16379 the command-line setting of -mlong-calls on a function-by-function 16380 basis. 16381 * -mfp64, which previously required a 64-bit target, now works with 16382 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 16383 mipsisa*-sde-elf* configurations provide suitable library support. 16384 * GCC now recognizes the -mdmx and -mmt options and passes them down 16385 to the assembler. It does nothing else with the options at present. 16386 16387 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 16388 (BEA) 16389 16390 * Support has been added for this new architecture. 16391 16392 RS6000 (POWER/PowerPC) 16393 16394 * Support for the PowerPC 750CL paired-single instructions has been 16395 added with a new powerpc-*-linux*paired* target configuration. It 16396 is enabled by an associated -mpaired option and can be accessed 16397 using new built-in functions. 16398 * Support for auto-detecting architecture and system configuration to 16399 auto-select processor optimization tuning. 16400 * Support for VMX on AIX 5.3 has been added. 16401 * Support for AIX Version 6.1 has been added. 16402 16403 S/390, zSeries and System z9 16404 16405 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 16406 added. When using the -march=z9-ec option, the compiler will 16407 generate code making use of instructions provided by the decimal 16408 floating point facility and the floating point conversion facility 16409 (pfpo). Besides the instructions used to implement decimal floating 16410 point operations these facilities also contain instructions to move 16411 between general purpose and floating point registers and to modify 16412 and copy the sign-bit of floating point values. 16413 * When the -march=z9-ec option is used the new 16414 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 16415 decimal floating point hardware instructions will be used or not. 16416 If none of them is given the hardware support is enabled by 16417 default. 16418 * The -mstack-guard option can now be omitted when using stack 16419 checking via -mstack-size in order to let GCC choose a sensible 16420 stack guard value according to the frame size of each function. 16421 * Various changes to improve performance of generated code have been 16422 implemented, including: 16423 + The condition code set by an add logical with carry 16424 instruction is now available for overflow checks like: a + b + 16425 carry < b. 16426 + The test data class instruction is now used to implement 16427 sign-bit and infinity checks of binary and decimal floating 16428 point numbers. 16429 16430 SPARC 16431 16432 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 16433 added. 16434 16435 Xtensa 16436 16437 * Stack unwinding for exception handling now uses by default a 16438 specialized version of DWARF unwinding. This is not 16439 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 16440 Xtensa with previous versions of GCC. 16441 * For Xtensa processors that include the Conditional Store option, 16442 the built-in functions for atomic memory access are now implemented 16443 using S32C1I instructions. 16444 * If the Xtensa NSA option is available, GCC will use it to implement 16445 the __builtin_ctz and __builtin_clz functions. 16446 16447 Documentation improvements 16448 16449 * Existing libstdc++ documentation has been edited and restructured 16450 into a single DocBook XML manual. The results can be viewed online 16451 [21]here. 16452 16453 Other significant improvements 16454 16455 * The compiler's --help command-line option has been extended so that 16456 it now takes an optional set of arguments. These arguments restrict 16457 the information displayed to specific classes of command-line 16458 options, and possibly only a subset of those options. It is also 16459 now possible to replace the descriptive text associated with each 16460 displayed option with an indication of its current value, or for 16461 binary options, whether it has been enabled or disabled. 16462 Here are some examples. The following will display all the options 16463 controlling warning messages: 16464 --help=warnings 16465 16466 Whereas this will display all the undocumented, target specific 16467 options: 16468 --help=target,undocumented 16469 16470 This sequence of commands will display the binary optimizations 16471 that are enabled by -O3: 16472 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 16473 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 16474 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 16475 16476 * The configure options --with-pkgversion and --with-bugurl have been 16477 added. These allow distributors of GCC to include a 16478 distributor-specific string in manuals and --version output and to 16479 specify the URL for reporting bugs in their versions of GCC. 16480 16481 [22]GCC 4.3.1 16482 16483 This is the [23]list of problem reports (PRs) from GCC's bug tracking 16484 system that are known to be fixed in the 4.3.1 release. This list might 16485 not be complete (that is, it is possible that some PRs that have been 16486 fixed are not listed here). 16487 16488 Target Specific Changes 16489 16490 IA-32/x86-64 16491 16492 ABI changes 16493 16494 * Starting with GCC 4.3.1, decimal floating point variables are 16495 aligned to their natural boundaries when they are passed on the 16496 stack for i386. 16497 16498 Command-line changes 16499 16500 * Starting with GCC 4.3.1, the -mcld option has been added to 16501 automatically generate a cld instruction in the prologue of 16502 functions that use string instructions. This option is used for 16503 backward compatibility on some operating systems and can be enabled 16504 by default for 32-bit x86 targets by configuring GCC with the 16505 --enable-cld configure option. 16506 16507 [24]GCC 4.3.2 16508 16509 This is the [25]list of problem reports (PRs) from GCC's bug tracking 16510 system that are known to be fixed in the 4.3.2 release. This list might 16511 not be complete (that is, it is possible that some PRs that have been 16512 fixed are not listed here). 16513 16514 [26]GCC 4.3.3 16515 16516 This is the [27]list of problem reports (PRs) from GCC's bug tracking 16517 system that are known to be fixed in the 4.3.3 release. This list might 16518 not be complete (that is, it is possible that some PRs that have been 16519 fixed are not listed here). 16520 16521 [28]GCC 4.3.4 16522 16523 This is the [29]list of problem reports (PRs) from GCC's bug tracking 16524 system that are known to be fixed in the 4.3.4 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 [30]GCC 4.3.5 16529 16530 This is the [31]list of problem reports (PRs) from GCC's bug tracking 16531 system that are known to be fixed in the 4.3.5 release. This list might 16532 not be complete (that is, it is possible that some PRs that have been 16533 fixed are not listed here). 16534 16535 [32]GCC 4.3.6 16536 16537 This is the [33]list of problem reports (PRs) from GCC's bug tracking 16538 system that are known to be fixed in the 4.3.6 release. This list might 16539 not be complete (that is, it is possible that some PRs that have been 16540 fixed are not listed here). 16541 16542 16543 For questions related to the use of GCC, please consult these web 16544 pages and the [34]GCC manuals. If that fails, the 16545 [35]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16546 web pages and the development of GCC are welcome on our developer 16547 list at [36]gcc (a] gcc.gnu.org. All of [37]our lists have public 16548 archives. 16549 16550 Copyright (C) [38]Free Software Foundation, Inc. Verbatim copying and 16551 distribution of this entire article is permitted in any medium, 16552 provided this notice is preserved. 16553 16554 These pages are [39]maintained by the GCC team. Last modified 16555 2025-01-31. 16556 16557 References 16558 16559 1. https://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 16560 2. https://gmplib.org/ 16561 3. https://www.mpfr.org/ 16562 4. https://gcc.gnu.org/install/prerequisites.html 16563 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 16564 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 16565 7. https://gcc.gnu.org/gcc-4.3/porting_to.html 16566 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 16567 9. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 16568 10. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 16569 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 16570 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 16571 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 16572 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 16573 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 16574 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 16575 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 16576 18. https://gcc.gnu.org/gcc-4.3/changes.html#targets 16577 19. https://gcc.gnu.org/gcc-4.3/changes.html#cris 16578 20. https://gcc.gnu.org/gcc-4.3/changes.html#m68k 16579 21. https://gcc.gnu.org/onlinedocs/libstdc++/ 16580 22. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.1 16581 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 16582 24. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.2 16583 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 16584 26. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.3 16585 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 16586 28. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.4 16587 29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 16588 30. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.5 16589 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 16590 32. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.6 16591 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 16592 34. https://gcc.gnu.org/onlinedocs/ 16593 35. mailto:gcc-help (a] gcc.gnu.org 16594 36. mailto:gcc (a] gcc.gnu.org 16595 37. https://gcc.gnu.org/lists.html 16596 38. https://www.fsf.org/ 16597 39. https://gcc.gnu.org/about.html 16598 ====================================================================== 16599 http://gcc.gnu.org/gcc-4.2/index.html 16600 16601 GCC 4.2 Release Series 16602 16603 (This release series is no longer supported.) 16604 16605 May 19, 2008 16606 16607 The [1]GNU project and the GCC developers are pleased to announce the 16608 release of GCC 4.2.4. 16609 16610 This release is a bug-fix release, containing fixes for regressions in 16611 GCC 4.2.3 relative to previous releases of GCC. 16612 16613 Release History 16614 16615 GCC 4.2.4 16616 May 19, 2008 ([2]changes) 16617 16618 GCC 4.2.3 16619 February 1, 2008 ([3]changes) 16620 16621 GCC 4.2.2 16622 October 7, 2007 ([4]changes) 16623 16624 GCC 4.2.1 16625 July 18, 2007 ([5]changes) 16626 16627 GCC 4.2.0 16628 May 13, 2007 ([6]changes) 16629 16630 References and Acknowledgements 16631 16632 GCC used to stand for the GNU C Compiler, but since the compiler 16633 supports several other languages aside from C, it now stands for the 16634 GNU Compiler Collection. 16635 16636 A list of [7]successful builds is updated as new information becomes 16637 available. 16638 16639 The GCC developers would like to thank the numerous people that have 16640 contributed new features, improvements, bug fixes, and other changes as 16641 well as test results to GCC. This [8]amazing group of volunteers is 16642 what makes GCC successful. 16643 16644 For additional information about GCC please refer to the [9]GCC project 16645 web site or contact the [10]GCC development mailing list. 16646 16647 To obtain GCC please use [11]our mirror sites or [12]our version 16648 control system. 16649 16650 16651 For questions related to the use of GCC, please consult these web 16652 pages and the [13]GCC manuals. If that fails, the 16653 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16654 web pages and the development of GCC are welcome on our developer 16655 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 16656 archives. 16657 16658 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 16659 distribution of this entire article is permitted in any medium, 16660 provided this notice is preserved. 16661 16662 These pages are [18]maintained by the GCC team. Last modified 16663 2022-10-26. 16664 16665 References 16666 16667 1. http://www.gnu.org/ 16668 2. https://gcc.gnu.org/gcc-4.2/changes.html 16669 3. https://gcc.gnu.org/gcc-4.2/changes.html 16670 4. https://gcc.gnu.org/gcc-4.2/changes.html 16671 5. https://gcc.gnu.org/gcc-4.2/changes.html 16672 6. https://gcc.gnu.org/gcc-4.2/changes.html 16673 7. https://gcc.gnu.org/gcc-4.2/buildstat.html 16674 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16675 9. https://gcc.gnu.org/index.html 16676 10. mailto:gcc (a] gcc.gnu.org 16677 11. https://gcc.gnu.org/mirrors.html 16678 12. https://gcc.gnu.org/git.html 16679 13. https://gcc.gnu.org/onlinedocs/ 16680 14. mailto:gcc-help (a] gcc.gnu.org 16681 15. mailto:gcc (a] gcc.gnu.org 16682 16. https://gcc.gnu.org/lists.html 16683 17. https://www.fsf.org/ 16684 18. https://gcc.gnu.org/about.html 16685 ====================================================================== 16686 http://gcc.gnu.org/gcc-4.2/changes.html 16687 16688 GCC 4.2 Release Series 16689 Changes, New Features, and Fixes 16690 16691 Caveats 16692 16693 * GCC no longer accepts the -fshared-data option. This option has had 16694 no effect in any GCC 4 release; the targets to which the option 16695 used to apply had been removed before GCC 4.0. 16696 16697 General Optimizer Improvements 16698 16699 * New command-line options specify the possible relationships among 16700 parameters and between parameters and global data. For example, 16701 -fargument-noalias-anything specifies that arguments do not alias 16702 any other storage. 16703 Each language will automatically use whatever option is required by 16704 the language standard. You should not need to use these options 16705 yourself. 16706 16707 New Languages and Language specific improvements 16708 16709 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 16710 * New command-line options -fstrict-overflow and -Wstrict-overflow 16711 have been added. -fstrict-overflow tells the compiler that it may 16712 assume that the program follows the strict signed overflow 16713 semantics permitted for the language: for C and C++ this means that 16714 the compiler may assume that signed overflow does not occur. For 16715 example, a loop like 16716 for (i = 1; i > 0; i *= 2) 16717 16718 is presumably intended to continue looping until i overflows. With 16719 -fstrict-overflow, the compiler may assume that signed overflow 16720 will not occur, and transform this into an infinite loop. 16721 -fstrict-overflow is turned on by default at -O2, and may be 16722 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 16723 be used to warn about cases where the compiler assumes that signed 16724 overflow will not occur. It takes five different levels: 16725 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 16726 -Wstrict-overflow=1 is enabled by -Wall. 16727 * The new command-line option -fno-toplevel-reorder directs GCC to 16728 emit top-level functions, variables, and asm statements in the same 16729 order that they appear in the input file. This is intended to 16730 support existing code which relies on a particular ordering (for 16731 example, code which uses top-level asm statements to switch 16732 sections). For new code, it is generally better to use function and 16733 variable attributes. The -fno-toplevel-reorder option may be used 16734 for most cases which currently use -fno-unit-at-a-time. The 16735 -fno-unit-at-a-time option will be removed in some future version 16736 of GCC. If you know of a case which requires -fno-unit-at-a-time 16737 which is not fixed by -fno-toplevel-reorder, please open a bug 16738 report. 16739 16740 C family 16741 16742 * The pragma redefine_extname will now macro expand its tokens for 16743 compatibility with SunPRO. 16744 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 16745 GCC to handle inline functions as specified in the C99 standard. In 16746 preparation for this, GCC 4.2 will warn about any use of non-static 16747 inline functions in gnu99 or c99 mode. This new warning may be 16748 disabled with the new gnu_inline function attribute or the new 16749 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 16750 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 16751 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 16752 in the current compilation. 16753 * A new command-line option -Waddress has been added to warn about 16754 suspicious uses of memory addresses as, for example, using the 16755 address of a function in a conditional expression, and comparisons 16756 against the memory address of a string literal. This warning is 16757 enabled by -Wall. 16758 16759 C++ 16760 16761 * C++ visibility handling has been overhauled. 16762 Restricted visiblity is propagated from classes to members, from 16763 functions to local statics, and from templates and template 16764 arguments to instantiations, unless the latter has explicitly 16765 declared visibility. 16766 The visibility attribute for a class must come between the 16767 class-key and the name, not after the closing brace. 16768 Attributes are now allowed for enums and elaborated-type-specifiers 16769 that only declare a type. 16770 Members of the anonymous namespace are now local to a particular 16771 translation unit, along with any other declarations which use them, 16772 though they are still treated as having external linkage for 16773 language semantics. 16774 * The (undocumented) extension which permitted templates with default 16775 arguments to be bound to template template parameters with fewer 16776 parameters has been removed. For example: 16777 template <template <typename> class C> 16778 void f(C<double>) {} 16779 16780 template <typename T, typename U = int> 16781 struct S {}; 16782 16783 template void f(S<double>); 16784 16785 is no longer accepted by G++. The reason this code is not accepted 16786 is that S is a template with two parameters; therefore, it cannot 16787 be bound to C which has only one parameter. 16788 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 16789 releases, have been removed. 16790 * The command-line option -fconst-strings, deprecated in previous GCC 16791 releases, has been removed. 16792 * The configure variable enable-__cxa_atexit is now enabled by 16793 default for more targets. Enabling this variable is necessary in 16794 order for static destructors to be executed in the correct order, 16795 but it depends upon the presence of a non-standard C library in the 16796 target library in order to work. The variable is now enabled for 16797 more targets which are known to have suitable C libraries. 16798 * -Wextra will produce warnings for if statements with a semicolon as 16799 the only body, to catch code like: 16800 if (a); 16801 return 1; 16802 return 0; 16803 16804 To suppress the warning in valid cases, use { } instead. 16805 * The C++ front end now also produces strict aliasing warnings when 16806 -fstrict-aliasing -Wstrict-aliasing is in effect. 16807 16808 Runtime Library (libstdc++) 16809 16810 * Added support for TR1 <random>, <complex>, and C compatibility 16811 headers. In addition, a lock-free version of shared_ptr was 16812 contributed as part of Phillip Jordan's Google Summer of Code 16813 project on lock-free containers. 16814 * In association with the Summer of Code work on lock-free 16815 containers, the interface for atomic builtins was adjusted, 16816 creating simpler alternatives for non-threaded code paths. Also, 16817 usage was consolidated and all elements were moved from namespace 16818 std to namespace__gnu_cxx. Affected interfaces are the functions 16819 __exchange_and_add, __atomic_add, and the objects __mutex, 16820 __recursive_mutex, and __scoped_lock. 16821 * Support for versioning weak symbol names via namespace association 16822 was added. However, as this changes the names of exported symbols, 16823 this is turned off by default in the current ABI. Intrepid users 16824 can enable this feature by using 16825 --enable-symvers=gnu-versioned-namespace during configuration. 16826 * Revised, simplified, and expanded policy-based associative 16827 containers, including data types for tree and trie forms 16828 (basic_tree, tree, trie), lists (list_update), and both 16829 collision-chaining and probing hash-based containers 16830 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 16831 the [3]documentation. 16832 * The implementation of the debug mode was modified, whereby the 16833 debug namespaces were nested inside of namespace std and namespace 16834 __gnu_cxx in order to resolve some long standing corner cases 16835 involving name lookup. Debug functionality from the policy-based 16836 data structures was consolidated and enabled with the single macro, 16837 _GLIBCXX_DEBUG. See PR 26142 for more information. 16838 * Added extensions for type traits: __conditional_type, 16839 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 16840 * Added a typelist implementation for compile-time meta-programming. 16841 Elements for typelist construction and operation can be found 16842 within namespace __gnu_cxx::typelist. 16843 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 16844 exception-safety. 16845 * Enabled library-wide visibility control, allowing -fvisibility to 16846 be used. 16847 * Consolidated all nested namespaces and the conversion of 16848 __gnu_internal implementation-private details to anonymous 16849 namespaces whenever possible. 16850 * Implemented LWG resolutions DR 431 and DR 538. 16851 16852 Fortran 16853 16854 * Support for allocatable components has been added (TR 15581 and 16855 Fortran 2003). 16856 * Support for the Fortran 2003 streaming IO extension has been added. 16857 * The GNU Fortran compiler now uses 4-byte record markers by default 16858 for unformatted files to be compatible with g77 and most other 16859 compilers. The implementation allows for records greater than 2 GB 16860 and is compatible with several other compilers. Older versions of 16861 gfortran used 8-byte record markers by default (on most systems). 16862 In order to change the length of the record markers, e.g. to read 16863 unformatted files created by older gfortran versions, the 16864 [4]-frecord-marker=8 option can be used. 16865 16866 Java (GCJ) 16867 16868 * A new command-line option -static-libgcj has been added for targets 16869 that use a linker compatible with GNU Binutils. As its name 16870 implies, this causes libgcj to be linked statically. In some cases 16871 this causes the resulting executable to start faster and use less 16872 memory than if the shared version of libgcj were used. However 16873 caution should be used as it can also cause essential parts of the 16874 library to be omitted. Some of these issues are discussed in: 16875 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 16876 * fastjar is no longer bundled with GCC. To build libgcj, you will 16877 need either InfoZIP (both zip and unzip) or an external jar 16878 program. In the former case, the GCC build will install a jar shell 16879 script that is based on InfoZIP and provides the same functionality 16880 as fastjar. 16881 16882 [6]New Targets and Target Specific Improvements 16883 16884 IA-32/x86-64 16885 16886 * -mtune=generic can now be used to generate code running well on 16887 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 16888 Pentium-M, Intel Pentium 4 and Intel Core 2. 16889 * -mtune=native and -march=native will produce code optimized for the 16890 host architecture as detected using the cpuid instruction. 16891 * Added a new command-line option -fstackrealign and and 16892 __attribute__ ((force_align_arg_pointer)) to realign the stack at 16893 runtime. This allows functions compiled with a vector-aligned stack 16894 to be invoked from legacy objects that keep only word-alignment. 16895 16896 SPARC 16897 16898 * The default CPU setting has been changed from V7 to V9 in 32-bit 16899 mode on Solaris 7 and above. This is already the case in 64-bit 16900 mode. It can be overridden by specifying --with-cpu at configure 16901 time. 16902 * Back-end support of built-in functions for atomic memory access has 16903 been implemented. 16904 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 16905 added. 16906 16907 M32C 16908 16909 * Various bug fixes have made some functions (notably, functions 16910 returning structures) incompatible with previous releases. 16911 Recompiling all libraries is recommended. Note that code quality 16912 has considerably improved since 4.1, making a recompile even more 16913 beneficial. 16914 16915 MIPS 16916 16917 * Added support for the Broadcom SB-1A core. 16918 16919 IA-64 16920 16921 * Added support for IA-64 data and control speculation. By default 16922 speculation is enabled only during second scheduler pass. A number 16923 of machine flags was introduced to control the usage of speculation 16924 for both scheduler passes. 16925 16926 HPPA 16927 16928 * Added Java language support (libffi and libjava) for 32-bit HP-UX 16929 11 target. 16930 16931 [7]Obsolete Systems 16932 16933 Documentation improvements 16934 16935 PDF Documentation 16936 16937 * A make pdf target has been added to the top-level makefile, 16938 enabling automated production of PDF documentation files. 16939 (Front-ends external to GCC should modify their Make-lang.in file 16940 to add a lang.pdf: target.) 16941 16942 Other significant improvements 16943 16944 Build system improvements 16945 16946 * All the components of the compiler are now bootstrapped by default. 16947 This improves the resilience to bugs in the system compiler or 16948 binary compatibility problems, as well as providing better testing 16949 of GCC 4.2 itself. In addition, if you build the compiler from a 16950 combined tree, the assembler, linker, etc. will also be 16951 bootstrapped (i.e. built with themselves). 16952 You can disable this behavior, and go back to the pre-GCC 4.2 set 16953 up, by configuring GCC with --disable-bootstrap. 16954 * The rules that configure follows to find target tools resemble more 16955 closely the locations that the built compiler will search. In 16956 addition, you can use the new configure option --with-target-tools 16957 to specify where to find the target tools used during the build, 16958 without affecting what the built compiler will use. 16959 This can be especially useful when building packages of GCC. For 16960 example, you may want to build GCC with GNU as or ld, even if the 16961 resulting compiler to work with the native assembler and linker. To 16962 do so, you can use --with-target-tools to point to the native 16963 tools. 16964 16965 Incompatible changes to the build system 16966 16967 * Front-ends external to GCC should modify their Make-lang.in file to 16968 replace double-colon rules (e.g. dvi::) with normal rules (like 16969 lang.dvi:). Front-end makefile hooks do not use double-colon rules 16970 anymore. 16971 * Up to GCC 4.1, a popular way to specify the target tools used 16972 during the build was to create directories named gas, binutils, 16973 etc. in the build tree, and create links to the tools from there. 16974 This does not work any more when the compiler is bootstrapped. The 16975 new configure option --with-target-tools provides a better way to 16976 achieve the same effect, and works for all native and cross 16977 settings. 16978 16979 16980 For questions related to the use of GCC, please consult these web 16981 pages and the [8]GCC manuals. If that fails, the 16982 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16983 web pages and the development of GCC are welcome on our developer 16984 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 16985 archives. 16986 16987 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 16988 distribution of this entire article is permitted in any medium, 16989 provided this notice is preserved. 16990 16991 These pages are [13]maintained by the GCC team. Last modified 16992 2025-01-31. 16993 16994 References 16995 16996 1. https://gcc.gnu.org/projects/gomp/ 16997 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 16998 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 16999 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 17000 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 17001 6. https://gcc.gnu.org/gcc-4.2/changes.html#targets 17002 7. https://gcc.gnu.org/gcc-4.2/changes.html#obsolete_systems 17003 8. https://gcc.gnu.org/onlinedocs/ 17004 9. mailto:gcc-help (a] gcc.gnu.org 17005 10. mailto:gcc (a] gcc.gnu.org 17006 11. https://gcc.gnu.org/lists.html 17007 12. https://www.fsf.org/ 17008 13. https://gcc.gnu.org/about.html 17009 ====================================================================== 17010 http://gcc.gnu.org/gcc-4.1/index.html 17011 17012 GCC 4.1 Release Series 17013 17014 (This release series is no longer supported.) 17015 17016 February 13, 2007 17017 17018 The [1]GNU project and the GCC developers are pleased to announce the 17019 release of GCC 4.1.2. 17020 17021 This release is a bug-fix release, containing fixes for regressions in 17022 GCC 4.1.1 relative to previous releases of GCC. 17023 17024 Release History 17025 17026 GCC 4.1.2 17027 February 13, 2007 ([2]changes) 17028 17029 GCC 4.1.1 17030 May 24, 2006 ([3]changes) 17031 17032 GCC 4.1.0 17033 February 28, 2006 ([4]changes) 17034 17035 References and Acknowledgements 17036 17037 GCC used to stand for the GNU C Compiler, but since the compiler 17038 supports several other languages aside from C, it now stands for the 17039 GNU Compiler Collection. 17040 17041 A list of [5]successful builds is updated as new information becomes 17042 available. 17043 17044 The GCC developers would like to thank the numerous people that have 17045 contributed new features, improvements, bug fixes, and other changes as 17046 well as test results to GCC. This [6]amazing group of volunteers is 17047 what makes GCC successful. 17048 17049 For additional information about GCC please refer to the [7]GCC project 17050 web site or contact the [8]GCC development mailing list. 17051 17052 To obtain GCC please use [9]our mirror sites or [10]our version control 17053 system. 17054 17055 17056 For questions related to the use of GCC, please consult these web 17057 pages and the [11]GCC manuals. If that fails, the 17058 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17059 web pages and the development of GCC are welcome on our developer 17060 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 17061 archives. 17062 17063 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 17064 distribution of this entire article is permitted in any medium, 17065 provided this notice is preserved. 17066 17067 These pages are [16]maintained by the GCC team. Last modified 17068 2022-10-26. 17069 17070 References 17071 17072 1. http://www.gnu.org/ 17073 2. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 17074 3. https://gcc.gnu.org/gcc-4.1/changes.html 17075 4. https://gcc.gnu.org/gcc-4.1/changes.html 17076 5. https://gcc.gnu.org/gcc-4.1/buildstat.html 17077 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17078 7. https://gcc.gnu.org/index.html 17079 8. mailto:gcc (a] gcc.gnu.org 17080 9. https://gcc.gnu.org/mirrors.html 17081 10. https://gcc.gnu.org/git.html 17082 11. https://gcc.gnu.org/onlinedocs/ 17083 12. mailto:gcc-help (a] gcc.gnu.org 17084 13. mailto:gcc (a] gcc.gnu.org 17085 14. https://gcc.gnu.org/lists.html 17086 15. https://www.fsf.org/ 17087 16. https://gcc.gnu.org/about.html 17088 ====================================================================== 17089 http://gcc.gnu.org/gcc-4.1/changes.html 17090 17091 GCC 4.1 Release Series 17092 Changes, New Features, and Fixes 17093 17094 The latest release in the 4.1 release series is [1]GCC 4.1.2. 17095 17096 Caveats 17097 17098 General Optimizer Improvements 17099 17100 * GCC now has infrastructure for inter-procedural optimizations and 17101 the following inter-procedural optimizations are implemented: 17102 + Profile guided inlining. When doing profile feedback guided 17103 optimization, GCC can now use the profile to make better 17104 informed decisions on whether inlining of a function is 17105 profitable or not. This means that GCC will no longer inline 17106 functions at call sites that are not executed very often, and 17107 that functions at hot call sites are more likely to be 17108 inlined. 17109 A new parameter min-inline-recursive-probability is also now 17110 available to throttle recursive inlining of functions with 17111 small average recursive depths. 17112 + Discovery of pure and const functions, a form of side-effects 17113 analysis. While older GCC releases could also discover such 17114 special functions, the new IPA-based pass runs earlier so that 17115 the results are available to more optimizers. The pass is also 17116 simply more powerful than the old one. 17117 + Analysis of references to static variables and type escape 17118 analysis, also forms of side-effects analysis. The results of 17119 these passes allow the compiler to be less conservative about 17120 call-clobbered variables and references. This results in more 17121 redundant loads being eliminated and in making static 17122 variables candidates for register promotion. 17123 + Improvement of RTL-based alias analysis. The results of type 17124 escape analysis are fed to the RTL type-based alias analyzer, 17125 allowing it to disambiguate more memory references. 17126 + Interprocedural constant propagation and function versioning. 17127 This pass looks for functions that are always called with the 17128 same constant value for one or more of the function arguments, 17129 and propagates those constants into those functions. 17130 + GCC will now eliminate static variables whose usage was 17131 optimized out. 17132 + -fwhole-program --combine can now be used to make all 17133 functions in program static allowing whole program 17134 optimization. As an exception, the main function and all 17135 functions marked with the new externally_visible attribute are 17136 kept global so that programs can link with runtime libraries. 17137 * GCC can now do a form of partial dead code elimination (PDCE) that 17138 allows code motion of expressions to the paths where the result of 17139 the expression is actually needed. This is not always a win, so the 17140 pass has been limited to only consider profitable cases. Here is an 17141 example: 17142 int foo (int *, int *); 17143 int 17144 bar (int d) 17145 { 17146 int a, b, c; 17147 b = d + 1; 17148 c = d + 2; 17149 a = b + c; 17150 if (d) 17151 { 17152 foo (&b, &c); 17153 a = b + c; 17154 } 17155 printf ("%d\n", a); 17156 } 17157 17158 The a = b + c can be sunk to right before the printf. Normal code 17159 sinking will not do this, it will sink the first one above into the 17160 else-branch of the conditional jump, which still gives you two 17161 copies of the code. 17162 * GCC now has a value range propagation pass. This allows the 17163 compiler to eliminate bounds checks and branches. The results of 17164 the pass can also be used to accurately compute branch 17165 probabilities. 17166 * The pass to convert PHI nodes to straight-line code (a form of 17167 if-conversion for GIMPLE) has been improved significantly. The two 17168 most significant improvements are an improved algorithm to 17169 determine the order in which the PHI nodes are considered, and an 17170 improvement that allow the pass to consider if-conversions of basic 17171 blocks with more than two predecessors. 17172 * Alias analysis improvements. GCC can now differentiate between 17173 different fields of structures in Tree-SSA's virtual operands form. 17174 This lets stores/loads from non-overlapping structure fields not 17175 conflict. A new algorithm to compute points-to sets was contributed 17176 that can allows GCC to see now that p->a and p->b, where p is a 17177 pointer to a structure, can never point to the same field. 17178 * Various enhancements to auto-vectorization: 17179 + Incrementally preserve SSA form when vectorizing. 17180 + Incrementally preserve loop-closed form when vectorizing. 17181 + Improvements to peeling for alignment: generate better code 17182 when the misalignment of an access is known at compile time, 17183 or when different accesses are known to have the same 17184 misalignment, even if the misalignment amount itself is 17185 unknown. 17186 + Consider dependence distance in the vectorizer. 17187 + Externalize generic parts of data reference analysis to make 17188 this analysis available to other passes. 17189 + Vectorization of conditional code. 17190 + Reduction support. 17191 * GCC can now partition functions in sections of hot and cold code. 17192 This can significantly improve performance due to better 17193 instruction cache locality. This feature works best together with 17194 profile feedback driven optimization. 17195 * A new pass to avoid saving of unneeded arguments to the stack in 17196 vararg functions if the compiler can prove that they will not be 17197 needed. 17198 * Transition of basic block profiling to tree level implementation 17199 has been completed. The new implementation should be considerably 17200 more reliable (hopefully avoiding profile mismatch errors when 17201 using -fprofile-use or -fbranch-probabilities) and can be used to 17202 drive higher level optimizations, such as inlining. 17203 The -ftree-based-profiling command-line option was removed and 17204 -fprofile-use now implies disabling old RTL level loop optimizer 17205 (-fno-loop-optimize). Speculative prefetching optimization 17206 (originally enabled by -fspeculative-prefetching) was removed. 17207 17208 New Languages and Language specific improvements 17209 17210 C and Objective-C 17211 17212 * The old Bison-based C and Objective-C parser has been replaced by a 17213 new, faster hand-written recursive-descent parser. 17214 17215 Ada 17216 17217 * The build infrastructure for the Ada runtime library and tools has 17218 been changed to be better integrated with the rest of the build 17219 infrastructure of GCC. This should make doing cross builds of Ada a 17220 bit easier. 17221 17222 [2]C++ 17223 17224 * ARM-style name-injection of friend declarations is no longer the 17225 default. For example: 17226 struct S { 17227 friend void f(); 17228 }; 17229 17230 void g() { f(); } 17231 will not be accepted; instead a declaration of f will need to be 17232 present outside of the scope of S. The new -ffriend-injection 17233 option will enable the old behavior. 17234 * The (undocumented) extension which permitted templates with default 17235 arguments to be bound to template template parameters with fewer 17236 parameters has been deprecated, and will be removed in the next 17237 major release of G++. For example: 17238 template <template <typename> class C> 17239 void f(C<double>) {} 17240 17241 template <typename T, typename U = int> 17242 struct S {}; 17243 17244 template void f(S<double>); 17245 17246 makes use of the deprecated extension. The reason this code is not 17247 valid ISO C++ is that S is a template with two parameters; 17248 therefore, it cannot be bound to C which has only one parameter. 17249 17250 Runtime Library (libstdc++) 17251 17252 * Optimization work: 17253 + A new implementation of std::search_n is provided, better 17254 performing in case of random access iterators. 17255 + Added further efficient specializations of istream functions, 17256 i.e., character array and string extractors. 17257 + Other smaller improvements throughout. 17258 * Policy-based associative containers, designed for high-performance, 17259 flexibility and semantic safety are delivered in ext/pb_assoc. 17260 * A versatile string class, __gnu_cxx::__versa_string, providing 17261 facilities conforming to the standard requirements for 17262 basic_string, is delivered in <ext/vstring.h>. In particular: 17263 + Two base classes are provided: the default one avoids 17264 reference counting and is optimized for short strings; the 17265 alternate one, still uses it while improving in a few low 17266 level areas (e.g., alignment). See vstring_fwd.h for some 17267 useful typedefs. 17268 + Various algorithms have been rewritten (e.g., replace), the 17269 code streamlined and simple optimizations added. 17270 + Option 3 of DR 431 is implemented for both available bases, 17271 thus improving the support for stateful allocators. 17272 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 17273 libstdc++/23953) and LWG resolutions put into effect for the first 17274 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 17275 Issue 6.19). The implementation status of TR1 is now tracked in the 17276 docs in tr1.html. 17277 17278 Objective-C++ 17279 17280 * A new language front end for Objective-C++ has been added. This 17281 language allows users to mix the object oriented features of 17282 Objective-C with those of C++. 17283 17284 Java (GCJ) 17285 17286 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 17287 features (plus some 0.20 bug-fixes) 17288 + Networking 17289 o The java.net.HttpURLConnection implementation no longer 17290 buffers the entire response body in memory. This means 17291 that response bodies larger than available memory can now 17292 be handled. 17293 + (N)IO 17294 o NIO FileChannel.map implementation, fast bulk put 17295 implementation for DirectByteBuffer (speeds up this 17296 method 10x). 17297 o FileChannel.lock() and FileChannel.force() implemented. 17298 + XML 17299 o gnu.xml fix for nodes created outside a namespace 17300 context. 17301 o Add support for output indenting and 17302 cdata-section-elements output instruction in 17303 xml.transform. 17304 o xml.xpath corrections for cases where elements/attributes 17305 might have been created in non-namespace-aware mode. 17306 Corrections to handling of XSL variables and minor 17307 conformance updates. 17308 + AWT 17309 o GNU JAWT implementation, the AWT Native Interface, which 17310 allows direct access to native screen resources from 17311 within a Canvas's paint method. GNU Classpath Examples 17312 comes with a Demo, see libjava/classpath/examples/README. 17313 o awt.datatransfer updated to 1.5 with support for 17314 FlavorEvents. The gtk+ awt peers now allow copy/paste of 17315 text, images, URIs/files and serialized objects with 17316 other applications and tracking clipboard change events 17317 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 17318 objects are supported). A GNU Classpath Examples 17319 datatransfer Demo was added to show the new 17320 functionality. 17321 o Split gtk+ awt peers event handling in two threads and 17322 improve gdk lock handling (solves several awt lock ups). 17323 o Speed up awt Image loading. 17324 o Better gtk+ scrollbar peer implementation when using gtk+ 17325 >= 2.6. 17326 o Handle image loading errors correctly for gdkpixbuf and 17327 MediaTracker. 17328 o Better handle GDK lock. Properly prefix gtkpeer native 17329 functions (cp_gtk). 17330 o GdkGraphics2D has been updated to use Cairo 0.5.x or 17331 higher. 17332 o BufferedImage and GtkImage rewrites. All image drawing 17333 operations should now work correctly (flipping requires 17334 gtk+ >= 2.6) 17335 o When gtk+ 2.6 or higher is installed the default log 17336 handler will produce stack traces whenever a WARNING, 17337 CRITICAL or ERROR message is produced. 17338 + Free Swing 17339 o The RepaintManager has been reworked for more efficient 17340 painting, especially for large GUIs. 17341 o The layout manager OverlayLayout has been implemented, 17342 the BoxLayout has been rewritten to make use of the 17343 SizeRequirements utility class and caching for more 17344 efficient layout. 17345 o Improved accessibility support. 17346 o Significant progress has been made in the implementation 17347 of the javax.swing.plaf.metal package, with most UI 17348 delegates in a working state now. Please test this with 17349 your own applications and provide feedback that will help 17350 us to improve this package. 17351 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 17352 extended to highlight various features in our Free Swing 17353 implementation. And it includes a look and feel switcher 17354 for Metal (default), Ocean and GNU themes. 17355 o The javax.swing.plaf.multi package is now implemented. 17356 o Editing and several key actions for JTree and JTable were 17357 implemented. 17358 o Lots of icons and look and feel improvements for Free 17359 Swing basic and metal themes were added. Try running the 17360 GNU Classpath Swing Demo in examples 17361 (gnu.classpath.examples.swing.Demo) with: 17362 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 17363 l or 17364 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 17365 l 17366 o Start of styled text capabilites for java.swing.text. 17367 o DefaultMutableTreeNode pre-order, post-order, depth-first 17368 and breadth-first traversal enumerations implemented. 17369 o JInternalFrame colors and titlebar draw properly. 17370 o JTree is working up to par (icons, selection and keyboard 17371 traversal). 17372 o JMenus were made more compatible in visual and 17373 programmatic behavior. 17374 o JTable changeSelection and multiple selections 17375 implemented. 17376 o JButton and JToggleButton change states work properly 17377 now. 17378 o JFileChooser fixes. 17379 o revalidate() and repaint() fixes which make Free Swing 17380 much more responsive. 17381 o MetalIconFactory implemented. 17382 o Free Swing Top-Level Compatibility. JFrame, JDialog, 17383 JApplet, JInternalFrame, and JWindow are now 1.5 17384 compatible in the sense that you can call add() and 17385 setLayout() directly on them, which will have the same 17386 effect as calling getContentPane().add() and 17387 getContentPane().setLayout(). 17388 o The JTree interface has been completed. JTrees now 17389 recognizes mouse clicks and selections work. 17390 o BoxLayout works properly now. 17391 o Fixed GrayFilter to actually work. 17392 o Metal SplitPane implemented. 17393 o Lots of Free Swing text and editor stuff work now. 17394 + Free RMI and Corba 17395 o Andrew Watson, Vice President and Technical Director of 17396 the Object Management Group, has officially assigned us 17397 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 17398 mark remote classpath-specific system exceptions. 17399 Obtaining the VMCID means that GNU Classpath now is a 17400 recogniseable type of node in a highly interoperable 17401 CORBA world. 17402 o GNU Classpath now includes the first working draft to 17403 support the RMI over IIOP protocol. The current 17404 implementation is capable of remote invocations, 17405 transferring various Serializables and Externalizables 17406 via RMI-IIOP protocol. It can flatten graphs and, at 17407 least for the simple cases, is interoperable with 1.5 17408 JDKs. 17409 o org.omg.PortableInterceptor and related functionality in 17410 other packages is now implemented: 17411 # The sever and client interceptors work as required 17412 since 1.4. 17413 # The IOR interceptor works as needed for 1.5. 17414 o The org.omg.DynamicAny package is completed and passes 17415 the prepared tests. 17416 o The Portable Object Adapter should now support the output 17417 of the recent IDL to java compilers. These compilers now 17418 generate servants and not CORBA objects as before, making 17419 the output depend on the existing POA implementation. 17420 Completing POA means that such code can already be tried 17421 to run on Classpath. Our POA is tested for the following 17422 usager scenarios: 17423 # POA converts servant to the CORBA object. 17424 # Servant provides to the CORBA object. 17425 # POA activates new CORBA object with the given Object 17426 Id (byte array) that is later accessible for the 17427 servant. 17428 # During the first call, the ServantActivator provides 17429 servant for this and all subsequent calls on the 17430 current object. 17431 # During each call, the ServantLocator provides 17432 servant for this call only. 17433 # ServantLocator or ServantActivator forwards call to 17434 another server. 17435 # POA has a single servant, responsible for all 17436 objects. 17437 # POA has a default servant, but some objects are 17438 explicitly connected to they specific servants. 17439 The POA is verified using tests from the former 17440 cost.omg.org. 17441 o The CORBA implementation is now a working prototype that 17442 should support features up to 1.3 inclusive. We invite 17443 groups writing CORBA dependent applications to try 17444 Classpath implementation, reporting any possible bugs. 17445 The CORBA prototype is interoperable with Sun's 17446 implementation v 1.4, transferring object references, 17447 primitive types, narrow and wide strings, arrays, 17448 structures, trees, abstract interfaces and value types 17449 (feature of CORBA 2.3) between these two platforms. 17450 Remote exceptions are transferred and handled correctly. 17451 The stringified object references (IORs) from various 17452 sources are parsed as required. The transient (for 17453 current session) and permanent (till jre restart) 17454 redirections work. Both Little and Big Endian encoded 17455 messages are accepted. The implementation is verified 17456 using tests from the former cost.omg.org. The current 17457 release includes working examples (see the examples 17458 directory), demonstrating the client-server 17459 communication, using either CORBA Request or IDL-based 17460 stub (usually generated by a IDL to java compiler). These 17461 examples also show how to use the Classpath CORBA naming 17462 service. The IDL to java compiler is not yet written, but 17463 as our library must be compatible, it naturally accepts 17464 the output of other idlj implementations. 17465 + Misc 17466 o Updated TimeZone data against Olson tzdata2005l. 17467 o Make zip and jar packages UTF-8 clean. 17468 o "native" code builds and compiles (warning free) on 17469 Darwin and Solaris. 17470 o java.util.logging.FileHandler now rotates files. 17471 o Start of a generic JDWP framework in gnu/classpath/jdwp. 17472 This is unfinished, but feedback (at classpath (a] gnu.org) 17473 from runtime hackers is greatly appreciated. Although 17474 most of the work is currently being done around gcj/gij 17475 we want this framework to be as VM neutral as possible. 17476 Early design is described in: 17477 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 17478 o QT4 AWT peers, enable by giving configure 17479 --enable-qt-peer. Included, but not ready for production 17480 yet. They are explicitly disabled and not supported. But 17481 if you want to help with the development of these new 17482 features we are interested in feedback. You will have to 17483 explicitly enable them to try them out (and they will 17484 most likely contain bugs). 17485 o Documentation fixes all over the place. See 17486 [4]https://developer.classpath.org/doc/ 17487 17488 [5]New Targets and Target Specific Improvements 17489 17490 IA-32/x86-64 17491 17492 * The x86-64 medium model (that allows building applications whose 17493 data segment exceeds 4GB) was redesigned to match latest ABI draft. 17494 New implementation split large datastructures into separate segment 17495 improving performance of accesses to small datastructures and also 17496 allows linking of small model libraries into medium model programs 17497 as long as the libraries are not accessing the large datastructures 17498 directly. Medium model is also supported in position independent 17499 code now. 17500 The ABI change results in partial incompatibility among medium 17501 model objects. Linking medium model libraries (or objects) compiled 17502 with new compiler into medium model program compiled with older 17503 will likely result in exceeding ranges of relocations. 17504 Binutils 2.16.91 or newer are required for compiling medium model 17505 now. 17506 17507 RS6000 (POWER/PowerPC) 17508 17509 * The AltiVec vector primitives in <altivec.h> are now implemented in 17510 a way that puts a smaller burden on the preprocessor, instead 17511 processing the "overloading" in the front ends. This should benefit 17512 compilation speed on AltiVec vector code. 17513 * AltiVec initializers now are generated more efficiently. 17514 * The popcountb instruction available on POWER5 now is generated. 17515 * The floating point round to integer instructions available on 17516 POWER5+ now is generated. 17517 * Floating point divides can be synthesized using the floating point 17518 reciprocal estimate instructions. 17519 * Double precision floating point constants are initialized as single 17520 precision values if they can be represented exactly. 17521 17522 S/390, zSeries and System z9 17523 17524 * Support for the IBM System z9 109 processor has been added. When 17525 using the -march=z9-109 option, the compiler will generate code 17526 making use of instructions provided by the extended immediate 17527 facility. 17528 * Support for 128-bit IEEE floating point has been added. When using 17529 the -mlong-double-128 option, the compiler will map the long double 17530 data type to 128-bit IEEE floating point. Using this option 17531 constitutes an ABI change, and requires glibc support. 17532 * Various changes to improve performance of generated code have been 17533 implemented, including: 17534 + In functions that do not require a literal pool, register %r13 17535 (which is traditionally reserved as literal pool pointer), can 17536 now be freely used for other purposes by the compiler. 17537 + More precise tracking of register use allows the compiler to 17538 generate more efficient function prolog and epilog code in 17539 certain cases. 17540 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 17541 instructions are now used to implement C string functions. 17542 + The MOVE CHARACTER instruction with single byte overlap is now 17543 used to implement the memset function with non-zero fill byte. 17544 + The LOAD ZERO instructions are now used where appropriate. 17545 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 17546 and INSERT IMMEDIATE instructions are now used more frequently 17547 to optimize bitfield operations. 17548 + The BRANCH ON COUNT instruction is now used more frequently. 17549 In particular, the fact that a loop contains a subroutine call 17550 no longer prevents the compiler from using this instruction. 17551 + The compiler is now aware that all shift and rotate 17552 instructions implicitly truncate the shift count to six bits. 17553 * Back-end support for the following generic features has been 17554 implemented: 17555 + The full set of [6]built-in functions for atomic memory 17556 access. 17557 + The -fstack-protector feature. 17558 + The optimization pass avoiding unnecessary stores of incoming 17559 argument registers in functions with variable argument list. 17560 17561 SPARC 17562 17563 * The default code model in 64-bit mode has been changed from 17564 Medium/Anywhere to Medium/Middle on Solaris. 17565 * TLS support is disabled by default on Solaris prior to release 10. 17566 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 17567 and later) by specifying --enable-tls at configure time. 17568 17569 MorphoSys 17570 17571 * Support has been added for this new architecture. 17572 17573 [7]Obsolete Systems 17574 17575 Documentation improvements 17576 17577 Other significant improvements 17578 17579 * GCC can now emit code for protecting applications from 17580 stack-smashing attacks. The protection is realized by buffer 17581 overflow detection and reordering of stack variables to avoid 17582 pointer corruption. 17583 * Some built-in functions have been fortified to protect them against 17584 various buffer overflow (and format string) vulnerabilities. 17585 Compared to the mudflap bounds checking feature, the safe builtins 17586 have far smaller overhead. This means that programs built using 17587 safe builtins should not experience any measurable slowdown. 17588 17589 [8]GCC 4.1.2 17590 17591 This is the [9]list of problem reports (PRs) from GCC's bug tracking 17592 system that are known to be fixed in the 4.1.2 release. This list might 17593 not be complete (that is, it is possible that some PRs that have been 17594 fixed are not listed here). 17595 17596 When generating code for a shared library, GCC now recognizes that 17597 global functions may be replaced when the program runs. Therefore, it 17598 is now more conservative in deducing information from the bodies of 17599 functions. For example, in this example: 17600 void f() {} 17601 void g() { 17602 try { f(); } 17603 catch (...) { 17604 cout << "Exception"; 17605 } 17606 } 17607 17608 G++ would previously have optimized away the catch clause, since it 17609 would have concluded that f cannot throw exceptions. Because users may 17610 replace f with another function in the main body of the program, this 17611 optimization is unsafe, and is no longer performed. If you wish G++ to 17612 continue to optimize as before, you must add a throw() clause to the 17613 declaration of f to make clear that it does not throw exceptions. 17614 17615 17616 For questions related to the use of GCC, please consult these web 17617 pages and the [10]GCC manuals. If that fails, the 17618 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17619 web pages and the development of GCC are welcome on our developer 17620 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 17621 archives. 17622 17623 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 17624 distribution of this entire article is permitted in any medium, 17625 provided this notice is preserved. 17626 17627 These pages are [15]maintained by the GCC team. Last modified 17628 2025-01-31. 17629 17630 References 17631 17632 1. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 17633 2. https://gcc.gnu.org/gcc-4.1/changes.html#cplusplus 17634 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 17635 4. https://developer.classpath.org/doc/ 17636 5. https://gcc.gnu.org/gcc-4.1/changes.html#targets 17637 6. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 17638 7. https://gcc.gnu.org/gcc-4.1/changes.html#obsolete_systems 17639 8. https://gcc.gnu.org/gcc-4.1/changes.html#GCC4.1.2 17640 9. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 17641 10. https://gcc.gnu.org/onlinedocs/ 17642 11. mailto:gcc-help (a] gcc.gnu.org 17643 12. mailto:gcc (a] gcc.gnu.org 17644 13. https://gcc.gnu.org/lists.html 17645 14. https://www.fsf.org/ 17646 15. https://gcc.gnu.org/about.html 17647 ====================================================================== 17648 http://gcc.gnu.org/gcc-4.0/index.html 17649 17650 GCC 4.0 Release Series 17651 17652 (This release series is no longer supported.) 17653 17654 January 31, 2007 17655 17656 The [1]GNU project and the GCC developers are pleased to announce the 17657 release of GCC 4.0.4. 17658 17659 This release is a bug-fix release, containing fixes for regressions in 17660 GCC 4.0.3 relative to previous releases of GCC. 17661 17662 Release History 17663 17664 GCC 4.0.4 17665 January 31, 2007 ([2]changes) 17666 17667 GCC 4.0.3 17668 March 10, 2006 ([3]changes) 17669 17670 GCC 4.0.2 17671 September 28, 2005 ([4]changes) 17672 17673 GCC 4.0.1 17674 July 7, 2005 ([5]changes) 17675 17676 GCC 4.0.0 17677 April 20, 2005 ([6]changes) 17678 17679 References and Acknowledgements 17680 17681 GCC used to stand for the GNU C Compiler, but since the compiler 17682 supports several other languages aside from C, it now stands for the 17683 GNU Compiler Collection. 17684 17685 A list of [7]successful builds is updated as new information becomes 17686 available. 17687 17688 The GCC developers would like to thank the numerous people that have 17689 contributed new features, improvements, bug fixes, and other changes as 17690 well as test results to GCC. This [8]amazing group of volunteers is 17691 what makes GCC successful. 17692 17693 For additional information about GCC please refer to the [9]GCC project 17694 web site or contact the [10]GCC development mailing list. 17695 17696 To obtain GCC please use [11]our mirror sites, or [12]our version 17697 control system. 17698 17699 17700 For questions related to the use of GCC, please consult these web 17701 pages and the [13]GCC manuals. If that fails, the 17702 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17703 web pages and the development of GCC are welcome on our developer 17704 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 17705 archives. 17706 17707 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 17708 distribution of this entire article is permitted in any medium, 17709 provided this notice is preserved. 17710 17711 These pages are [18]maintained by the GCC team. Last modified 17712 2022-10-26. 17713 17714 References 17715 17716 1. http://www.gnu.org/ 17717 2. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 17718 3. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 17719 4. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 17720 5. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 17721 6. https://gcc.gnu.org/gcc-4.0/changes.html 17722 7. https://gcc.gnu.org/gcc-4.0/buildstat.html 17723 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17724 9. https://gcc.gnu.org/index.html 17725 10. mailto:gcc (a] gcc.gnu.org 17726 11. https://gcc.gnu.org/mirrors.html 17727 12. https://gcc.gnu.org/git.html 17728 13. https://gcc.gnu.org/onlinedocs/ 17729 14. mailto:gcc-help (a] gcc.gnu.org 17730 15. mailto:gcc (a] gcc.gnu.org 17731 16. https://gcc.gnu.org/lists.html 17732 17. https://www.fsf.org/ 17733 18. https://gcc.gnu.org/about.html 17734 ====================================================================== 17735 http://gcc.gnu.org/gcc-4.0/changes.html 17736 17737 GCC 4.0 Release Series 17738 Changes, New Features, and Fixes 17739 17740 The latest release in the 4.0 release series is [1]GCC 4.0.4. 17741 17742 Caveats 17743 17744 * GCC now generates location lists by default when compiling with 17745 debug info and optimization. 17746 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 17747 or later is needed to debug binaries containing location 17748 lists. 17749 + When you are trying to view a value of a variable in a part of 17750 a function where it has no location (for example when the 17751 variable is no longer used and thus its location was used for 17752 something else) GDB will say that it is not available. 17753 You can disable generating location lists by -fno-var-tracking. 17754 * GCC no longer accepts the -fwritable-strings option. Use named 17755 character arrays when you need a writable string. 17756 * The options -freduce-all-givs and -fmove-all-movables have been 17757 discontinued. They were used to circumvent a shortcoming in the 17758 heuristics of the old loop optimization code with respect to common 17759 Fortran constructs. The new (tree) loop optimizer works differently 17760 and doesn't need those work-arounds. 17761 * The graph-coloring register allocator, formerly enabled by the 17762 option -fnew-ra, has been discontinued. 17763 * -I- has been deprecated. -iquote is meant to replace the need for 17764 this option. 17765 * The MIPS -membedded-pic and -mrnames options have been removed. 17766 * All MIPS targets now require the GNU assembler. In particular, IRIX 17767 configurations can no longer use the MIPSpro assemblers, although 17768 they do still support the MIPSpro linkers. 17769 * The SPARC option -mflat has been removed. 17770 * English-language diagnostic messages will now use Unicode quotation 17771 marks in UTF-8 locales. (Non-English messages already used the 17772 quotes appropriate for the language in previous releases.) If your 17773 terminal does not support UTF-8 but you are using a UTF-8 locale 17774 (such locales are the default on many GNU/Linux systems) then you 17775 should set LC_CTYPE=C in the environment to disable that locale. 17776 Programs that parse diagnostics and expect plain ASCII 17777 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 17778 explanation of Unicode quotation marks for more information. 17779 * The specs file is no longer installed on most platforms. Most users 17780 will be totally unaffected. However, if you are accustomed to 17781 editing the specs file yourself, you will now have to use the 17782 -dumpspecs option to generate the specs file, and then edit the 17783 resulting file. 17784 17785 General Optimizer Improvements 17786 17787 * The [3]tree ssa branch has been merged. This merge has brought in a 17788 completely new optimization framework based on a higher level 17789 intermediate representation than the existing RTL representation. 17790 Numerous new code transformations based on the new framework are 17791 available in GCC 4.0, including: 17792 + Scalar replacement of aggregates 17793 + Constant propagation 17794 + Value range propagation 17795 + Partial redundancy elimination 17796 + Load and store motion 17797 + Strength reduction 17798 + Dead store elimination 17799 + Dead and unreachable code elimination 17800 + [4]Autovectorization 17801 + Loop interchange 17802 + Tail recursion by accumulation 17803 Many of these passes outperform their counterparts from previous 17804 GCC releases. 17805 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 17806 scheduling optimization intended for loops that perform heavy 17807 computations. 17808 17809 New Languages and Language specific improvements 17810 17811 C family 17812 17813 * The sentinel attribute has been added to GCC. This function 17814 attribute allows GCC to warn when variadic functions such as execl 17815 are not NULL terminated. See the GCC manual for a complete 17816 description of its behavior. 17817 * Given __attribute__((alias("target"))) it is now an error if target 17818 is not a symbol, defined in the same translation unit. This also 17819 applies to aliases created by #pragma weak alias=target. This is 17820 because it's meaningless to define an alias to an undefined symbol. 17821 On Solaris, the native assembler would have caught this error, but 17822 GNU as does not. 17823 17824 C and Objective-C 17825 17826 * The -Wstrict-aliasing=2 option has been added. This warning catches 17827 all unsafe cases, but it may also give a warning for some cases 17828 that are safe. 17829 * The cast-as-lvalue, conditional-expression-as-lvalue and 17830 compound-expression-as-lvalue extensions, which were deprecated in 17831 3.3.4 and 3.4, have been removed. 17832 * The -fwritable-strings option, which was deprecated in 3.4, has 17833 been removed. 17834 * #pragma pack() semantics have been brought closer to those used by 17835 other compilers. This also applies to C++. 17836 * Taking the address of a variable with register storage is invalid 17837 in C. GCC now issues an error instead of a warning. 17838 * Arrays of incomplete element type are invalid in C. GCC now issues 17839 an error for such arrays. Declarations such as extern struct s x[]; 17840 (where struct s has not been defined) can be moved after the 17841 definition of struct s. Function parameters declared as arrays of 17842 incomplete type can instead be declared as pointers. 17843 17844 C++ 17845 17846 * When compiling without optimizations (-O0), the C++ front end is 17847 much faster than in any previous versions of GCC. Independent 17848 testers have measured speed-ups up to 25% in real-world production 17849 code, compared to the 3.4 family (which was already the fastest 17850 version to date). Upgrading from older versions might show even 17851 bigger improvements. 17852 * ELF visibility attributes can now be applied to a class type, so 17853 that it affects every member function of a class at once, without 17854 having to specify each individually: 17855 class __attribute__ ((visibility("hidden"))) Foo 17856 { 17857 int foo1(); 17858 void foo2(); 17859 }; 17860 The syntax is deliberately similar to the __declspec() system used 17861 by Microsoft Windows based compilers, allowing cross-platform 17862 projects to easily reuse their existing macro system for denoting 17863 exports and imports. By explicitly marking internal classes never 17864 used outside a binary as hidden, one can completely avoid PLT 17865 indirection overheads during their usage by the compiler. You can 17866 find out more about the advantages of this at 17867 [6]https://www.akkadia.org/drepper/dsohowto.pdf 17868 * The -fvisibility-inlines-hidden option has been added which marks 17869 all inlineable functions as having hidden ELF visibility, thus 17870 removing their symbol and typeinfo from the exported symbol table 17871 of the output ELF binary. Using this option can reduce the exported 17872 symbol count of template-heavy code by up to 40% with no code 17873 change at all, thus notably improving link and load times for the 17874 binary as well as a reduction in size of up to 10%. Also, check the 17875 new [7]-fvisibility option. 17876 * The compiler now uses the library interface specified by the [8]C++ 17877 ABI for thread-safe initialization of function-scope static 17878 variables. Most users should leave this alone, but embedded 17879 programmers may want to disable this by specifying 17880 -fno-threadsafe-statics for a small savings in code size. 17881 * Taking the address of an explicit register variable is no longer 17882 supported. Note that C++ allows taking the address of variables 17883 with register storage so this will continue to compile with a 17884 warning. For example, assuming that r0 is a machine register: 17885 register int foo asm ("r0"); 17886 register int bar; 17887 &foo; // error, no longer accepted 17888 &bar; // OK, with a warning 17889 * G++ has an undocumented extension to virtual function covariancy 17890 rules that allowed the overrider to return a type that was 17891 implicitly convertable to the overridden function's return type. 17892 For instance a function returning void * could be overridden by a 17893 function returning T *. This is now deprecated and will be removed 17894 in a future release. 17895 * The G++ minimum and maximum operators (<? and >?) and their 17896 compound forms (<?=) and >?=) have been deprecated and will be 17897 removed in a future version. Code using these operators should be 17898 modified to use std::min and std::max instead. 17899 * Declaration of nested classes of class templates as friends are 17900 supported: 17901 template <typename T> struct A { 17902 class B {}; 17903 }; 17904 class C { 17905 template <typename T> friend class A<T>::B; 17906 }; 17907 This complements the feature member functions of class templates as 17908 friends introduced in GCC 3.4.0. 17909 * When declaring a friend class using an unqualified name, classes 17910 outside the innermost non-class scope are not searched: 17911 class A; 17912 namespace N { 17913 class B { 17914 friend class A; // Refer to N::A which has not been declared yet 17915 // because name outside namespace N are not searched 17916 friend class ::A; // Refer to ::A 17917 }; 17918 } 17919 Hiding the friend name until declaration is still not implemented. 17920 * Friends of classes defined outside their namespace are correctly 17921 handled: 17922 namespace N { 17923 class A; 17924 } 17925 class N::A { 17926 friend class B; // Refer to N::B in GCC 4.0.0 17927 // but ::B in earlier versions of GCC 17928 }; 17929 17930 Runtime Library (libstdc++) 17931 17932 * Optimization work: 17933 + Added efficient specializations of istream functions for char 17934 and wchar_t. 17935 + Further performance tuning of strings, in particular wrt 17936 single-char append and getline. 17937 + iter_swap - and therefore most of the mutating algorithms - 17938 now makes an unqualified call to swap when the value_type of 17939 the two iterators is the same. 17940 * A large subset of the features in Technical Report 1 (TR1 for 17941 short) is experimentally delivered (i.e., no guarantees about the 17942 implementation are provided. In particular it is not promised that 17943 the library will remain link-compatible when code using TR1 is 17944 used): 17945 + General utilities such as reference_wrapper and shared_ptr. 17946 + Function objects, i.e., result_of, mem_fn, bind, function. 17947 + Support for metaprogramming. 17948 + New containers such as tuple, array, unordered_set, 17949 unordered_map, unordered_multiset, unordered_multimap. 17950 * As usual, many bugs have been fixed and LWG resolutions implemented 17951 for the first time (e.g., DR 409). 17952 17953 Java 17954 17955 * In order to prevent naming conflicts with other implementations of 17956 these tools, some GCJ binaries have been renamed: 17957 + rmic is now grmic, 17958 + rmiregistry is now grmiregistry, and 17959 + jar is now fastjar. 17960 In particular, these names were problematic for the jpackage.org 17961 packaging conventions which install symlinks in /usr/bin that point 17962 to the preferred versions of these tools. 17963 * The -findirect-dispatch argument to the compiler now works and 17964 generates code following a new "binary compatibility" ABI. Code 17965 compiled this way follows the binary compatibility rules of the 17966 Java Language Specification. 17967 * libgcj now has support for using GCJ as a JIT, using the 17968 gnu.gcj.jit family of system properties. 17969 * libgcj can now find a shared library corresponding to the bytecode 17970 representation of a class. See the documentation for the new 17971 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 17972 property. 17973 * There have been many improvements to the class library. Here are 17974 some highlights: 17975 + Much more of AWT and Swing exist. 17976 + Many new packages and classes were added, including 17977 java.util.regex, java.net.URI, javax.crypto, 17978 javax.crypto.interfaces, javax.crypto.spec, javax.net, 17979 javax.net.ssl, javax.security.auth, 17980 javax.security.auth.callback, javax.security.auth.login, 17981 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 17982 javax.imageio, javax.imageio.event, javax.imageio.spi, 17983 javax.print, javax.print.attribute, 17984 javax.print.attribute.standard, javax.print.event, and 17985 javax.xml 17986 + Updated SAX and DOM, and imported GNU JAXP 17987 17988 Fortran 17989 17990 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 17991 front end. The new front end supports Fortran 90 and Fortran 95. It 17992 may not yet be as stable as the old Fortran front end. 17993 17994 Ada 17995 17996 * Ada (with tasking and Zero Cost Exceptions) is now available on 17997 many more targets, including but not limited to: alpha-linux, 17998 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 17999 s390x-linux, sparc-linux. 18000 * Some of the new Ada 2005 features are now implemented like 18001 Wide_Wide_Character and Ada.Containers. 18002 * Many bugs have been fixed, tools and documentation improved. 18003 * To compile Ada from the sources, install an older working Ada 18004 compiler and then use --enable-languages=ada at configuration time, 18005 since the Ada front end is not currently activated by default. See 18006 the [10]Installing GCC for details. 18007 18008 [11]New Targets and Target Specific Improvements 18009 18010 H8/300 18011 18012 * The frame layout has changed. In the new layout, the prologue of a 18013 function first saves registers and then allocate space for locals, 18014 resulting in an 1% improvement on code size. 18015 18016 IA-32/x86-64 (AMD64) 18017 18018 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 18019 log1p, log2, logb and tan mathematical builtins (and their float 18020 and long double variants) are now implemented as inline x87 18021 intrinsics when using -ffast-math. 18022 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 18023 (and their float and long double variants) are now implemented as 18024 inline x87 intrinsics when using -ffast-math. 18025 * The x87's fsincos instruction is now used automatically with 18026 -ffast-math when calculating both the sin and cos of the same 18027 argument. 18028 * Instruction selection for multiplication and division by constants 18029 has been improved. 18030 18031 IA-64 18032 18033 * Floating point division, integer division and sqrt are now inlined, 18034 resulting in significant performance improvements on some codes. 18035 18036 MIPS 18037 18038 * Division by zero checks now use conditional traps if the target 18039 processor supports them. This decreases code size by one word per 18040 division operation. The old behavior (branch and break) can be 18041 obtained either at configure time by passing --with-divide=breaks 18042 to configure or at runtime by passing -mdivide-breaks to GCC. 18043 * Support for MIPS64 paired-single instructions has been added. It is 18044 enabled by -mpaired-single and can be accessed using both the 18045 target-independent vector extensions and new MIPS-specific built-in 18046 functions. 18047 * Support for the MIPS-3D ASE has been added. It is enabled by 18048 -mips3d and provides new MIPS-3D-specific built-in functions. 18049 * The -mexplicit-relocs option now supports static n64 code (as is 18050 used, for example, in 64-bit linux kernels). -mexplicit-relocs 18051 should now be feature-complete and is enabled by default when GCC 18052 is configured to use a compatible assembler. 18053 * Support for the NEC VR4130 series has been added. This support 18054 includes the use of VR-specific instructions and a new VR4130 18055 scheduler. Full VR4130 support can be selected with -march=vr4130 18056 while code for any ISA can be tuned for the VR4130 using 18057 -mtune=vr4130. There is also a new -mvr4130-align option that 18058 produces better schedules at the cost of increased code size. 18059 * Support for the Broadcom SB-1 has been extended. There is now an 18060 SB-1 scheduler as well as support for the SB-1-specific 18061 paired-single instructions. Full SB-1 support can be selected with 18062 -march=sb1 while code for any ISA can be optimized for the SB-1 18063 using -mtune=sb1. 18064 * The compiler can now work around errata in R4000, R4400, VR4120 and 18065 VR4130 processors. These workarounds are enabled by -mfix-r4000, 18066 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 18067 and VR4130 workarounds need binutils 2.16 or above. 18068 * IRIX shared libraries are now installed into the standard library 18069 directories: o32 libraries go into lib/, n32 libraries go into 18070 lib32/ and n64 libraries go into lib64/. 18071 * The compiler supports a new -msym32 option. It can be used to 18072 optimize n64 code in which all symbols are known to have 32-bit 18073 values. 18074 18075 S/390 and zSeries 18076 18077 * New command-line options help to generate code intended to run in 18078 an environment where stack space is restricted, e.g. Linux kernel 18079 code: 18080 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 18081 warnings for single functions that require large or dynamic 18082 stack frames. 18083 + -mstack-size and -mstack-guard generate code that checks for 18084 stack overflow at run time. 18085 + -mpacked-stack generates code that reduces the stack frame 18086 size of many functions by reusing unneeded parts of the stack 18087 bias area. 18088 * The -msoft-float option now ensures that generated code never 18089 accesses floating point registers. 18090 * The s390x-ibm-tpf target now fully supports C++, including 18091 exceptions and threads. 18092 * Various changes to improve performance of the generated code have 18093 been implemented, including: 18094 + GCC now uses sibling calls where possible. 18095 + Condition code handling has been optimized, allowing GCC to 18096 omit redundant comparisons in certain cases. 18097 + The cost function guiding many optimizations has been refined 18098 to more accurately represent the z900 and z990 processors. 18099 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 18100 instructions are now used to avoid conditional branches in 18101 certain cases. 18102 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 18103 optimize address arithmetic required to access large stack 18104 frames. 18105 + GCC now makes more efficient use of memory-to-memory type 18106 instructions (MVC, CLC, ...). 18107 + More precise tracking of special register use allows better 18108 instruction scheduling, in particular of the function prologue 18109 and epilogue sequences. 18110 + The Java front end now generates inline code to implement 18111 integer division, instead of calling library routines. 18112 18113 SPARC 18114 18115 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 18116 -mf934 have been removed. They have been replaced with -mcpu=xxx. 18117 * The internal model used to estimate the relative cost of each 18118 instruction has been updated. It is expected to give better results 18119 on recent UltraSPARC processors. 18120 * Code generation for function prologues and epilogues has been 18121 improved, resulting in better scheduling and allowing multiple exit 18122 points in functions. 18123 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 18124 It is enabled by -mvis and provides new built-in functions for VIS 18125 instructions on UltraSPARC processors. 18126 * The option -mapp-regs has been turned on by default on Solaris too. 18127 18128 NetWare 18129 18130 * Novell NetWare (on ix86, no other hardware platform was ever really 18131 supported by this OS) has been re-enabled and the ABI supported by 18132 GCC has been brought into sync with that of MetroWerks CodeWarrior 18133 (the ABI previously supported was that of some Unix systems, which 18134 NetWare never tried to support). 18135 18136 [12]Obsolete Systems 18137 18138 Support for a number of older systems has been declared obsolete in GCC 18139 4.0. Unless there is activity to revive them, the next release of GCC 18140 will have their sources permanently removed. 18141 18142 All GCC ports for the following processor architectures have been 18143 declared obsolete: 18144 * Intel i860 18145 * Ubicom IP2022 18146 * National Semiconductor NS32K (ns32k) 18147 * Texas Instruments TMS320C[34]x 18148 18149 Also, those for some individual systems have been obsoleted: 18150 * SPARC family 18151 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 18152 sparc86x-*-elf) 18153 + OpenBSD 32-bit (sparc-*-openbsd*) 18154 18155 Documentation improvements 18156 18157 Other significant improvements 18158 18159 * Location lists are now generated by default when compiling with 18160 debug info and optimization. Location lists provide more accurate 18161 debug info about locations of variables and they allow debugging 18162 code compiled with -fomit-frame-pointer. 18163 * The -fvisibility option has been added which allows the default ELF 18164 visibility of all symbols to be set per compilation and the new 18165 #pragma GCC visibility preprocessor command allows the setting of 18166 default ELF visibility for a region of code. Using 18167 -fvisibility=hidden especially in combination with the new 18168 -fvisibility-inlines-hidden can yield substantial improvements in 18169 output binary quality including avoiding PLT indirection overheads, 18170 reduction of the exported symbol count by up to 60% (with resultant 18171 improvements to link and load times), better scope for the 18172 optimizer to improve code and up to a 20% reduction in binary size. 18173 Using these options correctly yields a binary with a similar symbol 18174 count to a Windows DLL. 18175 Perhaps more importantly, this new feature finally allows (with 18176 careful planning) complete avoidance of symbol clashes when 18177 manually loading shared objects with RTLD_GLOBAL, thus finally 18178 solving problems many projects such as python were forced to use 18179 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 18180 can find more information about using these options at 18181 [13]https://gcc.gnu.org/wiki/Visibility. 18182 __________________________________________________________________ 18183 18184 [14]GCC 4.0.1 18185 18186 This is the [15]list of problem reports (PRs) from GCC's bug tracking 18187 system that are known to be fixed in the 4.0.1 release. This list might 18188 not be complete (that is, it is possible that some PRs that have been 18189 fixed are not listed here). 18190 18191 [16]GCC 4.0.2 18192 18193 This is the [17]list of problem reports (PRs) from GCC's bug tracking 18194 system that are known to be fixed in the 4.0.2 release. This list might 18195 not be complete (that is, it is possible that some PRs that have been 18196 fixed are not listed here). 18197 18198 Unfortunately, due to a release engineering failure, this release has a 18199 regression on Solaris that will affect some C++ programs. We suggest 18200 that Solaris users apply a [18]patch that corrects the problem. Users 18201 who do not wish to apply the patch should explicitly link C++ programs 18202 with the -pthreads option, even if they do not use threads. This 18203 problem has been corrected in the current 4.0 branch sources and will 18204 not be present in GCC 4.0.3. 18205 18206 [19]GCC 4.0.3 18207 18208 Starting with this release, the function getcontext is recognized by 18209 the compiler as having the same semantics as the setjmp function. In 18210 particular, the compiler will ensure that all registers are dead before 18211 calling such a function and will emit a warning about the variables 18212 that may be clobbered after the second return from the function. 18213 18214 [20]GCC 4.0.4 18215 18216 This is the [21]list of problem reports (PRs) from GCC's bug tracking 18217 system that are known to be fixed in the 4.0.4 release. This list might 18218 not be complete (that is, it is possible that some PRs that have been 18219 fixed are not listed here). 18220 18221 The 4.0.4 release is provided for those that require a high degree of 18222 binary compatibility with previous 4.0.x releases. For most users, the 18223 GCC team recommends that version 4.1.1 or later be used instead." 18224 18225 18226 For questions related to the use of GCC, please consult these web 18227 pages and the [22]GCC manuals. If that fails, the 18228 [23]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18229 web pages and the development of GCC are welcome on our developer 18230 list at [24]gcc (a] gcc.gnu.org. All of [25]our lists have public 18231 archives. 18232 18233 Copyright (C) [26]Free Software Foundation, Inc. Verbatim copying and 18234 distribution of this entire article is permitted in any medium, 18235 provided this notice is preserved. 18236 18237 These pages are [27]maintained by the GCC team. Last modified 18238 2025-01-31. 18239 18240 References 18241 18242 1. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 18243 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 18244 3. https://gcc.gnu.org/projects/tree-ssa/ 18245 4. https://gcc.gnu.org/projects/tree-ssa/vectorization.html 18246 5. https://gcc.gnu.org/news/sms.html 18247 6. https://www.akkadia.org/drepper/dsohowto.pdf 18248 7. https://gcc.gnu.org/gcc-4.0/changes.html#visibility 18249 8. https://itanium-cxx-abi.github.io/cxx-abi/ 18250 9. https://gcc.gnu.org/fortran/ 18251 10. https://gcc.gnu.org/install/ 18252 11. https://gcc.gnu.org/gcc-4.0/changes.html#targets 18253 12. https://gcc.gnu.org/gcc-4.0/changes.html#obsolete_systems 18254 13. https://gcc.gnu.org/wiki/Visibility 18255 14. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.1 18256 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 18257 16. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.2 18258 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 18259 18. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 18260 19. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.3 18261 20. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.4 18262 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 18263 22. https://gcc.gnu.org/onlinedocs/ 18264 23. mailto:gcc-help (a] gcc.gnu.org 18265 24. mailto:gcc (a] gcc.gnu.org 18266 25. https://gcc.gnu.org/lists.html 18267 26. https://www.fsf.org/ 18268 27. https://gcc.gnu.org/about.html 18269 ====================================================================== 18270 http://gcc.gnu.org/gcc-3.4/index.html 18271 18272 GCC 3.4 Release Series 18273 18274 (This release series is no longer supported.) 18275 18276 May 26, 2006 18277 18278 The [1]GNU project and the GCC developers are pleased to announce the 18279 release of GCC 3.4.6. 18280 18281 This release is a bug-fix release, containing fixes for regressions in 18282 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 18283 3.4.x series. 18284 18285 The GCC 3.4 release series includes numerous [2]new features, 18286 improvements, bug fixes, and other changes, thanks to an [3]amazing 18287 group of volunteers. 18288 18289 Release History 18290 18291 GCC 3.4.6 18292 March 6, 2006 ([4]changes) 18293 18294 GCC 3.4.5 18295 November 30, 2005 ([5]changes) 18296 18297 GCC 3.4.4 18298 May 18, 2005 ([6]changes) 18299 18300 GCC 3.4.3 18301 November 4, 2004 ([7]changes) 18302 18303 GCC 3.4.2 18304 September 6, 2004 ([8]changes) 18305 18306 GCC 3.4.1 18307 July 1, 2004 ([9]changes) 18308 18309 GCC 3.4.0 18310 April 18, 2004 ([10]changes) 18311 18312 References and Acknowledgements 18313 18314 GCC used to stand for the GNU C Compiler, but since the compiler 18315 supports several other languages aside from C, it now stands for the 18316 GNU Compiler Collection. 18317 18318 A list of [11]successful builds is updated as new information becomes 18319 available. 18320 18321 The GCC developers would like to thank the numerous people that have 18322 contributed new features, improvements, bug fixes, and other changes as 18323 well as test results to GCC. This [12]amazing group of volunteers is 18324 what makes GCC successful. 18325 18326 For additional information about GCC please refer to the [13]GCC 18327 project web site or contact the [14]GCC development mailing list. 18328 18329 To obtain GCC please use [15]our mirror sites, or [16]our version 18330 control system. 18331 18332 18333 For questions related to the use of GCC, please consult these web 18334 pages and the [17]GCC manuals. If that fails, the 18335 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18336 web pages and the development of GCC are welcome on our developer 18337 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 18338 archives. 18339 18340 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 18341 distribution of this entire article is permitted in any medium, 18342 provided this notice is preserved. 18343 18344 These pages are [22]maintained by the GCC team. Last modified 18345 2022-10-26. 18346 18347 References 18348 18349 1. http://www.gnu.org/ 18350 2. https://gcc.gnu.org/gcc-3.4/changes.html 18351 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18352 4. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 18353 5. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 18354 6. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 18355 7. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 18356 8. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 18357 9. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 18358 10. https://gcc.gnu.org/gcc-3.4/changes.html 18359 11. https://gcc.gnu.org/gcc-3.4/buildstat.html 18360 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18361 13. https://gcc.gnu.org/index.html 18362 14. mailto:gcc (a] gcc.gnu.org 18363 15. https://gcc.gnu.org/mirrors.html 18364 16. https://gcc.gnu.org/git.html 18365 17. https://gcc.gnu.org/onlinedocs/ 18366 18. mailto:gcc-help (a] gcc.gnu.org 18367 19. mailto:gcc (a] gcc.gnu.org 18368 20. https://gcc.gnu.org/lists.html 18369 21. https://www.fsf.org/ 18370 22. https://gcc.gnu.org/about.html 18371 ====================================================================== 18372 http://gcc.gnu.org/gcc-3.4/changes.html 18373 18374 GCC 3.4 Release Series 18375 Changes, New Features, and Fixes 18376 18377 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 18378 is now closed. 18379 18380 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 18381 a bug, please make sure it's really GCC, and not your code, that is 18382 broken. 18383 18384 Caveats 18385 18386 * GNU Make is now required to build GCC. 18387 * With -nostdinc the preprocessor used to ignore both standard 18388 include paths and include paths contained in environment variables. 18389 It was neither documented nor intended that environment variable 18390 paths be ignored, so this has been corrected. 18391 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 18392 -fvolatile-static. It is unlikely that they worked correctly in any 18393 3.x release. 18394 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 18395 * Support for all the systems [3]obsoleted in GCC 3.3 has been 18396 removed from GCC 3.4. See below for a [4]list of systems which are 18397 obsoleted in this release. 18398 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 18399 compilers will not work. 18400 * The implementation of the [5]MIPS ABIs has changed. As a result, 18401 the code generated for certain MIPS targets will not be binary 18402 compatible with earlier releases. 18403 * In previous releases, the MIPS port had a fake "hilo" register with 18404 the user-visible name accum. This register has been removed. 18405 * The implementation of the [6]SPARC ABIs has changed. As a result, 18406 the code generated will not be binary compatible with earlier 18407 releases in certain cases. 18408 * The configure option --enable-threads=pthreads has been removed; 18409 use --enable-threads=posix instead, which should have the same 18410 effect. 18411 * Code size estimates used by inlining heuristics for C, Objective-C, 18412 C++ and Java have been redesigned significantly. As a result the 18413 parameters of -finline-insns, --param max-inline-insns-single and 18414 --param max-inline-insns-auto need to be reconsidered. 18415 * --param max-inline-slope and --param min-inline-insns have been 18416 removed; they are not needed for the new bottom-up inlining 18417 heuristics. 18418 * The new unit-at-a-time compilation scheme has several compatibility 18419 issues: 18420 + The order in which functions, variables, and top-level asm 18421 statements are emitted may have changed. Code relying on some 18422 particular ordering needs to be updated. The majority of such 18423 top-level asm statements can be replaced by section 18424 attributes. 18425 + Unreferenced static variables and functions are removed. This 18426 may result in undefined references when an asm statement 18427 refers to the variable/function directly. In that case either 18428 the variable/function shall be listed in asm statement operand 18429 or in the case of top-level asm statements the attribute used 18430 shall be used to force function/variable to be always output 18431 and considered as a possibly used by unknown code. 18432 For variables the attribute is accepted only by GCC 3.4 and 18433 newer, while for earlier versions it is sufficient to use 18434 unused to silence warnings about the variables not being 18435 referenced. To keep code portable across different GCC 18436 versions, you can use appropriate preprocessor conditionals. 18437 + Static functions now can use non-standard passing conventions 18438 that may break asm statements calling functions directly. 18439 Again the attribute used shall be used to prevent this 18440 behavior. 18441 As a temporary workaround, -fno-unit-at-a-time can be used, but 18442 this scheme may not be supported by future releases of GCC. 18443 * GCC 3.4 automatically places zero-initialized variables in the .bss 18444 section on some operating systems. Versions of GNU Emacs up to (and 18445 including) 21.3 will not work correctly when using this 18446 optimization; you can use -fno-zero-initialized-in-bss to disable 18447 it. 18448 * If GCC 3.4 is configured with --enable-threads=posix (the default 18449 on most targets that support pthreads) then _REENTRANT will be 18450 defined unconditionally by some libstdc++ headers. C++ code which 18451 relies on that macro to detect whether multi-threaded code is being 18452 compiled might change in meaning, possibly resulting in linker 18453 errors for single-threaded programs. Affected users of [7]Boost 18454 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 18455 See Bugzilla for [8]more information. 18456 18457 General Optimizer Improvements 18458 18459 * Usability of the profile feedback and coverage testing has been 18460 improved. 18461 + Performance of profiled programs has been improved by faster 18462 profile merging code. 18463 + Better use of the profile feedback for optimization (loop 18464 unrolling and loop peeling). 18465 + File locking support allowing fork() calls and parallel runs 18466 of profiled programs. 18467 + Coverage file format has been redesigned. 18468 + gcov coverage tool has been improved. 18469 + make profiledbootstrap available to build a faster compiler. 18470 Experiments made on i386 hardware showed an 11% speedup on -O0 18471 and a 7.5% speedup on -O2 compilation of a [9]large C++ 18472 testcase. 18473 + New value profiling pass enabled via -fprofile-values 18474 + New value profile transformations pass enabled via -fvpt aims 18475 to optimize some code sequences by exploiting knowledge about 18476 value ranges or other properties of the operands. At the 18477 moment a conversion of expensive divisions into cheaper 18478 operations has been implemented. 18479 + New -fprofile-generate and -fprofile-use command-line options 18480 to simplify the use of profile feedback. 18481 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 18482 Java which is enabled via -funit-at-a-time (and implied by -O2). In 18483 this scheme a whole file is parsed first and optimized later. The 18484 following basic inter-procedural optimizations are implemented: 18485 + Removal of unreachable functions and variables 18486 + Discovery of local functions (functions with static linkage 18487 whose address is never taken) 18488 + On i386, these local functions use register parameter passing 18489 conventions. 18490 + Reordering of functions in topological order of the call graph 18491 to enable better propagation of optimizing hints (such as the 18492 stack alignments needed by functions) in the back end. 18493 + Call graph based out-of-order inlining heuristics which allows 18494 to limit overall compilation unit growth (--param 18495 inline-unit-growth). 18496 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 18497 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 18498 CPU). 18499 * More realistic code size estimates used by inlining for C, 18500 Objective-C, C++ and Java. The growth of large functions can now be 18501 limited via --param large-function-insns and --param 18502 large-function-growth. 18503 * A new cfg-level loop optimizer pass replaces the old loop unrolling 18504 pass and adds two other loop transformations -- loop peeling and 18505 loop unswitching -- and also uses the profile feedback to limit 18506 code growth. (The three optimizations are enabled by 18507 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 18508 respectively). 18509 The old loop unroller still can be enabled by -fold-unroll-loops 18510 and may produce better code in some cases, especially when the 18511 webizer optimization pass is not run. 18512 * A new web construction pass enabled via -fweb (and implied by -O3) 18513 improves the quality of register allocation, CSE, first scheduling 18514 pass and some other optimization passes by avoiding re-use of 18515 pseudo registers with non-overlapping live ranges. The pass almost 18516 always improves code quality but does make debugging difficult and 18517 thus is not enabled by default by -O2 18518 The pass is especially effective as cleanup after code duplication 18519 passes, such as the loop unroller or the tracer. 18520 * Experimental implementations of superblock or trace scheduling in 18521 the second scheduling pass can be enabled via 18522 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 18523 18524 New Languages and Language specific improvements 18525 18526 Ada 18527 18528 * The Ada front end has been updated to include numerous bug fixes 18529 and enhancements. These include: 18530 + Improved project file support 18531 + Additional set of warnings about potential wrong code 18532 + Improved error messages 18533 + Improved code generation 18534 + Improved cross reference information 18535 + Improved inlining 18536 + Better run-time check elimination 18537 + Better error recovery 18538 + More efficient implementation of unbounded strings 18539 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 18540 ... 18541 + New GNAT.xxxx packages (e.g. GNAT.Strings, 18542 GNAT.Exception_Action) 18543 + New pragmas 18544 + New -gnatS switch replacing gnatpsta 18545 + Implementation of new Ada features (in particular limited 18546 with, limited aggregates) 18547 18548 C/Objective-C/C++ 18549 18550 * Precompiled headers are now supported. Precompiled headers can 18551 dramatically speed up compilation of some projects. There are some 18552 known defects in the current precompiled header implementation that 18553 will result in compiler crashes in relatively rare situations. 18554 Therefore, precompiled headers should be considered a "technology 18555 preview" in this release. Read the manual for details about how to 18556 use precompiled headers. 18557 * File handling in the preprocessor has been rewritten. GCC no longer 18558 gets confused by symlinks and hardlinks, and now has a correct 18559 implementation of #import and #pragma once. These two directives 18560 have therefore been un-deprecated. 18561 * The undocumented extension that allowed C programs to have a label 18562 at the end of a compound statement, which has been deprecated since 18563 GCC 3.0, has been removed. 18564 * The cast-as-lvalue extension has been removed for C++ and 18565 deprecated for C and Objective-C. In particular, code like this: 18566 int i; 18567 (char) i = 5; 18568 18569 or this: 18570 char *p; 18571 ((int *) p)++; 18572 18573 is no longer accepted for C++ and will not be accepted for C and 18574 Objective-C in a future version. 18575 * The conditional-expression-as-lvalue extension has been deprecated 18576 for C and Objective-C. In particular, code like this: 18577 int a, b, c; 18578 (a ? b : c) = 2; 18579 18580 will not be accepted for C and Objective-C in a future version. 18581 * The compound-expression-as-lvalue extension has been deprecated for 18582 C and Objective-C. In particular, code like this: 18583 int a, b; 18584 (a, b) = 2; 18585 18586 will not be accepted for C and Objective-C in a future version. A 18587 possible non-intrusive workaround is the following: 18588 (*(a, &b)) = 2; 18589 18590 * Several [10]built-in functions such as __builtin_popcount for 18591 counting bits, finding the highest and lowest bit in a word, and 18592 parity have been added. 18593 * The -fwritable-strings option has been deprecated and will be 18594 removed. 18595 * Many C math library functions are now recognized as built-ins and 18596 optimized. 18597 * The C, C++, and Objective-C compilers can now handle source files 18598 written in any character encoding supported by the host C library. 18599 The default input character set is taken from the current locale, 18600 and may be overridden with the -finput-charset command line option. 18601 In the future we will add support for inline encoding markers. 18602 18603 [11]C++ 18604 18605 * G++ is now much closer to full conformance to the ISO/ANSI C++ 18606 standard. This means, among other things, that a lot of invalid 18607 constructs which used to be accepted in previous versions will now 18608 be rejected. It is very likely that existing C++ code will need to 18609 be fixed. This document lists some of the most common issues. 18610 * A hand-written recursive-descent C++ parser has replaced the 18611 YACC-derived C++ parser from previous GCC releases. The new parser 18612 contains much improved infrastructure needed for better parsing of 18613 C++ source codes, handling of extensions, and clean separation 18614 (where possible) between proper semantics analysis and parsing. The 18615 new parser fixes many bugs that were found in the old parser. 18616 * You must now use the typename and template keywords to disambiguate 18617 dependent names, as required by the C++ standard. 18618 struct K { 18619 typedef int mytype_t; 18620 }; 18621 18622 template <class T1> struct A { 18623 template <class T2> struct B { 18624 void callme(void); 18625 }; 18626 18627 template <int N> void bar(void) 18628 { 18629 // Use 'typename' to tell the parser that T1::mytype_t names 18630 // a type. This is needed because the name is dependent (in 18631 // this case, on template parameter T1). 18632 typename T1::mytype_t x; 18633 x = 0; 18634 } 18635 }; 18636 18637 template <class T> void template_func(void) 18638 { 18639 // Use 'template' to prefix member templates within 18640 // dependent types (a has type A<T>, which depends on 18641 // the template parameter T). 18642 A<T> a; 18643 a.template bar<0>(); 18644 18645 // Use 'template' to tell the parser that B is a nested 18646 // template class (dependent on template parameter T), and 18647 // 'typename' because the whole A<T>::B<int> is 18648 // the name of a type (again, dependent). 18649 typename A<T>::template B<int> b; 18650 b.callme(); 18651 } 18652 18653 void non_template_func(void) 18654 { 18655 // Outside of any template class or function, no names can be 18656 // dependent, so the use of the keyword 'typename' and 'template' 18657 // is not needed (and actually forbidden). 18658 A<K> a; 18659 a.bar<0>(); 18660 A<K>::B<float> b; 18661 b.callme(); 18662 } 18663 * In a template definition, unqualified names will no longer find 18664 members of a dependent base (as specified by [temp.dep]/3 in the 18665 C++ standard). For example, 18666 template <typename T> struct B { 18667 int m; 18668 int n; 18669 int f (); 18670 int g (); 18671 }; 18672 int n; 18673 int g (); 18674 template <typename T> struct C : B<T> { 18675 void h () 18676 { 18677 m = 0; // error 18678 f (); // error 18679 n = 0; // ::n is modified 18680 g (); // ::g is called 18681 } 18682 }; 18683 You must make the names dependent, e.g. by prefixing them with 18684 this->. Here is the corrected definition of C<T>::h, 18685 template <typename T> void C<T>::h () 18686 { 18687 this->m = 0; 18688 this->f (); 18689 this->n = 0 18690 this->g (); 18691 } 18692 As an alternative solution (unfortunately not backwards compatible 18693 with GCC 3.3), you may use using declarations instead of this->: 18694 template <typename T> struct C : B<T> { 18695 using B<T>::m; 18696 using B<T>::f; 18697 using B<T>::n; 18698 using B<T>::g; 18699 void h () 18700 { 18701 m = 0; 18702 f (); 18703 n = 0; 18704 g (); 18705 } 18706 }; 18707 * In templates, all non-dependent names are now looked up and bound 18708 at definition time (while parsing the code), instead of later when 18709 the template is instantiated. For instance: 18710 void foo(int); 18711 18712 template <int> struct A { 18713 static void bar(void){ 18714 foo('a'); 18715 } 18716 }; 18717 18718 void foo(char); 18719 18720 int main() 18721 { 18722 A<0>::bar(); // Calls foo(int), used to call foo(char). 18723 } 18724 18725 * In an explicit instantiation of a class template, you must use 18726 class or struct before the template-id: 18727 template <int N> 18728 class A {}; 18729 18730 template A<0>; // error, not accepted anymore 18731 template class A<0>; // OK 18732 * The "named return value" and "implicit typename" extensions have 18733 been removed. 18734 * Default arguments in function types have been deprecated and will 18735 be removed. 18736 * ARM-style name-injection of friend declarations has been deprecated 18737 and will be removed. For example: struct S { friend void f(); }; 18738 void g() { f(); } will not be accepted by future versions of G++; 18739 instead a declaration of "f" will need to be present outside of the 18740 scope of "S". 18741 * Covariant returns are implemented for all but varadic functions 18742 that require an adjustment. 18743 * When -pedantic is used, G++ now issues errors about spurious 18744 semicolons. For example, 18745 namespace N {}; // Invalid semicolon. 18746 void f() {}; // Invalid semicolon. 18747 * G++ no longer accepts attributes for a declarator after the 18748 initializer associated with that declarator. For example, 18749 X x(1) __attribute__((...)); 18750 is no longer accepted. Instead, use: 18751 X x __attribute__((...)) (1); 18752 * Inside the scope of a template class, the name of the class itself 18753 can be treated as either a class or a template. So GCC used to 18754 accept the class name as argument of type template, and template 18755 template parameter. However this is not C++ standard compliant. Now 18756 the name is not treated as a valid template template argument 18757 unless you qualify the name by its scope. For example, the code 18758 below no longer compiles. 18759 template <template <class> class TT> class X {}; 18760 template <class T> class Y { 18761 X<Y> x; // Invalid, Y is always a type template parameter. 18762 }; 18763 The valid code for the above example is 18764 X< ::Y> x; // Valid. 18765 (Notice the space between < and : to prevent GCC to interpret this 18766 as a digraph for [.) 18767 * Friend declarations that refer to template specializations are 18768 rejected if the template has not already been declared. For 18769 example, 18770 template <typename T> 18771 class C { 18772 friend void f<> (C&); 18773 }; 18774 is rejected. You must first declare f as a template, 18775 template <typename T> 18776 void f(T); 18777 * In case of friend declarations, every name used in the friend 18778 declaration must be accessible at the point of that declaration. 18779 Previous versions of G++ used to be less strict about this and 18780 allowed friend declarations for private class members, for example. 18781 See the ISO C++ Standard Committee's [12]defect report #209 for 18782 details. 18783 * Declaration of member functions of class templates as friends are 18784 supported. For example, 18785 template <typename T> struct A { 18786 void f(); 18787 }; 18788 class C { 18789 template <typename T> friend void A<T>::f(); 18790 }; 18791 * You must use template <> to introduce template specializations, as 18792 required by the standard. For example, 18793 template <typename T> 18794 struct S; 18795 18796 struct S<int> { }; 18797 is rejected. You must write, 18798 template <> struct S<int> {}; 18799 * G++ used to accept code like this, 18800 struct S { 18801 int h(); 18802 void f(int i = g()); 18803 int g(int i = h()); 18804 }; 18805 This behavior is not mandated by the standard. Now G++ issues an 18806 error about this code. To avoid the error, you must move the 18807 declaration of g before the declaration of f. The default arguments 18808 for g must be visible at the point where it is called. 18809 * The C++ ABI Section 3.3.3 specifications for the array construction 18810 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 18811 NULL when the allocator argument returns NULL. These changes are 18812 incorporated into the libstdc++ runtime library. 18813 * Using a name introduced by a typedef in a friend declaration or in 18814 an explicit instantiation is now rejected, as specified by the ISO 18815 C++ standard. 18816 class A; 18817 typedef A B; 18818 class C { 18819 friend class B; // error, no typedef name here 18820 friend B; // error, friend always needs class/struct/enum 18821 friend class A; // OK 18822 }; 18823 18824 template <int> class Q {}; 18825 typedef Q<0> R; 18826 template class R; // error, no typedef name here 18827 template class Q<0>; // OK 18828 * When allocating an array with a new expression, GCC used to allow 18829 parentheses around the type name. This is actually ill-formed and 18830 it is now rejected: 18831 int* a = new (int)[10]; // error, not accepted anymore 18832 int* a = new int[10]; // OK 18833 * When binding an rvalue of class type to a reference, the copy 18834 constructor of the class must be accessible. For instance, consider 18835 the following code: 18836 class A 18837 { 18838 public: 18839 A(); 18840 18841 private: 18842 A(const A&); // private copy ctor 18843 }; 18844 18845 A makeA(void); 18846 void foo(const A&); 18847 18848 void bar(void) 18849 { 18850 foo(A()); // error, copy ctor is not accessible 18851 foo(makeA()); // error, copy ctor is not accessible 18852 18853 A a1; 18854 foo(a1); // OK, a1 is a lvalue 18855 } 18856 This might be surprising at first sight, especially since most 18857 popular compilers do not correctly implement this rule ([13]further 18858 details). 18859 * When forming a pointer to member or a pointer to member function, 18860 access checks for class visibility (public, protected, private) are 18861 now performed using the qualifying scope of the name itself. This 18862 is better explained with an example: 18863 class A 18864 { 18865 public: 18866 void pub_func(); 18867 protected: 18868 void prot_func(); 18869 private: 18870 void priv_func(); 18871 }; 18872 18873 class B : public A 18874 { 18875 public: 18876 void foo() 18877 { 18878 &A::pub_func; // OK, pub_func is accessible through A 18879 &A::prot_func; // error, cannot access prot_func through A 18880 &A::priv_func; // error, cannot access priv_func through A 18881 18882 &B::pub_func; // OK, pub_func is accessible through B 18883 &B::prot_func; // OK, can access prot_func through B (within B) 18884 &B::priv_func; // error, cannot access priv_func through B 18885 } 18886 }; 18887 18888 Runtime Library (libstdc++) 18889 18890 * Optimization work: 18891 + Streamlined streambuf, filebuf, separate synched with C 18892 Standard I/O streambuf. 18893 + All formatted I/O now uses cached locale information. 18894 + STL optimizations (memory/speed for list, red-black trees as 18895 used by sets and maps). 18896 + More use of GCC builtins. 18897 + String optimizations (avoid contention on 18898 increment/decrement-and-test of the reference count in the 18899 empty-string object, constructor from input_iterators 18900 speedup). 18901 * Static linkage size reductions. 18902 * Large File Support (files larger than 2 GB on 32-bit systems). 18903 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 18904 * Generic character traits. 18905 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 18906 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 18907 * The allocator class is now standard-conformant, and two additional 18908 extension allocators have been added, mt_alloc and 18909 bitmap_allocator. 18910 * PCH support: -include bits/stdc++.h (2x compile speedup). 18911 * Rewrote __cxa_demangle with support for C++ style allocators. 18912 * New debug modes for STL containers and iterators. 18913 * Testsuite rewrite: five times as many tests, plus increasingly 18914 sophisticated tests, including I/O, MT, multi-locale, wide and 18915 narrow characters. 18916 * Use current versions of GNU "autotools" for build/configuration. 18917 18918 Objective-C 18919 18920 * The Objective-C front end has been updated to include the numerous 18921 bug fixes and enhancements previously available only in Apple's 18922 version of GCC. These include: 18923 + Structured exception (@try... @catch... @finally, @throw) and 18924 synchronization (@synchronized) support. These are accessible 18925 via the -fobjc-exceptions switch; as of this writing, they may 18926 only be used in conjunction with -fnext-runtime on Mac OS X 18927 10.3 and later. See [14]Options Controlling Objective-C 18928 Dialect for more information. 18929 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 18930 may now be encoded as 'B'. In addition, the back-end/codegen 18931 dependencies have been removed. 18932 + An overhaul of message dispatch construction, ensuring that 18933 the various receiver types (and casts thereof) are handled 18934 properly, and that correct diagnostics are issued. 18935 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 18936 (-freplace-objc-classes) debugging modes, currently available 18937 on Mac OS X 10.3 and later. See [15]Options Controlling 18938 Objective-C Dialect for more information. 18939 + Access to optimized runtime entry points (-fno-nil-receivers ) 18940 on the assumption that message receivers are never nil. This 18941 is currently available on Mac OS X 10.3 and later. See 18942 [16]Options Controlling Objective-C Dialect for more 18943 information. 18944 18945 Java 18946 18947 * Compiling a .jar file will now cause non-.class entries to be 18948 automatically compiled as resources. 18949 * libgcj has been ported to Darwin. 18950 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 18951 to gcj. 18952 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 18953 code from shared libraries. 18954 * libgcj has been much more completely merged with [17]GNU Classpath. 18955 * Class loading is now much more correct; in particular the caller's 18956 class loader is now used when that is required. 18957 * [18]Eclipse 2.x will run out of the box using gij. 18958 * Parts of java.nio have been implemented. Direct and indirect 18959 buffers work, as do fundamental file and socket operations. 18960 * java.awt has been improved, though it is still not ready for 18961 general use. 18962 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 18963 method. 18964 * The MinGW port has matured. Enhancements include socket timeout 18965 support, thread interruption, improved Runtime.exec() handling and 18966 support for accented characters in filenames. 18967 18968 Fortran 18969 18970 * Fortran improvements are listed in the [19]Fortran documentation. 18971 18972 New Targets and Target Specific Improvements 18973 18974 Alpha 18975 18976 * Several [20]built-in functions have been added such as 18977 __builtin_alpha_zap to allow utilizing the more obscure 18978 instructions of the CPU. 18979 * Parameter passing of complex arguments has changed to match the 18980 ABI. This change is incompatible with previous GCC versions, but 18981 does fix compatibility with the Tru64 compiler and several corner 18982 cases where GCC was incompatible with itself. 18983 18984 ARM 18985 18986 * Nicolas Pitre has contributed his hand-coded floating-point support 18987 code for ARM. It is both significantly smaller and faster than the 18988 existing C-based implementation, even when building applications 18989 for Thumb. The arm-elf configuration has been converted to use the 18990 new code. 18991 * Support for the Intel's iWMMXt architecture, a second generation 18992 XScale processor, has been added. Enabled at run time with the 18993 -mcpu=iwmmxt command line switch. 18994 * A new ARM target has been added: arm-wince-pe. This is similar to 18995 the arm-pe target, but it defaults to using the APCS32 ABI. 18996 * The existing ARM pipeline description has been converted to the use 18997 the [21]DFA processor pipeline model. There is not much change in 18998 code performance, but the description is now [22]easier to 18999 understand. 19000 * Support for the Cirrus EP9312 Maverick floating point co-processor 19001 added. Enabled at run time with the -mcpu=ep9312 command line 19002 switch. Note however that the multilibs to support this chip are 19003 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 19004 enable their production you will have to uncomment the entries in 19005 that file. 19006 19007 H8/300 19008 19009 * Support for long long has been added. 19010 * Support for saveall attribute has been added. 19011 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 19012 for H8/300H and H8S, which is much faster than the previous 19013 implementation. 19014 * A lot of small performance improvements. 19015 19016 IA-32/AMD64 (x86-64) 19017 19018 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 19019 -march=k8 and -mcpu=k8. 19020 * Scalar SSE code generation carefully avoids reformatting penalties, 19021 hidden dependencies and minimizes the number of uops generated on 19022 both Intel and AMD CPUs. 19023 * Vector MMX and SSE operands are now passed in registers to improve 19024 performance and match the argument passing convention used by the 19025 Intel C++ Compiler. As a result it is not possible to call 19026 functions accepting vector arguments compiled by older GCC version. 19027 * Conditional jump elimination is now more aggressive on modern CPUs. 19028 * The Athlon ports has been converted to use the DFA processor 19029 pipeline description. 19030 * Optimization of indirect tail calls is now possible in a similar 19031 fashion as direct sibcall optimization. 19032 * Further small performance improvements. 19033 * -m128bit-long-double is now less buggy. 19034 * __float128 support in 64-bit compilation. 19035 * Support for data structures exceeding 2GB in 64-bit mode. 19036 * -mcpu has been renamed to -mtune. 19037 19038 IA-64 19039 19040 * Tuning code for the Itanium 2 processor has been added. The 19041 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 19042 enabled by default now. To generate code tuned for Itanium 1 the 19043 option -mtune=itanium1 should be used. 19044 * [23]DFA processor pipeline descriptions for the IA-64 processors 19045 have been added. This resulted in about 3% improvement on the 19046 SPECInt2000 benchmark for Itanium 2. 19047 * Instruction bundling for the IA-64 processors has been rewritten 19048 using the DFA pipeline hazard recognizer. It resulted in about 60% 19049 compiler speedup on the SPECInt2000 C programs. 19050 19051 M32R 19052 19053 * Support for the M32R/2 processor has been added by Renesas. 19054 * Support for an M32R GNU/Linux target and PIC code generation has 19055 been added by Renesas. 19056 19057 M68000 19058 19059 * Bernardo Innocenti (Develer S.r.l.) has contributed the 19060 m68k-uclinux target, based on former work done by Paul Dale 19061 (SnapGear Inc.). Code generation for the ColdFire processors family 19062 has been enhanced and extended to support the MCF 53xx and MCF 54xx 19063 cores, integrating former work done by Peter Barada (Motorola). 19064 19065 MIPS 19066 19067 Processor-specific changes 19068 19069 * Support for the RM7000 and RM9000 processors has been added. It can 19070 be selected using the -march compiler option and should work with 19071 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 19072 * Support for revision 2 of the MIPS32 ISA has been added. It can be 19073 selected with the command-line option -march=mips32r2. 19074 * There is a new option, -mfix-sb1, to work around certain SB-1 19075 errata. 19076 19077 Configuration 19078 19079 * It is possible to customize GCC using the following configure-time 19080 options: 19081 + --with-arch, which specifies the default value of the -march 19082 option. 19083 + --with-tune, which specifies the default value of the -mtune 19084 option. 19085 + --with-abi, which specifies the default ABI. 19086 + --with-float=soft, which tells GCC to use software floating 19087 point by default. 19088 + --with-float=hard, which tells GCC to use hardware floating 19089 point by default. 19090 * A 64-bit GNU/Linux port has been added. The associated 19091 configurations are mips64-linux-gnu and mips64el-linux-gnu. 19092 * The 32-bit GNU/Linux port now supports Java. 19093 * The IRIX 6 configuration now supports the o32 ABI and will build 19094 o32 multilibs by default. This support is compatible with both 19095 binutils and the SGI tools, but note that several features, 19096 including debugging information and DWARF2 exception handling, are 19097 only available when using the GNU assembler. Use of the GNU 19098 assembler and linker (version 2.15 or above) is strongly 19099 recommended. 19100 * The IRIX 6 configuration now supports 128-bit long doubles. 19101 * There are two new RTEMS-specific configurations, mips-rtems and 19102 mipsel-rtems. 19103 * There are two new *-elf configurations, mipsisa32r2-elf and 19104 mipsisa32r2el-elf. 19105 19106 General 19107 19108 * Several [24]ABI bugs have been fixed. Unfortunately, these changes 19109 will break binary compatibility with earlier releases. 19110 * GCC can now use explicit relocation operators when generating 19111 -mabicalls code. This behavior is controlled by -mexplicit-relocs 19112 and can have several performance benefits. For example: 19113 + It allows for more optimization of GOT accesses, including 19114 better scheduling and redundancy elimination. 19115 + It allows sibling calls to be implemented as jumps. 19116 + n32 and n64 leaf functions can use a call-clobbered global 19117 pointer instead of $28. 19118 + The code to set up $gp can be removed from functions that 19119 don't need it. 19120 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 19121 option is equivalent to the assembler's -xgot option and should be 19122 used instead of -Wa,-xgot. 19123 * Frame pointer elimination is now supported when generating 64-bit 19124 MIPS16 code. 19125 * Inline block moves have been optimized to take more account of 19126 alignment information. 19127 * Many internal changes have been made to the MIPS port, mostly aimed 19128 at reducing the reliance on assembler macros. 19129 19130 PowerPC 19131 19132 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 19133 [25]ABI incompatibilities regarding the way parameters are passed 19134 during functions calls. These changes may result in incompatibility 19135 between code compiled with GCC 3.3 and GCC 3.4. 19136 19137 PowerPC Darwin 19138 19139 * Support for shared/dylib gcc libraries has been added. It is 19140 enabled by default on powerpc-apple-darwin7.0.0 and up. 19141 * Libgcj is enabled by default. On systems older than 19142 powerpc-apple-darwin7.0.0 you need to install dlcompat. 19143 * 128-bit IBM extended precision format support added for long 19144 double. 19145 19146 PowerPC64 GNU/Linux 19147 19148 * By default, PowerPC64 GNU/Linux now uses natural alignment of 19149 structure elements. The old four byte alignment for double, with 19150 special rules for a struct starting with a double, can be chosen 19151 with -malign-power. This change may result in incompatibility 19152 between code compiled with GCC 3.3 and GCC 3.4. 19153 * -mabi=altivec is now the default rather than -mabi=no-altivec. 19154 * 128-bit IBM extended precision format support added for long 19155 double. 19156 19157 S/390 and zSeries 19158 19159 * New command-line options allow to specify the intended execution 19160 environment for generated code: 19161 + -mesa/-mzarch allows to specify whether to generate code 19162 running in ESA/390 mode or in z/Architecture mode (this is 19163 applicable to 31-bit code only). 19164 + -march allows to specify a minimum processor architecture 19165 level (g5, g6, z900, or z990). 19166 + -mtune allows to specify which processor to tune for. 19167 * It is possible to customize GCC using the following configure-time 19168 options: 19169 + --with-mode, which specifies whether to default to assuming 19170 ESA/390 or z/Architecture mode. 19171 + --with-arch, which specifies the default value of the -march 19172 option. 19173 + --with-tune, which specifies the default value of the -mtune 19174 option. 19175 * Support for the z990 processor has been added, and can be selected 19176 using -march=z990 or -mtune=z990. This includes instruction 19177 scheduling tuned for the superscalar instruction pipeline of the 19178 z990 processor as well as support for all new instructions provided 19179 by the long-displacement facility. 19180 * Support to generate 31-bit code optimized for zSeries processors 19181 (running in ESA/390 or in z/Architecture mode) has been added. This 19182 can be selected using -march=z900 and -mzarch respectively. 19183 * Instruction scheduling for the z900 and z990 processors now uses 19184 the DFA pipeline hazard recognizer. 19185 * GCC no longer generates code to maintain a stack backchain, 19186 previously used to generate stack backtraces for debugging 19187 purposes. As replacement that does not incur runtime overhead, 19188 DWARF-2 call frame information is provided by GCC; this is 19189 supported by GDB 6.1. The old behavior can be restored using the 19190 -mbackchain option. 19191 * The stack frame size of functions may now exceed 2 GB in 64-bit 19192 code. 19193 * A port for the 64-bit IBM TPF operating system has been added; the 19194 configuration is s390x-ibm-tpf. This configuration is supported as 19195 cross-compilation target only. 19196 * Various changes to improve the generated code have been 19197 implemented, including: 19198 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 19199 instructions to significantly speed up many floating-point 19200 applications. 19201 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 19202 WITH BORROW instructions to speed up long long arithmetic. 19203 + GCC now uses the SEARCH STRING instruction to implement 19204 strlen(). 19205 + In many cases, function call overhead for 31-bit code has been 19206 reduced by placing the literal pool after the function code 19207 instead of after the function prolog. 19208 + Register 14 is no longer reserved in 64-bit code. 19209 + Handling of global register variables has been improved. 19210 19211 SPARC 19212 19213 * The option -mflat is deprecated. 19214 * Support for large (> 2GB) frames has been added to the 64-bit port. 19215 * Several [26]ABI bugs have been fixed. Unfortunately, these changes 19216 will break binary compatibility with earlier releases. 19217 * The default debugging format has been switched from STABS to 19218 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 19219 the default debugging format for 64-bit code on Solaris. 19220 19221 SuperH 19222 19223 * Support for the SH2E processor has been added. Enabled at run time 19224 with the -m2e command line switch, or at configure time by 19225 specifying sh2e as the machine part of the target triple. 19226 19227 V850 19228 19229 * Support for the Mitsubishi V850E1 processor has been added. This is 19230 a variant of the V850E processor with some additional debugging 19231 instructions. 19232 19233 Xtensa 19234 19235 * Several ABI bugs have been fixed. Unfortunately, these changes 19236 break binary compatibility with earlier releases. 19237 + For big-endian processors, the padding of aggregate return 19238 values larger than a word has changed. If the size of an 19239 aggregate return value is not a multiple of 32 bits, previous 19240 versions of GCC inserted padding in the most-significant bytes 19241 of the first return value register. Aggregates larger than a 19242 word are now padded in the least-significant bytes of the last 19243 return value register used. Aggregates smaller than a word are 19244 still padded in the most-significant bytes. The return value 19245 padding has not changed for little-endian processors. 19246 + Function arguments with 16-byte alignment are now properly 19247 aligned. 19248 + The implementation of the va_list type has changed. A va_list 19249 value created by va_start from a previous release cannot be 19250 used with va_arg from this release, or vice versa. 19251 * More processor configuration options for Xtensa processors are 19252 supported: 19253 + the ABS instruction is now optional; 19254 + the ADDX* and SUBX* instructions are now optional; 19255 + an experimental CONST16 instruction can be used to synthesize 19256 constants instead of loading them from constant pools. 19257 These and other Xtensa processor configuration options can no 19258 longer be enabled or disabled by command-line options; the 19259 processor configuration must be specified by the xtensa-config.h 19260 header file when building GCC. Additionally, the 19261 -mno-serialize-volatile option is no longer supported. 19262 19263 [27]Obsolete Systems 19264 19265 Support for a number of older systems has been declared obsolete in GCC 19266 3.4. Unless there is activity to revive them, the next release of GCC 19267 will have their sources permanently removed. 19268 19269 All configurations of the following processor architectures have been 19270 declared obsolete: 19271 * Mitsubishi D30V, d30v-* 19272 * AT&T DSP1600 and DSP1610, dsp16xx-* 19273 * Intel 80960, i960 19274 19275 Also, some individual systems have been obsoleted: 19276 * ARM Family 19277 + Support for generating code for operation in APCS/26 mode 19278 (-mapcs-26). 19279 * IBM ESA/390 19280 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 19281 maintained and supported.) 19282 * Intel 386 family 19283 + MOSS, i?86-moss-msdos and i?86-*-moss* 19284 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 19285 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 19286 i?86-*-freebsd2* 19287 + GNU/Linux with a.out object format, i?86-linux*aout* 19288 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 19289 + Interix versions before Interix 3, i?86-*-interix 19290 + Mach microkernel, i?86-mach* 19291 + SCO UnixWare with UDK, i?86-*-udk* 19292 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 19293 + VSTa microkernel, i386-*-vsta 19294 * Motorola M68000 family 19295 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 19296 + NetBSD with a.out object format (before NetBSD 1.4), 19297 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 19298 + Generic System V r.4, m68k-*-sysv4* 19299 * VAX 19300 + Generic VAX, vax-*-* (This is generic VAX only; we have not 19301 obsoleted any VAX triples for specific operating systems.) 19302 19303 Documentation improvements 19304 19305 Other significant improvements 19306 19307 * The build system has undergone several significant cleanups. 19308 Subdirectories will only be configured if they are being built, and 19309 all subdirectory configures are run from the make command. The top 19310 level has been autoconfiscated. 19311 * Building GCC no longer writes to its source directory. This should 19312 help those wishing to share a read-only source directory over NFS 19313 or build from a CD. The exceptions to this feature are if you 19314 configure with either --enable-maintainer-mode or 19315 --enable-generated-files-in-srcdir. 19316 * The -W warning option has been renamed to -Wextra, which is more 19317 easily understood. The older spelling will be retained for 19318 backwards compatibility. 19319 * Substantial improvements in compile time have been made, 19320 particularly for non-optimizing compilations. 19321 __________________________________________________________________ 19322 19323 [28]GCC 3.4.0 19324 19325 Bug Fixes 19326 19327 A vast number of bugs have been fixed in 3.4.0, too many to publish a 19328 complete list here. [29]Follow this link to query the Bugzilla database 19329 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 19330 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 19331 regressions. 19332 __________________________________________________________________ 19333 19334 [30]GCC 3.4.1 19335 19336 Bug Fixes 19337 19338 This section lists the problem reports (PRs) from GCC's bug tracking 19339 system that are known to be fixed in the 3.4.1 release. This list might 19340 not be complete (that is, it is possible that some PRs that have been 19341 fixed are not listed here). 19342 19343 Bootstrap failures 19344 19345 * [31]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 19346 emitted - PIC related 19347 * [32]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 19348 * [33]14760 A bug in configure.in prevents using both 19349 --program-suffix and --program-prefix 19350 * [34]14671 [hppa64] bootstrap fails: ICE in 19351 save_call_clobbered_regs, in caller_save.c 19352 * [35]15093 [alpha][Java] make bootstrap fails to configure libffi on 19353 Alpha 19354 * [36]15178 Solaris 9/x86 fails linking after stage 3 19355 19356 Multi-platform internal compiler errors (ICEs) 19357 19358 * [37]12753 (preprocessor) Memory corruption in preprocessor on bad 19359 input 19360 * [38]13985 ICE in gcc.c-torture/compile/930621-1.c 19361 * [39]14810 (c++) tree check failures with invalid code involving 19362 templates 19363 * [40]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 19364 cp/parser.c 19365 * [41]15044 (c++) ICE on syntax error, template header 19366 * [42]15057 (c++) Compiling of conditional value throw constructs 19367 cause a segmentation violation 19368 * [43]15064 (c++) typeid of template parameter gives ICE 19369 * [44]15142 (c++) ICE when passing a string where a char* is expected 19370 in a throw statement 19371 * [45]15159 ICE in rtl_verify_flow_info_1 19372 * [46]15165 (c++) ICE in instantiate_template 19373 * [47]15193 Unary minus using pointer to V4SF vector causes 19374 -fforce-mem to exhaust all memory 19375 * [48]15209 (c++) Runs out of memory with packed structs 19376 * [49]15227 (c++) Trouble with invalid function definition 19377 * [50]15285 (c++) instantiate_type ICE when forming pointer to 19378 template function 19379 * [51]15299 (c++) ICE in resolve_overloaded_unification 19380 * [52]15329 (c++) ICE on constructor of member template 19381 * [53]15550 ICE in extract_insn, in recog.c 19382 * [54]15554 (c++) ICE in tsubst_copy, in cp/pt.c 19383 * [55]15640 (c++) ICE on invalid code in arg_assoc, in 19384 cp/name-lookup.c 19385 * [56]15666 [unit-at-a-time] Gcc abort on valid code 19386 * [57]15696 (c++) ICE with bad pointer-to-member code 19387 * [58]15701 (c++) ICE with friends and template template parameter 19388 * [59]15761 ICE in do_SUBST, in combine.c 19389 * [60]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 19390 19391 Ada 19392 19393 * [61]14538 All RTEMS targets broken for gnat 19394 19395 C front end 19396 19397 * [62]12391 missing warning about assigning to an incomplete type 19398 * [63]14649 atan(1.0) should not be a constant expression 19399 * [64]15004 [unit-at-a-time] no warning for unused paramater in 19400 static function 19401 * [65]15749 --pedantic-errors behaves differently from --pedantic 19402 with C-compiler on GNU/Linux 19403 19404 C++ compiler and library 19405 19406 * [66]10646 non-const reference is incorrectly matched in a "const T" 19407 partial specialization 19408 * [67]12077 wcin.rdbuf()->in_avail() return value too high 19409 * [68]13598 enc_filebuf doesn't work 19410 * [69]14211 const_cast returns lvalue but should be rvalue 19411 * [70]14220 num_put::do_put() undesired float/double behavior 19412 * [71]14245 problem with user-defined allocators in std::basic_string 19413 * [72]14340 libstdc++ Debug mode: failure to convert iterator to 19414 const_iterator 19415 * [73]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 19416 FILE* 19417 * [74]14668 no warning anymore for reevaluation of declaration 19418 * [75]14775 LFS (large file support) tests missing 19419 * [76]14821 Duplicate namespace alias declaration should not conflict 19420 * [77]14930 Friend declaration ignored 19421 * [78]14932 cannot use offsetof to get offsets of array elements in 19422 g++ 3.4.0 19423 * [79]14950 [non unit-at-a-time] always_inline does not mix with 19424 templates and -O0 19425 * [80]14962 g++ ignores #pragma redefine_extname 19426 * [81]14975 Segfault on low-level write error during imbue 19427 * [82]15002 Linewise stream input is unusably slow (std::string slow) 19428 * [83]15025 compiler accepts redeclaration of template as 19429 non-template 19430 * [84]15046 [arm] Math functions misdetected by cross configuration 19431 * [85]15069 a bit test on a variable of enum type is miscompiled 19432 * [86]15074 g++ -lsupc++ still links against libstdc++ 19433 * [87]15083 spurious "statement has no effect" warning 19434 * [88]15096 parse error with templates and pointer to const member 19435 * [89]15287 combination of operator[] and operator .* fails in 19436 templates 19437 * [90]15317 __attribute__ unused in first parameter of constructor 19438 gives error 19439 * [91]15337 sizeof on incomplete type diagnostic 19440 * [92]15361 bitset<>::_Find_next fails 19441 * [93]15412 _GLIBCXX_ symbols symbols defined and used in different 19442 namespaces 19443 * [94]15427 valid code results in incomplete type error 19444 * [95]15471 Incorrect member pointer offsets in anonymous 19445 structs/unions 19446 * [96]15503 nested template problem 19447 * [97]15507 compiler hangs while laying out union 19448 * [98]15542 operator & and template definitions 19449 * [99]15565 SLES9: leading + sign for unsigned int with showpos 19450 * [100]15625 friend defined inside a template fails to find static 19451 function 19452 * [101]15629 Function templates, overloads, and friend name injection 19453 * [102]15742 'noreturn' attribute ignored in method of template 19454 functions. 19455 * [103]15775 Allocator::pointer consistently ignored 19456 * [104]15821 Duplicate namespace alias within namespace rejected 19457 * [105]15862 'enum yn' fails (confict with undeclared builtin) 19458 * [106]15875 rejects pointer to member in template 19459 * [107]15877 valid code using templates and anonymous enums is 19460 rejected 19461 * [108]15947 Puzzling error message for wrong destructor declaration 19462 in template class 19463 * [109]16020 cannot copy __gnu_debug::bitset 19464 * [110]16154 input iterator concept too restrictive 19465 * [111]16174 deducing top-level consts 19466 19467 Java 19468 19469 * [112]14315 Java compiler is not parallel make safe 19470 19471 Fortran 19472 19473 * [113]15151 [g77] incorrect logical i/o in 64-bit mode 19474 19475 Objective-C 19476 19477 * [114]7993 private variables cannot be shadowed in subclasses 19478 19479 Optimization bugs 19480 19481 * [115]15228 useless copies of floating point operands 19482 * [116]15345 [non-unit-at-a-time] unreferenced nested inline 19483 functions not optimized away 19484 * [117]15945 Incorrect floating point optimization 19485 * [118]15526 ftrapv aborts on 0 * (-1) 19486 * [119]14690 Miscompiled POOMA tests 19487 * [120]15112 GCC generates code to write to unchanging memory 19488 19489 Preprocessor 19490 19491 * [121]15067 Minor glitch in the source of cpp 19492 19493 Main driver program bugs 19494 19495 * [122]1963 collect2 interprets -oldstyle_liblookup as -o 19496 ldstyle_liblookup 19497 19498 x86-specific (Intel/AMD) 19499 19500 * [123]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 19501 section} 19502 19503 HPPA-specific 19504 19505 * [124]14782 GCC produces an unaligned data access at -O2 19506 * [125]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 19507 * [126]15202 ICE in reload_cse_simplify_operands, in postreload.c 19508 19509 IA64-specific 19510 19511 * [127]14610 __float80 constants incorrectly emitted 19512 * [128]14813 init_array sections are initialized in the wrong order 19513 * [129]14857 GCC segfault on duplicated asm statement 19514 * [130]15598 Gcc 3.4 ICE on valid code 19515 * [131]15653 Gcc 3.4 ICE on valid code 19516 19517 MIPS-specific 19518 19519 * [132]15189 wrong filling of delay slot with -march=mips1 -G0 19520 -mno-split-addresses -mno-explicit-relocs 19521 * [133]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 19522 2.14.91 19523 * [134]16144 Bogus reference to __divdf3 when -O1 19524 * [135]16176 Miscompilation of unaligned data in MIPS backend 19525 19526 PowerPC-specific 19527 19528 * [136]11591 ICE in gcc.dg/altivec-5.c 19529 * [137]12028 powerpc-eabispe produces bad sCOND operation 19530 * [138]14478 rs6000 geu/ltu patterns generate incorrect code 19531 * [139]14567 long double and va_arg complex args 19532 * [140]14715 Altivec stack layout may overlap gpr save with stack 19533 temps 19534 * [141]14902 (libstdc++) Stream checking functions fail when -pthread 19535 option is used. 19536 * [142]14924 Compiler ICE on valid code 19537 * [143]14960 -maltivec affects vector return with -mabi=no-altivec 19538 * [144]15106 vector varargs failure passing from altivec to 19539 non-altivec code for -m32 19540 * [145]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 19541 half-word operation 19542 * [146]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 19543 and stvx 19544 * [147]15662 Segmentation fault when an exception is thrown - even if 19545 try and catch are specified 19546 19547 s390-specific 19548 19549 * [148]15054 Bad code due to overlapping stack temporaries 19550 19551 SPARC-specific 19552 19553 * [149]15783 ICE with union assignment in 64-bit mode 19554 * [150]15626 GCC 3.4 emits "ld: warning: relocation error: 19555 R_SPARC_UA32" 19556 19557 x86-64-specific 19558 19559 * [151]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 19560 * [152]14723 Backported -march=nocona from mainline 19561 * [153]15290 __float128 failed to pass to function properly 19562 19563 Cygwin/Mingw32-specific 19564 19565 * [154]15250 Option -mms-bitfields support on GCC 3.4 is not 19566 conformant to MS layout 19567 * [155]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 19568 worker on windows32 targets 19569 19570 Bugs specific to embedded processors 19571 19572 * [156]8309 [m68k] -m5200 produces erroneous SImode set of short 19573 varaible on stack 19574 * [157]13250 [SH] Gcc code for rotation clobbers the register, but 19575 gcc continues to use the register as if it was not clobbered 19576 * [158]13803 [coldfire] movqi operand constraints too restrictivefor 19577 TARGET_COLDFIRE 19578 * [159]14093 [SH] ICE for code when using -mhitachi option in SH 19579 * [160]14457 [m6811hc] ICE with simple c++ source 19580 * [161]14542 [m6811hc] ICE on simple source 19581 * [162]15100 [SH] cc1plus got hang-up on 19582 libstdc++-v3/testsuite/abi_check.cc 19583 * [163]15296 [CRIS] Delayed branch scheduling causing invalid code on 19584 cris-* 19585 * [164]15396 [SH] ICE with -O2 -fPIC 19586 * [165]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 19587 ColdFire 19588 19589 Testsuite problems (compiler not affected) 19590 19591 * [166]11610 libstdc++ testcases 27_io/* don't work properly remotely 19592 * [167]15488 (libstdc++) possibly insufficient file permissions for 19593 executing test suite 19594 * [168]15489 (libstdc++) testsuite_files determined incorrectly 19595 19596 Documentation bugs 19597 19598 * [169]13928 (libstdc++) no whatis info in some man pages generated 19599 by doxygen 19600 * [170]14150 Ada documentation out of date 19601 * [171]14949 (c++) Need to document method visibility changes 19602 * [172]15123 libstdc++-doc: Allocators.3 manpage is empty 19603 __________________________________________________________________ 19604 19605 [173]GCC 3.4.2 19606 19607 Bug Fixes 19608 19609 This section lists the problem reports (PRs) from GCC's bug tracking 19610 system that are known to be fixed in the 3.4.2 release. This list might 19611 not be complete (that is, it is possible that some PRs that have been 19612 fixed are not listed here). 19613 19614 Bootstrap failures and issues 19615 19616 * [174]16469 [mips-sgi-irix5.3] bootstrap fails in 19617 libstdc++-v3/testsuite 19618 * [175]16344 [hppa-linux-gnu] libstdc++'s PCH built by 19619 profiledbootstrap does not work with the built compiler 19620 * [176]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 19621 19622 Multi-platform internal compiler errors (ICEs) 19623 19624 * [177]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 19625 cp_parser_class_specifier, in cp/parser.c 19626 * [178]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 19627 * [179]15461 (c++) ICE due to NRV and inlining 19628 * [180]15890 (c++) ICE in c_expand_expr, in c-common.c 19629 * [181]16180 ICE: segmentation fault in RTL optimization 19630 * [182]16224 (c++) ICE in write_unscoped_name (template/namespace) 19631 * [183]16408 ICE: in delete_insn, in cfgrtl.c 19632 * [184]16529 (c++) ICE for: namespace-alias shall not be declared as 19633 the name of any other entity 19634 * [185]16698 (c++) ICE with exceptions and declaration of __cxa_throw 19635 * [186]16706 (c++) ICE in finish_member_declaration, in 19636 cp/semantics.c 19637 * [187]16810 (c++) Legal C++ program with cast gives ICE in 19638 build_ptrmemfunc 19639 * [188]16851 (c++) ICE when throwing a comma expression 19640 * [189]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 19641 * [190]16904 (c++) ICE in finish_class_member_access_expr, in 19642 cp/typeck.c 19643 * [191]16905 (c++) ICE (segfault) with exceptions 19644 * [192]16964 (c++) ICE in cp_parser_class_specifier due to 19645 redefinition 19646 * [193]17068 (c++) ICE: tree check: expected class 'd', have 'x' 19647 (identifier_node) in dependent_template_p, in cp/pt.c 19648 19649 Preprocessor bugs 19650 19651 * [194]16366 Preprocessor option -remap causes memory corruption 19652 19653 Optimization 19654 19655 * [195]15345 unreferenced nested inline functions not optimized away 19656 * [196]16590 Incorrect execution when compiling with -O2 19657 * [197]16693 Bitwise AND is lost when used within a cast to an enum 19658 of the same precision 19659 * [198]17078 Jump into if(0) substatement fails 19660 19661 Problems in generated debug information 19662 19663 * [199]13956 incorrect stabs for nested local variables 19664 19665 C front end bugs 19666 19667 * [200]16684 GCC should not warn about redundant redeclarations of 19668 built-ins 19669 19670 C++ compiler and library 19671 19672 * [201]12658 Thread safety problems in locale::global() and 19673 locale::locale() 19674 * [202]13092 g++ accepts invalid pointer-to-member conversion 19675 * [203]15320 Excessive memory consumption 19676 * [204]16246 Incorrect template argument deduction 19677 * [205]16273 Memory exhausted when using nested classes and virtual 19678 functions 19679 * [206]16401 ostringstream in gcc 3.4.x very slow for big data 19680 * [207]16411 undefined reference to 19681 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 19682 >::file() 19683 * [208]16489 G++ incorrectly rejects use of a null constant integral 19684 expression as a null constant pointer 19685 * [209]16618 offsetof fails with constant member 19686 * [210]16637 syntax error reported for valid input code 19687 * [211]16717 __attribute__((constructor)) broken in C++ 19688 * [212]16813 compiler error in DEBUG version of range insertion 19689 std::map::insert 19690 * [213]16853 pointer-to-member initialization from incompatible one 19691 accepted 19692 * [214]16889 ambiguity is not detected 19693 * [215]16959 Segmentation fault in ios_base::sync_with_stdio 19694 19695 Java compiler and library 19696 19697 * [216]7587 direct threaded interpreter not thread-safe 19698 * [217]16473 ServerSocket accept() leaks file descriptors 19699 * [218]16478 Hash synchronization deadlock with finalizers 19700 19701 Alpha-specific 19702 19703 * [219]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 19704 * [220]16974 could not split insn (ice in final_scan_insn, in 19705 final.c) 19706 19707 x86-specific 19708 19709 * [221]16298 ICE in output_operand 19710 * [222]17113 ICE with SSE2 intrinsics 19711 19712 x86-64 specific 19713 19714 * [223]14697 libstdc++ couldn't find 32bit libgcc_s 19715 19716 MIPS-specific 19717 19718 * [224]15869 [mips64] No NOP after LW (with -mips1 -O0) 19719 * [225]16325 [mips64] value profiling clobbers gp on mips 19720 * [226]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 19721 char[]s 19722 * [227]16380 [mips64] Use of uninitialised register after dbra 19723 conversion 19724 * [228]16407 [mips64] Unaligned access to local variables 19725 * [229]16643 [mips64] verify_local_live_at_start ICE after 19726 crossjumping & cfgcleanup 19727 19728 ARM-specific 19729 19730 * [230]15927 THUMB -O2: strength-reduced iteration variable ends up 19731 off by 1 19732 * [231]15948 THUMB: ICE with non-commutative cbranch 19733 * [232]17019 THUMB: bad switch statement in md code for 19734 addsi3_cbranch_scratch 19735 19736 IA64-specific 19737 19738 * [233]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 19739 (-mtune=merced) 19740 * [234]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 19741 (-mtune=itanium) 19742 * [235]16278 Gcc failed to build Linux kernel with -mtune=merced 19743 * [236]16414 ICE on valid code: typo in comparison of asm_noperands 19744 result 19745 * [237]16445 ICE on valid code: don't count ignored insns 19746 * [238]16490 ICE (segfault) while compiling with -fprofile-use 19747 * [239]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 19748 19749 PowerPC-specific 19750 19751 * [240]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 19752 * [241]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 19753 issue) 19754 19755 SPARC-specific 19756 19757 * [242]16199 ICE while compiling apache 2.0.49 19758 * [243]16416 -m64 doesn't imply -mcpu=v9 anymore 19759 * [244]16430 ICE when returning non-C aggregates larger than 16 bytes 19760 19761 Bugs specific to embedded processors 19762 19763 * [245]16379 [m32r] can't output large model function call of memcpy 19764 * [246]17093 [m32r] ICE with -msdata=use -O0 19765 * [247]17119 [m32r] ICE at switch case 0x8000 19766 19767 DJGPP-specific 19768 19769 * [248]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 19770 19771 Alpha Tru64-specific 19772 19773 * [249]16210 libstdc++ gratuitously omits "long long" I/O 19774 19775 Testsuite, documentation issues (compiler is not affected): 19776 19777 * [250]15488 (libstdc++) possibly insufficient file permissions for 19778 executing test suite 19779 * [251]16250 ada/doctools runs makeinfo even in release tarball 19780 __________________________________________________________________ 19781 19782 [252]GCC 3.4.3 19783 19784 This is the [253]list of problem reports (PRs) from GCC's bug tracking 19785 system that are known to be fixed in the 3.4.3 release. This list might 19786 not be complete (that is, it is possible that some PRs that have been 19787 fixed are not listed here). 19788 19789 Bootstrap failures 19790 19791 * [254]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 19792 * [255]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 19793 when undeclared 19794 19795 Internal compiler errors (ICEs) affecting multiple platforms 19796 19797 * [256]13948 (java) GCJ segmentation fault while compiling GL4Java 19798 .class files 19799 * [257]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 19800 * [258]16301 (c++) ICE when "strong" attribute is attached to a using 19801 directive 19802 * [259]16566 ICE with flexible arrays 19803 * [260]17023 ICE with nested functions in parameter declaration 19804 * [261]17027 ICE with noreturn function in loop at -O2 19805 * [262]17524 ICE in grokdeclarator, in cp/decl.c 19806 * [263]17826 (c++) ICE in cp_tree_equal 19807 19808 C and optimization bugs 19809 19810 * [264]15526 -ftrapv aborts on 0 * (-1) 19811 * [265]16999 #ident stopped working 19812 * [266]17503 quadratic behaviour in invalid_mode_change_p 19813 * [267]17581 Long long arithmetic fails inside a switch/case 19814 statement when compiled with -O2 19815 * [268]18129 -fwritable-strings doesn't work 19816 19817 C++ compiler and library bugs 19818 19819 * [269]10975 incorrect initial ostringstream::tellp() 19820 * [270]11722 Unbuffered filebuf::sgetn is slow 19821 * [271]14534 Unrecognizing static function as a template parameter 19822 when its return value is also templated 19823 * [272]15172 Copy constructor optimization in aggregate 19824 initialization 19825 * [273]15786 Bad error message for frequently occuring error. 19826 * [274]16162 Rejects valid member-template-definition 19827 * [275]16612 empty basic_strings can't live in shared memory 19828 * [276]16715 std::basic_iostream is instantiated when used, even 19829 though instantiations are already contained in libstdc++ 19830 * [277]16848 code in /ext/demangle.h appears broken 19831 * [278]17132 GCC fails to eliminate function template specialization 19832 when argument deduction fails 19833 * [279]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 19834 in ropeimpl.h 19835 * [280]17327 use of `enumeral_type' in template type unification 19836 * [281]17393 "unused variable '._0'" warning with -Wall 19837 * [282]17501 Confusion with member templates 19838 * [283]17537 g++ not passing -lstdc++ to linker when all command line 19839 arguments are libraries 19840 * [284]17585 usage of unqualified name of static member from within 19841 class not allowed 19842 * [285]17821 Poor diagnostic for using "." instead of "->" 19843 * [286]17829 wrong error: call of overloaded function is ambiguous 19844 * [287]17851 Misleading diagnostic for invalid function declarations 19845 with undeclared types 19846 * [288]17976 Destructor is called twice 19847 * [289]18020 rejects valid definition of enum value in template 19848 * [290]18093 bogus conflict in namespace aliasing 19849 * [291]18140 C++ parser bug when using >> in templates 19850 19851 Fortran 19852 19853 * [292]17541 data statements with double precision constants fail 19854 19855 x86-specific 19856 19857 * [293]17853 -O2 ICE for MMX testcase 19858 19859 SPARC-specific 19860 19861 * [294]17245 ICE compiling gsl-1.5 statistics/lag1.c 19862 19863 Darwin-specific 19864 19865 * [295]17167 FATAL:Symbol L_foo$stub already defined. 19866 19867 AIX-specific 19868 19869 * [296]17277 could not catch an exception when specified -maix64 19870 19871 Solaris-specific 19872 19873 * [297]17505 <cmath> calls acosf(), ceilf(), and other functions 19874 missing from system libraries 19875 19876 HP/UX specific: 19877 19878 * [298]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 19879 19880 ARM-specific 19881 19882 * [299]17384 ICE with mode attribute on structures 19883 19884 MIPS-specific 19885 19886 * [300]17770 No NOP after LWL with -mips1 19887 19888 Other embedded target specific 19889 19890 * [301]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 19891 * [302]14064 [avr-elf] -fdata-sections triggers ICE 19892 * [303]14678 [m68hc11-elf] gcc ICE 19893 * [304]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 19894 * [305]15790 [i686-coff] Alignment error building gcc with i686-coff 19895 target 19896 * [306]15886 [SH] Miscompilation with -O2 -fPIC 19897 * [307]16884 [avr-elf] [fweb related] bug while initializing 19898 variables 19899 19900 Bugs relating to debugger support 19901 19902 * [308]13841 missing debug info for _Complex function arguments 19903 * [309]15860 [big-endian targets] No DW_AT_location debug info is 19904 emitted for formal arguments to a function that uses "register" 19905 qualifiers 19906 19907 Testsuite issues (compiler not affected) 19908 19909 * [310]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 19910 * [311]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 19911 * [312]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 19912 testsuite 19913 19914 Documentation 19915 19916 * [313]15498 typo in gcc manual: non-existing locale example en_UK, 19917 should be en_GB 19918 * [314]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 19919 document broken shell 19920 * [315]16406 USE_LD_AS_NEEDED undocumented 19921 __________________________________________________________________ 19922 19923 [316]GCC 3.4.4 19924 19925 This is the [317]list of problem reports (PRs) from GCC's bug tracking 19926 system that are known to be fixed in the 3.4.4 release. This list might 19927 not be complete (that is, it is possible that some PRs that have been 19928 fixed are not listed here). 19929 __________________________________________________________________ 19930 19931 [318]GCC 3.4.5 19932 19933 This is the [319]list of problem reports (PRs) from GCC's bug tracking 19934 system that are known to be fixed in the 3.4.5 release. This list might 19935 not be complete (that is, it is possible that some PRs that have been 19936 fixed are not listed here). 19937 19938 Bootstrap issues 19939 19940 * [320]24688 sco_math fixincl breaks math.h 19941 19942 C compiler bugs 19943 19944 * [321]17188 struct Foo { } redefinition 19945 * [322]20187 wrong code for ((unsigned char)(unsigned long 19946 long)((a?a:1)&(a*b)))?0:1) 19947 * [323]21873 infinite warning loop on bad array initializer 19948 * [324]21899 enum definition accepts values to be overriden 19949 * [325]22061 ICE in find_function_data, in function.c 19950 * [326]22308 Failure to diagnose violation of constraint 6.516p2 19951 * [327]22458 ICE on missing brace 19952 * [328]22589 ICE casting to long long 19953 * [329]24101 Segfault with preprocessed source 19954 19955 C++ compiler and library bugs 19956 19957 * [330]10611 operations on vector mode not recognized in C++ 19958 * [331]13377 unexpected behavior of namespace usage directive 19959 * [332]16002 Strange error message with new parser 19960 * [333]17413 local classes as template argument 19961 * [334]17609 spurious error message after using keyword 19962 * [335]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 19963 * [336]18124 ICE with invalid template template parameter 19964 * [337]18155 typedef in template declaration not rejected 19965 * [338]18177 ICE with const_cast for undeclared variable 19966 * [339]18368 C++ error message regression 19967 * [340]16378 ICE when returning a copy of a packed member 19968 * [341]18466 int ::i; accepted 19969 * [342]18512 ICE on invalid usage of template base class 19970 * [343]18454 ICE when returning undefined type 19971 * [344]18738 typename not allowed with non-dependent qualified name 19972 * [345]18803 rejects access to operator() in template 19973 * [346]19004 ICE in uses_template_parms, in cp/pt.c 19974 * [347]19208 Spurious error about variably modified type 19975 * [348]18253 bad error message / ICE for invalid template parameter 19976 * [349]19608 ICE after friend function definition in local class 19977 * [350]19884 ICE on explicit instantiation of a non-template 19978 constructor 19979 * [351]20153 ICE when C++ template function contains anonymous union 19980 * [352]20563 Infinite loop in diagnostic (and ice after error 19981 message) 19982 * [353]20789 ICE with incomplete type in template 19983 * [354]21336 Internal compiler error when using custom new operators 19984 * [355]21768 ICE in error message due to violation of coding 19985 conventions 19986 * [356]21853 constness of pointer to data member ignored 19987 * [357]21903 Default argument of template function causes a 19988 compile-time error 19989 * [358]21983 multiple diagnostics 19990 * [359]21987 New testsuite failure 19991 g++.dg/warn/conversion-function-1.C 19992 * [360]22153 ICE on invalid template specialization 19993 * [361]22172 Internal compiler error, seg fault. 19994 * [362]21286 filebuf::xsgetn vs pipes 19995 * [363]22233 ICE with wrong number of template parameters 19996 * [364]22508 ICE after invalid operator new 19997 * [365]22545 ICE with pointer to class member & user defined 19998 conversion operator 19999 * [366]23528 Wrong default allocator in ext/hash_map 20000 * [367]23550 char_traits requirements/1.cc test bad math 20001 * [368]23586 Bad diagnostic for invalid namespace-name 20002 * [369]23624 ICE in invert_truthvalue, in fold-const.c 20003 * [370]23639 Bad error message: not a member of '<declaration error>' 20004 * [371]23797 ICE on typename outside template 20005 * [372]23965 Bogus error message: no matching function for call to 20006 'foo(<type error>)' 20007 * [373]24052 &#`label_decl' not supported by dump_expr#<expression 20008 error> 20009 * [374]24580 virtual base class cause exception not to be caught 20010 20011 Problems in generated debug information 20012 20013 * [375]24267 Bad DWARF for altivec vectors 20014 20015 Optimizations issues 20016 20017 * [376]17810 ICE in verify_local_live_at_start 20018 * [377]17860 Wrong generated code for loop with varying bound 20019 * [378]21709 ICE on compile-time complex NaN 20020 * [379]21964 broken tail call at -O2 or more 20021 * [380]22167 Strange optimization bug when using -Os 20022 * [381]22619 Compilation failure for real_const_1.f and 20023 real_const_2.f90 20024 * [382]23241 Invalid code generated for comparison of uchar to 255 20025 * [383]23478 Miscompilation due to reloading of a var that is also 20026 used in EH pad 20027 * [384]24470 segmentation fault in cc1plus when compiling with -O 20028 * [385]24950 ICE in operand_subword_force 20029 20030 Precompiled headers problems 20031 20032 * [386]14400 Cannot compile qt-x11-free-3.3.0 20033 * [387]14940 PCH largefile test fails on various platforms 20034 20035 Preprocessor bugs 20036 20037 * [388]20239 ICE on empty preprocessed input 20038 * [389]15220 "gcc -E -MM -MG" reports missing system headers in 20039 source directory 20040 20041 Testsuite issues 20042 20043 * [390]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 20044 i686-pc-linux-gnu 20045 20046 Alpha specific 20047 20048 * [391]21888 bootstrap failure with linker relaxation enabled 20049 20050 ARM specific 20051 20052 * [392]15342 [arm-linux]: ICE in verify_local_live_at_start 20053 * [393]23985 Memory aliasing information incorrect in inlined memcpy 20054 20055 ColdFile specific 20056 20057 * [394]16719 Illegal move of byte into address register causes 20058 compiler to ICE 20059 20060 HPPA specific 20061 20062 * [395]21723 ICE while building libgfortran 20063 * [396]21841 -mhp-ld/-mgnu-ld documentation 20064 20065 IA-64 specific 20066 20067 * [397]23644 IA-64 hardware models and configuration options 20068 documentation error 20069 * [398]24718 Shared libgcc not used for linking by default 20070 20071 M68000 specific 20072 20073 * [399]18421 ICE in reload_cse_simplify_operands, in postreload.c 20074 20075 MIPS specific 20076 20077 * [400]20621 ICE in change_address_1, in emit-rtl.c 20078 20079 PowerPC and PowerPC64 specific 20080 20081 * [401]18583 error on valid code: const 20082 __attribute__((altivec(vector__))) doesn't work in arrays 20083 * [402]20191 ICE in reload_cse_simplify_operands 20084 * [403]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 20085 * [404]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 20086 * [405]23404 gij trashes args of functions with more than 8 fp args 20087 * [406]23539 C & C++ compiler generating misaligned references 20088 regardless of compiler flags 20089 * [407]24102 floatdisf2_internal2 broken 20090 * [408]24465 -mminimal-toc miscompilation of __thread vars 20091 20092 Solaris specific 20093 20094 * [409]19933 Problem with define of HUGE_VAL in math_c99 20095 * [410]21889 Native Solaris assembler cannot grok DTP-relative debug 20096 symbols 20097 20098 SPARC specific 20099 20100 * [411]19300 PCH failures on sparc-linux 20101 * [412]20301 Assembler labels have a leading "-" 20102 * [413]20673 C PCH testsuite assembly comparison failure 20103 20104 x86 and x86_64 specific 20105 20106 * [414]18582 ICE with arrays of type V2DF 20107 * [415]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 20108 -fsched2-use-traces 20109 * [416]21716 ICE in reg-stack.c's swap_rtx_condition 20110 * [417]24315 amd64 fails -fpeephole2 20111 __________________________________________________________________ 20112 20113 [418]GCC 3.4.6 20114 20115 This is the [419]list of problem reports (PRs) from GCC's bug tracking 20116 system that are known to be fixed in the 3.4.6 release. This list might 20117 not be complete (that is, it is possible that some PRs that have been 20118 fixed are not listed here). 20119 20120 20121 For questions related to the use of GCC, please consult these web 20122 pages and the [420]GCC manuals. If that fails, the 20123 [421]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20124 web pages and the development of GCC are welcome on our developer 20125 list at [422]gcc (a] gcc.gnu.org. All of [423]our lists have public 20126 archives. 20127 20128 Copyright (C) [424]Free Software Foundation, Inc. Verbatim copying and 20129 distribution of this entire article is permitted in any medium, 20130 provided this notice is preserved. 20131 20132 These pages are [425]maintained by the GCC team. Last modified 20133 2025-01-31. 20134 20135 References 20136 20137 1. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 20138 2. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 20139 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 20140 4. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 20141 5. https://gcc.gnu.org/gcc-3.4/mips-abi.html 20142 6. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 20143 7. https://www.boost.org/ 20144 8. https://gcc.gnu.org/PR11953 20145 9. https://gcc.gnu.org/PR8361 20146 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 20147 11. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 20148 12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 20149 13. https://gcc.gnu.org/bugs/#cxx_rvalbind 20150 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20151 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20152 16. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 20153 17. https://www.gnu.org/software/classpath/ 20154 18. https://www.eclipse.org/ 20155 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 20156 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 20157 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 20158 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 20159 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 20160 24. https://gcc.gnu.org/gcc-3.4/mips-abi.html 20161 25. https://gcc.gnu.org/gcc-3.4/powerpc-abi.html 20162 26. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 20163 27. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 20164 28. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.0 20165 29. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 20166 30. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.1 20167 31. https://gcc.gnu.org/PR10129 20168 32. https://gcc.gnu.org/PR14576 20169 33. https://gcc.gnu.org/PR14760 20170 34. https://gcc.gnu.org/PR14671 20171 35. https://gcc.gnu.org/PR15093 20172 36. https://gcc.gnu.org/PR15178 20173 37. https://gcc.gnu.org/PR12753 20174 38. https://gcc.gnu.org/PR13985 20175 39. https://gcc.gnu.org/PR14810 20176 40. https://gcc.gnu.org/PR14883 20177 41. https://gcc.gnu.org/PR15044 20178 42. https://gcc.gnu.org/PR15057 20179 43. https://gcc.gnu.org/PR15064 20180 44. https://gcc.gnu.org/PR15142 20181 45. https://gcc.gnu.org/PR15159 20182 46. https://gcc.gnu.org/PR15165 20183 47. https://gcc.gnu.org/PR15193 20184 48. https://gcc.gnu.org/PR15209 20185 49. https://gcc.gnu.org/PR15227 20186 50. https://gcc.gnu.org/PR15285 20187 51. https://gcc.gnu.org/PR15299 20188 52. https://gcc.gnu.org/PR15329 20189 53. https://gcc.gnu.org/PR15550 20190 54. https://gcc.gnu.org/PR15554 20191 55. https://gcc.gnu.org/PR15640 20192 56. https://gcc.gnu.org/PR15666 20193 57. https://gcc.gnu.org/PR15696 20194 58. https://gcc.gnu.org/PR15701 20195 59. https://gcc.gnu.org/PR15761 20196 60. https://gcc.gnu.org/PR15829 20197 61. https://gcc.gnu.org/PR14538 20198 62. https://gcc.gnu.org/PR12391 20199 63. https://gcc.gnu.org/PR14649 20200 64. https://gcc.gnu.org/PR15004 20201 65. https://gcc.gnu.org/PR15749 20202 66. https://gcc.gnu.org/PR10646 20203 67. https://gcc.gnu.org/PR12077 20204 68. https://gcc.gnu.org/PR13598 20205 69. https://gcc.gnu.org/PR14211 20206 70. https://gcc.gnu.org/PR14220 20207 71. https://gcc.gnu.org/PR14245 20208 72. https://gcc.gnu.org/PR14340 20209 73. https://gcc.gnu.org/PR14600 20210 74. https://gcc.gnu.org/PR14668 20211 75. https://gcc.gnu.org/PR14775 20212 76. https://gcc.gnu.org/PR14821 20213 77. https://gcc.gnu.org/PR14930 20214 78. https://gcc.gnu.org/PR14932 20215 79. https://gcc.gnu.org/PR14950 20216 80. https://gcc.gnu.org/PR14962 20217 81. https://gcc.gnu.org/PR14975 20218 82. https://gcc.gnu.org/PR15002 20219 83. https://gcc.gnu.org/PR15025 20220 84. https://gcc.gnu.org/PR15046 20221 85. https://gcc.gnu.org/PR15069 20222 86. https://gcc.gnu.org/PR15074 20223 87. https://gcc.gnu.org/PR15083 20224 88. https://gcc.gnu.org/PR15096 20225 89. https://gcc.gnu.org/PR15287 20226 90. https://gcc.gnu.org/PR15317 20227 91. https://gcc.gnu.org/PR15337 20228 92. https://gcc.gnu.org/PR15361 20229 93. https://gcc.gnu.org/PR15412 20230 94. https://gcc.gnu.org/PR15427 20231 95. https://gcc.gnu.org/PR15471 20232 96. https://gcc.gnu.org/PR15503 20233 97. https://gcc.gnu.org/PR15507 20234 98. https://gcc.gnu.org/PR15542 20235 99. https://gcc.gnu.org/PR15565 20236 100. https://gcc.gnu.org/PR15625 20237 101. https://gcc.gnu.org/PR15629 20238 102. https://gcc.gnu.org/PR15742 20239 103. https://gcc.gnu.org/PR15775 20240 104. https://gcc.gnu.org/PR15821 20241 105. https://gcc.gnu.org/PR15862 20242 106. https://gcc.gnu.org/PR15875 20243 107. https://gcc.gnu.org/PR15877 20244 108. https://gcc.gnu.org/PR15947 20245 109. https://gcc.gnu.org/PR16020 20246 110. https://gcc.gnu.org/PR16154 20247 111. https://gcc.gnu.org/PR16174 20248 112. https://gcc.gnu.org/PR14315 20249 113. https://gcc.gnu.org/PR15151 20250 114. https://gcc.gnu.org/PR7993 20251 115. https://gcc.gnu.org/PR15228 20252 116. https://gcc.gnu.org/PR15345 20253 117. https://gcc.gnu.org/PR15945 20254 118. https://gcc.gnu.org/PR15526 20255 119. https://gcc.gnu.org/PR14690 20256 120. https://gcc.gnu.org/PR15112 20257 121. https://gcc.gnu.org/PR15067 20258 122. https://gcc.gnu.org/PR1963 20259 123. https://gcc.gnu.org/PR15717 20260 124. https://gcc.gnu.org/PR14782 20261 125. https://gcc.gnu.org/PR14828 20262 126. https://gcc.gnu.org/PR15202 20263 127. https://gcc.gnu.org/PR14610 20264 128. https://gcc.gnu.org/PR14813 20265 129. https://gcc.gnu.org/PR14857 20266 130. https://gcc.gnu.org/PR15598 20267 131. https://gcc.gnu.org/PR15653 20268 132. https://gcc.gnu.org/PR15189 20269 133. https://gcc.gnu.org/PR15331 20270 134. https://gcc.gnu.org/PR16144 20271 135. https://gcc.gnu.org/PR16176 20272 136. https://gcc.gnu.org/PR11591 20273 137. https://gcc.gnu.org/PR12028 20274 138. https://gcc.gnu.org/PR14478 20275 139. https://gcc.gnu.org/PR14567 20276 140. https://gcc.gnu.org/PR14715 20277 141. https://gcc.gnu.org/PR14902 20278 142. https://gcc.gnu.org/PR14924 20279 143. https://gcc.gnu.org/PR14960 20280 144. https://gcc.gnu.org/PR15106 20281 145. https://gcc.gnu.org/PR16026 20282 146. https://gcc.gnu.org/PR15191 20283 147. https://gcc.gnu.org/PR15662 20284 148. https://gcc.gnu.org/PR15054 20285 149. https://gcc.gnu.org/PR15783 20286 150. https://gcc.gnu.org/PR15626 20287 151. https://gcc.gnu.org/PR14326 20288 152. https://gcc.gnu.org/PR14723 20289 153. https://gcc.gnu.org/PR15290 20290 154. https://gcc.gnu.org/PR15250 20291 155. https://gcc.gnu.org/PR15551 20292 156. https://gcc.gnu.org/PR8309 20293 157. https://gcc.gnu.org/PR13250 20294 158. https://gcc.gnu.org/PR13803 20295 159. https://gcc.gnu.org/PR14093 20296 160. https://gcc.gnu.org/PR14457 20297 161. https://gcc.gnu.org/PR14542 20298 162. https://gcc.gnu.org/PR15100 20299 163. https://gcc.gnu.org/PR15296 20300 164. https://gcc.gnu.org/PR15396 20301 165. https://gcc.gnu.org/PR15782 20302 166. https://gcc.gnu.org/PR11610 20303 167. https://gcc.gnu.org/PR15488 20304 168. https://gcc.gnu.org/PR15489 20305 169. https://gcc.gnu.org/PR13928 20306 170. https://gcc.gnu.org/PR14150 20307 171. https://gcc.gnu.org/PR14949 20308 172. https://gcc.gnu.org/PR15123 20309 173. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.2 20310 174. https://gcc.gnu.org/PR16469 20311 175. https://gcc.gnu.org/PR16344 20312 176. https://gcc.gnu.org/PR16842 20313 177. https://gcc.gnu.org/PR12608 20314 178. https://gcc.gnu.org/PR14492 20315 179. https://gcc.gnu.org/PR15461 20316 180. https://gcc.gnu.org/PR15890 20317 181. https://gcc.gnu.org/PR16180 20318 182. https://gcc.gnu.org/PR16224 20319 183. https://gcc.gnu.org/PR16408 20320 184. https://gcc.gnu.org/PR16529 20321 185. https://gcc.gnu.org/PR16698 20322 186. https://gcc.gnu.org/PR16706 20323 187. https://gcc.gnu.org/PR16810 20324 188. https://gcc.gnu.org/PR16851 20325 189. https://gcc.gnu.org/PR16870 20326 190. https://gcc.gnu.org/PR16904 20327 191. https://gcc.gnu.org/PR16905 20328 192. https://gcc.gnu.org/PR16964 20329 193. https://gcc.gnu.org/PR17068 20330 194. https://gcc.gnu.org/PR16366 20331 195. https://gcc.gnu.org/PR15345 20332 196. https://gcc.gnu.org/PR16590 20333 197. https://gcc.gnu.org/PR16693 20334 198. https://gcc.gnu.org/PR17078 20335 199. https://gcc.gnu.org/PR13956 20336 200. https://gcc.gnu.org/PR16684 20337 201. https://gcc.gnu.org/PR12658 20338 202. https://gcc.gnu.org/PR13092 20339 203. https://gcc.gnu.org/PR15320 20340 204. https://gcc.gnu.org/PR16246 20341 205. https://gcc.gnu.org/PR16273 20342 206. https://gcc.gnu.org/PR16401 20343 207. https://gcc.gnu.org/PR16411 20344 208. https://gcc.gnu.org/PR16489 20345 209. https://gcc.gnu.org/PR16618 20346 210. https://gcc.gnu.org/PR16637 20347 211. https://gcc.gnu.org/PR16717 20348 212. https://gcc.gnu.org/PR16813 20349 213. https://gcc.gnu.org/PR16853 20350 214. https://gcc.gnu.org/PR16889 20351 215. https://gcc.gnu.org/PR16959 20352 216. https://gcc.gnu.org/PR7587 20353 217. https://gcc.gnu.org/PR16473 20354 218. https://gcc.gnu.org/PR16478 20355 219. https://gcc.gnu.org/PR10695 20356 220. https://gcc.gnu.org/PR16974 20357 221. https://gcc.gnu.org/PR16298 20358 222. https://gcc.gnu.org/PR17113 20359 223. https://gcc.gnu.org/PR14697 20360 224. https://gcc.gnu.org/PR15869 20361 225. https://gcc.gnu.org/PR16325 20362 226. https://gcc.gnu.org/PR16357 20363 227. https://gcc.gnu.org/PR16380 20364 228. https://gcc.gnu.org/PR16407 20365 229. https://gcc.gnu.org/PR16643 20366 230. https://gcc.gnu.org/PR15927 20367 231. https://gcc.gnu.org/PR15948 20368 232. https://gcc.gnu.org/PR17019 20369 233. https://gcc.gnu.org/PR16130 20370 234. https://gcc.gnu.org/PR16142 20371 235. https://gcc.gnu.org/PR16278 20372 236. https://gcc.gnu.org/PR16414 20373 237. https://gcc.gnu.org/PR16445 20374 238. https://gcc.gnu.org/PR16490 20375 239. https://gcc.gnu.org/PR16683 20376 240. https://gcc.gnu.org/PR16195 20377 241. https://gcc.gnu.org/PR16239 20378 242. https://gcc.gnu.org/PR16199 20379 243. https://gcc.gnu.org/PR16416 20380 244. https://gcc.gnu.org/PR16430 20381 245. https://gcc.gnu.org/PR16379 20382 246. https://gcc.gnu.org/PR17093 20383 247. https://gcc.gnu.org/PR17119 20384 248. https://gcc.gnu.org/PR15928 20385 249. https://gcc.gnu.org/PR16210 20386 250. https://gcc.gnu.org/PR15488 20387 251. https://gcc.gnu.org/PR16250 20388 252. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.3 20389 253. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 20390 254. https://gcc.gnu.org/PR17369 20391 255. https://gcc.gnu.org/PR17850 20392 256. https://gcc.gnu.org/PR13948 20393 257. https://gcc.gnu.org/PR14492 20394 258. https://gcc.gnu.org/PR16301 20395 259. https://gcc.gnu.org/PR16566 20396 260. https://gcc.gnu.org/PR17023 20397 261. https://gcc.gnu.org/PR17027 20398 262. https://gcc.gnu.org/PR17524 20399 263. https://gcc.gnu.org/PR17826 20400 264. https://gcc.gnu.org/PR15526 20401 265. https://gcc.gnu.org/PR16999 20402 266. https://gcc.gnu.org/PR17503 20403 267. https://gcc.gnu.org/PR17581 20404 268. https://gcc.gnu.org/PR18129 20405 269. https://gcc.gnu.org/PR10975 20406 270. https://gcc.gnu.org/PR11722 20407 271. https://gcc.gnu.org/PR14534 20408 272. https://gcc.gnu.org/PR15172 20409 273. https://gcc.gnu.org/PR15786 20410 274. https://gcc.gnu.org/PR16162 20411 275. https://gcc.gnu.org/PR16612 20412 276. https://gcc.gnu.org/PR16715 20413 277. https://gcc.gnu.org/PR16848 20414 278. https://gcc.gnu.org/PR17132 20415 279. https://gcc.gnu.org/PR17259 20416 280. https://gcc.gnu.org/PR17327 20417 281. https://gcc.gnu.org/PR17393 20418 282. https://gcc.gnu.org/PR17501 20419 283. https://gcc.gnu.org/PR17537 20420 284. https://gcc.gnu.org/PR17585 20421 285. https://gcc.gnu.org/PR17821 20422 286. https://gcc.gnu.org/PR17829 20423 287. https://gcc.gnu.org/PR17851 20424 288. https://gcc.gnu.org/PR17976 20425 289. https://gcc.gnu.org/PR18020 20426 290. https://gcc.gnu.org/PR18093 20427 291. https://gcc.gnu.org/PR18140 20428 292. https://gcc.gnu.org/PR17541 20429 293. https://gcc.gnu.org/PR17853 20430 294. https://gcc.gnu.org/PR17245 20431 295. https://gcc.gnu.org/PR17167 20432 296. https://gcc.gnu.org/PR17277 20433 297. https://gcc.gnu.org/PR17505 20434 298. https://gcc.gnu.org/PR17684 20435 299. https://gcc.gnu.org/PR17384 20436 300. https://gcc.gnu.org/PR17770 20437 301. https://gcc.gnu.org/PR11476 20438 302. https://gcc.gnu.org/PR14064 20439 303. https://gcc.gnu.org/PR14678 20440 304. https://gcc.gnu.org/PR15583 20441 305. https://gcc.gnu.org/PR15790 20442 306. https://gcc.gnu.org/PR15886 20443 307. https://gcc.gnu.org/PR16884 20444 308. https://gcc.gnu.org/PR13841 20445 309. https://gcc.gnu.org/PR15860 20446 310. https://gcc.gnu.org/PR17465 20447 311. https://gcc.gnu.org/PR17469 20448 312. https://gcc.gnu.org/PR18138 20449 313. https://gcc.gnu.org/PR15498 20450 314. https://gcc.gnu.org/PR15747 20451 315. https://gcc.gnu.org/PR16406 20452 316. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.4 20453 317. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 20454 318. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.5 20455 319. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 20456 320. https://gcc.gnu.org/PR24688 20457 321. https://gcc.gnu.org/PR17188 20458 322. https://gcc.gnu.org/PR20187 20459 323. https://gcc.gnu.org/PR21873 20460 324. https://gcc.gnu.org/PR21899 20461 325. https://gcc.gnu.org/PR22061 20462 326. https://gcc.gnu.org/PR22208 20463 327. https://gcc.gnu.org/PR22458 20464 328. https://gcc.gnu.org/PR22589 20465 329. https://gcc.gnu.org/PR24101 20466 330. https://gcc.gnu.org/PR10611 20467 331. https://gcc.gnu.org/PR13377 20468 332. https://gcc.gnu.org/PR16002 20469 333. https://gcc.gnu.org/PR17413 20470 334. https://gcc.gnu.org/PR17609 20471 335. https://gcc.gnu.org/PR17618 20472 336. https://gcc.gnu.org/PR18124 20473 337. https://gcc.gnu.org/PR18155 20474 338. https://gcc.gnu.org/PR18177 20475 339. https://gcc.gnu.org/PR18368 20476 340. https://gcc.gnu.org/PR18378 20477 341. https://gcc.gnu.org/PR18466 20478 342. https://gcc.gnu.org/PR18512 20479 343. https://gcc.gnu.org/PR18545 20480 344. https://gcc.gnu.org/PR18738 20481 345. https://gcc.gnu.org/PR18803 20482 346. https://gcc.gnu.org/PR19004 20483 347. https://gcc.gnu.org/PR19208 20484 348. https://gcc.gnu.org/PR19253 20485 349. https://gcc.gnu.org/PR19608 20486 350. https://gcc.gnu.org/PR19884 20487 351. https://gcc.gnu.org/PR20153 20488 352. https://gcc.gnu.org/PR20563 20489 353. https://gcc.gnu.org/PR20789 20490 354. https://gcc.gnu.org/PR21336 20491 355. https://gcc.gnu.org/PR21768 20492 356. https://gcc.gnu.org/PR21853 20493 357. https://gcc.gnu.org/PR21903 20494 358. https://gcc.gnu.org/PR21983 20495 359. https://gcc.gnu.org/PR21987 20496 360. https://gcc.gnu.org/PR22153 20497 361. https://gcc.gnu.org/PR22172 20498 362. https://gcc.gnu.org/PR21286 20499 363. https://gcc.gnu.org/PR22233 20500 364. https://gcc.gnu.org/PR22508 20501 365. https://gcc.gnu.org/PR22545 20502 366. https://gcc.gnu.org/PR23528 20503 367. https://gcc.gnu.org/PR23550 20504 368. https://gcc.gnu.org/PR23586 20505 369. https://gcc.gnu.org/PR23624 20506 370. https://gcc.gnu.org/PR23639 20507 371. https://gcc.gnu.org/PR23797 20508 372. https://gcc.gnu.org/PR23965 20509 373. https://gcc.gnu.org/PR24052 20510 374. https://gcc.gnu.org/PR24580 20511 375. https://gcc.gnu.org/PR24267 20512 376. https://gcc.gnu.org/PR17810 20513 377. https://gcc.gnu.org/PR17860 20514 378. https://gcc.gnu.org/PR21709 20515 379. https://gcc.gnu.org/PR21964 20516 380. https://gcc.gnu.org/PR22167 20517 381. https://gcc.gnu.org/PR22619 20518 382. https://gcc.gnu.org/PR23241 20519 383. https://gcc.gnu.org/PR23478 20520 384. https://gcc.gnu.org/PR24470 20521 385. https://gcc.gnu.org/PR24950 20522 386. https://gcc.gnu.org/PR14400 20523 387. https://gcc.gnu.org/PR14940 20524 388. https://gcc.gnu.org/PR20239 20525 389. https://gcc.gnu.org/PR15220 20526 390. https://gcc.gnu.org/PR19275 20527 391. https://gcc.gnu.org/PR21888 20528 392. https://gcc.gnu.org/PR15342 20529 393. https://gcc.gnu.org/PR23985 20530 394. https://gcc.gnu.org/PR16719 20531 395. https://gcc.gnu.org/PR21723 20532 396. https://gcc.gnu.org/PR21841 20533 397. https://gcc.gnu.org/PR23644 20534 398. https://gcc.gnu.org/PR24718 20535 399. https://gcc.gnu.org/PR18421 20536 400. https://gcc.gnu.org/PR20621 20537 401. https://gcc.gnu.org/PR18583 20538 402. https://gcc.gnu.org/PR20191 20539 403. https://gcc.gnu.org/PR22083 20540 404. https://gcc.gnu.org/PR23070 20541 405. https://gcc.gnu.org/PR23404 20542 406. https://gcc.gnu.org/PR23539 20543 407. https://gcc.gnu.org/PR24102 20544 408. https://gcc.gnu.org/PR24465 20545 409. https://gcc.gnu.org/PR19933 20546 410. https://gcc.gnu.org/PR21889 20547 411. https://gcc.gnu.org/PR19300 20548 412. https://gcc.gnu.org/PR20301 20549 413. https://gcc.gnu.org/PR20673 20550 414. https://gcc.gnu.org/PR18582 20551 415. https://gcc.gnu.org/PR19340 20552 416. https://gcc.gnu.org/PR21716 20553 417. https://gcc.gnu.org/PR24315 20554 418. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.6 20555 419. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 20556 420. https://gcc.gnu.org/onlinedocs/ 20557 421. mailto:gcc-help (a] gcc.gnu.org 20558 422. mailto:gcc (a] gcc.gnu.org 20559 423. https://gcc.gnu.org/lists.html 20560 424. https://www.fsf.org/ 20561 425. https://gcc.gnu.org/about.html 20562 ====================================================================== 20563 http://gcc.gnu.org/gcc-3.3/index.html 20564 20565 GCC 3.3 Release Series 20566 20567 (This release series is no longer supported.) 20568 20569 May 03, 2005 20570 20571 The [1]GNU project and the GCC developers are pleased to announce the 20572 release of GCC 3.3.6. 20573 20574 This release is a bug-fix release, containing fixes for regressions in 20575 GCC 3.3.5 relative to previous releases of GCC. 20576 20577 This release is the last of the series 3.3.x. 20578 20579 The GCC 3.3 release series includes numerous [2]new features, 20580 improvements, bug fixes, and other changes, thanks to an [3]amazing 20581 group of volunteers. 20582 20583 Release History 20584 20585 GCC 3.3.6 20586 May 3, 2005 ([4]changes) 20587 20588 GCC 3.3.5 20589 September 30, 2004 ([5]changes) 20590 20591 GCC 3.3.4 20592 May 31, 2004 ([6]changes) 20593 20594 GCC 3.3.3 20595 February 14, 2004 ([7]changes) 20596 20597 GCC 3.3.2 20598 October 16, 2003 ([8]changes) 20599 20600 GCC 3.3.1 20601 August 8, 2003 ([9]changes) 20602 20603 GCC 3.3 20604 May 14, 2003 ([10]changes) 20605 20606 References and Acknowledgements 20607 20608 GCC used to stand for the GNU C Compiler, but since the compiler 20609 supports several other languages aside from C, it now stands for the 20610 GNU Compiler Collection. 20611 20612 A list of [11]successful builds is updated as new information becomes 20613 available. 20614 20615 The GCC developers would like to thank the numerous people that have 20616 contributed new features, improvements, bug fixes, and other changes as 20617 well as test results to GCC. This [12]amazing group of volunteers is 20618 what makes GCC successful. 20619 20620 For additional information about GCC please refer to the [13]GCC 20621 project web site or contact the [14]GCC development mailing list. 20622 20623 To obtain GCC please use [15]our mirror sites, or our CVS server. 20624 20625 20626 For questions related to the use of GCC, please consult these web 20627 pages and the [16]GCC manuals. If that fails, the 20628 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20629 web pages and the development of GCC are welcome on our developer 20630 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 20631 archives. 20632 20633 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 20634 distribution of this entire article is permitted in any medium, 20635 provided this notice is preserved. 20636 20637 These pages are [21]maintained by the GCC team. Last modified 20638 2022-10-26. 20639 20640 References 20641 20642 1. http://www.gnu.org/ 20643 2. https://gcc.gnu.org/gcc-3.3/changes.html 20644 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20645 4. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 20646 5. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 20647 6. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 20648 7. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 20649 8. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 20650 9. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 20651 10. https://gcc.gnu.org/gcc-3.3/changes.html 20652 11. https://gcc.gnu.org/gcc-3.3/buildstat.html 20653 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20654 13. https://gcc.gnu.org/index.html 20655 14. mailto:gcc (a] gcc.gnu.org 20656 15. https://gcc.gnu.org/mirrors.html 20657 16. https://gcc.gnu.org/onlinedocs/ 20658 17. mailto:gcc-help (a] gcc.gnu.org 20659 18. mailto:gcc (a] gcc.gnu.org 20660 19. https://gcc.gnu.org/lists.html 20661 20. https://www.fsf.org/ 20662 21. https://gcc.gnu.org/about.html 20663 ====================================================================== 20664 http://gcc.gnu.org/gcc-3.3/changes.html 20665 20666 GCC 3.3 Release Series 20667 Changes, New Features, and Fixes 20668 20669 The latest release in the 3.3 release series is [1]GCC 3.3.6. 20670 20671 Caveats 20672 20673 * The preprocessor no longer accepts multi-line string literals. They 20674 were deprecated in 3.0, 3.1, and 3.2. 20675 * The preprocessor no longer supports the -A- switch when appearing 20676 alone. -A- followed by an assertion is still supported. 20677 * Support for all the systems [2]obsoleted in GCC 3.1 has been 20678 removed from GCC 3.3. See below for a [3]list of systems which are 20679 obsoleted in this release. 20680 * Checking for null format arguments has been decoupled from the rest 20681 of the format checking mechanism. Programs which use the format 20682 attribute may regain this functionality by using the new [4]nonnull 20683 function attribute. Note that all functions for which GCC has a 20684 built-in format attribute, an appropriate built-in nonnull 20685 attribute is also applied. 20686 * The DWARF (version 1) debugging format has been deprecated and will 20687 be removed in a future version of GCC. Version 2 of the DWARF 20688 debugging format will continue to be supported for the foreseeable 20689 future. 20690 * The C and Objective-C compilers no longer accept the "Naming Types" 20691 extension (typedef foo = bar); it was already unavailable in C++. 20692 Code which uses it will need to be changed to use the "typeof" 20693 extension instead: typedef typeof(bar) foo. (We have removed this 20694 extension without a period of deprecation because it has caused the 20695 compiler to crash since version 3.0 and no one noticed until very 20696 recently. Thus we conclude it is not in widespread use.) 20697 * The -traditional C compiler option has been removed. It was 20698 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 20699 available.) The <varargs.h> header, used for writing variadic 20700 functions in traditional C, still exists but will produce an error 20701 message if used. 20702 * GCC 3.3.1 automatically places zero-initialized variables in the 20703 .bss section on some operating systems. Versions of GNU Emacs up to 20704 (and including) 21.3 will not work correctly when using this 20705 optimization; you can use -fno-zero-initialized-in-bss to disable 20706 it. 20707 20708 General Optimizer Improvements 20709 20710 * A new scheme for accurately describing processor pipelines, the 20711 [5]DFA scheduler, has been added. 20712 * Pavel Nejedly, Charles University Prague, has contributed new file 20713 format used by the edge coverage profiler (-fprofile-arcs). 20714 The new format is robust and diagnoses common mistakes where 20715 profiles from different versions (or compilations) of the program 20716 are combined resulting in nonsensical profiles and slow code to 20717 produced with profile feedback. Additionally this format allows 20718 extra data to be gathered. Currently, overall statistics are 20719 produced helping optimizers to identify hot spots of a program 20720 globally replacing the old intra-procedural scheme and resulting in 20721 better code. Note that the gcov tool from older GCC versions will 20722 not be able to parse the profiles generated by GCC 3.3 and vice 20723 versa. 20724 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 20725 pass enabled using -ftracer. This pass simplifies the control flow 20726 of functions allowing other optimizations to do better job. 20727 He also contributed the function reordering pass 20728 (-freorder-functions) to optimize function placement using profile 20729 feedback. 20730 20731 New Languages and Language specific improvements 20732 20733 C/ObjC/C++ 20734 20735 * The preprocessor now accepts directives within macro arguments. It 20736 processes them just as if they had not been within macro arguments. 20737 * The separate ISO and traditional preprocessors have been completely 20738 removed. The front end handles either type of preprocessed output 20739 if necessary. 20740 * In C99 mode preprocessor arithmetic is done in the precision of the 20741 target's intmax_t, as required by that standard. 20742 * The preprocessor can now copy comments inside macros to the output 20743 file when the macro is expanded. This feature, enabled using the 20744 -CC option, is intended for use by applications which place 20745 metadata or directives inside comments, such as lint. 20746 * The method of constructing the list of directories to be searched 20747 for header files has been revised. If a directory named by a -I 20748 option is a standard system include directory, the option is 20749 ignored to ensure that the default search order for system 20750 directories and the special treatment of system header files are 20751 not defeated. 20752 * A few more [6]ISO C99 features now work correctly. 20753 * A new function attribute, nonnull, has been added which allows 20754 pointer arguments to functions to be specified as requiring a 20755 non-null value. The compiler currently uses this information to 20756 issue a warning when it detects a null value passed in such an 20757 argument slot. 20758 * A new type attribute, may_alias, has been added. Accesses to 20759 objects with types with this attribute are not subjected to 20760 type-based alias analysis, but are instead assumed to be able to 20761 alias any other type of objects, just like the char type. 20762 20763 C++ 20764 20765 * Type based alias analysis has been implemented for C++ aggregate 20766 types. 20767 20768 Objective-C 20769 20770 * Generate an error if Objective-C objects are passed by value in 20771 function and method calls. 20772 * When -Wselector is used, check the whole list of selectors at the 20773 end of compilation, and emit a warning if a @selector() is not 20774 known. 20775 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 20776 * No longer need to include objc/objc-class.h to compile self calls 20777 in class methods (NeXT runtime only). 20778 * New -Wundeclared-selector option. 20779 * Removed selector bloating which was causing object files to be 10% 20780 bigger on average (GNU runtime only). 20781 * Using at run time @protocol() objects has been fixed in certain 20782 situations (GNU runtime only). 20783 * Type checking has been fixed and improved in many situations 20784 involving protocols. 20785 20786 Java 20787 20788 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 20789 1.4) API. 20790 * The JDK 1.4 assert facility has been implemented. 20791 * The bytecode interpreter is now direct threaded and thus faster. 20792 20793 Fortran 20794 20795 * Fortran improvements are listed in [7]the Fortran documentation. 20796 20797 Ada 20798 20799 * Ada tasking now works with glibc 2.3.x threading libraries. 20800 20801 New Targets and Target Specific Improvements 20802 20803 * The following changes have been made to the HP-PA port: 20804 + The port now defaults to scheduling for the PA8000 series of 20805 processors. 20806 + Scheduling support for the PA7300 processor has been added. 20807 + The 32-bit port now supports weak symbols under HP-UX 11. 20808 + The handling of initializers and finalizers has been improved 20809 under HP-UX 11. The 64-bit port no longer uses collect2. 20810 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 20811 + ABI fixes to correct the passing of small structures by value. 20812 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 20813 use the DFA processor pipeline description. 20814 * The following NetBSD configurations for the SuperH processor family 20815 have been added: 20816 + SH3, big-endian, sh-*-netbsdelf* 20817 + SH3, little-endian, shle-*-netbsdelf* 20818 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 20819 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 20820 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 20821 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 20822 * The following changes have been made to the IA-32/x86-64 port: 20823 + SSE2 and 3dNOW! intrinsics are now supported. 20824 + Support for thread local storage has been added to the IA-32 20825 and x86-64 ports. 20826 + The x86-64 port has been significantly improved. 20827 * The following changes have been made to the MIPS port: 20828 + All configurations now accept the -mabi switch. Note that you 20829 will need appropriate multilibs for this option to work 20830 properly. 20831 + ELF configurations will always pass an ABI flag to the 20832 assembler, except when the MIPS EABI is selected. 20833 + -mabi=64 no longer selects MIPS IV code. 20834 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 20835 been removed from this release. 20836 + -march now changes the core ISA level. In previous releases, 20837 it would change the use of processor-specific extensions, but 20838 would leave the core ISA unchanged. For example, mips64-elf 20839 -march=r8000 will now generate MIPS IV code. 20840 + Under most configurations, -mipsN now acts as a synonym for 20841 -march. 20842 + There are some new preprocessor macros to describe the -march 20843 and -mtune settings. See the documentation of those options 20844 for details. 20845 + Support for the NEC VR-Series processors has been added. This 20846 includes the 54xx, 5500, and 41xx series. 20847 + Support for the Sandcraft sr71k processor has been added. 20848 * The following changes have been made to the S/390 port: 20849 + Support to build the Java runtime libraries has been added. 20850 Java is now enabled by default on s390-*-linux* and 20851 s390x-*-linux* targets. 20852 + Multilib support for the s390x-*-linux* target has been added; 20853 this allows to build 31-bit binaries using the -m31 option. 20854 + Support for thread local storage has been added. 20855 + Inline assembler code may now use the 'Q' constraint to 20856 specify memory operands without index register. 20857 + Various platform-specific performance improvements have been 20858 implemented; in particular, the compiler now uses the BRANCH 20859 ON COUNT family of instructions and makes more frequent use of 20860 the TEST UNDER MASK family of instructions. 20861 * The following changes have been made to the PowerPC port: 20862 + Support for IBM Power4 processor added. 20863 + Support for Motorola e500 SPE added. 20864 + Support for AIX 5.2 added. 20865 + Function and Data sections now supported on AIX. 20866 + Sibcall optimizations added. 20867 * The support for H8 Tiny is added to the H8/300 port with -mn. 20868 20869 [8]Obsolete Systems 20870 20871 Support for a number of older systems has been declared obsolete in GCC 20872 3.3. Unless there is activity to revive them, the next release of GCC 20873 will have their sources permanently removed. 20874 20875 All configurations of the following processor architectures have been 20876 declared obsolete: 20877 * Matsushita MN10200, mn10200-*-* 20878 * Motorola 88000, m88k-*-* 20879 * IBM ROMP, romp-*-* 20880 20881 Also, some individual systems have been obsoleted: 20882 * Alpha 20883 + Interix, alpha*-*-interix* 20884 + Linux libc1, alpha*-*-linux*libc1* 20885 + Linux ECOFF, alpha*-*-linux*ecoff* 20886 * ARM 20887 + Generic a.out, arm*-*-aout* 20888 + Conix, arm*-*-conix* 20889 + "Old ABI," arm*-*-oabi 20890 + StrongARM/COFF, strongarm-*-coff* 20891 * HPPA (PA-RISC) 20892 + Generic OSF, hppa1.0-*-osf* 20893 + Generic BSD, hppa1.0-*-bsd* 20894 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 20895 + HiUX, hppa*-*-hiux* 20896 + Mach Lites, hppa*-*-lites* 20897 * Intel 386 family 20898 + Windows NT 3.x, i?86-*-win32 20899 * MC68000 family 20900 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 20901 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 20902 m68k-sun-mach* 20903 + AT&T systems, m68000-att-sysv* 20904 + Atari systems, m68k-atari-sysv* 20905 + Motorola systems, m68k-motorola-sysv* 20906 + NCR systems, m68k-ncr-sysv* 20907 + Plexus systems, m68k-plexus-sysv* 20908 + Commodore systems, m68k-cbm-sysv* 20909 + Citicorp TTI, m68k-tti-* 20910 + Unos, m68k-crds-unos* 20911 + Concurrent RTU, m68k-ccur-rtu* 20912 + Linux a.out, m68k-*-linux*aout* 20913 + Linux libc1, m68k-*-linux*libc1* 20914 + pSOS, m68k-*-psos* 20915 * MIPS 20916 + Generic ECOFF, mips*-*-ecoff* 20917 + SINIX, mips-sni-sysv4 20918 + Orion RTEMS, mips64orion-*-rtems* 20919 * National Semiconductor 32000 20920 + OpenBSD, ns32k-*-openbsd* 20921 * POWER (aka RS/6000) and PowerPC 20922 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 20923 + Bull BOSX, rs6000-bull-bosx 20924 + Generic Mach, rs6000-*-mach* 20925 + Generic SysV, powerpc*-*-sysv* 20926 + Linux libc1, powerpc*-*-linux*libc1* 20927 * Sun SPARC 20928 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 20929 sparclite-*-aout*, and sparc86x-*-aout* 20930 + NetBSD a.out, sparc-*-netbsd*aout* 20931 + Generic BSD, sparc-*-bsd* 20932 + ChorusOS, sparc-*-chorusos* 20933 + Linux a.out, sparc-*-linux*aout* 20934 + Linux libc1, sparc-*-linux*libc1* 20935 + LynxOS, sparc-*-lynxos* 20936 + Solaris on HAL hardware, sparc-hal-solaris2* 20937 + SunOS versions 3 and 4, sparc-*-sunos[34]* 20938 * NEC V850 20939 + RTEMS, v850-*-rtems* 20940 * VAX 20941 + VMS, vax-*-vms* 20942 20943 Documentation improvements 20944 20945 Other significant improvements 20946 20947 * Almost all front-end dependencies in the compiler have been 20948 separated out into a set of language hooks. This should make adding 20949 a new front end clearer and easier. 20950 * One effect of removing the separate preprocessor is a small 20951 increase in the robustness of the compiler in general, and the 20952 maintainability of target descriptions. Previously target-specific 20953 built-in macros and others, such as __FAST_MATH__, had to be 20954 handled with so-called specs that were hard to maintain. Often they 20955 would fail to behave properly when conflicting options were 20956 supplied on the command line, and define macros in the user's 20957 namespace even when strict ISO compliance was requested. 20958 Integrating the preprocessor has cleanly solved these issues. 20959 * The Makefile suite now supports redirection of make install by 20960 means of the variable DESTDIR. 20961 __________________________________________________________________ 20962 20963 GCC 3.3 20964 20965 Detailed release notes for the GCC 3.3 release follow. 20966 20967 Bug Fixes 20968 20969 bootstrap failures 20970 20971 * [9]10140 cross compiler build failures: missing __mempcpy (DUP: 20972 [10]10198,[11]10338) 20973 20974 Internal compiler errors (multi-platform) 20975 20976 * [12]3581 large string causes segmentation fault in cc1 20977 * [13]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 20978 * [14]5533 (c++) ICE when processing std::accumulate(begin, end, 20979 init, invalid_op) 20980 * [15]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 20981 * [16]6412 (c++) ICE in retrieve_specialization 20982 * [17]6620 (c++) partial template specialization causes an ICE 20983 (segmentation fault) 20984 * [18]6663 (c++) ICE with attribute aligned 20985 * [19]7068 ICE with incomplete types 20986 * [20]7083 (c++) ICE using -gstabs with dodgy class derivation 20987 * [21]7647 (c++) ICE when data member has the name of the enclosing 20988 class 20989 * [22]7675 ICE in fixup_var_refs_1 20990 * [23]7718 'complex' template instantiation causes ICE 20991 * [24]8116 (c++) ICE in member template function 20992 * [25]8358 (ada) Ada compiler accesses freed memory, crashes 20993 * [26]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 20994 fault 20995 * [27]8564 (c++) ICE in find_function_data, in function.c 20996 * [28]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 20997 * [29]8766 (c++) ICE after failed initialization of static template 20998 variable 20999 * [30]8803 ICE in instantiate_virtual_regs_1, in function.c 21000 * [31]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 21001 * [32]8906 (c++) ICE (Segmentation fault) when parsing nested-class 21002 definition 21003 * [33]9216 (c++) ICE on missing template parameter 21004 * [34]9261 (c++) ICE in arg_assoc, in cp/decl2.c 21005 * [35]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 21006 loop 21007 * [36]9429 (c++) ICE in template instantiation with a pointered new 21008 operator 21009 * [37]9516 Internal error when using a big array 21010 * [38]9600 (c++) ICE with typedefs in template class 21011 * [39]9629 (c++) virtual inheritance segfault 21012 * [40]9672 (c++) ICE: Error reporting routines re-entered 21013 * [41]9749 (c++) ICE in write_expression on invalid function 21014 prototype 21015 * [42]9794 (fortran) ICE: floating point exception during constant 21016 folding 21017 * [43]9829 (c++) Missing colon in nested namespace usage causes ICE 21018 * [44]9916 (c++) ICE with noreturn function in ?: statement 21019 * [45]9936 ICE with local function and variable-length 2d array 21020 * [46]10262 (c++) cc1plus crashes with large generated code 21021 * [47]10278 (c++) ICE in parser for invalid code 21022 * [48]10446 (c++) ICE on definition of nonexistent member function of 21023 nested class in a class template 21024 * [49]10451 (c++) ICE in grokdeclarator on spurious mutable 21025 declaration 21026 * [50]10506 (c++) ICE in build_new at cp/init.c with 21027 -fkeep-inline-functions and multiple inheritance 21028 * [51]10549 (c++) ICE in store_bit_field on bitfields that exceed the 21029 precision of the declared type 21030 21031 Optimization bugs 21032 21033 * [52]2001 Inordinately long compile times in reload CSE regs 21034 * [53]2391 Exponential compilation time explosion in combine 21035 * [54]2960 Duplicate loop conditions even with -Os 21036 * [55]4046 redundant conditional branch 21037 * [56]6405 Loop-unrolling related performance regressions 21038 * [57]6798 very long compile time with large case-statement 21039 * [58]6871 const objects shouldn't be moved to .bss 21040 * [59]6909 problem w/ -Os on modified loop-2c.c test case 21041 * [60]7189 gcc -O2 -Wall does not print ``control reaches end of 21042 non-void function'' warning 21043 * [61]7642 optimization problem with signbit() 21044 * [62]8634 incorrect code for inlining of memcpy under -O2 21045 * [63]8750 Cygwin prolog generation erroneously emitting __alloca as 21046 regular function call 21047 21048 C front end 21049 21050 * [64]2161 long if-else cascade overflows parser stack 21051 * [65]4319 short accepted on typedef'd char 21052 * [66]8602 incorrect line numbers in warning messages when using 21053 inline functions 21054 * [67]9177 -fdump-translation-unit: C front end deletes function_decl 21055 AST nodes and breaks debugging dumps 21056 * [68]9853 miscompilation of non-constant structure initializer 21057 21058 c++ compiler and library 21059 21060 * [69]45 legal template specialization code is rejected (DUP: 21061 [70]3784) 21062 * [71]764 lookup failure: friend operator and dereferencing a pointer 21063 and templates (DUP: [72]5116) 21064 * [73]2862 gcc accepts invalid explicit instantiation syntax (DUP: 21065 2863) 21066 * [74]3663 G++ doesn't check access control during template 21067 instantiation 21068 * [75]3797 gcc fails to emit explicit specialization of a template 21069 member 21070 * [76]3948 Two destructors are called when no copy destructor is 21071 defined (ABI change) 21072 * [77]4137 Conversion operator within template is not accepted 21073 * [78]4361 bogus ambiguity taking the address of a member template 21074 * [79]4802 g++ accepts illegal template code (access to private 21075 member; DUP: [80]5837) 21076 * [81]4803 inline function is used but never defined, and g++ does 21077 not object 21078 * [82]5094 Partial specialization cannot be friend? 21079 * [83]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 21080 * [84]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 21081 time 21082 * [85]7015 certain __asm__ constructs rejected 21083 * [86]7086 compile time regression (quadratic behavior in 21084 fixup_var_refs) 21085 * [87]7099 G++ doesn't set the noreturn attribute on std::exit and 21086 std::abort 21087 * [88]7247 copy constructor missing when inlining enabled (invalid 21088 optimization?) 21089 * [89]7441 string array initialization compilation time regression 21090 from seconds to minutes 21091 * [90]7768 __PRETTY_FUNCTION__ for template destructor is wrong 21092 * [91]7804 bad printing of floating point constant in warning message 21093 * [92]8099 Friend classes and template specializations 21094 * [93]8117 member function pointers and multiple inheritance 21095 * [94]8205 using declaration and multiple inheritance 21096 * [95]8645 unnecessary non-zero checks in stl_tree.h 21097 * [96]8724 explicit destructor call for incomplete class allowed 21098 * [97]8805 compile time regression with many member variables 21099 * [98]8691 -O3 and -fno-implicit-templates are incompatible 21100 * [99]8700 unhelpful error message for binding temp to reference 21101 * [100]8724 explicit destructor call for incomplete class allowed 21102 * [101]8949 numeric_limits<>::denorm_min() and is_iec559 problems 21103 * [102]9016 Failure to consistently constant fold "constant" C++ 21104 objects 21105 * [103]9053 g++ confused about ambiguity of overloaded function 21106 templates 21107 * [104]9152 undefined virtual thunks 21108 * [105]9182 basic_filebuf<> does not report errors in codecvt<>::out 21109 * [106]9297 data corruption due to codegen bug (when copying.) 21110 * [107]9318 i/ostream::operator>>/<<(streambuf*) broken 21111 * [108]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 21112 * [109]9400 bogus -Wshadow warning: shadowed declaration of this in 21113 local classes 21114 * [110]9424 i/ostream::operator>>/<<(streambuf*) drops characters 21115 * [111]9425 filebuf::pbackfail broken (DUP: [112]9439) 21116 * [113]9474 GCC freezes in compiling a weird code mixing <iostream> 21117 and <iostream.h> 21118 * [114]9548 Incorrect results from setf(ios::fixed) and precision(-1) 21119 [115][DR 231] 21120 * [116]9555 ostream inserters fail to set badbit on exception 21121 * [117]9561 ostream inserters rethrow exception of wrong type 21122 * [118]9563 ostream::sentry returns true after a failed preparation 21123 * [119]9582 one-definition rule violation in std::allocator 21124 * [120]9622 __PRETTY_FUNCTION__ incorrect in template destructors 21125 * [121]9683 bug in initialization chains for static const variables 21126 from template classes 21127 * [122]9791 -Woverloaded-virtual reports hiding of destructor 21128 * [123]9817 collate::compare doesn't handle nul characters 21129 * [124]9825 filebuf::sputbackc breaks sbumpc 21130 * [125]9826 operator>>(basic_istream, basic_string) fails to compile 21131 with custom traits 21132 * [126]9924 Multiple using statements for builtin functions not 21133 allowed 21134 * [127]9946 destructor is not called for temporary object 21135 * [128]9964 filebuf::close() sometimes fails to close file 21136 * [129]9988 filebuf::overflow writes EOF to file 21137 * [130]10033 optimization breaks polymorphic references w/ typeid 21138 operator 21139 * [131]10097 filebuf::underflow drops characters 21140 * [132]10132 filebuf destructor can throw exceptions 21141 * [133]10180 gcc fails to warn about non-inlined function 21142 * [134]10199 method parametrized by template does not work everywhere 21143 * [135]10300 use of array-new (nothrow) in segfaults on NULL return 21144 * [136]10427 Stack corruption with variable-length automatic arrays 21145 and virtual destructors 21146 * [137]10503 Compilation never stops in fixed_type_or_null 21147 21148 Objective-C 21149 21150 * [138]5956 selectors aren't matched properly when added to the 21151 selector table 21152 21153 Fortran compiler and library 21154 21155 * [139]1832 list directed i/o overflow hangs, -fbounds-check doesn't 21156 detect 21157 * [140]3924 g77 generates code that is rejected by GAS if COFF debug 21158 info requested 21159 * [141]5634 doc: explain that configure --prefix=~/... does not work 21160 * [142]6367 multiple repeat counts confuse namelist read into array 21161 * [143]6491 Logical operations error on logicals when using 21162 -fugly-logint 21163 * [144]6742 Generation of C++ Prototype for FORTRAN and extern "C" 21164 * [145]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 21165 on irix6.5 21166 * [146]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 21167 assume a direct access file 21168 * [147]7278 g77 "bug"; the executable misbehaves (with -O2 21169 -fno-automatic) 21170 * [148]7384 DATE_AND_TIME milliseconds field inactive on Windows 21171 * [149]7388 Incorrect output with 0-based array of characters 21172 * [150]8587 Double complex zero ** double precision number -> NaN 21173 instead of zero 21174 * [151]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 21175 unknown register name line-length-none 21176 * [152]10197 Direct access files not unformatted by default 21177 21178 Java compiler and library 21179 21180 * [153]6005 gcj fails to build rhug on alpha 21181 * [154]6389 System.getProperty("") should always throw an 21182 IllegalArgumentException 21183 * [155]6576 java.util.ResourceBundle.getResource ignores locale 21184 * [156]6652 new java.io.File("").getCanonicalFile() throws exception 21185 * [157]7060 getMethod() doesn't search super interface 21186 * [158]7073 bytecode interpreter gives wrong answer for interface 21187 getSuperclass() 21188 * [159]7180 possible bug in 21189 javax.naming.spi.NamingManager.getPlusPath() 21190 * [160]7416 java.security startup refs "GNU libgcj.security" 21191 * [161]7570 Runtime.exec with null envp: child doesn't inherit parent 21192 env (DUP: [162]7578) 21193 * [163]7611 Internal error while compiling libjava with -O 21194 * [164]7709 NullPointerException in _Jv_ResolvePoolEntry 21195 * [165]7766 ZipInputStream.available returns 0 immediately after 21196 construction 21197 * [166]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 21198 * [167]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 21199 * [168]8142 '$' in class names vs. dlopen 'dynamic string tokens' 21200 * [169]8234 ZipInputStream chokes when InputStream.read() returns 21201 small chunks 21202 * [170]8415 reflection bug: exception info for Method 21203 * [171]8481 java.Random.nextInt(int) may return negative 21204 * [172]8593 Error reading GZIPped files with BufferedReader 21205 * [173]8759 java.beans.Introspector has no flushCaches() or 21206 flushFromCaches() methods 21207 * [174]8997 spin() calls Thread.sleep 21208 * [175]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 21209 instead of the root content of C: 21210 * [176]9254 java::lang::Object::wait(), threads-win32.cc returns 21211 wrong return codes 21212 * [177]9271 Severe bias in java.security.SecureRandom 21213 21214 Ada compiler and library 21215 21216 * [178]6767 make gnatlib-shared fails on -laddr2line 21217 * [179]9911 gnatmake fails to link when GCC configured with 21218 --with-sjlj-exceptions=yes 21219 * [180]10020 Can't bootstrap gcc on AIX with Ada enabled 21220 * [181]10546 Ada tasking not working on Red Hat 9 21221 21222 preprocessor 21223 21224 * [182]7029 preprocessor should ignore #warning with -M 21225 21226 ARM-specific 21227 21228 * [183]2903 [arm] Optimization bug with long long arithmetic 21229 * [184]7873 arm-linux-gcc fails when assigning address to a bit field 21230 21231 FreeBSD-specific 21232 21233 * [185]7680 float functions undefined in math.h/cmath with #define 21234 _XOPEN_SOURCE 21235 21236 HP-UX or HP-PA-specific 21237 21238 * [186]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 21239 * [187]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 21240 fputc_unlocked 21241 * [188]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 21242 21243 m68hc11-specific 21244 21245 * [189]6744 Bad assembler code generated: reference to pseudo 21246 register z 21247 * [190]7361 Internal compiler error in reload_cse_simplify_operands, 21248 in reload1.c 21249 21250 MIPS-specific 21251 21252 * [191]9496 [mips-linux] bug in optimizer? 21253 21254 PowerPC-specific 21255 21256 * [192]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 21257 space 21258 * [193]8480 reload ICEs for LAPACK code on powerpc64-linux 21259 * [194]8784 [AIX] Internal compiler error in simplify_gen_subreg 21260 * [195]10315 [powerpc] ICE: in extract_insn, in recog.c 21261 21262 SPARC-specific 21263 21264 * [196]10267 (documentation) Wrong build instructions for 21265 *-*-solaris2* 21266 21267 x86-specific (Intel/AMD) 21268 21269 * [197]7916 ICE in instantiate_virtual_register_1 21270 * [198]7926 (c++) i486 instructions in header files make c++ programs 21271 crash on i386 21272 * [199]8555 ICE in gen_split_1231 21273 * [200]8994 ICE with -O -march=pentium4 21274 * [201]9426 ICE with -fssa -funroll-loops -fprofile-arcs 21275 * [202]9806 ICE in inline assembly with -fPIC flag 21276 * [203]10077 gcc -msse2 generates movd to move dwords between xmm 21277 regs 21278 * [204]10233 64-bit comparison only comparing bottom 32-bits 21279 * [205]10286 type-punning doesn't work with __m64 and -O 21280 * [206]10308 [x86] ICE with -O -fgcse or -O2 21281 __________________________________________________________________ 21282 21283 [207]GCC 3.3.1 21284 21285 Bug Fixes 21286 21287 This section lists the problem reports (PRs) from GCC's bug tracking 21288 system that are known to be fixed in the 3.3.1 release. This list might 21289 not be complete (that is, it is possible that some PRs that have been 21290 fixed are not listed here). 21291 21292 Bootstrap failures 21293 21294 * [208]11272 [Solaris] make bootstrap fails while building libstdc++ 21295 21296 Internal compiler errors (multi-platform) 21297 21298 * [209]5754 ICE on invalid nested template class 21299 * [210]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 21300 and --enable-checking 21301 * [211]6949 (c++) ICE in tsubst_decl, in cp/pt.c 21302 * [212]7053 (c++) ICE when declaring a function already defined as a 21303 friend method of a template class 21304 * [213]8164 (c++) ICE when using different const expressions as 21305 template parameter 21306 * [214]8384 (c++) ICE in is_base_type, in dwarf2out.c 21307 * [215]9559 (c++) ICE with invalid initialization of a static const 21308 * [216]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 21309 when redeclaring a static member variable 21310 * [217]9864 (fortran) ICE in add_abstract_origin_attribute, in 21311 dwarfout.c with -g -O -finline-functions 21312 * [218]10432 (c++) ICE in poplevel, in cp/decl.c 21313 * [219]10475 ICE in subreg_highpart_offset for code with long long 21314 * [220]10635 (c++) ICE when dereferencing an incomplete type casted 21315 from a void pointer 21316 * [221]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 21317 instantiating static member variables 21318 * [222]10700 ICE in copy_to_mode_reg on 64-bit targets 21319 * [223]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 21320 * [224]10796 (c++) ICE when defining an enum with two values: -1 and 21321 MAX_INT_64BIT 21322 * [225]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 21323 sched.c 21324 * [226]10939 (c++) ICE with template code 21325 * [227]10956 (c++) ICE when specializing a template member function 21326 of a template class, in tsubst, in cp/pt.c 21327 * [228]11041 (c++) ICE: const myclass &x = *x; (when operator*() 21328 defined) 21329 * [229]11059 (c++) ICE with empty union 21330 * [230]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 21331 -O2 -fnon-call-exceptions 21332 * [231]11105 (c++) ICE in mangle_conv_op_name_for_type 21333 * [232]11149 (c++) ICE on error when instantiation with call function 21334 of a base type 21335 * [233]11228 (c++) ICE on new-expression using array operator new and 21336 default-initialization 21337 * [234]11282 (c++) Infinite memory usage after syntax error 21338 * [235]11301 (fortran) ICE with -fno-globals 21339 * [236]11308 (c++) ICE when using an enum type name as if it were a 21340 class or namespace 21341 * [237]11473 (c++) ICE with -gstabs when empty struct inherits from 21342 an empty struct 21343 * [238]11503 (c++) ICE when instantiating template with ADDR_EXPR 21344 * [239]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 21345 template member functions 21346 21347 Optimization bugs 21348 21349 * [240]11198 -O2 -frename-registers generates wrong code (aliasing 21350 problem) 21351 * [241]11304 Wrong code production with -fomit-frame-pointer 21352 * [242]11381 volatile memory access optimized away 21353 * [243]11536 [strength-reduce] -O2 optimization produces wrong code 21354 * [244]11557 constant folding bug generates wrong code 21355 21356 C front end 21357 21358 * [245]5897 No warning for statement after return 21359 * [246]11279 DWARF-2 output mishandles large enums 21360 21361 Preprocessor bugs 21362 21363 * [247]11022 no warning for non-compatible macro redefinition 21364 21365 C++ compiler and library 21366 21367 * [248]2330 static_cast<>() to a private base is allowed 21368 * [249]5388 Incorrect message "operands to ?: have different types" 21369 * [250]5390 Libiberty fails to demangle multi-digit template 21370 parameters 21371 * [251]7877 Incorrect parameter passing to specializations of member 21372 function templates 21373 * [252]9393 Anonymous namespaces and compiling the same file twice 21374 * [253]10032 -pedantic converts some errors to warnings 21375 * [254]10468 const typeof(x) is non-const, but only in templates 21376 * [255]10527 confused error message with "new int()" parameter 21377 initializer 21378 * [256]10679 parameter MIN_INLINE_INSNS is not honored 21379 * [257]10682 gcc chokes on a typedef for an enum inside a class 21380 template 21381 * [258]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 21382 0. 21383 * [259]10845 template member function (with nested template as 21384 parameter) cannot be called anymore if another unrelated template 21385 member function is defined 21386 * [260]10849 Cannot define an out-of-class specialization of a 21387 private nested template class 21388 * [261]10888 Suppress -Winline warnings for system headers 21389 * [262]10929 -Winline warns about functions for which no definition 21390 is visible 21391 * [263]10931 valid conversion static_cast<const unsigned 21392 int&>(lvalue-of-type-int) is rejected 21393 * [264]10940 Bad code with explicit specialization 21394 * [265]10968 If member function implicitly instantiated, explicit 21395 instantiation of class fails to instantiate it 21396 * [266]10990 Cannot convert with dynamic_cast<> to a private base 21397 class from within a member function 21398 * [267]11039 Bad interaction between implicit typename deprecation 21399 and friendship 21400 * [268]11062 (libstdc++) avoid __attribute__ ((unused)); say 21401 "__unused__" instead 21402 * [269]11095 C++ iostream manipulator causes segfault when called 21403 with negative argument 21404 * [270]11098 g++ doesn't emit complete debugging information for 21405 local variables in destructors 21406 * [271]11137 GNU/Linux shared library constructors not called unless 21407 there's one global object 21408 * [272]11154 spurious ambiguity report for template class 21409 specialization 21410 * [273]11329 Compiler cannot find user defined implicit typecast 21411 * [274]11332 Spurious error with casts in ?: expression 21412 * [275]11431 static_cast behavior with subclasses when default 21413 constructor available 21414 * [276]11528 money_get facet does not accept "$.00" as valid 21415 * [277]11546 Type lookup problems in out-of-line definition of a 21416 class doubly nested from a template class 21417 * [278]11567 C++ code containing templated member function with same 21418 name as pure virtual member function results in linking failure 21419 * [279]11645 Failure to deal with using and private inheritance 21420 21421 Java compiler and library 21422 21423 * [280]5179 Qualified static field access doesn't initialize its 21424 class 21425 * [281]8204 gcj -O2 to native reorders certain instructions 21426 improperly 21427 * [282]10838 java.io.ObjectInputStream syntax error 21428 * [283]10886 The RMI registry that comes with GCJ does not work 21429 correctly 21430 * [284]11349 JNDI URL context factories not located correctly 21431 21432 x86-specific (Intel/AMD) 21433 21434 * [285]4823 ICE on inline assembly code 21435 * [286]8878 miscompilation with -O and SSE 21436 * [287]9815 (c++ library) atomicity.h - fails to compile with -O3 21437 -masm=intel 21438 * [288]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 21439 in reload1.c 21440 * [289]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 21441 * [290]10673 ICE for x86-64 on freebsd libc vfprintf.c source 21442 * [291]11044 [x86] out of range loop instructions for FP code on K6 21443 * [292]11089 ICE: instantiate_virtual_regs_lossage while using SSE 21444 built-ins 21445 * [293]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 21446 is used 21447 21448 SPARC- or Solaris- specific 21449 21450 * [294]9362 solaris 'as' dies when fed .s and "-gstabs" 21451 * [295]10142 [SPARC64] gcc produces wrong code when passing 21452 structures by value 21453 * [296]10663 New configure check aborts with Sun tools. 21454 * [297]10835 combinatorial explosion in scheduler on HyperSPARC 21455 * [298]10876 ICE in calculate_giv_inc when building KDE 21456 * [299]10955 wrong code at -O3 for structure argument in context of 21457 structure return 21458 * [300]11018 -mcpu=ultrasparc busts tar-1.13.25 21459 * [301]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 21460 Linux kernel 21461 21462 ia64 specific 21463 21464 * [302]10907 gcc violates the ia64 ABI (GP must be preserved) 21465 * [303]11320 scheduler bug (in machine depended reorganization pass) 21466 * [304]11599 bug with conditional and __builtin_prefetch 21467 21468 PowerPC specific 21469 21470 * [305]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 21471 during loop) 21472 * [306]10871 error in rs6000_stack_info save_size computation 21473 * [307]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 21474 cures it 21475 21476 m68k-specific 21477 21478 * [308]7594 [m68k] ICE on legal code associated with simplify-rtx 21479 * [309]10557 [m68k] ICE in subreg_offset_representable_p 21480 * [310]11054 [m68k] ICE in reg_overlap_mentioned_p 21481 21482 ARM-specific 21483 21484 * [311]10834 [arm] GCC 3.3 still generates incorrect instructions for 21485 functions with __attribute__ ((interrupt ("IRQ"))) 21486 * [312]10842 [arm] Clobbered link register is copied to pc under 21487 certain circumstances 21488 * [313]11052 [arm] noce_process_if_block() can lose REG_INC notes 21489 * [314]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 21490 (3.4) 21491 21492 MIPS-specific 21493 21494 * [315]11084 ICE in propagate_one_insn, in flow.c 21495 21496 SH-specific 21497 21498 * [316]10331 can't compile c++ part of gcc cross compiler for sh-elf 21499 * [317]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 21500 * [318]11096 i686-linux to sh-linux cross compiler fails to compile 21501 C++ files 21502 21503 GNU/Linux (or Hurd?) specific 21504 21505 * [319]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 21506 21507 UnixWare specific 21508 21509 * [320]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 21510 7.1.1 21511 21512 Cygwin (or mingw) specific 21513 21514 * [321]5287 ICE with dllimport attribute 21515 * [322]10148 [MingW/CygWin] Compiler dumps core 21516 21517 DJGPP specific 21518 21519 * [323]8787 GCC fails to emit .intel_syntax when invoked with 21520 -masm=intel on DJGPP 21521 21522 Darwin (and MacOS X) specific 21523 21524 * [324]10900 trampolines crash 21525 21526 Documentation 21527 21528 * [325]1607 (c++) Format attributes on methods undocumented 21529 * [326]4252 Invalid option `-fdump-translation-unit' 21530 * [327]4490 Clarify restrictions on -m96bit-long-double, 21531 -m128bit-long-double 21532 * [328]10355 document an issue with regparm attribute on some systems 21533 (e.g. Solaris) 21534 * [329]10726 (fortran) Documentation for function "IDate Intrinsic 21535 (Unix)" is wrong 21536 * [330]10805 document bug in old version of Sun assembler 21537 * [331]10815 warn against GNU binutils on AIX 21538 * [332]10877 document need for newer binutils on i?86-*-linux-gnu 21539 * [333]11280 Manual incorrect with respect to -freorder-blocks 21540 * [334]11466 Document -mlittle-endian and its restrictions for the 21541 sparc64 port 21542 21543 Testsuite bugs (compiler itself is not affected) 21544 21545 * [335]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 21546 report failure 21547 * [336]10810 gcc-3.3 fails make check: buffer overrun in 21548 test_demangle.c 21549 __________________________________________________________________ 21550 21551 [337]GCC 3.3.2 21552 21553 Bug Fixes 21554 21555 This section lists the problem reports (PRs) from GCC's bug tracker 21556 that are known to be fixed in the 3.3.2 release. This list might not be 21557 complete (that is, it is possible that some PRs that have been fixed 21558 are not listed here). 21559 21560 Bootstrap failures and problems 21561 21562 * [338]8336 [SCO5] bootstrap config still tries to use COFF options 21563 * [339]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 21564 --enable-threads=posix 21565 * [340]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 21566 * [341]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 21567 7.1.1) 21568 * [342]11687 xstormy16-elf build fails in libf2c 21569 * [343]12263 [SGI IRIX] bootstrap fails during compile of 21570 libf2c/libI77/backspace.c 21571 * [344]12490 buffer overflow in scan-decls.c (during Solaris 9 21572 fix-header processing) 21573 21574 Internal compiler errors (multi-platform) 21575 21576 * [345]7277 Casting integers to vector types causes ICE 21577 * [346]7939 (c++) ICE on invalid function template specialization 21578 * [347]11063 (c++) ICE on parsing initialization list of const array 21579 member 21580 * [348]11207 ICE with negative index in array element designator 21581 * [349]11522 (fortran) g77 dwarf-2 ICE in 21582 add_abstract_origin_attribute 21583 * [350]11595 (c++) ICE on duplicate label definition 21584 * [351]11646 (c++) ICE in commit_one_edge_insertion with 21585 -fnon-call-exceptions -fgcse -O 21586 * [352]11665 ICE in struct initializer when taking address 21587 * [353]11852 (c++) ICE with bad struct initializer. 21588 * [354]11878 (c++) ICE in cp_expr_size 21589 * [355]11883 ICE with any -O on mercury-generated C code 21590 * [356]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 21591 cp/typeck2.c when applying typeid operator to template template 21592 parameter 21593 * [357]12146 ICE in lookup_template_function, in cp/pt.c 21594 * [358]12215 ICE in make_label_edge with -fnon-call-exceptions 21595 -fno-gcse -O2 21596 * [359]12369 (c++) ICE with templates and friends 21597 * [360]12446 ICE in emit_move_insn on complicated array reference 21598 * [361]12510 ICE in final_scan_insn 21599 * [362]12544 ICE with large parameters used in nested functions 21600 21601 C and optimization bugs 21602 21603 * [363]9862 spurious warnings with -W -finline-functions 21604 * [364]10962 lookup_field is a linear search on a linked list (can be 21605 slow if large struct) 21606 * [365]11370 -Wunreachable-code gives false complaints 21607 * [366]11637 invalid assembly with -fnon-call-exceptions 21608 * [367]11885 Problem with bitfields in packed structs 21609 * [368]12082 Inappropriate unreachable code warnings 21610 * [369]12180 Inline optimization fails for variadic function 21611 * [370]12340 loop unroller + gcse produces wrong code 21612 21613 C++ compiler and library 21614 21615 * [371]3907 nested template parameter collides with member name 21616 * [372]5293 confusing message when binding a temporary to a reference 21617 * [373]5296 [DR115] Pointers to functions and to template functions 21618 behave differently in deduction 21619 * [374]7939 ICE on function template specialization 21620 * [375]8656 Unable to assign function with __attribute__ and pointer 21621 return type to an appropriate variable 21622 * [376]10147 Confusing error message for invalid template function 21623 argument 21624 * [377]11400 std::search_n() makes assumptions about Size parameter 21625 * [378]11409 issues with using declarations, overloading, and 21626 built-in functions 21627 * [379]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 21628 multiple bits in mask 21629 * [380]11786 operator() call on variable in other namespace not 21630 recognized 21631 * [381]11867 static_cast ignores ambiguity 21632 * [382]11928 bug with conversion operators that are typedefs 21633 * [383]12114 Uninitialized memory accessed in dtor 21634 * [384]12163 static_cast + explicit constructor regression 21635 * [385]12181 Wrong code with comma operator and c++ 21636 * [386]12236 regparm and fastcall messes up parameters 21637 * [387]12266 incorrect instantiation of unneeded template during 21638 overload resolution 21639 * [388]12296 istream::peek() doesn't set eofbit 21640 * [389]12298 [sjlj exceptions] Stack unwind destroys 21641 not-yet-constructed object 21642 * [390]12369 ICE with templates and friends 21643 * [391]12337 apparently infinite loop in g++ 21644 * [392]12344 stdcall attribute ignored if function returns a pointer 21645 * [393]12451 missing(late) class forward declaration in cxxabi.h 21646 * [394]12486 g++ accepts invalid use of a qualified name 21647 21648 x86 specific (Intel/AMD) 21649 21650 * [395]8869 [x86 MMX] ICE with const variable optimization and MMX 21651 builtins 21652 * [396]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 21653 -O2 21654 * [397]11689 g++3.3 emits un-assembleable code for k6 architecture 21655 * [398]12116 [k6] Invalid assembly output values with X-MAME code 21656 * [399]12070 ICE converting between double and long double with 21657 -msoft-float 21658 21659 ia64-specific 21660 21661 * [400]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 21662 * [401]11535 __builtin_return_address may not work on ia64 21663 * [402]11693 [ia64] ICE in gen_nop_type 21664 * [403]12224 [ia64] Thread-local storage doesn't work 21665 21666 PowerPC-specific 21667 21668 * [404]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 21669 kernel 21670 * [405]11319 loop miscompiled on ppc32 21671 * [406]11949 ICE Compiler segfault with ffmpeg -maltivec code 21672 21673 SPARC-specific 21674 21675 * [407]11662 wrong code for expr. with cast to long long and 21676 exclusive or 21677 * [408]11965 invalid assembler code for a shift < 32 operation 21678 * [409]12301 (c++) stack corruption when a returned expression throws 21679 an exception 21680 21681 Alpha-specific 21682 21683 * [410]11717 [alpha-linux] unrecognizable insn compiling for.c of 21684 kernel 2.4.22-pre8 21685 21686 HPUX-specific 21687 21688 * [411]11313 problem with #pragma weak and static inline functions 21689 * [412]11712 __STDC_EXT__ not defined for C++ by default anymore? 21690 21691 Solaris specific 21692 21693 * [413]12166 Profiled programs crash if PROFDIR is set 21694 21695 Solaris-x86 specific 21696 21697 * [414]12101 i386 Solaris no longer works with GNU as? 21698 21699 Miscellaneous embedded target-specific bugs 21700 21701 * [415]10988 [m32r-elf] wrong blockmove code with -O3 21702 * [416]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 21703 -O2 21704 * [417]11902 [sh4] spec file improperly inserts rpath even when none 21705 needed 21706 * [418]11903 [sh4] -pthread fails to link due to error in spec file 21707 on sh4 21708 __________________________________________________________________ 21709 21710 [419]GCC 3.3.3 21711 21712 Minor features 21713 21714 In addition to the bug fixes documented below, this release contains 21715 few minor features such as: 21716 * Support for --with-sysroot 21717 * Support for automatic detection of executable stacks 21718 * Support for SSE3 instructions 21719 * Support for thread local storage debugging under GDB on S390 21720 21721 Bug Fixes 21722 21723 This section lists the problem reports (PRs) from GCC's bug tracker 21724 that are known to be fixed in the 3.3.3 release. This list might not be 21725 complete (that is, it is possible that some PRs that have been fixed 21726 are not listed here). 21727 21728 Bootstrap failures and issues 21729 21730 * [420]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 21731 * [421]12399 boehm-gc fails (when building a cross compiler): libtool 21732 unable to infer tagged configuration 21733 * [422]13068 mklibgcc.in doesn't handle multi-level multilib 21734 subdirectories properly 21735 21736 Internal compiler errors (multi-platform) 21737 21738 * [423]10060 ICE (stack overflow) on huge file (300k lines) due to 21739 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 21740 * [424]10555 (c++) ICE on undefined template argument 21741 * [425]10706 (c++) ICE in mangle_class_name_for_template 21742 * [426]11496 (fortran) error in flow_loops_find when -funroll-loops 21743 active 21744 * [427]11741 ICE in pre_insert_copy_insn, in gcse.c 21745 * [428]12440 GCC crashes during compilation of quicktime4linux 2.0.0 21746 * [429]12632 (fortran) -fbounds-check ICE 21747 * [430]12712 (c++) ICE on short legit C++ code fragment with gcc 21748 3.3.2 21749 * [431]12726 (c++) ICE (segfault) on trivial code 21750 * [432]12890 (c++) ICE on compilation of class with throwing method 21751 * [433]12900 (c++) ICE in rtl_verify_flow_info_1 21752 * [434]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 21753 correct code with -O2 -fno-force-mem 21754 * [435]13289 (c++) ICE in regenerate_decl_from_template on recursive 21755 template 21756 * [436]13318 ICE: floating point exception in the loop optimizer 21757 * [437]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 21758 except.c 21759 * [438]13574 (c++) invalid array default initializer in class lets 21760 gcc consume all memory and die 21761 * [439]13475 ICE on SIMD variables with partial value initialization 21762 * [440]13797 (c++) ICE on invalid template parameter 21763 * [441]13824 (java) gcj SEGV with simple .java program 21764 21765 C and optimization bugs 21766 21767 * [442]8776 loop invariants are not removed (most likely) 21768 * [443]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 21769 strncmp by memcmp 21770 * [444]11350 undefined labels with -Os -fPIC 21771 * [445]12826 Optimizer removes reference through volatile pointer 21772 * [446]12500 stabs debug info: void no longer a predefined / builtin 21773 type 21774 * [447]12941 builtin-bitops-1.c miscompilation (latent bug) 21775 * [448]12953 tree inliner bug (in inline_forbidden_p) and fix 21776 * [449]13041 linux-2.6/sound/core/oss/rate.c miscompiled 21777 * [450]13507 spurious printf format warning 21778 * [451]13382 Type information for const pointer disappears during 21779 optimization. 21780 * [452]13394 noreturn attribute ignored on recursive invokation 21781 * [453]13400 Compiled code crashes storing to read-only location 21782 * [454]13521 Endless loop in calculate_global_regs_live 21783 21784 C++ compiler and library 21785 21786 Some of the bug fixes in this list were made to implement decisions 21787 that the ISO C++ standards committee has made concerning several defect 21788 reports (DRs). Links in the list below point to detailed discussion of 21789 the relevant defect report. 21790 * [455]2094 unimplemented: use of `ptrmem_cst' in template type 21791 unification 21792 * [456]2294 using declaration confusion 21793 * [457]5050 template instantiation depth exceeds limit: recursion 21794 problem? 21795 * [458]9371 Bad exception handling in 21796 i/ostream::operator>>/<<(streambuf*) 21797 * [459]9546 bad exception handling in ostream members 21798 * [460]10081 basic_ios::_M_cache_locale leaves NULL members in the 21799 face of unknown locales 21800 * [461]10093 [462][DR 61] Setting failbit in exceptions doesn't work 21801 * [463]10095 istream::operator>>(int&) sets ios::badbit when 21802 ios::failbit is set. 21803 * [464]11554 Warning about reordering of initializers doesn't mention 21804 location of constructor 21805 * [465]12297 istream::sentry::sentry() handles eof() incorrectly. 21806 * [466]12352 Exception safety problems in src/localename.cc 21807 * [467]12438 Memory leak in locale::combine() 21808 * [468]12540 Memory leak in locale::locale(const char*) 21809 * [469]12594 DRs [470]60 [TC] and [471]63 [TC] not implemented 21810 * [472]12657 Resolution of [473]DR 292 (WP) still unimplemented 21811 * [474]12696 memory eating infinite loop in diagnostics (error 21812 recovery problem) 21813 * [475]12815 Code compiled with optimization behaves unexpectedly 21814 * [476]12862 Conflicts between typedefs/enums and namespace member 21815 declarations 21816 * [477]12926 Wrong value after assignment in initialize list using 21817 bit-fields 21818 * [478]12967 Resolution of [479]DR 300 [WP] still unimplemented 21819 * [480]12971 Resolution of [481]DR 328 [WP] still unimplemented 21820 * [482]13007 basic_streambuf::pubimbue, imbue wrong 21821 * [483]13009 Implicitly-defined assignment operator writes to wrong 21822 memory 21823 * [484]13057 regparm attribute not applied to destructor 21824 * [485]13070 -Wformat option ignored in g++ 21825 * [486]13081 forward template declarations in <complex> let inlining 21826 fail 21827 * [487]13239 Assertion does not seem to work correctly anymore 21828 * [488]13262 "xxx is private within this context" when initializing a 21829 self-contained template class 21830 * [489]13290 simple typo in concept checking for std::generate_n 21831 * [490]13323 Template code does not compile in presence of typedef 21832 * [491]13369 __verify_grouping (and __add_grouping?) not correct 21833 * [492]13371 infinite loop with packed struct and inlining 21834 * [493]13445 Template argument replacement "dereferences" a typedef 21835 * [494]13461 Fails to access protected-ctor from public constant 21836 * [495]13462 Non-standard-conforming type set::pointer 21837 * [496]13478 gcc uses wrong constructor to initialize a const 21838 reference 21839 * [497]13544 "conflicting types" for enums in different scopes 21840 * [498]13650 string::compare should not (always) use 21841 traits_type::length() 21842 * [499]13683 bogus warning about passing non-PODs through ellipsis 21843 * [500]13688 Derived class is denied access to protected base class 21844 member class 21845 * [501]13774 Member variable cleared in virtual multiple inheritance 21846 class 21847 * [502]13884 Protect sstream.tcc from extern template use 21848 21849 Java compiler and library 21850 21851 * [503]10746 [win32] garbage collection crash in GCJ 21852 21853 Objective-C compiler and library 21854 21855 * [504]11433 Crash due to dereferencing null pointer when querying 21856 protocol 21857 21858 Fortran compiler and library 21859 21860 * [505]12633 logical expression gives incorrect result with 21861 -fugly-logint option 21862 * [506]13037 [gcse-lm] g77 generates incorrect code 21863 * [507]13213 Hex constant problem when compiling with -fugly-logint 21864 and -ftypeless-boz 21865 21866 x86-specific (Intel/AMD) 21867 21868 * [508]4490 ICE with -m128bit-long-double 21869 * [509]12292 [x86_64] ICE: RTL check: expected code `const_int', have 21870 `reg' in make_field_assignment, in combine.c 21871 * [510]12441 ICE: can't find a register to spill 21872 * [511]12943 array static-init failure under -fpic, -fPIC 21873 * [512]13608 Incorrect code with -O3 -ffast-math 21874 21875 PowerPC-specific 21876 21877 * [513]11598 testcase gcc.dg/20020118-1.c fails runtime check of 21878 __attribute__((aligned(16))) 21879 * [514]11793 ICE in extract_insn, in recog.c (const_vector's) 21880 * [515]12467 vmsumubm emitted when vmsummbm appropriate (typo in 21881 altivec.md) 21882 * [516]12537 g++ generates writeable text sections 21883 21884 SPARC-specific 21885 21886 * [517]12496 wrong result for __atomic_add(&value, -1) when using -O0 21887 -m64 21888 * [518]12865 mprotect call to make trampoline executable may fail 21889 * [519]13354 ICE in sparc_emit_set_const32 21890 21891 ARM-specific 21892 21893 * [520]10467 [arm] ICE in pre_insert_copy_insn, 21894 21895 ia64-specific 21896 21897 * [521]11226 ICE passing struct arg with two floats 21898 * [522]11227 ICE for _Complex float, _Complex long double args 21899 * [523]12644 GCC 3.3.2 fails to compile glibc on ia64 21900 * [524]13149 build gcc-3.3.2 1305 error:unrecognizable insn 21901 * Various fixes for libunwind 21902 21903 Alpha-specific 21904 21905 * [525]12654 Incorrect comparison code generated for Alpha 21906 * [526]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 21907 * [527]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 21908 21909 HPPA-specific 21910 21911 * [528]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 21912 * [529]12158 [hppa] compilation does not terminate at -O1 21913 21914 S390-specific 21915 21916 * [530]11992 Wrong built-in code for memcmp with length 1<<24: only 21917 (1<<24)-1 possible for CLCL-Instruction 21918 21919 SH-specific 21920 21921 * [531]9365 segfault in gen_far_branch (config/sh/sh.c) 21922 * [532]10392 optimizer generates faulty array indexing 21923 * [533]11322 SH profiler outputs multiple definitions of symbol 21924 * [534]13069 gcc/config/sh/rtems.h broken 21925 * [535]13302 Putting a va_list in a struct causes seg fault 21926 * [536]13585 Incorrect optimization of call to sfunc 21927 * Fix inappropriately exported libgcc functions from the shared 21928 library 21929 21930 Other embedded target specific 21931 21932 * [537]8916 [mcore] unsigned char assign gets hosed. 21933 * [538]11576 [h8300] ICE in change_address_1, in emit-rtl.c 21934 * [539]13122 [h8300] local variable gets corrupted by function call 21935 when -fomit-frame-pointer is given 21936 * [540]13256 [cris] strict_low_part mistreated in delay slots 21937 * [541]13373 [mcore] optimization with -frerun-cse-after-loop 21938 -fexpensive-optimizations produces wrong code on mcore 21939 21940 GNU HURD-specific 21941 21942 * [542]12561 gcc/config/t-gnu needs updating to work with 21943 --with-sysroot 21944 21945 Tru64 Unix specific 21946 21947 * [543]6243 testsuite fails almost all tests due to no libintl in 21948 LD_LIBRARY_PATH during test. 21949 * [544]11397 weak aliases broken on Tru64 UNIX 21950 21951 AIX-specific 21952 21953 * [545]12505 build failure due to defines of uchar in cpphash.h and 21954 sys/types.h 21955 * [546]13150 WEAK symbols not exported by collect2 21956 21957 IRIX-specific 21958 21959 * [547]12666 fixincludes problem on IRIX 6.5.19m 21960 21961 Solaris-specific 21962 21963 * [548]12969 Including sys/byteorder.h breaks configure checks 21964 21965 Testsuite problems (compiler is not affected) 21966 21967 * [549]10819 testsuite creates CR+LF on compiler version lines in 21968 test summary files 21969 * [550]11612 abi_check not finding correct libgcc_s.so.1 21970 21971 Miscellaneous 21972 21973 * [551]13211 using -###, incorrect warnings about unused linker file 21974 are produced 21975 __________________________________________________________________ 21976 21977 [552]GCC 3.3.4 21978 21979 This is the [553]list of problem reports (PRs) from GCC's bug tracking 21980 system that are known to be fixed in the 3.3.4 release. This list might 21981 not be complete (that is, it is possible that some PRs that have been 21982 fixed are not listed here). 21983 __________________________________________________________________ 21984 21985 [554]GCC 3.3.5 21986 21987 This is the [555]list of problem reports (PRs) from GCC's bug tracking 21988 system that are known to be fixed in the 3.3.5 release. This list might 21989 not be complete (that is, it is possible that some PRs that have been 21990 fixed are not listed here). 21991 __________________________________________________________________ 21992 21993 [556]GCC 3.3.6 21994 21995 This is the [557]list of problem reports (PRs) from GCC's bug tracking 21996 system that are known to be fixed in the 3.3.6 release. This list might 21997 not be complete (that is, it is possible that some PRs that have been 21998 fixed are not listed here). 21999 22000 22001 For questions related to the use of GCC, please consult these web 22002 pages and the [558]GCC manuals. If that fails, the 22003 [559]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22004 web pages and the development of GCC are welcome on our developer 22005 list at [560]gcc (a] gcc.gnu.org. All of [561]our lists have public 22006 archives. 22007 22008 Copyright (C) [562]Free Software Foundation, Inc. Verbatim copying and 22009 distribution of this entire article is permitted in any medium, 22010 provided this notice is preserved. 22011 22012 These pages are [563]maintained by the GCC team. Last modified 22013 2025-01-31. 22014 22015 References 22016 22017 1. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 22018 2. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 22019 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 22020 4. https://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 22021 5. https://gcc.gnu.org/news/dfa.html 22022 6. https://gcc.gnu.org/c99status.html 22023 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 22024 8. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 22025 9. https://gcc.gnu.org/PR10140 22026 10. https://gcc.gnu.org/PR10198 22027 11. https://gcc.gnu.org/PR10338 22028 12. https://gcc.gnu.org/PR3581 22029 13. https://gcc.gnu.org/PR4382 22030 14. https://gcc.gnu.org/PR5533 22031 15. https://gcc.gnu.org/PR6387 22032 16. https://gcc.gnu.org/PR6412 22033 17. https://gcc.gnu.org/PR6620 22034 18. https://gcc.gnu.org/PR6663 22035 19. https://gcc.gnu.org/PR7068 22036 20. https://gcc.gnu.org/PR7083 22037 21. https://gcc.gnu.org/PR7647 22038 22. https://gcc.gnu.org/PR7675 22039 23. https://gcc.gnu.org/PR7718 22040 24. https://gcc.gnu.org/PR8116 22041 25. https://gcc.gnu.org/PR8358 22042 26. https://gcc.gnu.org/PR8511 22043 27. https://gcc.gnu.org/PR8564 22044 28. https://gcc.gnu.org/PR8660 22045 29. https://gcc.gnu.org/PR8766 22046 30. https://gcc.gnu.org/PR8803 22047 31. https://gcc.gnu.org/PR8846 22048 32. https://gcc.gnu.org/PR8906 22049 33. https://gcc.gnu.org/PR9216 22050 34. https://gcc.gnu.org/PR9261 22051 35. https://gcc.gnu.org/PR9263 22052 36. https://gcc.gnu.org/PR9429 22053 37. https://gcc.gnu.org/PR9516 22054 38. https://gcc.gnu.org/PR9600 22055 39. https://gcc.gnu.org/PR9629 22056 40. https://gcc.gnu.org/PR9672 22057 41. https://gcc.gnu.org/PR9749 22058 42. https://gcc.gnu.org/PR9794 22059 43. https://gcc.gnu.org/PR9829 22060 44. https://gcc.gnu.org/PR9916 22061 45. https://gcc.gnu.org/PR9936 22062 46. https://gcc.gnu.org/PR10262 22063 47. https://gcc.gnu.org/PR10278 22064 48. https://gcc.gnu.org/PR10446 22065 49. https://gcc.gnu.org/PR10451 22066 50. https://gcc.gnu.org/PR10506 22067 51. https://gcc.gnu.org/PR10549 22068 52. https://gcc.gnu.org/PR2001 22069 53. https://gcc.gnu.org/PR2391 22070 54. https://gcc.gnu.org/PR2960 22071 55. https://gcc.gnu.org/PR4046 22072 56. https://gcc.gnu.org/PR6405 22073 57. https://gcc.gnu.org/PR6798 22074 58. https://gcc.gnu.org/PR6871 22075 59. https://gcc.gnu.org/PR6909 22076 60. https://gcc.gnu.org/PR7189 22077 61. https://gcc.gnu.org/PR7642 22078 62. https://gcc.gnu.org/PR8634 22079 63. https://gcc.gnu.org/PR8750 22080 64. https://gcc.gnu.org/PR2161 22081 65. https://gcc.gnu.org/PR4319 22082 66. https://gcc.gnu.org/PR8602 22083 67. https://gcc.gnu.org/PR9177 22084 68. https://gcc.gnu.org/PR9853 22085 69. https://gcc.gnu.org/PR45 22086 70. https://gcc.gnu.org/PR3784 22087 71. https://gcc.gnu.org/PR764 22088 72. https://gcc.gnu.org/PR5116 22089 73. https://gcc.gnu.org/PR2862 22090 74. https://gcc.gnu.org/PR3663 22091 75. https://gcc.gnu.org/PR3797 22092 76. https://gcc.gnu.org/PR3948 22093 77. https://gcc.gnu.org/PR4137 22094 78. https://gcc.gnu.org/PR4361 22095 79. https://gcc.gnu.org/PR4802 22096 80. https://gcc.gnu.org/PR5837 22097 81. https://gcc.gnu.org/PR4803 22098 82. https://gcc.gnu.org/PR5094 22099 83. https://gcc.gnu.org/PR5730 22100 84. https://gcc.gnu.org/PR6713 22101 85. https://gcc.gnu.org/PR7015 22102 86. https://gcc.gnu.org/PR7086 22103 87. https://gcc.gnu.org/PR7099 22104 88. https://gcc.gnu.org/PR7247 22105 89. https://gcc.gnu.org/PR7441 22106 90. https://gcc.gnu.org/PR7768 22107 91. https://gcc.gnu.org/PR7804 22108 92. https://gcc.gnu.org/PR8099 22109 93. https://gcc.gnu.org/PR8117 22110 94. https://gcc.gnu.org/PR8205 22111 95. https://gcc.gnu.org/PR8645 22112 96. https://gcc.gnu.org/PR8724 22113 97. https://gcc.gnu.org/PR8805 22114 98. https://gcc.gnu.org/PR8691 22115 99. https://gcc.gnu.org/PR8700 22116 100. https://gcc.gnu.org/PR8724 22117 101. https://gcc.gnu.org/PR8949 22118 102. https://gcc.gnu.org/PR9016 22119 103. https://gcc.gnu.org/PR9053 22120 104. https://gcc.gnu.org/PR9152 22121 105. https://gcc.gnu.org/PR9182 22122 106. https://gcc.gnu.org/PR9297 22123 107. https://gcc.gnu.org/PR9318 22124 108. https://gcc.gnu.org/PR9320 22125 109. https://gcc.gnu.org/PR9400 22126 110. https://gcc.gnu.org/PR9424 22127 111. https://gcc.gnu.org/PR9425 22128 112. https://gcc.gnu.org/PR9439 22129 113. https://gcc.gnu.org/PR9474 22130 114. https://gcc.gnu.org/PR9548 22131 115. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 22132 116. https://gcc.gnu.org/PR9555 22133 117. https://gcc.gnu.org/PR9561 22134 118. https://gcc.gnu.org/PR9563 22135 119. https://gcc.gnu.org/PR9582 22136 120. https://gcc.gnu.org/PR9622 22137 121. https://gcc.gnu.org/PR9683 22138 122. https://gcc.gnu.org/PR9791 22139 123. https://gcc.gnu.org/PR9817 22140 124. https://gcc.gnu.org/PR9825 22141 125. https://gcc.gnu.org/PR9826 22142 126. https://gcc.gnu.org/PR9924 22143 127. https://gcc.gnu.org/PR9946 22144 128. https://gcc.gnu.org/PR9964 22145 129. https://gcc.gnu.org/PR9988 22146 130. https://gcc.gnu.org/PR10033 22147 131. https://gcc.gnu.org/PR10097 22148 132. https://gcc.gnu.org/PR10132 22149 133. https://gcc.gnu.org/PR10180 22150 134. https://gcc.gnu.org/PR10199 22151 135. https://gcc.gnu.org/PR10300 22152 136. https://gcc.gnu.org/PR10427 22153 137. https://gcc.gnu.org/PR10503 22154 138. https://gcc.gnu.org/PR5956 22155 139. https://gcc.gnu.org/PR1832 22156 140. https://gcc.gnu.org/PR3924 22157 141. https://gcc.gnu.org/PR5634 22158 142. https://gcc.gnu.org/PR6367 22159 143. https://gcc.gnu.org/PR6491 22160 144. https://gcc.gnu.org/PR6742 22161 145. https://gcc.gnu.org/PR7113 22162 146. https://gcc.gnu.org/PR7236 22163 147. https://gcc.gnu.org/PR7278 22164 148. https://gcc.gnu.org/PR7384 22165 149. https://gcc.gnu.org/PR7388 22166 150. https://gcc.gnu.org/PR8587 22167 151. https://gcc.gnu.org/PR9038 22168 152. https://gcc.gnu.org/PR10197 22169 153. https://gcc.gnu.org/PR6005 22170 154. https://gcc.gnu.org/PR6389 22171 155. https://gcc.gnu.org/PR6576 22172 156. https://gcc.gnu.org/PR6652 22173 157. https://gcc.gnu.org/PR7060 22174 158. https://gcc.gnu.org/PR7073 22175 159. https://gcc.gnu.org/PR7180 22176 160. https://gcc.gnu.org/PR7416 22177 161. https://gcc.gnu.org/PR7570 22178 162. https://gcc.gnu.org/PR7578 22179 163. https://gcc.gnu.org/PR7611 22180 164. https://gcc.gnu.org/PR7709 22181 165. https://gcc.gnu.org/PR7766 22182 166. https://gcc.gnu.org/PR7785 22183 167. https://gcc.gnu.org/PR7786 22184 168. https://gcc.gnu.org/PR8142 22185 169. https://gcc.gnu.org/PR8234 22186 170. https://gcc.gnu.org/PR8415 22187 171. https://gcc.gnu.org/PR8481 22188 172. https://gcc.gnu.org/PR8593 22189 173. https://gcc.gnu.org/PR8759 22190 174. https://gcc.gnu.org/PR8997 22191 175. https://gcc.gnu.org/PR9253 22192 176. https://gcc.gnu.org/PR9254 22193 177. https://gcc.gnu.org/PR9271 22194 178. https://gcc.gnu.org/PR6767 22195 179. https://gcc.gnu.org/PR9911 22196 180. https://gcc.gnu.org/PR10020 22197 181. https://gcc.gnu.org/PR10546 22198 182. https://gcc.gnu.org/PR7029 22199 183. https://gcc.gnu.org/PR2903 22200 184. https://gcc.gnu.org/PR7873 22201 185. https://gcc.gnu.org/PR7680 22202 186. https://gcc.gnu.org/PR8705 22203 187. https://gcc.gnu.org/PR9986 22204 188. https://gcc.gnu.org/PR10056 22205 189. https://gcc.gnu.org/PR6744 22206 190. https://gcc.gnu.org/PR7361 22207 191. https://gcc.gnu.org/PR9496 22208 192. https://gcc.gnu.org/PR7067 22209 193. https://gcc.gnu.org/PR8480 22210 194. https://gcc.gnu.org/PR8784 22211 195. https://gcc.gnu.org/PR10315 22212 196. https://gcc.gnu.org/PR10267 22213 197. https://gcc.gnu.org/PR7916 22214 198. https://gcc.gnu.org/PR7926 22215 199. https://gcc.gnu.org/PR8555 22216 200. https://gcc.gnu.org/PR8994 22217 201. https://gcc.gnu.org/PR9426 22218 202. https://gcc.gnu.org/PR9806 22219 203. https://gcc.gnu.org/PR10077 22220 204. https://gcc.gnu.org/PR10233 22221 205. https://gcc.gnu.org/PR10286 22222 206. https://gcc.gnu.org/PR10308 22223 207. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.1 22224 208. https://gcc.gnu.org/PR11272 22225 209. https://gcc.gnu.org/PR5754 22226 210. https://gcc.gnu.org/PR6597 22227 211. https://gcc.gnu.org/PR6949 22228 212. https://gcc.gnu.org/PR7053 22229 213. https://gcc.gnu.org/PR8164 22230 214. https://gcc.gnu.org/PR8384 22231 215. https://gcc.gnu.org/PR9559 22232 216. https://gcc.gnu.org/PR9649 22233 217. https://gcc.gnu.org/PR9864 22234 218. https://gcc.gnu.org/PR10432 22235 219. https://gcc.gnu.org/PR10475 22236 220. https://gcc.gnu.org/PR10635 22237 221. https://gcc.gnu.org/PR10661 22238 222. https://gcc.gnu.org/PR10700 22239 223. https://gcc.gnu.org/PR10712 22240 224. https://gcc.gnu.org/PR10796 22241 225. https://gcc.gnu.org/PR10890 22242 226. https://gcc.gnu.org/PR10939 22243 227. https://gcc.gnu.org/PR10956 22244 228. https://gcc.gnu.org/PR11041 22245 229. https://gcc.gnu.org/PR11059 22246 230. https://gcc.gnu.org/PR11083 22247 231. https://gcc.gnu.org/PR11105 22248 232. https://gcc.gnu.org/PR11149 22249 233. https://gcc.gnu.org/PR11228 22250 234. https://gcc.gnu.org/PR11282 22251 235. https://gcc.gnu.org/PR11301 22252 236. https://gcc.gnu.org/PR11308 22253 237. https://gcc.gnu.org/PR11473 22254 238. https://gcc.gnu.org/PR11503 22255 239. https://gcc.gnu.org/PR11513 22256 240. https://gcc.gnu.org/PR11198 22257 241. https://gcc.gnu.org/PR11304 22258 242. https://gcc.gnu.org/PR11381 22259 243. https://gcc.gnu.org/PR11536 22260 244. https://gcc.gnu.org/PR11557 22261 245. https://gcc.gnu.org/PR5897 22262 246. https://gcc.gnu.org/PR11279 22263 247. https://gcc.gnu.org/PR11022 22264 248. https://gcc.gnu.org/PR2330 22265 249. https://gcc.gnu.org/PR5388 22266 250. https://gcc.gnu.org/PR5390 22267 251. https://gcc.gnu.org/PR7877 22268 252. https://gcc.gnu.org/PR9393 22269 253. https://gcc.gnu.org/PR10032 22270 254. https://gcc.gnu.org/PR10468 22271 255. https://gcc.gnu.org/PR10527 22272 256. https://gcc.gnu.org/PR10679 22273 257. https://gcc.gnu.org/PR10682 22274 258. https://gcc.gnu.org/PR10689 22275 259. https://gcc.gnu.org/PR10845 22276 260. https://gcc.gnu.org/PR10849 22277 261. https://gcc.gnu.org/PR10888 22278 262. https://gcc.gnu.org/PR10929 22279 263. https://gcc.gnu.org/PR10931 22280 264. https://gcc.gnu.org/PR10940 22281 265. https://gcc.gnu.org/PR10968 22282 266. https://gcc.gnu.org/PR10990 22283 267. https://gcc.gnu.org/PR11039 22284 268. https://gcc.gnu.org/PR11062 22285 269. https://gcc.gnu.org/PR11095 22286 270. https://gcc.gnu.org/PR11098 22287 271. https://gcc.gnu.org/PR11137 22288 272. https://gcc.gnu.org/PR11154 22289 273. https://gcc.gnu.org/PR11329 22290 274. https://gcc.gnu.org/PR11332 22291 275. https://gcc.gnu.org/PR11431 22292 276. https://gcc.gnu.org/PR11528 22293 277. https://gcc.gnu.org/PR11546 22294 278. https://gcc.gnu.org/PR11567 22295 279. https://gcc.gnu.org/PR11645 22296 280. https://gcc.gnu.org/PR5179 22297 281. https://gcc.gnu.org/PR8204 22298 282. https://gcc.gnu.org/PR10838 22299 283. https://gcc.gnu.org/PR10886 22300 284. https://gcc.gnu.org/PR11349 22301 285. https://gcc.gnu.org/PR4823 22302 286. https://gcc.gnu.org/PR8878 22303 287. https://gcc.gnu.org/PR9815 22304 288. https://gcc.gnu.org/PR10402 22305 289. https://gcc.gnu.org/PR10504 22306 290. https://gcc.gnu.org/PR10673 22307 291. https://gcc.gnu.org/PR11044 22308 292. https://gcc.gnu.org/PR11089 22309 293. https://gcc.gnu.org/PR11420 22310 294. https://gcc.gnu.org/PR9362 22311 295. https://gcc.gnu.org/PR10142 22312 296. https://gcc.gnu.org/PR10663 22313 297. https://gcc.gnu.org/PR10835 22314 298. https://gcc.gnu.org/PR10876 22315 299. https://gcc.gnu.org/PR10955 22316 300. https://gcc.gnu.org/PR11018 22317 301. https://gcc.gnu.org/PR11556 22318 302. https://gcc.gnu.org/PR10907 22319 303. https://gcc.gnu.org/PR11320 22320 304. https://gcc.gnu.org/PR11599 22321 305. https://gcc.gnu.org/PR9745 22322 306. https://gcc.gnu.org/PR10871 22323 307. https://gcc.gnu.org/PR11440 22324 308. https://gcc.gnu.org/PR7594 22325 309. https://gcc.gnu.org/PR10557 22326 310. https://gcc.gnu.org/PR11054 22327 311. https://gcc.gnu.org/PR10834 22328 312. https://gcc.gnu.org/PR10842 22329 313. https://gcc.gnu.org/PR11052 22330 314. https://gcc.gnu.org/PR11183 22331 315. https://gcc.gnu.org/PR11084 22332 316. https://gcc.gnu.org/PR10331 22333 317. https://gcc.gnu.org/PR10413 22334 318. https://gcc.gnu.org/PR11096 22335 319. https://gcc.gnu.org/PR2873 22336 320. https://gcc.gnu.org/PR3163 22337 321. https://gcc.gnu.org/PR5287 22338 322. https://gcc.gnu.org/PR10148 22339 323. https://gcc.gnu.org/PR8787 22340 324. https://gcc.gnu.org/PR10900 22341 325. https://gcc.gnu.org/PR1607 22342 326. https://gcc.gnu.org/PR4252 22343 327. https://gcc.gnu.org/PR4490 22344 328. https://gcc.gnu.org/PR10355 22345 329. https://gcc.gnu.org/PR10726 22346 330. https://gcc.gnu.org/PR10805 22347 331. https://gcc.gnu.org/PR10815 22348 332. https://gcc.gnu.org/PR10877 22349 333. https://gcc.gnu.org/PR11280 22350 334. https://gcc.gnu.org/PR11466 22351 335. https://gcc.gnu.org/PR10737 22352 336. https://gcc.gnu.org/PR10810 22353 337. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.2 22354 338. https://gcc.gnu.org/PR8336 22355 339. https://gcc.gnu.org/PR9330 22356 340. https://gcc.gnu.org/PR9631 22357 341. https://gcc.gnu.org/PR9877 22358 342. https://gcc.gnu.org/PR11687 22359 343. https://gcc.gnu.org/PR12263 22360 344. https://gcc.gnu.org/PR12490 22361 345. https://gcc.gnu.org/PR7277 22362 346. https://gcc.gnu.org/PR7939 22363 347. https://gcc.gnu.org/PR11063 22364 348. https://gcc.gnu.org/PR11207 22365 349. https://gcc.gnu.org/PR11522 22366 350. https://gcc.gnu.org/PR11595 22367 351. https://gcc.gnu.org/PR11646 22368 352. https://gcc.gnu.org/PR11665 22369 353. https://gcc.gnu.org/PR11852 22370 354. https://gcc.gnu.org/PR11878 22371 355. https://gcc.gnu.org/PR11883 22372 356. https://gcc.gnu.org/PR11991 22373 357. https://gcc.gnu.org/PR12146 22374 358. https://gcc.gnu.org/PR12215 22375 359. https://gcc.gnu.org/PR12369 22376 360. https://gcc.gnu.org/PR12446 22377 361. https://gcc.gnu.org/PR12510 22378 362. https://gcc.gnu.org/PR12544 22379 363. https://gcc.gnu.org/PR9862 22380 364. https://gcc.gnu.org/PR10962 22381 365. https://gcc.gnu.org/PR11370 22382 366. https://gcc.gnu.org/PR11637 22383 367. https://gcc.gnu.org/PR11885 22384 368. https://gcc.gnu.org/PR12082 22385 369. https://gcc.gnu.org/PR12180 22386 370. https://gcc.gnu.org/PR12340 22387 371. https://gcc.gnu.org/PR3907 22388 372. https://gcc.gnu.org/PR5293 22389 373. https://gcc.gnu.org/PR5296 22390 374. https://gcc.gnu.org/PR7939 22391 375. https://gcc.gnu.org/PR8656 22392 376. https://gcc.gnu.org/PR10147 22393 377. https://gcc.gnu.org/PR11400 22394 378. https://gcc.gnu.org/PR11409 22395 379. https://gcc.gnu.org/PR11740 22396 380. https://gcc.gnu.org/PR11786 22397 381. https://gcc.gnu.org/PR11867 22398 382. https://gcc.gnu.org/PR11928 22399 383. https://gcc.gnu.org/PR12114 22400 384. https://gcc.gnu.org/PR12163 22401 385. https://gcc.gnu.org/PR12181 22402 386. https://gcc.gnu.org/PR12236 22403 387. https://gcc.gnu.org/PR12266 22404 388. https://gcc.gnu.org/PR12296 22405 389. https://gcc.gnu.org/PR12298 22406 390. https://gcc.gnu.org/PR12369 22407 391. https://gcc.gnu.org/PR12337 22408 392. https://gcc.gnu.org/PR12344 22409 393. https://gcc.gnu.org/PR12451 22410 394. https://gcc.gnu.org/PR12486 22411 395. https://gcc.gnu.org/PR8869 22412 396. https://gcc.gnu.org/PR9786 22413 397. https://gcc.gnu.org/PR11689 22414 398. https://gcc.gnu.org/PR12116 22415 399. https://gcc.gnu.org/PR12070 22416 400. https://gcc.gnu.org/PR11184 22417 401. https://gcc.gnu.org/PR11535 22418 402. https://gcc.gnu.org/PR11693 22419 403. https://gcc.gnu.org/PR12224 22420 404. https://gcc.gnu.org/PR11087 22421 405. https://gcc.gnu.org/PR11319 22422 406. https://gcc.gnu.org/PR11949 22423 407. https://gcc.gnu.org/PR11662 22424 408. https://gcc.gnu.org/PR11965 22425 409. https://gcc.gnu.org/PR12301 22426 410. https://gcc.gnu.org/PR11717 22427 411. https://gcc.gnu.org/PR11313 22428 412. https://gcc.gnu.org/PR11712 22429 413. https://gcc.gnu.org/PR12166 22430 414. https://gcc.gnu.org/PR12101 22431 415. https://gcc.gnu.org/PR10988 22432 416. https://gcc.gnu.org/PR11805 22433 417. https://gcc.gnu.org/PR11902 22434 418. https://gcc.gnu.org/PR11903 22435 419. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.3 22436 420. https://gcc.gnu.org/PR11890 22437 421. https://gcc.gnu.org/PR12399 22438 422. https://gcc.gnu.org/PR13068 22439 423. https://gcc.gnu.org/PR10060 22440 424. https://gcc.gnu.org/PR10555 22441 425. https://gcc.gnu.org/PR10706 22442 426. https://gcc.gnu.org/PR11496 22443 427. https://gcc.gnu.org/PR11741 22444 428. https://gcc.gnu.org/PR12440 22445 429. https://gcc.gnu.org/PR12632 22446 430. https://gcc.gnu.org/PR12712 22447 431. https://gcc.gnu.org/PR12726 22448 432. https://gcc.gnu.org/PR12890 22449 433. https://gcc.gnu.org/PR12900 22450 434. https://gcc.gnu.org/PR13060 22451 435. https://gcc.gnu.org/PR13289 22452 436. https://gcc.gnu.org/PR13318 22453 437. https://gcc.gnu.org/PR13392 22454 438. https://gcc.gnu.org/PR13574 22455 439. https://gcc.gnu.org/PR13475 22456 440. https://gcc.gnu.org/PR13797 22457 441. https://gcc.gnu.org/PR13824 22458 442. https://gcc.gnu.org/PR8776 22459 443. https://gcc.gnu.org/PR10339 22460 444. https://gcc.gnu.org/PR11350 22461 445. https://gcc.gnu.org/PR12826 22462 446. https://gcc.gnu.org/PR12500 22463 447. https://gcc.gnu.org/PR12941 22464 448. https://gcc.gnu.org/PR12953 22465 449. https://gcc.gnu.org/PR13041 22466 450. https://gcc.gnu.org/PR13507 22467 451. https://gcc.gnu.org/PR13382 22468 452. https://gcc.gnu.org/PR13394 22469 453. https://gcc.gnu.org/PR13400 22470 454. https://gcc.gnu.org/PR13521 22471 455. https://gcc.gnu.org/PR2094 22472 456. https://gcc.gnu.org/PR2294 22473 457. https://gcc.gnu.org/PR5050 22474 458. https://gcc.gnu.org/PR9371 22475 459. https://gcc.gnu.org/PR9546 22476 460. https://gcc.gnu.org/PR10081 22477 461. https://gcc.gnu.org/PR10093 22478 462. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 22479 463. https://gcc.gnu.org/PR10095 22480 464. https://gcc.gnu.org/PR11554 22481 465. https://gcc.gnu.org/PR12297 22482 466. https://gcc.gnu.org/PR12352 22483 467. https://gcc.gnu.org/PR12438 22484 468. https://gcc.gnu.org/PR12540 22485 469. https://gcc.gnu.org/PR12594 22486 470. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 22487 471. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 22488 472. https://gcc.gnu.org/PR12657 22489 473. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 22490 474. https://gcc.gnu.org/PR12696 22491 475. https://gcc.gnu.org/PR12815 22492 476. https://gcc.gnu.org/PR12862 22493 477. https://gcc.gnu.org/PR12926 22494 478. https://gcc.gnu.org/PR12967 22495 479. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 22496 480. https://gcc.gnu.org/PR12971 22497 481. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 22498 482. https://gcc.gnu.org/PR13007 22499 483. https://gcc.gnu.org/PR13009 22500 484. https://gcc.gnu.org/PR13057 22501 485. https://gcc.gnu.org/PR13070 22502 486. https://gcc.gnu.org/PR13081 22503 487. https://gcc.gnu.org/PR13239 22504 488. https://gcc.gnu.org/PR13262 22505 489. https://gcc.gnu.org/PR13290 22506 490. https://gcc.gnu.org/PR13323 22507 491. https://gcc.gnu.org/PR13369 22508 492. https://gcc.gnu.org/PR13371 22509 493. https://gcc.gnu.org/PR13445 22510 494. https://gcc.gnu.org/PR13461 22511 495. https://gcc.gnu.org/PR13462 22512 496. https://gcc.gnu.org/PR13478 22513 497. https://gcc.gnu.org/PR13544 22514 498. https://gcc.gnu.org/PR13650 22515 499. https://gcc.gnu.org/PR13683 22516 500. https://gcc.gnu.org/PR13688 22517 501. https://gcc.gnu.org/PR13774 22518 502. https://gcc.gnu.org/PR13884 22519 503. https://gcc.gnu.org/PR10746 22520 504. https://gcc.gnu.org/PR11433 22521 505. https://gcc.gnu.org/PR12633 22522 506. https://gcc.gnu.org/PR13037 22523 507. https://gcc.gnu.org/PR13213 22524 508. https://gcc.gnu.org/PR4490 22525 509. https://gcc.gnu.org/PR12292 22526 510. https://gcc.gnu.org/PR12441 22527 511. https://gcc.gnu.org/PR12943 22528 512. https://gcc.gnu.org/PR13608 22529 513. https://gcc.gnu.org/PR11598 22530 514. https://gcc.gnu.org/PR11793 22531 515. https://gcc.gnu.org/PR12467 22532 516. https://gcc.gnu.org/PR12537 22533 517. https://gcc.gnu.org/PR12496 22534 518. https://gcc.gnu.org/PR12865 22535 519. https://gcc.gnu.org/PR13354 22536 520. https://gcc.gnu.org/PR10467 22537 521. https://gcc.gnu.org/PR11226 22538 522. https://gcc.gnu.org/PR11227 22539 523. https://gcc.gnu.org/PR12644 22540 524. https://gcc.gnu.org/PR13149 22541 525. https://gcc.gnu.org/PR12654 22542 526. https://gcc.gnu.org/PR12965 22543 527. https://gcc.gnu.org/PR13031 22544 528. https://gcc.gnu.org/PR11634 22545 529. https://gcc.gnu.org/PR12158 22546 530. https://gcc.gnu.org/PR11992 22547 531. https://gcc.gnu.org/PR9365 22548 532. https://gcc.gnu.org/PR10392 22549 533. https://gcc.gnu.org/PR11322 22550 534. https://gcc.gnu.org/PR13069 22551 535. https://gcc.gnu.org/PR13302 22552 536. https://gcc.gnu.org/PR13585 22553 537. https://gcc.gnu.org/PR8916 22554 538. https://gcc.gnu.org/PR11576 22555 539. https://gcc.gnu.org/PR13122 22556 540. https://gcc.gnu.org/PR13256 22557 541. https://gcc.gnu.org/PR13373 22558 542. https://gcc.gnu.org/PR12561 22559 543. https://gcc.gnu.org/PR6243 22560 544. https://gcc.gnu.org/PR11397 22561 545. https://gcc.gnu.org/PR12505 22562 546. https://gcc.gnu.org/PR13150 22563 547. https://gcc.gnu.org/PR12666 22564 548. https://gcc.gnu.org/PR12969 22565 549. https://gcc.gnu.org/PR10819 22566 550. https://gcc.gnu.org/PR11612 22567 551. https://gcc.gnu.org/PR13211 22568 552. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.4 22569 553. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 22570 554. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.5 22571 555. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 22572 556. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.6 22573 557. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 22574 558. https://gcc.gnu.org/onlinedocs/ 22575 559. mailto:gcc-help (a] gcc.gnu.org 22576 560. mailto:gcc (a] gcc.gnu.org 22577 561. https://gcc.gnu.org/lists.html 22578 562. https://www.fsf.org/ 22579 563. https://gcc.gnu.org/about.html 22580 ====================================================================== 22581 http://gcc.gnu.org/gcc-3.2/index.html 22582 22583 GCC 3.2 Release Series 22584 22585 (This release series is no longer supported.) 22586 22587 April 25, 2003 22588 22589 The [1]GNU project and the GCC developers are pleased to announce the 22590 release of GCC 3.2.3. 22591 22592 The purpose of the GCC 3.2 release series is to provide a stable 22593 platform for OS distributors to use building their next releases. A 22594 primary objective was to stabilize the C++ ABI; we believe that the 22595 interface to the compiler and the C++ standard library are now 22596 relatively stable. 22597 22598 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 22599 interoperate with code compiled by GCC 3.1.1 or earlier. 22600 22601 Please refer to our [2]detailed list of news, caveats, and bug-fixes 22602 for further information. 22603 22604 Release History 22605 22606 GCC 3.2.3 22607 April 25, 2003 ([3]changes) 22608 22609 GCC 3.2.2 22610 February 5, 2003 ([4]changes) 22611 22612 GCC 3.2.1 22613 November 19, 2002 ([5]changes) 22614 22615 GCC 3.2 22616 August 14, 2002 ([6]changes) 22617 22618 References and Acknowledgements 22619 22620 GCC used to stand for the GNU C Compiler, but since the compiler 22621 supports several other languages aside from C, it now stands for the 22622 GNU Compiler Collection. 22623 22624 A list of [7]successful builds is updated as new information becomes 22625 available. 22626 22627 The GCC developers would like to thank the numerous people that have 22628 contributed new features, improvements, bug fixes, and other changes as 22629 well as test results to GCC. This [8]amazing group of volunteers is 22630 what makes GCC successful. 22631 22632 For additional information about GCC please refer to the [9]GCC project 22633 web site or contact the [10]GCC development mailing list. 22634 22635 To obtain GCC please use [11]our mirror sites, or our CVS server. 22636 22637 22638 For questions related to the use of GCC, please consult these web 22639 pages and the [12]GCC manuals. If that fails, the 22640 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22641 web pages and the development of GCC are welcome on our developer 22642 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 22643 archives. 22644 22645 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 22646 distribution of this entire article is permitted in any medium, 22647 provided this notice is preserved. 22648 22649 These pages are [17]maintained by the GCC team. Last modified 22650 2022-10-26. 22651 22652 References 22653 22654 1. http://www.gnu.org/ 22655 2. https://gcc.gnu.org/gcc-3.2/changes.html 22656 3. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 22657 4. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 22658 5. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 22659 6. https://gcc.gnu.org/gcc-3.2/changes.html#3.2 22660 7. https://gcc.gnu.org/gcc-3.2/buildstat.html 22661 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 22662 9. https://gcc.gnu.org/index.html 22663 10. mailto:gcc (a] gcc.gnu.org 22664 11. https://gcc.gnu.org/mirrors.html 22665 12. https://gcc.gnu.org/onlinedocs/ 22666 13. mailto:gcc-help (a] gcc.gnu.org 22667 14. mailto:gcc (a] gcc.gnu.org 22668 15. https://gcc.gnu.org/lists.html 22669 16. https://www.fsf.org/ 22670 17. https://gcc.gnu.org/about.html 22671 ====================================================================== 22672 http://gcc.gnu.org/gcc-3.2/changes.html 22673 22674 GCC 3.2 Release Series 22675 Changes, New Features, and Fixes 22676 22677 The latest release in the 3.2 release series is [1]GCC 3.2.3. 22678 22679 Caveats and New Features 22680 22681 Caveats 22682 22683 * The C++ compiler does not correctly zero-initialize 22684 pointers-to-data members. You must explicitly initialize them. For 22685 example: int S::*m(0); will work, but depending on 22686 default-initialization to zero will not work. This bug cannot be 22687 fixed in GCC 3.2 without inducing unacceptable risks. It will be 22688 fixed in GCC 3.3. 22689 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 22690 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 22691 a number of C++ ABI fixes which make its C++ compiler generate 22692 binary code which is incompatible with the C++ compilers found in 22693 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 22694 22695 Frontend Enhancements 22696 22697 C/C++/Objective-C 22698 22699 * The method of constructing the list of directories to be searched 22700 for header files has been revised. If a directory named by a -I 22701 option is a standard system include directory, the option is 22702 ignored to ensure that the default search order for system 22703 directories and the special treatment of system header files are 22704 not defeated. 22705 * The C and Objective-C compilers no longer accept the "Naming Types" 22706 extension (typedef foo = bar); it was already unavailable in C++. 22707 Code which uses it will need to be changed to use the "typeof" 22708 extension instead: typedef typeof(bar) foo. (We have removed this 22709 extension without a period of deprecation because it has caused the 22710 compiler to crash since version 3.0 and no one noticed until very 22711 recently. Thus we conclude it is not in widespread use.) 22712 22713 C++ 22714 22715 * GCC 3.2 fixed several differences between the C++ ABI implemented 22716 in GCC and the multi-vendor standard, but more have been found 22717 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 22718 code which is affected by these bugs. We will fix these bugs in 22719 some future release, once we are confident that all have been 22720 found; until then, it is our intention to make changes to the ABI 22721 only if they are necessary for correct compilation of C++, as 22722 opposed to conformance to the ABI documents. 22723 * For details on how to build an ABI compliant compiler for GNU/Linux 22724 systems, check the [3]common C++ ABI page. 22725 22726 New Targets and Target Specific Improvements 22727 22728 IA-32 22729 22730 * Fixed a number of bugs in SSE and MMX intrinsics. 22731 * Fixed common compiler crashes with SSE instruction set enabled 22732 (implied by -march=pentium3, pentium4, athlon-xp) 22733 * __m128 and __m128i is not 128bit aligned when used in structures. 22734 22735 x86-64 22736 22737 * A bug whereby the compiler could generate bad code for bzero has 22738 been fixed. 22739 * ABI fixes (implying ABI incompatibilities with previous version in 22740 some corner cases) 22741 * Fixed prefetch code generation 22742 __________________________________________________________________ 22743 22744 [4]GCC 3.2.3 22745 22746 3.2.3 is a bug fix release only; there are no new features that were 22747 not present in GCC 3.2.2. 22748 22749 Bug Fixes 22750 22751 This section lists the problem reports (PRs) from GCC's bug tracking 22752 system that are known to be fixed in the 3.2.3 release. This list might 22753 not be complete (that is, it is possible that some PRs that have been 22754 fixed are not listed here), and some of the titles have been changed to 22755 make them more clear. 22756 22757 Internal Compiler Errors (multi-platform) 22758 22759 * [5]3782: (c++) -quiet -fstats produces a segmentation fault in 22760 cc1plus 22761 * [6]6440: (c++) template specializations cause ICE 22762 * [7]7050: (c++) ICE on: (i ? get_string() : throw) 22763 * [8]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 22764 * [9]7982: (c++) ICE due to infinite recursion (using STL set) 22765 * [10]8068: exceedingly high (infinite) memory usage 22766 * [11]8178: ICE with __builtin_ffs 22767 * [12]8396: ICE in copy_to_mode_reg, in explow.c 22768 * [13]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 22769 * [14]9768: ICE when optimizing inline code at -O2 22770 * [15]9798: (c++) Infinite recursion (segfault) in 22771 cp/decl.c:push_using_directive with recursive using directives 22772 * [16]9799: mismatching structure initializer with nested flexible 22773 array member: ICE 22774 * [17]9928: ICE on duplicate enum declaration 22775 * [18]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 22776 sparc, alpha) 22777 * [19]10352: ICE in find_reloads_toplev 22778 * [20]10336: ICE with -Wunreachable-code 22779 22780 C/optimizer bugs: 22781 22782 * [21]8224: Incorrect joining of signed and unsigned division 22783 * [22]8613: -O2 produces wrong code with builtin strlen and 22784 postincrements 22785 * [23]8828: gcc reports some code is unreachable when it is not 22786 * [24]9226: GCSE breaking argument passing 22787 * [25]9853: miscompilation of non-constant structure initializer 22788 * [26]9797: C99-style struct initializers are miscompiled 22789 * [27]9967: Some standard C function calls should not be replaced 22790 when optimizing for size 22791 * [28]10116: ce2: invalid merge of join_bb in the context of switch 22792 statements 22793 * [29]10171: wrong code for inlined function 22794 * [30]10175: -Wunreachable-code doesn't work for single lines 22795 22796 C++ compiler and library: 22797 22798 * [31]8316: Confusing diagnostic for code that misuses conversion 22799 operators 22800 * [32]9169: filebuf output fails if codecvt<>::out returns noconv 22801 * [33]9420: incomplete type incorrectly reported 22802 * [34]9459: typeof in return type specification of template not 22803 supported 22804 * [35]9507: filebuf::open handles ios_base::ate incorrectly 22805 * [36]9538: Out-of-bounds memory access in streambuf::sputbackc 22806 * [37]9602: Total confusion about template/friend/virtual/abstract 22807 * [38]9993: destructor not called for local object created within and 22808 returned from infinite loop 22809 * [39]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 22810 system 22811 22812 Java compiler and library: 22813 22814 * [40]9652: libgcj build fails on irix6.5.1[78] 22815 * [41]10144: gas on solaris complains about bad .stabs lines for 22816 java, native as unaffected 22817 22818 x86-specific (Intel/AMD): 22819 22820 * [42]8746: gcc miscompiles Linux kernel ppa driver on x86 22821 * [43]9888: -mcpu=k6 -Os produces out of range loop instructions 22822 * [44]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 22823 failed 22824 * [45]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 22825 failed 22826 22827 SPARC-specific: 22828 22829 * [46]7784: [Sparc] ICE in extract_insn, in recog.c 22830 * [47]7796: sparc extra failure with -m64 on execute/930921-1.c in 22831 unroll.c 22832 * [48]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 22833 * [49]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 22834 execute/loop-2d.c 22835 * [50]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 22836 * [51]9414: Scheduling bug on Ultrasparc 22837 * [52]10067: GCC-3.2.2 outputs invalid asm on sparc64 22838 22839 m68k-specific: 22840 22841 * [53]7248: broken "inclusive or" code 22842 * [54]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 22843 22844 PowerPC-specific: 22845 22846 * [55]9732: Wrong code with -O2 -fPIC 22847 * [56]10073: ICE: powerpc cannot split insn 22848 22849 Alpha-specific: 22850 22851 * [57]7702: optimization problem on a DEC alpha under OSF1 22852 * [58]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 22853 22854 HP-specific: 22855 22856 * [59]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 22857 * [60]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 22858 (missing symbol) 22859 * [61]10271: Floating point args don't get reloaded across function 22860 calls with -O2 22861 22862 MIPS specific: 22863 22864 * [62]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 22865 compile/920501-4.c 22866 22867 CRIS specific: 22868 22869 * [63]10377: gcc-3.2.2 creates bad assembler code for cris 22870 22871 Miscellaneous and minor bugs: 22872 22873 * [64]6955: collect2 says "core dumped" when there is no core 22874 __________________________________________________________________ 22875 22876 [65]GCC 3.2.2 22877 22878 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 22879 install by means of the DESTDIR variable. Parts of the GCC tree have 22880 featured that support long before, but now it is available even from 22881 the top level. 22882 22883 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 22884 features that were not present in GCC 3.2.1. 22885 22886 Bug Fixes 22887 22888 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 22889 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 22890 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 22891 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 22892 change, and thus restores ABI-compatibility with previous releases 22893 (except GCC 3.2.1) on these platforms. 22894 22895 This section lists the problem reports (PRs) from GCC's bug tracking 22896 system that are known to be fixed in the 3.2.2 release. This list might 22897 not be complete (that is, it is possible that some PRs that have been 22898 fixed are not listed here) and some of the titles have been changed to 22899 make them more clear. 22900 22901 Internal Compiler Errors (multi-platform) 22902 22903 * [66]5919: (c++) ICE when passing variable array to template 22904 function 22905 * [67]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 22906 * [68]7507: ICE with -O2 when address of called function is a 22907 complicated expression 22908 * [69]7622: ICE with nested inline functions if function's address is 22909 taken 22910 * [70]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 22911 [71]9258) 22912 * [72]8031: (c++) ICE in code comparing typeids and casting from 22913 virtual base 22914 * [73]8275: ICE in simplify_subreg 22915 * [74]8332: (c++) builtin strlen/template interaction causes ICE 22916 * [75]8372: (c++) ICE on explicit call of destructor 22917 * [76]8439: (c, not c++) empty struct causes ICE 22918 * [77]8442: (c++) ICE with nested template classes 22919 * [78]8518: ICE when compiling mplayer ("extern inline" issue) 22920 * [79]8615: (c++) ICE with out-of-range character constant template 22921 argument 22922 * [80]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 22923 * [81]8799: (c++) ICE: error reporting routines re-entered 22924 * [82]9328: (c++) ICE with typeof(X) for overloaded X 22925 * [83]9465: (preprocessor) cpp -traditional ICE on null bytes 22926 22927 C++ (compiler and library) bugs 22928 22929 * [84]47: scoping in nested classes is broken 22930 * [85]6745: problems with iostream rdbuf() member function 22931 * [86]8214: conversion from const char* const to char* sometimes 22932 accepted illegally 22933 * [87]8493: builtin strlen and overload resolution (same bug as 22934 [88]8332) 22935 * [89]8503: strange behaviour of function types 22936 * [90]8727: compiler confused by inheritance from an anonymous struct 22937 * [91]7445: poor performance of std::locale::classic() in 22938 multi-threaded applications 22939 * [92]8230: mishandling of overflow in vector<T>::resize 22940 * [93]8399: sync_with_stdio(false) breaks unformatted input 22941 * [94]8662: illegal access of private member of unnamed class is 22942 accepted 22943 * [95]8707: "make distclean" fails in libstdc++-v3 directory 22944 * [96]8708: __USE_MALLOC doesn't work 22945 * [97]8790: Use of non-thread-safe strtok in src/localename.cc 22946 * [98]8887: Bug in date formats with --enable-clocale=generic 22947 * [99]9076: Call Frame Instructions are not handled correctly during 22948 unwind operation 22949 * [100]9151: std::setprecision limited to 16 digits when outputting a 22950 double to a stream 22951 * [101]9168: codecvt<char, char, mbstate_t> overwrites output buffers 22952 * [102]9269: libstdc++ headers: explicit specialization of function 22953 must precede its first use 22954 * [103]9322: return value of basic_streambuf<>::getloc affected by 22955 locale::global 22956 * [104]9433: segfault in runtime support for dynamic_cast 22957 22958 C and optimizer bugs 22959 22960 * [105]8032: GCC incorrectly initializes static structs that have 22961 flexible arrays 22962 * [106]8639: simple arithmetic expression broken 22963 * [107]8794: optimization improperly eliminates certain expressions 22964 * [108]8832: traditional "asm volatile" code is illegally optimized 22965 * [109]8988: loop optimizer bug: with -O2, code is generated that 22966 segfaults (found on i386, bug present for all platforms) 22967 * [110]9492: structure copy clobbers subsequent stores to structure 22968 22969 Objective-C bugs 22970 22971 * [111]9267: Objective-C parser won't build with newer bison versions 22972 (e.g. 1.875) 22973 22974 Ada bugs 22975 22976 * [112]8344: Ada build problem due to conflict between gcc/final.o, 22977 gcc/ada/final.o 22978 22979 Preprocessor bugs 22980 22981 * [113]8524: _Pragma within macros is improperly expanded 22982 * [114]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 22983 -fshort-wchar 22984 22985 ARM-specific 22986 22987 * [115]9090: arm ICE with >= -O2; regression from gcc-2.95 22988 22989 x86-specific (Intel/AMD) 22990 22991 * [116]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 22992 * [117]8599: loop unroll bug with -march=k6-3 22993 * [118]9506: ABI breakage in structure return (affects BSD and 22994 Cygwin, but not GNU/Linux) 22995 22996 FreeBSD 5.0 specific 22997 22998 * [119]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 22999 23000 RTEMS-specific 23001 23002 * [120]9292: hppa1.1-rtems configurery problems 23003 * [121]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 23004 * [122]9295: [mips-rtems] config/mips/rtems.h init/fini issue 23005 * [123]9296: gthr-rtems regression 23006 * [124]9316: powerpc-rtems: extending multilibs 23007 23008 HP-PA specific 23009 23010 * [125]9493: ICE with -O2 when building a simple function 23011 23012 Documentation 23013 23014 * [126]7341: hyperlink to gcov in GCC documentation doesn't work 23015 * [127]8947: Please add a warning about "-malign-double" in docs 23016 * [128]7448, [129]8882: typo cleanups 23017 __________________________________________________________________ 23018 23019 [130]GCC 3.2.1 23020 23021 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 23022 generates code that is known not to be binary-compatible with the 23023 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 23024 in the distribution, for details. 23025 23026 This release also removes an old GCC extension, "naming types", and the 23027 documentation now directs users to use a different GCC extension, 23028 __typeof__, instead. The feature had evidently been broken for a while. 23029 23030 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 23031 the new warning there are no new features that were not present in GCC 23032 3.2. 23033 23034 In addition, the previous fix for [131]PR 7445 (poor performance of 23035 std::locale::classic() in multi-threaded applications) was reverted 23036 ("unfixed"), because the "fix" was not thread-safe. 23037 23038 Bug Fixes 23039 23040 This section lists the problem reports (PRs) from GCC's bug tracking 23041 system that are known to be fixed in the 3.2.1 release. This list might 23042 not be complete (that is, it is possible that some PRs that have been 23043 fixed are not listed here). As you can see, the number of bug fixes is 23044 quite large, so it is strongly recommended that users of earlier GCC 23045 3.x releases upgrade to GCC 3.2.1. 23046 23047 Internal Compiler Errors (multi-platform) 23048 23049 * [132]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 23050 * [133]5661: (c++) ICE instantiating template on array of unknown 23051 size (bad code) 23052 * [134]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 23053 64-bit platforms 23054 * [135]6994: (c++) ICE in find_function_data 23055 * [136]7150: preprocessor: GCC -dM -E gives an ICE 23056 * [137]7160: ICE when optimizing branches without a return value 23057 * [138]7228: (c++) ICE when using member template and template 23058 function 23059 * [139]7266: (c++) ICE with -pedantic on missing typename 23060 * [140]7353: ICE from use of "Naming Types" extension, see above 23061 * [141]7411: ICE in instantiate_virtual_regs_1, in function.c 23062 * [142]7478: (c++) ICE on static_cast inside template 23063 * [143]7526: preprocessor core dump when _Pragma implies #pragma 23064 dependency 23065 * [144]7721: (c++) ICE on simple (but incorrect) template ([145]7803 23066 is a duplicate) 23067 * [146]7754: (c++) ICE on union with template parameter 23068 * [147]7788: (c++) redeclaring a definition as an incomplete class 23069 causes ICE 23070 * [148]8031: (c++) ICE in comptypes, in cp/typeck.c 23071 * [149]8055: preprocessor dies with SIG11 when building FreeBSD 23072 kernel 23073 * [150]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 23074 variables 23075 * [151]8134: (c++) ICE in force_store_init_value on legal code 23076 * [152]8149: (c++) ICE on incomplete type 23077 * [153]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 23078 initialization 23079 23080 C++ (compiler and library) bugs 23081 23082 * [154]5607: No pointer adjustment in covariant return types 23083 * [155]6579: Infinite loop with statement expressions in member 23084 initialization 23085 * [156]6803: Default copy constructor bug in GCC 3.1 23086 * [157]7176: g++ confused by friend and static member with same name 23087 * [158]7188: Segfault with template class and recursive (incorrect) 23088 initializer list 23089 * [159]7306: Regression: GCC 3.x fails to compile code with virtual 23090 inheritance if a method has a variable number of arguments 23091 * [160]7461: ctype<char>::classic_table() returns offset array on 23092 Cygwin 23093 * [161]7524: f(const float arg[3]) fails 23094 * [162]7584: Erroneous ambiguous base error on using declaration 23095 * [163]7676: Member template overloading problem 23096 * [164]7679: infinite loop when a right parenthesis is missing 23097 * [165]7811: default locale not taken from environment 23098 * [166]7961: compare( char *) implemented incorrectly in 23099 basic_string<> 23100 * [167]8071: basic_ostream::operator<<(streambuf*) loops forever if 23101 streambuf::underflow() leaves gptr() NULL (dups: [168]8127, 23102 [169]6745) 23103 * [170]8096: deque::at() throws std::range_error instead of 23104 std::out_of_range 23105 * [171]8127: cout << cin.rdbuf() infinite loop 23106 * [172]8218: Excessively large memory consumed for classes with large 23107 array members 23108 * [173]8287: GCC 3.2: Destructor called for non-constructed local 23109 object 23110 * [174]8347: empty vector range used in string construction causes 23111 core dump 23112 * [175]8348: fail() flag is set in istringstream when eof() flag is 23113 set 23114 * [176]8391: regression: infinite loop in cp/decl2.c(finish_file) 23115 23116 C and optimizer bugs 23117 23118 * [177]6627: -fno-align-functions doesn't seem to disable function 23119 alignment 23120 * [178]6631: life_analysis misoptimizes code to initialize fields of 23121 a structure 23122 * [179]7102: unsigned char division results in floating exception 23123 * [180]7120: Run once loop should *always* be unrolled 23124 (pessimization) 23125 * [181]7209: Bug involving array referencing and ?: operator 23126 * [182]7515: invalid inlining of global function with -O3 23127 * [183]7814: incorrect scheduling for glibc-2.2.92 strcpy test 23128 * [184]8467: bug in sibling call optimization 23129 23130 Preprocessor bugs 23131 23132 * [185]4890: incorrect line markers from the traditional preprocessor 23133 * [186]7357: -M option omits system headers files (making it the same 23134 as -MM) 23135 * [187]7358: Changes to Sun's make Dependencies 23136 * [188]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 23137 C headers 23138 * [189]7862: Interrupting GCC -MD removes .d file but not .o 23139 * [190]8190: Failed compilation deletes -MD dependency file 23140 * [191]8524: _Pragma within macro is improperly expanded 23141 23142 x86 specific (Intel/AMD) 23143 23144 * [192]5351: (i686-only) function pass-by-value structure copy 23145 corrupts stack ([193]7591 is a duplicate) 23146 * [194]6845, [195]7034, [196]7124, [197]7174: ICE's with 23147 -march=pentium3/pentium2/athlon (these are all the same underlying 23148 bug, in MMX register use) 23149 * [198]7134, [199]7375, [200]7390: ICE with -march=athlon (maybe same 23150 as above?) 23151 * [201]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 23152 * [202]6981: wrong code in 64-bit manipulation on x86 23153 * [203]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 23154 macro 23155 * [204]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 23156 intrinsics are broken 23157 * [205]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 23158 -march=pentium4 23159 * [206]7693: Typo in i386 mmintrin.h header 23160 * [207]7723: ICE - Pentium3 sse - GCC 3.2 23161 * [208]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 23162 * [209]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 23163 23164 PowerPC specific 23165 23166 * [210]5967: GCC bug when profiling nested functions on powerpc 23167 * [211]6984: wrong code generated with -O2, -O3, -Os for do-while 23168 loop on PowerPC 23169 * [212]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 23170 * [213]7130: miscompiled code for GCC-3.1 on 23171 powerpc-unknown-linux-gnu with -funroll-all-loops 23172 * [214]7133: PowerPC ICE: unrecognizable insn 23173 * [215]7380: ICE in extract_insn, at recog.c:2148 23174 * [216]8252: ICE on Altivec code with optimization turned on 23175 * [217]8451: Altivec ICE in GCC 3.2 23176 23177 HP/PA specific 23178 23179 * [218]7250: __ashrdi3 returns wrong value on 32 bit hppa 23180 23181 SPARC specific 23182 23183 * [219]6668: when using --disable-multilib, libgcc_s.so is installed 23184 in the wrong place on sparc-solaris 23185 * [220]7151: ICE when compiling for UltraSPARC 23186 * [221]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 23187 double and -O1 23188 * [222]7842: [REGRESSION] SPARC code gen bug 23189 23190 ARM specific 23191 23192 * [223]7856: [arm] invalid offset in constant pool reference 23193 * [224]7967: optimization produces wrong code (ARM) 23194 23195 Alpha specific 23196 23197 * [225]7374: __builtin_fabsl broken on alpha 23198 23199 IBM s390 specific 23200 23201 * [226]7370: ICE in fixup_var_refs_1 on s390x 23202 * [227]7409: loop optimization bug on s390x-linux-gnu 23203 * [228]8232: s390x: ICE when using bcmp with int length argument 23204 23205 SCO specific 23206 23207 * [229]7623: SCO OpenServer build fails with machmode.def: undefined 23208 symbol: BITS_PER_UNIT 23209 23210 m68k/Coldfire specific 23211 23212 * [230]8314: crtbegin, crtend need to be multilib'ed for this 23213 platform 23214 23215 Documentation 23216 23217 * [231]761: Document some undocumented options 23218 * [232]5610: Fix documentation about invoking SSE instructions 23219 (-mfpmath=sse) 23220 * [233]7484: List -Wmissing-declarations as C-only option 23221 * [234]7531: -mcmodel not documented for x86-64 23222 * [235]8120: Update documentation of bad use of ## 23223 __________________________________________________________________ 23224 23225 [236]GCC 3.2 23226 23227 3.2 is a small bug fix release, but there is a change to the 23228 application binary interface (ABI), hence the change to the second part 23229 of the version number. 23230 23231 The main purpose of the 3.2 release is to correct a couple of problems 23232 in the C++ ABI, with the intention of providing a stable interface 23233 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 23234 23235 Bug Fixes 23236 23237 C++ 23238 23239 * [237]7320: g++ 3.2 relocation problem 23240 * [238]7470: vtable: virtual function pointers not in declaration 23241 order 23242 23243 libstdc++ 23244 23245 * [239]6410: Trouble with non-ASCII monetary symbols and wchar_t 23246 * [240]6503, [241]6642, [242]7186: Problems with comparing or 23247 subtracting various types of const and non-const iterators 23248 * [243]7216: ambiguity with basic_iostream::traits_type 23249 * [244]7220: problem with basic_istream::ignore(0,delimiter) 23250 * [245]7222: locale::operator==() doesn't work on std::locale("") 23251 * [246]7286: placement operator delete issue 23252 * [247]7442: cxxabi.h does not match the C++ ABI 23253 * [248]7445: poor performance of std::locale::classic() in 23254 multi-threaded applications 23255 23256 x86-64 specific 23257 23258 * [249]7291: off-by-one in generated inline bzero code for x86-64 23259 23260 23261 For questions related to the use of GCC, please consult these web 23262 pages and the [250]GCC manuals. If that fails, the 23263 [251]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23264 web pages and the development of GCC are welcome on our developer 23265 list at [252]gcc (a] gcc.gnu.org. All of [253]our lists have public 23266 archives. 23267 23268 Copyright (C) [254]Free Software Foundation, Inc. Verbatim copying and 23269 distribution of this entire article is permitted in any medium, 23270 provided this notice is preserved. 23271 23272 These pages are [255]maintained by the GCC team. Last modified 23273 2025-04-08. 23274 23275 References 23276 23277 1. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 23278 2. https://gcc.gnu.org/gcc-3.1/changes.html 23279 3. https://gcc.gnu.org/gcc-3.2/c++-abi.html 23280 4. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.3 23281 5. https://gcc.gnu.org/PR3782 23282 6. https://gcc.gnu.org/PR6440 23283 7. https://gcc.gnu.org/PR7050 23284 8. https://gcc.gnu.org/PR7741 23285 9. https://gcc.gnu.org/PR7982 23286 10. https://gcc.gnu.org/PR8068 23287 11. https://gcc.gnu.org/PR8178 23288 12. https://gcc.gnu.org/PR8396 23289 13. https://gcc.gnu.org/PR8674 23290 14. https://gcc.gnu.org/PR9768 23291 15. https://gcc.gnu.org/PR9798 23292 16. https://gcc.gnu.org/PR9799 23293 17. https://gcc.gnu.org/PR9928 23294 18. https://gcc.gnu.org/PR10114 23295 19. https://gcc.gnu.org/PR10352 23296 20. https://gcc.gnu.org/PR10336 23297 21. https://gcc.gnu.org/PR8224 23298 22. https://gcc.gnu.org/PR8613 23299 23. https://gcc.gnu.org/PR8828 23300 24. https://gcc.gnu.org/PR9226 23301 25. https://gcc.gnu.org/PR9853 23302 26. https://gcc.gnu.org/PR9797 23303 27. https://gcc.gnu.org/PR9967 23304 28. https://gcc.gnu.org/PR10116 23305 29. https://gcc.gnu.org/PR10171 23306 30. https://gcc.gnu.org/PR10175 23307 31. https://gcc.gnu.org/PR8316 23308 32. https://gcc.gnu.org/PR9169 23309 33. https://gcc.gnu.org/PR9420 23310 34. https://gcc.gnu.org/PR9459 23311 35. https://gcc.gnu.org/PR9507 23312 36. https://gcc.gnu.org/PR9538 23313 37. https://gcc.gnu.org/PR9602 23314 38. https://gcc.gnu.org/PR9993 23315 39. https://gcc.gnu.org/PR10167 23316 40. https://gcc.gnu.org/PR9652 23317 41. https://gcc.gnu.org/PR10144 23318 42. https://gcc.gnu.org/PR8746 23319 43. https://gcc.gnu.org/PR9888 23320 44. https://gcc.gnu.org/PR9638 23321 45. https://gcc.gnu.org/PR9954 23322 46. https://gcc.gnu.org/PR7784 23323 47. https://gcc.gnu.org/PR7796 23324 48. https://gcc.gnu.org/PR8281 23325 49. https://gcc.gnu.org/PR8366 23326 50. https://gcc.gnu.org/PR8726 23327 51. https://gcc.gnu.org/PR9414 23328 52. https://gcc.gnu.org/PR10067 23329 53. https://gcc.gnu.org/PR7248 23330 54. https://gcc.gnu.org/PR8343 23331 55. https://gcc.gnu.org/PR9732 23332 56. https://gcc.gnu.org/PR10073 23333 57. https://gcc.gnu.org/PR7702 23334 58. https://gcc.gnu.org/PR9671 23335 59. https://gcc.gnu.org/PR8694 23336 60. https://gcc.gnu.org/PR9953 23337 61. https://gcc.gnu.org/PR10271 23338 62. https://gcc.gnu.org/PR6362 23339 63. https://gcc.gnu.org/PR10377 23340 64. https://gcc.gnu.org/PR6955 23341 65. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.2 23342 66. https://gcc.gnu.org/PR5919 23343 67. https://gcc.gnu.org/PR7129 23344 68. https://gcc.gnu.org/PR7507 23345 69. https://gcc.gnu.org/PR7622 23346 70. https://gcc.gnu.org/PR7681 23347 71. https://gcc.gnu.org/PR9528 23348 72. https://gcc.gnu.org/PR8031 23349 73. https://gcc.gnu.org/PR8275 23350 74. https://gcc.gnu.org/PR8332 23351 75. https://gcc.gnu.org/PR8372 23352 76. https://gcc.gnu.org/PR8439 23353 77. https://gcc.gnu.org/PR8442 23354 78. https://gcc.gnu.org/PR8518 23355 79. https://gcc.gnu.org/PR8615 23356 80. https://gcc.gnu.org/PR8663 23357 81. https://gcc.gnu.org/PR8799 23358 82. https://gcc.gnu.org/PR9328 23359 83. https://gcc.gnu.org/PR9465 23360 84. https://gcc.gnu.org/PR47 23361 85. https://gcc.gnu.org/PR6745 23362 86. https://gcc.gnu.org/PR8214 23363 87. https://gcc.gnu.org/PR8493 23364 88. https://gcc.gnu.org/PR8332 23365 89. https://gcc.gnu.org/PR8503 23366 90. https://gcc.gnu.org/PR8727 23367 91. https://gcc.gnu.org/PR7445 23368 92. https://gcc.gnu.org/PR8230 23369 93. https://gcc.gnu.org/PR8399 23370 94. https://gcc.gnu.org/PR8662 23371 95. https://gcc.gnu.org/PR8707 23372 96. https://gcc.gnu.org/PR8708 23373 97. https://gcc.gnu.org/PR8790 23374 98. https://gcc.gnu.org/PR8887 23375 99. https://gcc.gnu.org/PR9076 23376 100. https://gcc.gnu.org/PR9151 23377 101. https://gcc.gnu.org/PR9168 23378 102. https://gcc.gnu.org/PR9269 23379 103. https://gcc.gnu.org/PR9322 23380 104. https://gcc.gnu.org/PR9433 23381 105. https://gcc.gnu.org/PR8032 23382 106. https://gcc.gnu.org/PR8639 23383 107. https://gcc.gnu.org/PR8794 23384 108. https://gcc.gnu.org/PR8832 23385 109. https://gcc.gnu.org/PR8988 23386 110. https://gcc.gnu.org/PR9492 23387 111. https://gcc.gnu.org/PR9267 23388 112. https://gcc.gnu.org/PR8344 23389 113. https://gcc.gnu.org/PR8524 23390 114. https://gcc.gnu.org/PR8880 23391 115. https://gcc.gnu.org/PR9090 23392 116. https://gcc.gnu.org/PR8588 23393 117. https://gcc.gnu.org/PR8599 23394 118. https://gcc.gnu.org/PR9506 23395 119. https://gcc.gnu.org/PR9484 23396 120. https://gcc.gnu.org/PR9292 23397 121. https://gcc.gnu.org/PR9293 23398 122. https://gcc.gnu.org/PR9295 23399 123. https://gcc.gnu.org/PR9296 23400 124. https://gcc.gnu.org/PR9316 23401 125. https://gcc.gnu.org/PR9493 23402 126. https://gcc.gnu.org/PR7341 23403 127. https://gcc.gnu.org/PR8947 23404 128. https://gcc.gnu.org/PR7448 23405 129. https://gcc.gnu.org/PR8882 23406 130. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.1 23407 131. https://gcc.gnu.org/PR7445 23408 132. https://gcc.gnu.org/PR2521 23409 133. https://gcc.gnu.org/PR5661 23410 134. https://gcc.gnu.org/PR6419 23411 135. https://gcc.gnu.org/PR6994 23412 136. https://gcc.gnu.org/PR7150 23413 137. https://gcc.gnu.org/PR7160 23414 138. https://gcc.gnu.org/PR7228 23415 139. https://gcc.gnu.org/PR7266 23416 140. https://gcc.gnu.org/PR7353 23417 141. https://gcc.gnu.org/PR7411 23418 142. https://gcc.gnu.org/PR7478 23419 143. https://gcc.gnu.org/PR7526 23420 144. https://gcc.gnu.org/PR7721 23421 145. https://gcc.gnu.org/PR7803 23422 146. https://gcc.gnu.org/PR7754 23423 147. https://gcc.gnu.org/PR7788 23424 148. https://gcc.gnu.org/PR8031 23425 149. https://gcc.gnu.org/PR8055 23426 150. https://gcc.gnu.org/PR8067 23427 151. https://gcc.gnu.org/PR8134 23428 152. https://gcc.gnu.org/PR8149 23429 153. https://gcc.gnu.org/PR8160 23430 154. https://gcc.gnu.org/PR5607 23431 155. https://gcc.gnu.org/PR6579 23432 156. https://gcc.gnu.org/PR6803 23433 157. https://gcc.gnu.org/PR7176 23434 158. https://gcc.gnu.org/PR7188 23435 159. https://gcc.gnu.org/PR7306 23436 160. https://gcc.gnu.org/PR7461 23437 161. https://gcc.gnu.org/PR7524 23438 162. https://gcc.gnu.org/PR7584 23439 163. https://gcc.gnu.org/PR7676 23440 164. https://gcc.gnu.org/PR7679 23441 165. https://gcc.gnu.org/PR7811 23442 166. https://gcc.gnu.org/PR7961 23443 167. https://gcc.gnu.org/PR8071 23444 168. https://gcc.gnu.org/PR8127 23445 169. https://gcc.gnu.org/PR6745 23446 170. https://gcc.gnu.org/PR8096 23447 171. https://gcc.gnu.org/PR8127 23448 172. https://gcc.gnu.org/PR8218 23449 173. https://gcc.gnu.org/PR8287 23450 174. https://gcc.gnu.org/PR8347 23451 175. https://gcc.gnu.org/PR8348 23452 176. https://gcc.gnu.org/PR8391 23453 177. https://gcc.gnu.org/PR6627 23454 178. https://gcc.gnu.org/PR6631 23455 179. https://gcc.gnu.org/PR7102 23456 180. https://gcc.gnu.org/PR7120 23457 181. https://gcc.gnu.org/PR7209 23458 182. https://gcc.gnu.org/PR7515 23459 183. https://gcc.gnu.org/PR7814 23460 184. https://gcc.gnu.org/PR8467 23461 185. https://gcc.gnu.org/PR4890 23462 186. https://gcc.gnu.org/PR7357 23463 187. https://gcc.gnu.org/PR7358 23464 188. https://gcc.gnu.org/PR7602 23465 189. https://gcc.gnu.org/PR7862 23466 190. https://gcc.gnu.org/PR8190 23467 191. https://gcc.gnu.org/PR8524 23468 192. https://gcc.gnu.org/PR5351 23469 193. https://gcc.gnu.org/PR7591 23470 194. https://gcc.gnu.org/PR6845 23471 195. https://gcc.gnu.org/PR7034 23472 196. https://gcc.gnu.org/PR7124 23473 197. https://gcc.gnu.org/PR7174 23474 198. https://gcc.gnu.org/PR7134 23475 199. https://gcc.gnu.org/PR7375 23476 200. https://gcc.gnu.org/PR7390 23477 201. https://gcc.gnu.org/PR6890 23478 202. https://gcc.gnu.org/PR6981 23479 203. https://gcc.gnu.org/PR7242 23480 204. https://gcc.gnu.org/PR7396 23481 205. https://gcc.gnu.org/PR7630 23482 206. https://gcc.gnu.org/PR7693 23483 207. https://gcc.gnu.org/PR7723 23484 208. https://gcc.gnu.org/PR7951 23485 209. https://gcc.gnu.org/PR8146 23486 210. https://gcc.gnu.org/PR5967 23487 211. https://gcc.gnu.org/PR6984 23488 212. https://gcc.gnu.org/PR7114 23489 213. https://gcc.gnu.org/PR7130 23490 214. https://gcc.gnu.org/PR7133 23491 215. https://gcc.gnu.org/PR7380 23492 216. https://gcc.gnu.org/PR8252 23493 217. https://gcc.gnu.org/PR8451 23494 218. https://gcc.gnu.org/PR7250 23495 219. https://gcc.gnu.org/PR6668 23496 220. https://gcc.gnu.org/PR7151 23497 221. https://gcc.gnu.org/PR7335 23498 222. https://gcc.gnu.org/PR7842 23499 223. https://gcc.gnu.org/PR7856 23500 224. https://gcc.gnu.org/PR7967 23501 225. https://gcc.gnu.org/PR7374 23502 226. https://gcc.gnu.org/PR7370 23503 227. https://gcc.gnu.org/PR7409 23504 228. https://gcc.gnu.org/PR8232 23505 229. https://gcc.gnu.org/PR7623 23506 230. https://gcc.gnu.org/PR8314 23507 231. https://gcc.gnu.org/PR761 23508 232. https://gcc.gnu.org/PR5610 23509 233. https://gcc.gnu.org/PR7484 23510 234. https://gcc.gnu.org/PR7531 23511 235. https://gcc.gnu.org/PR8120 23512 236. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2 23513 237. https://gcc.gnu.org/PR7320 23514 238. https://gcc.gnu.org/PR7470 23515 239. https://gcc.gnu.org/PR6410 23516 240. https://gcc.gnu.org/PR6503 23517 241. https://gcc.gnu.org/PR6642 23518 242. https://gcc.gnu.org/PR7186 23519 243. https://gcc.gnu.org/PR7216 23520 244. https://gcc.gnu.org/PR7220 23521 245. https://gcc.gnu.org/PR7222 23522 246. https://gcc.gnu.org/PR7286 23523 247. https://gcc.gnu.org/PR7442 23524 248. https://gcc.gnu.org/PR7445 23525 249. https://gcc.gnu.org/PR7291 23526 250. https://gcc.gnu.org/onlinedocs/ 23527 251. mailto:gcc-help (a] gcc.gnu.org 23528 252. mailto:gcc (a] gcc.gnu.org 23529 253. https://gcc.gnu.org/lists.html 23530 254. https://www.fsf.org/ 23531 255. https://gcc.gnu.org/about.html 23532 ====================================================================== 23533 http://gcc.gnu.org/gcc-3.1/index.html 23534 23535 GCC 3.1 23536 23537 (This release series is no longer supported.) 23538 23539 July 27, 2002 23540 23541 The GNU project and the GCC developers are pleased to announce the 23542 release of GCC 3.1.1. 23543 23544 The links below still apply to GCC 3.1.1. 23545 23546 May 15, 2002 23547 23548 The [1]GNU project and the GCC developers are pleased to announce the 23549 release of GCC 3.1. 23550 23551 GCC used to stand for the GNU C Compiler, but since the compiler 23552 supports several other languages aside from C, it now stands for the 23553 GNU Compiler Collection. 23554 23555 A list of [2]successful builds is updated as new information becomes 23556 available. 23557 23558 The GCC developers would like to thank the numerous people that have 23559 contributed [3]new features, improvements, bug fixes, and other changes 23560 as well as test results to GCC. This [4]amazing group of volunteers is 23561 what makes GCC successful. 23562 23563 For additional information about GCC please refer to the [5]GCC project 23564 web site or contact the [6]GCC development mailing list. 23565 23566 To obtain GCC please use [7]our mirror sites, or our CVS server. 23567 __________________________________________________________________ 23568 23569 23570 For questions related to the use of GCC, please consult these web 23571 pages and the [8]GCC manuals. If that fails, the 23572 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23573 web pages and the development of GCC are welcome on our developer 23574 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 23575 archives. 23576 23577 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 23578 distribution of this entire article is permitted in any medium, 23579 provided this notice is preserved. 23580 23581 These pages are [13]maintained by the GCC team. Last modified 23582 2024-07-04. 23583 23584 References 23585 23586 1. https://www.gnu.org/ 23587 2. https://gcc.gnu.org/gcc-3.1/buildstat.html 23588 3. https://gcc.gnu.org/gcc-3.1/changes.html 23589 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 23590 5. https://gcc.gnu.org/index.html 23591 6. mailto:gcc (a] gcc.gnu.org 23592 7. https://gcc.gnu.org/mirrors.html 23593 8. https://gcc.gnu.org/onlinedocs/ 23594 9. mailto:gcc-help (a] gcc.gnu.org 23595 10. mailto:gcc (a] gcc.gnu.org 23596 11. https://gcc.gnu.org/lists.html 23597 12. https://www.fsf.org/ 23598 13. https://gcc.gnu.org/about.html 23599 ====================================================================== 23600 http://gcc.gnu.org/gcc-3.1/changes.html 23601 23602 GCC 3.1 Release Series 23603 Changes, New Features, and Fixes 23604 23605 Additional changes in GCC 3.1.1 23606 23607 * A bug related to how structures and unions are returned has been 23608 fixed for powerpc-*-netbsd*. 23609 * An important bug in the implementation of -fprefetch-loop-arrays 23610 has been fixed. Previously the optimization prefetched random 23611 blocks of memory for most targets except for i386. 23612 * The Java compiler now compiles Java programs much faster and also 23613 works with parallel make. 23614 * Nested functions have been fixed for mips*-*-netbsd*. 23615 * Some missing floating point support routines have beed added for 23616 mips*-*-netbsd*. 23617 * This [1]message gives additional information about the bugs fixed 23618 in this release. 23619 23620 Caveats 23621 23622 * The -traditional C compiler option has been deprecated and will be 23623 removed in GCC 3.3. (It remains possible to preprocess non-C code 23624 with the traditional preprocessor.) 23625 * The default debugging format for most ELF platforms (including 23626 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 23627 from stabs to DWARF2. This requires GDB 5.1.1 or later. 23628 23629 General Optimizer Improvements 23630 23631 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 23632 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 23633 for profile driven optimizations. 23634 Options -fprofile-arcs and -fbranch-probabilities can now be used 23635 to improve speed of the generated code by profiling the actual 23636 program behaviour on typical runs. In the absence of profile info 23637 the compiler attempts to guess the profile statically. 23638 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 23639 monitor performance of the generated code. 23640 According to the SPECInt2000 results on an AMD Athlon CPU, the code 23641 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 23642 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 23643 is about 2.1% faster compared to 2.95.3. Tests were done using the 23644 -O2 -march=athlon command-line options. 23645 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 23646 infrastructure developed by CodeSourcery, LLC for the C++ front 23647 end, so that it is now used in the C front end too. Inlining 23648 functions as trees exposes them earlier to the compiler, giving it 23649 more opportunities for optimization. 23650 * Support for data prefetching instructions has been added to the GCC 23651 back end and several targets. A new __builtin_prefetch intrinsic is 23652 available to explicitly insert prefetch instructions and 23653 experimental support for loop array prefetching has been added (see 23654 -fprefetch-loop-array documentation). 23655 * Support for emitting debugging information for macros has been 23656 added for DWARF2. It is activated using -g3. 23657 23658 New Languages and Language specific improvements 23659 23660 C/C++ 23661 23662 * A few more [4]ISO C99 features. 23663 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 23664 * The preprocessor's symbol table has been merged with the symbol 23665 table of the C, C++ and Objective-C front ends. 23666 * The preprocessor consumes less memory than the preprocessor in GCC 23667 3.0, often significantly so. On normal input files, it typically 23668 consumes less memory than pre-3.0 cccp-based GCC, too. 23669 23670 C++ 23671 23672 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 23673 was a workaround to allow std compliant code to work with the 23674 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 23675 * The C++ ABI has been fixed so that void (A::*)() const is mangled 23676 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 23677 affects pointer to cv-qualified member function types. 23678 * The C++ ABI has been changed to correctly handle this code: 23679 struct A { 23680 void operator delete[] (void *, size_t); 23681 }; 23682 23683 struct B : public A { 23684 }; 23685 23686 new B[10]; 23687 23688 The amount of storage allocated for the array will be greater than 23689 it was in 3.0, in order to store the number of elements in the 23690 array, so that the correct size can be passed to operator delete[] 23691 when the array is deleted. Previously, the value passed to operator 23692 delete[] was unpredictable. 23693 This change will only affect code that declares a two-argument 23694 operator delete[] with a second parameter of type size_t in a base 23695 class, and does not override that definition in a derived class. 23696 * The C++ ABI has been changed so that: 23697 struct A { 23698 void operator delete[] (void *, size_t); 23699 void operator delete[] (void *); 23700 }; 23701 23702 does not cause unnecessary storage to be allocated when an array of 23703 A objects is allocated. 23704 This change will only affect code that declares both of these forms 23705 of operator delete[], and declared the two-argument form before the 23706 one-argument form. 23707 * The C++ ABI has been changed so that when a parameter is passed by 23708 value, any cleanup for that parameter is performed in the caller, 23709 as specified by the ia64 C++ ABI, rather than the called function 23710 as before. As a result, classes with a non-trivial destructor but a 23711 trivial copy constructor will be passed and returned by invisible 23712 reference, rather than by bitwise copy as before. 23713 * G++ now supports the "named return value optimization": for code 23714 like 23715 A f () { 23716 A a; 23717 ... 23718 return a; 23719 } 23720 23721 G++ will allocate a in the return value slot, so that the return 23722 becomes a no-op. For this to work, all return statements in the 23723 function must return the same variable. 23724 * Improvements to the C++ library are listed in [5]the libstdc++-v3 23725 FAQ. 23726 23727 Objective-C 23728 23729 * Annoying linker warnings (due to incorrect code being generated) 23730 have been fixed. 23731 * If a class method cannot be found, the compiler no longer issues a 23732 warning if a corresponding instance method exists in the root 23733 class. 23734 * Forward @protocol declarations have been fixed. 23735 * Loading of categories has been fixed in certain situations (GNU run 23736 time only). 23737 * The class lookup in the run-time library has been rewritten so that 23738 class method dispatch is more than twice as fast as it used to be 23739 (GNU run time only). 23740 23741 Java 23742 23743 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 23744 javax.transaction. 23745 * Property files and other system resources can be compiled into 23746 executables which use libgcj using the new gcj --resource feature. 23747 * libgcj has been ported to more platforms. In particular there is 23748 now a mostly-functional mingw32 (Windows) target port. 23749 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 23750 Java code can now be called from a C/C++ application. 23751 * gcj can now use builtin functions for certain known methods, for 23752 instance Math.cos. 23753 * gcj can now automatically remove redundant array-store checks in 23754 some common cases. 23755 * The --no-store-checks optimization option was added. This can be 23756 used to omit runtime store checks for code which is known not to 23757 throw ArrayStoreException 23758 * The following third party interface standards were added to libgcj: 23759 org.w3c.dom and org.xml.sax. 23760 * java.security has been merged with GNU Classpath. The new package 23761 is now JDK 1.2 compliant, and much more complete. 23762 * A bytecode verifier was added to the libgcj interpreter. 23763 * java.lang.Character was rewritten to comply with the Unicode 3.0 23764 standard, and improve performance. 23765 * Partial support for many more locales was added to libgcj. 23766 * Socket timeouts have been implemented. 23767 * libgcj has been merged into a single shared library. There are no 23768 longer separate shared libraries for the garbage collector and 23769 zlib. 23770 * Several performance improvements were made to gcj and libgcj: 23771 + Hash synchronization (thin locks) 23772 + A special allocation path for finalizer-free objects 23773 + Thread-local allocation 23774 + Parallel GC, and other GC tweaks 23775 23776 Fortran 23777 23778 Fortran improvements are listed in [6]the Fortran documentation. 23779 23780 Ada 23781 23782 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 23783 tools. The GNAT compiler fully implements the Ada language as defined 23784 by the ISO/IEC 8652 standard. 23785 23786 Please note that the integration of the Ada front end is still work in 23787 progress. 23788 23789 New Targets and Target Specific Improvements 23790 23791 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 23792 architecture used in new editions of Donald E. Knuth's The Art of 23793 Computer Programming. 23794 * Axis Communications has contributed its port to the CRIS CPU 23795 architecture, used in the ETRAX system-on-a-chip series. 23796 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 23797 SH5 64-bit RISC microprocessor architecture, extending the existing 23798 SH port. 23799 * UltraSPARC is fully supported in 64-bit mode. The option -m64 23800 enables it. 23801 * For compatibility with the Sun compiler #pragma redefine_extname 23802 has been implemented on Solaris. 23803 * The x86 back end has had some noticeable work done to it. 23804 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 23805 Jaeger have contributed a port to the AMD x86-64 architecture. 23806 For more information on x86-64 see http://www.x86-64.org. 23807 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 23808 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 23809 enable the respective instruction sets. Intel C++ compatible 23810 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 23811 will be added in next major release. 23812 + Following those improvements, targets for Pentium MMX, K6-2, 23813 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 23814 added. Refer to the documentation on -march= and -mcpu= 23815 options for details. 23816 + For those targets that support it, -mfpmath=sse will cause the 23817 compiler to generate SSE/SSE2 instructions for floating point 23818 math instead of x87 instructions. Usually, this will lead to 23819 quicker code -- especially on the Pentium 4. Note that only 23820 scalar floating point instructions are used and GCC does not 23821 exploit SIMD features yet. 23822 + Prefetch support has been added to the Pentium III, Pentium 4, 23823 K6-2, K6-3, and Athlon series. 23824 + Code generated for floating point to integer conversions has 23825 been improved leading to better performance of many 3D 23826 applications. 23827 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 23828 * C++ support for AIX has been improved. 23829 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 23830 PowerPC port supporting the AltiVec programming model (SIMD). The 23831 support, though presently useful, is experimental and is expected 23832 to stabilize for 3.2. The support is written to conform to 23833 Motorola's AltiVec specs. See -maltivec. 23834 23835 [8]Obsolete Systems 23836 23837 Support for a number of older systems has been declared obsolete in GCC 23838 3.1. Unless there is activity to revive them, the next release of GCC 23839 will have their sources permanently removed. 23840 23841 All configurations of the following processor architectures have been 23842 declared obsolete: 23843 * MIL-STD-1750A, 1750a-*-* 23844 * AMD A29k, a29k-*-* 23845 * Convex, c*-convex-* 23846 * Clipper, clipper-*-* 23847 * Elxsi, elxsi-*-* 23848 * Intel i860, i860-*-* 23849 * Sun picoJava, pj-*-* and pjl-*-* 23850 * Western Electric 32000, we32k-*-* 23851 23852 Most configurations of the following processor architectures have been 23853 declared obsolete, but we are preserving a few systems which may have 23854 active developers. It is unlikely that the remaining systems will 23855 survive much longer unless we see definite signs of port activity. 23856 * Motorola 88000 except 23857 + Generic a.out, m88k-*-aout* 23858 + Generic SVR4, m88k-*-sysv4 23859 + OpenBSD, m88k-*-openbsd* 23860 * NS32k except 23861 + NetBSD, ns32k-*-netbsd* 23862 + OpenBSD, ns32k-*-openbsd*. 23863 * ROMP except 23864 + OpenBSD, romp-*-openbsd*. 23865 23866 Finally, only some configurations of these processor architectures are 23867 being obsoleted. 23868 * Alpha: 23869 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 23870 alpha*-*-osf[45], are still supported.) 23871 * ARM: 23872 + RISCiX, arm-*-riscix*. 23873 * i386: 23874 + 386BSD, i?86-*-bsd* 23875 + Chorus, i?86-*-chorusos* 23876 + DG/UX, i?86-*-dgux* 23877 + FreeBSD 1.x, i?86-*-freebsd1.* 23878 + IBM AIX, i?86-*-aix* 23879 + ISC UNIX, i?86-*-isc* 23880 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 23881 + NEXTstep, i?86-next-* 23882 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 23883 + RTEMS/coff, i?86-*-rtemscoff* 23884 + RTEMS/go32, i?86-go32-rtems* 23885 + Sequent/BSD, i?86-sequent-bsd* 23886 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 23887 i?86-sequent-sysv3* 23888 + SunOS, i?86-*-sunos* 23889 * Motorola 68000: 23890 + Altos, m68[k0]*-altos-* 23891 + Apollo, m68[k0]*-apollo-* 23892 + Apple A/UX, m68[k0]*-apple-* 23893 + Bull, m68[k0]*-bull-* 23894 + Convergent, m68[k0]*-convergent-* 23895 + Generic SVR3, m68[k0]*-*-sysv3* 23896 + ISI, m68[k0]*-isi-* 23897 + LynxOS, m68[k0]*-*-lynxos* 23898 + NEXT, m68[k0]*-next-* 23899 + RTEMS/coff, m68[k0]*-*-rtemscoff* 23900 + Sony, m68[k0]*-sony-* 23901 * MIPS: 23902 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 23903 + Generic BSD, mips-*-bsd* 23904 + Generic System V, mips-*-sysv* 23905 + IRIX before version 5, mips-sgi-irix[1234]* 23906 + RiscOS, mips-*-riscos* 23907 + Sony, mips-sony-* 23908 + Tandem, mips-tandem-* 23909 * SPARC: 23910 + RTEMS/a.out, sparc-*-rtemsaout*. 23911 23912 Documentation improvements 23913 23914 * The old manual ("Using and Porting the GNU Compiler Collection") 23915 has been replaced by a users manual ("Using the GNU Compiler 23916 Collection") and a separate internals reference manual ("GNU 23917 Compiler Collection Internals"). 23918 * More complete and much improved documentation about GCC's internal 23919 representation used by the C and C++ front ends. 23920 * Many cleanups and improvements in general. 23921 23922 23923 For questions related to the use of GCC, please consult these web 23924 pages and the [9]GCC manuals. If that fails, the 23925 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23926 web pages and the development of GCC are welcome on our developer 23927 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 23928 archives. 23929 23930 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 23931 distribution of this entire article is permitted in any medium, 23932 provided this notice is preserved. 23933 23934 These pages are [14]maintained by the GCC team. Last modified 23935 2025-01-31. 23936 23937 References 23938 23939 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 23940 2. https://gcc.gnu.org/news/profiledriven.html 23941 3. https://gcc.gnu.org/benchmarks/ 23942 4. https://gcc.gnu.org/c99status.html 23943 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 23944 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 23945 7. https://www.adacore.com/ 23946 8. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 23947 9. https://gcc.gnu.org/onlinedocs/ 23948 10. mailto:gcc-help (a] gcc.gnu.org 23949 11. mailto:gcc (a] gcc.gnu.org 23950 12. https://gcc.gnu.org/lists.html 23951 13. https://www.fsf.org/ 23952 14. https://gcc.gnu.org/about.html 23953 ====================================================================== 23954 http://gcc.gnu.org/gcc-3.0/index.html 23955 23956 GCC 3.0.4 23957 23958 (This release series is no longer supported.) 23959 23960 February 20, 2002 23961 23962 The [1]GNU project and the GCC developers are pleased to announce the 23963 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 23964 series. 23965 23966 GCC used to stand for the GNU C Compiler, but since the compiler 23967 supports several other languages aside from C, it now stands for the 23968 GNU Compiler Collection. 23969 23970 GCC 3.0.x has several new optimizations, new targets, new languages and 23971 many other new features, relative to GCC 2.95.x. See the [2]new 23972 features page for a more complete list. 23973 23974 A list of [3]successful builds is updated as new information becomes 23975 available. 23976 23977 The GCC developers would like to thank the numerous people that have 23978 contributed new features, test results, bug fixes, etc to GCC. This 23979 [4]amazing group of volunteers is what makes GCC successful. 23980 23981 And finally, we can't in good conscience fail to mention some 23982 [5]caveats to using GCC 3.0.x. 23983 23984 For additional information about GCC please refer to the [6]GCC project 23985 web site or contact the [7]GCC development mailing list. 23986 23987 To obtain GCC please use [8]our mirror sites, or our CVS server. 23988 __________________________________________________________________ 23989 23990 Previous 3.0.x Releases 23991 23992 December 20, 2001: GCC 3.0.3 has been released. 23993 October 25, 2001: GCC 3.0.2 has been released. 23994 August 20, 2001: GCC 3.0.1 has been released. 23995 June 18, 2001: GCC 3.0 has been released. 23996 23997 23998 For questions related to the use of GCC, please consult these web 23999 pages and the [9]GCC manuals. If that fails, the 24000 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24001 web pages and the development of GCC are welcome on our developer 24002 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 24003 archives. 24004 24005 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 24006 distribution of this entire article is permitted in any medium, 24007 provided this notice is preserved. 24008 24009 These pages are [14]maintained by the GCC team. Last modified 24010 2024-07-04. 24011 24012 References 24013 24014 1. https://www.gnu.org/ 24015 2. https://gcc.gnu.org/gcc-3.0/features.html 24016 3. https://gcc.gnu.org/gcc-3.0/buildstat.html 24017 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24018 5. https://gcc.gnu.org/gcc-3.0/caveats.html 24019 6. https://gcc.gnu.org/index.html 24020 7. mailto:gcc (a] gcc.gnu.org 24021 8. https://gcc.gnu.org/mirrors.html 24022 9. https://gcc.gnu.org/onlinedocs/ 24023 10. mailto:gcc-help (a] gcc.gnu.org 24024 11. mailto:gcc (a] gcc.gnu.org 24025 12. https://gcc.gnu.org/lists.html 24026 13. https://www.fsf.org/ 24027 14. https://gcc.gnu.org/about.html 24028 ====================================================================== 24029 http://gcc.gnu.org/gcc-3.0/features.html 24030 24031 GCC 3.0 New Features 24032 24033 Additional changes in GCC 3.0.4 24034 24035 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 24036 system, which use the ELF object file format, on x86 processors. 24037 * Correct debugging information is generated from functions that have 24038 lines from multiple files (e.g. yacc output). 24039 * A fix for whitespace handling in the -traditional preprocessor, 24040 which can affect Fortran. 24041 * Fixes to the exception handling runtime. 24042 * More fixes for bad code generation in C++. 24043 * A fix for shared library generation under AIX 4.3. 24044 * Documentation updates. 24045 * Port of GCC to Tensilica's Xtensa processor contributed. 24046 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 24047 24048 Additional changes in GCC 3.0.3 24049 24050 * A fix to correct an accidental change to the PowerPC ABI. 24051 * Fixes for bad code generation on a variety of architectures. 24052 * Improvements to the debugging information generated for C++ 24053 classes. 24054 * Fixes for bad code generation in C++. 24055 * A fix to avoid crashes in the C++ demangler. 24056 * A fix to the C++ standard library to avoid buffer overflows. 24057 * Miscellaneous improvements for a variety of architectures. 24058 24059 Additional changes in GCC 3.0.2 24060 24061 * Fixes for bad code generation during loop unrolling. 24062 * Fixes for bad code generation by the sibling call optimization. 24063 * Minor improvements to x86 code generation. 24064 * Implementation of function descriptors in C++ vtables for IA64. 24065 * Numerous minor bug-fixes. 24066 24067 Additional changes in GCC 3.0.1 24068 24069 * C++ fixes for incorrect code-generation. 24070 * Improved cross-compiling support for the C++ standard library. 24071 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 24072 in GCC 3.0. 24073 * Fixes for various exception-handling bugs. 24074 * A port to the S/390 architecture. 24075 24076 General Optimizer Improvements 24077 24078 * [2]Basic block reordering pass. 24079 * New if-conversion pass with support for conditional (predicated) 24080 execution. 24081 * New tail call and sibling call elimination optimizations. 24082 * New register renaming pass. 24083 * New (experimental) [3]static single assignment (SSA) representation 24084 support. 24085 * New dead-code elimination pass implemented using the SSA 24086 representation. 24087 * [4]Global null pointer test elimination. 24088 * [5]Global code hoisting/unification. 24089 * More builtins and optimizations for stdio.h, string.h and old BSD 24090 functions, as well as for ISO C99 functions. 24091 * New builtin __builtin_expect for giving hints to the branch 24092 predictor. 24093 24094 New Languages and Language specific improvements 24095 24096 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 24097 and supported, including the run-time library containing most 24098 common non-GUI Java classes, a bytecode interpreter, and the Boehm 24099 conservative garbage collector. Many bugs have been fixed. GCJ can 24100 compile Java source or Java bytecodes to either native code or Java 24101 class files, and supports native methods written in either the 24102 standard JNI or the more efficient and convenient CNI. 24103 * Here is a [6]partial list of C++ improvements, both new features 24104 and those no longer supported. 24105 * New C++ ABI. On the IA-64 platform GCC is capable of 24106 inter-operating with other IA-64 compilers. 24107 * The new ABI also significantly reduces the size of symbol and debug 24108 information. 24109 * New C++ support library and many C++ bug fixes, vastly improving 24110 our conformance to the ISO C++ standard. 24111 * New [7]inliner for C++. 24112 * Rewritten C preprocessor, integrated into the C, C++ and Objective 24113 C compilers, with very many improvements including ISO C99 support 24114 and [8]improvements to dependency generation. 24115 * Support for more [9]ISO C99 features. 24116 * Many improvements to support for checking calls to format functions 24117 such as printf and scanf, including support for ISO C99 format 24118 features, extensions from the Single Unix Specification and GNU 24119 libc 2.2, checking of strfmon formats and features to assist in 24120 auditing for format string security bugs. 24121 * New warnings for C code that may have undefined semantics because 24122 of violations of sequence point rules in the C standard (such as a 24123 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 24124 * Additional warning option -Wfloat-equal. 24125 * Improvements to -Wtraditional. 24126 * Fortran improvements are listed in [10]the Fortran documentation. 24127 24128 New Targets and Target Specific Improvements 24129 24130 * New x86 back end, generating much improved code. 24131 * Support for a generic i386-elf target contributed. 24132 * New option to emit x86 assembly code using Intel style syntax 24133 (-mintel-syntax). 24134 * HPUX 11 support contributed. 24135 * Improved PowerPC code generation, including scheduled prologue and 24136 epilogue. 24137 * Port of GCC to Intel's IA-64 processor contributed. 24138 * Port of GCC to Motorola's MCore 210 and 340 contributed. 24139 * New unified back-end for Arm, Thumb and StrongArm contributed. 24140 * Port of GCC to Intel's XScale processor contributed. 24141 * Port of GCC to Atmel's AVR microcontrollers contributed. 24142 * Port of GCC to Mitsubishi's D30V processor contributed. 24143 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 24144 processor family) contributed. 24145 * Port of GCC to Fujitsu's FR30 processor contributed. 24146 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 24147 * Port of GCC to Sun's picoJava processor core contributed. 24148 24149 Documentation improvements 24150 24151 * Substantially rewritten and improved C preprocessor manual. 24152 * Many improvements to other documentation. 24153 * Manpages for gcc, cpp and gcov are now generated automatically from 24154 the master Texinfo manual, eliminating the problem of manpages 24155 being out of date. (The generated manpages are only extracts from 24156 the full manual, which is provided in Texinfo form, from which 24157 info, HTML, other formats and a printed manual can be generated.) 24158 * Generated info files are included in the release tarballs alongside 24159 their Texinfo sources, avoiding problems on some platforms with 24160 building makeinfo as part of the GCC distribution. 24161 24162 Other significant improvements 24163 24164 * Garbage collection used internally by the compiler for most memory 24165 allocation instead of obstacks. 24166 * Lengauer and Tarjan algorithm used for computing dominators in the 24167 CFG. This algorithm can be significantly faster and more space 24168 efficient than our older algorithm. 24169 * gccbug script provided to assist in submitting bug reports to our 24170 bug tracking system. (Bug reports previously submitted directly to 24171 our mailing lists, for which you received no bug tracking number, 24172 should be submitted again using gccbug if you can reproduce the 24173 problem with GCC 3.0.) 24174 * The internal libgcc library is [11]built as a shared library on 24175 systems that support it. 24176 * Extensive testsuite included with GCC, with many new tests. In 24177 addition to tests for GCC bugs that have been fixed, many tests 24178 have been added for language features, compiler warnings and 24179 builtin functions. 24180 * Additional language-independent warning options -Wpacked, -Wpadded, 24181 -Wunreachable-code and -Wdisabled-optimization. 24182 * Target-independent options -falign-functions, -falign-loops and 24183 -falign-jumps. 24184 24185 Plus a great many bug fixes and almost all the [12]features found in 24186 GCC 2.95. 24187 24188 24189 For questions related to the use of GCC, please consult these web 24190 pages and the [13]GCC manuals. If that fails, the 24191 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24192 web pages and the development of GCC are welcome on our developer 24193 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 24194 archives. 24195 24196 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 24197 distribution of this entire article is permitted in any medium, 24198 provided this notice is preserved. 24199 24200 These pages are [18]maintained by the GCC team. Last modified 24201 2022-10-26. 24202 24203 References 24204 24205 1. http://www.netbsd.org/ 24206 2. https://gcc.gnu.org/news/reorder.html 24207 3. https://gcc.gnu.org/news/ssa.html 24208 4. https://gcc.gnu.org/news/null.html 24209 5. https://gcc.gnu.org/news/unify.html 24210 6. https://gcc.gnu.org/gcc-3.0/c++features.html 24211 7. https://gcc.gnu.org/news/inlining.html 24212 8. https://gcc.gnu.org/news/dependencies.html 24213 9. https://gcc.gnu.org/c99status.html 24214 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 24215 11. https://gcc.gnu.org/gcc-3.0/libgcc.html 24216 12. https://gcc.gnu.org/gcc-2.95/features.html 24217 13. https://gcc.gnu.org/onlinedocs/ 24218 14. mailto:gcc-help (a] gcc.gnu.org 24219 15. mailto:gcc (a] gcc.gnu.org 24220 16. https://gcc.gnu.org/lists.html 24221 17. https://www.fsf.org/ 24222 18. https://gcc.gnu.org/about.html 24223 ====================================================================== 24224 http://gcc.gnu.org/gcc-3.0/caveats.html 24225 24226 GCC 3.0 Caveats 24227 24228 * -fstrict-aliasing is now part of -O2 and higher optimization 24229 levels. This allows the compiler to assume the strictest aliasing 24230 rules applicable to the language being compiled. For C and C++, 24231 this activates optimizations based on the type of expressions. This 24232 optimization may thus break old, non-compliant code. 24233 * Enumerations are now properly promoted to int in function 24234 parameters and function returns. Normally this change is not 24235 visible, but when using -fshort-enums this is an ABI change. 24236 * The undocumented extension that allowed C programs to have a label 24237 at the end of a compound statement has been deprecated and may be 24238 removed in a future version. Programs that now generate a warning 24239 about this may be fixed by adding a null statement (a single 24240 semicolon) after the label. 24241 * The poorly documented extension that allowed string constants in C, 24242 C++ and Objective C to contain unescaped newlines has been 24243 deprecated and may be removed in a future version. Programs using 24244 this extension may be fixed in several ways: the bare newline may 24245 be replaced by \n, or preceded by \n\, or string concatenation may 24246 be used with the bare newline preceded by \n" and " placed at the 24247 start of the next line. 24248 * The Chill compiler is not included in GCC 3.0, because of the lack 24249 of a volunteer to convert it to use garbage collection. 24250 * Certain non-standard iostream methods from earlier versions of 24251 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 24252 ostream::form, and istream::gets. 24253 * The new C++ ABI is not yet fully supported by current (as of 24254 2001-07-01) releases and development versions of GDB, or any 24255 earlier versions. There is a problem setting breakpoints by line 24256 number, and other related issues that have been fixed in GCC 3.0 24257 but not yet handled in GDB: 24258 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 24259 24260 24261 For questions related to the use of GCC, please consult these web 24262 pages and the [2]GCC manuals. If that fails, the 24263 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24264 web pages and the development of GCC are welcome on our developer 24265 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 24266 24267 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 24268 distribution of this entire article is permitted in any medium, 24269 provided this notice is preserved. 24270 24271 These pages are [7]maintained by the GCC team. Last modified 24272 2022-10-26. 24273 24274 References 24275 24276 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 24277 2. https://gcc.gnu.org/onlinedocs/ 24278 3. mailto:gcc-help (a] gcc.gnu.org 24279 4. mailto:gcc (a] gcc.gnu.org 24280 5. https://gcc.gnu.org/lists.html 24281 6. https://www.fsf.org/ 24282 7. https://gcc.gnu.org/about.html 24283 ====================================================================== 24284 http://gcc.gnu.org/gcc-2.95/index.html 24285 24286 GCC 2.95 24287 24288 (This release series is no longer supported.) 24289 24290 March 16, 2001: The GNU project and the GCC developers are pleased to 24291 announce the release of GCC version 2.95.3. 24292 24293 Release History 24294 24295 GCC 2.95.3 24296 March 16, 2001 24297 24298 GCC 2.95.2 24299 October 27, 1999 24300 24301 GCC 2.95.1 24302 August 19, 1999 24303 24304 GCC 2.95 24305 July 31, 1999. This is the first release of GCC since the April 24306 1999 GCC/EGCS reunification and includes nearly a year's worth 24307 of new development and bugfixes. 24308 24309 References and Acknowledgements 24310 24311 GCC used to stand for the GNU C Compiler, but since the compiler 24312 supports several other languages aside from C, it now stands for the 24313 GNU Compiler Collection. 24314 24315 The whole suite has been extensively [1]regression tested and 24316 [2]package tested. It should be reliable and suitable for widespread 24317 use. 24318 24319 The compiler has several new optimizations, new targets, new languages 24320 and other new features. See the [3]new features page for a more 24321 complete list of new features found in the GCC 2.95 releases. 24322 24323 The sources include installation instructions in both HTML and 24324 plaintext forms in the install directory in the distribution. However, 24325 the most up to date installation instructions and [4]build/test status 24326 are on the web pages. We will update those pages as new information 24327 becomes available. 24328 24329 The GCC developers would like to thank the numerous people that have 24330 contributed new features, test results, bugfixes, etc to GCC. This 24331 [5]amazing group of volunteers is what makes GCC successful. 24332 24333 And finally, we can't in good conscience fail to mention some 24334 [6]caveats to using GCC 2.95. 24335 24336 Download GCC 2.95 from one of our many [7]mirror sites. 24337 24338 For additional information about GCC please see the [8]GCC project web 24339 server or contact the [9]GCC development mailing list. 24340 24341 24342 For questions related to the use of GCC, please consult these web 24343 pages and the [10]GCC manuals. If that fails, the 24344 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24345 web pages and the development of GCC are welcome on our developer 24346 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 24347 archives. 24348 24349 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 24350 distribution of this entire article is permitted in any medium, 24351 provided this notice is preserved. 24352 24353 These pages are [15]maintained by the GCC team. Last modified 24354 2022-10-26. 24355 24356 References 24357 24358 1. https://gcc.gnu.org/gcc-2.95/regress.html 24359 2. https://gcc.gnu.org/gcc-2.95/othertest.html 24360 3. https://gcc.gnu.org/gcc-2.95/features.html 24361 4. https://gcc.gnu.org/gcc-2.95/buildstat.html 24362 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24363 6. https://gcc.gnu.org/gcc-2.95/caveats.html 24364 7. https://gcc.gnu.org/mirrors.html 24365 8. https://gcc.gnu.org/index.html 24366 9. mailto:gcc (a] gcc.gnu.org 24367 10. https://gcc.gnu.org/onlinedocs/ 24368 11. mailto:gcc-help (a] gcc.gnu.org 24369 12. mailto:gcc (a] gcc.gnu.org 24370 13. https://gcc.gnu.org/lists.html 24371 14. https://www.fsf.org/ 24372 15. https://gcc.gnu.org/about.html 24373 ====================================================================== 24374 http://gcc.gnu.org/gcc-2.95/features.html 24375 24376 GCC 2.95 New Features 24377 24378 * General Optimizer Improvements: 24379 + [1]Localized register spilling to improve speed and code 24380 density especially on small register class machines. 24381 + [2]Global CSE using lazy code motion algorithms. 24382 + [3]Improved global constant/copy propagation. 24383 + [4]Improved control flow graph analysis and manipulation. 24384 + [5]Local dead store elimination. 24385 + [6]Memory Load hoisting/store sinking in loops. 24386 + [7]Type based alias analysis is enabled by default. Note this 24387 feature will expose bugs in the Linux kernel. Please refer to 24388 the FAQ (as shipped with GCC 2.95) for additional information 24389 on this issue. 24390 + Major revamp of GIV detection, combination and simplification 24391 to improve loop performance. 24392 + Major improvements to register allocation and reloading. 24393 * New Languages and Language specific improvements 24394 + [8]Many C++ improvements. 24395 + [9]Many Fortran improvements. 24396 + [10]Java front-end has been integrated. A [11]runtime library 24397 is available separately. 24398 + [12]ISO C99 support 24399 + [13]Chill front-end and runtime has been integrated. 24400 + Boehm garbage collector support in libobjc. 24401 + More support for various pragmas which appear in vendor 24402 include files 24403 * New Targets and Target Specific Improvements 24404 + [14]SPARC back end rewrite. 24405 + -mschedule=8000 will optimize code for PA8000 class 24406 processors; -mpa-risc-2-0 will generate code for PA2.0 24407 processors 24408 + Various micro-optimizations for the ia32 port. K6 24409 optimizations 24410 + Compiler will attempt to align doubles in the stack on the 24411 ia32 port 24412 + Alpha EV6 support 24413 + PowerPC 750 24414 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 24415 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 24416 + c3x, c4x 24417 + HyperSPARC 24418 + SparcLite86x 24419 + sh4 24420 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 24421 arm-linux) 24422 + vxWorks targets include support for vxWorks threads 24423 + StrongARM 110 and ARM9 support added. ARM Scheduling 24424 parameters rewritten. 24425 + Various changes to the MIPS port to avoid assembler macros, 24426 which in turn improves performance 24427 + Various performance improvements to the i960 port. 24428 + Major rewrite of ns32k port 24429 * Other significant improvements 24430 + [15]Ability to dump cfg information and display it using vcg. 24431 + The new faster scheme for fixing vendor header files is 24432 enabled by default. 24433 + Experimental internationalization support. 24434 + multibyte character support 24435 + Some compile-time speedups for pathological problems 24436 + Better support for complex types 24437 * Plus the usual mountain of bugfixes 24438 * Core compiler is based on the gcc2 development tree from Sept 30, 24439 1998, so we have all of the [16]features found in GCC 2.8. 24440 24441 Additional Changes in GCC 2.95.1 24442 24443 * Generic bugfixes and improvements 24444 + Various documentation fixes related to the GCC/EGCS merger. 24445 + Fix memory management bug which could lead to spurious aborts, 24446 core dumps or random parsing errors in the compiler. 24447 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 24448 support. 24449 + Fix infinite loop in the CSE optimizer. 24450 + Avoid undefined behavior in compiler FP emulation code 24451 + Fix install problem when prefix is overridden on the make 24452 install command. 24453 + Fix problem with unwanted installation of assert.h on some 24454 systems. 24455 + Fix problem with finding the wrong assembler in a single tree 24456 build. 24457 + Avoid increasing the known alignment of a register that is 24458 already known to be a pointer. 24459 * Platform specific bugfixes and improvements 24460 + Codegen bugfix for prologue/epilogue for cpu32 target. 24461 + Fix long long code generation bug for the Coldfire target. 24462 + Fix various aborts in the SH compiler. 24463 + Fix bugs in libgcc support library for the SH. 24464 + Fix alpha ev6 code generation bug. 24465 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 24466 AIX platforms. 24467 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 24468 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 24469 targets. 24470 + Fix weak symbol handling for rs6000/ppc svr4 targets. 24471 + Fix various problems with 64bit code generation for the 24472 rs6000/ppc port. 24473 + Fix codegen bug which caused tetex to be mis-compiled on the 24474 x86. 24475 + Fix compiler abort in new cfg code exposed by x86 port. 24476 + Fix out of range array reference in code convert flat 24477 registers to the x87 stacked FP register file. 24478 + Fix minor vxworks configuration bug. 24479 + Fix return type of bsearch for SunOS 4.x. 24480 * Language & Runtime specific fixes. 24481 + The G++ signature extension has been deprecated. It will be 24482 removed in the next major release of G++. Use of signatures 24483 will result in a warning from the compiler. 24484 + Several bugs relating to templates and namespaces were fixed. 24485 + A bug that caused crashes when combining templates with -g on 24486 DWARF1 platforms was fixed. 24487 + Pointers-to-members, virtual functions, and multiple 24488 inheritance should now work together correctly. 24489 + Some code-generation bugs relating to function try blocks were 24490 fixed. 24491 + G++ is a little bit more lenient with certain archaic 24492 constructs than in GCC 2.95. 24493 + Fix to prevent shared library version #s from bring truncated 24494 to 1 digit 24495 + Fix missing std:: in the libstdc++ library. 24496 + Fix stream locking problems in libio. 24497 + Fix problem in java compiler driver. 24498 24499 Additional Changes in GCC 2.95.2 24500 24501 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 24502 the optimizations performed by -fstrict-aliasing are valid according to 24503 the C and C++ standards, the optimization have caused some problems, 24504 particularly with old non-conforming code. 24505 24506 The GCC developers are experimenting with ways to warn users about code 24507 which violates the C/C++ standards, but those warnings are not ready 24508 for widespread use at this time. Rather than wait for those warnings 24509 the GCC developers have chosen to disable -fstrict-aliasing by default 24510 for the GCC 2.95.2 release. 24511 24512 We strongly encourage developers to find and fix code which violates 24513 the C/C++ standards as -fstrict-aliasing may be enabled by default in 24514 future releases. Use the option -fstrict-aliasing to re-enable these 24515 optimizations. 24516 * Generic bugfixes and improvements 24517 + Fix incorrectly optimized memory reference in global common 24518 subexpression elimination (GCSE) optimization pass. 24519 + Fix code generation bug in regmove.c in which it could 24520 incorrectly change a "const" value. 24521 + Fix bug in optimization of conditionals involving volatile 24522 memory references. 24523 + Avoid over-allocation of stack space for some procedures. 24524 + Fixed bug in the compiler which caused incorrect optimization 24525 of an obscure series of bit manipulations, shifts and 24526 arithmetic. 24527 + Fixed register allocator bug which caused teTeX to be 24528 mis-compiled on SPARC targets. 24529 + Avoid incorrect optimization of degenerate case statements for 24530 certain targets such as the ARM. 24531 + Fix out of range memory reference in the jump optimizer. 24532 + Avoid dereferencing null pointer in fix-header. 24533 + Fix test for GCC specific features so that it is possible to 24534 bootstrap with gcc-2.6.2 and older versions of GCC. 24535 + Fix typo in scheduler which could potentially cause out of 24536 range memory accesses. 24537 + Avoid incorrect loop reversal which caused incorrect code for 24538 certain loops on PowerPC targets. 24539 + Avoid incorrect optimization of switch statements on certain 24540 targets (for example the ARM). 24541 * Platform specific bugfixes and improvements 24542 + Work around bug in Sun V5.0 compilers which caused bootstrap 24543 comparison failures on SPARC targets. 24544 + Fix SPARC back end bug which caused aborts in final.c. 24545 + Fix sparc-hal-solaris2* configuration fragments. 24546 + Fix bug in sparc block profiling. 24547 + Fix obscure code generation bug for the PARISC targets. 24548 + Define __STDC_EXT__ for HPUX configurations. 24549 + Various POWERPC64 code generation bugfixes. 24550 + Fix abort for PPC targets using ELF (ex GNU/Linux). 24551 + Fix collect2 problems for AIX targets. 24552 + Correct handling of .file directive for PPC targets. 24553 + Fix bug in fix_trunc x86 patterns. 24554 + Fix x86 port to correctly pop the FP stack for functions that 24555 return structures in memory. 24556 + Fix minor bug in strlen x86 pattern. 24557 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 24558 + Fix template repository code to handle leading underscore in 24559 mangled names. 24560 + Fix weak/weak alias support for OpenBSD. 24561 + GNU/Linux for the ARM has C++ compatible include files. 24562 * Language & Runtime specific fixes. 24563 + Fix handling of constructor attribute in the C front-end which 24564 caused problems building the Chill runtime library on some 24565 targets. 24566 + Fix minor problem merging type qualifiers in the C front-end. 24567 + Fix aliasing bug for pointers and references (C/C++). 24568 + Fix incorrect "non-constant initializer bug" when -traditional 24569 or -fwritable-strings is enabled. 24570 + Fix build error for Chill front-end on SunOS. 24571 + Do not complain about duplicate instantiations when using 24572 -frepo (C++). 24573 + Fix array bounds handling in C++ front-end which caused 24574 problems with dwarf debugging information in some 24575 circumstances. 24576 + Fix minor namespace problem. 24577 + Fix problem linking java programs. 24578 24579 Additional Changes in GCC 2.95.3 24580 24581 * Generic bugfixes and improvements 24582 + Fix numerous problems that caused incorrect optimization in 24583 the register reloading code. 24584 + Fix numerous problems that caused incorrect optimization in 24585 the loop optimizer. 24586 + Fix aborts in the functions build_insn_chain and scan_loops 24587 under some circumstances. 24588 + Fix an alias analysis bug. 24589 + Fix an infinite compilation bug in the combiner. 24590 + A few problems with complex number support have been fixed. 24591 + It is no longer possible for gcc to act as a fork bomb when 24592 installed incorrectly. 24593 + The -fpack-struct option should be recognized now. 24594 + Fixed a bug that caused incorrect code to be generated due to 24595 a lost stack adjustment. 24596 * Platform specific bugfixes and improvements 24597 + Support building ARM toolchains hosted on Windows. 24598 + Fix attribute calculations in ARM toolchains. 24599 + arm-linux support has been improved. 24600 + Fix a PIC failure on sparc targets. 24601 + On ix86 targets, the regparm attribute should now work 24602 reliably. 24603 + Several updates for the h8300 port. 24604 + Fix problem building libio with glibc 2.2. 24605 24606 24607 For questions related to the use of GCC, please consult these web 24608 pages and the [17]GCC manuals. If that fails, the 24609 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24610 web pages and the development of GCC are welcome on our developer 24611 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 24612 archives. 24613 24614 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 24615 distribution of this entire article is permitted in any medium, 24616 provided this notice is preserved. 24617 24618 These pages are [22]maintained by the GCC team. Last modified 24619 2023-09-02. 24620 24621 References 24622 24623 1. https://gcc.gnu.org/news/spill.html 24624 2. https://gcc.gnu.org/news/lcm.html 24625 3. https://gcc.gnu.org/news/cprop.html 24626 4. https://gcc.gnu.org/news/cfg.html 24627 5. https://gcc.gnu.org/news/dse.html 24628 6. https://gcc.gnu.org/news/hoist.html 24629 7. https://gcc.gnu.org/news/alias.html 24630 8. https://gcc.gnu.org/gcc-2.95/c++features.html 24631 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 24632 10. https://gcc.gnu.org/news/gcj-announce.txt 24633 11. https://gcc.gnu.org/news/javaannounce.html 24634 12. https://gcc.gnu.org/c99status.html 24635 13. https://gcc.gnu.org/news/chill.html 24636 14. https://gcc.gnu.org/news/sparc.html 24637 15. https://gcc.gnu.org/news/egcs-vcg.html 24638 16. https://gcc.gnu.org/egcs-1.0/features-2.8.html 24639 17. https://gcc.gnu.org/onlinedocs/ 24640 18. mailto:gcc-help (a] gcc.gnu.org 24641 19. mailto:gcc (a] gcc.gnu.org 24642 20. https://gcc.gnu.org/lists.html 24643 21. https://www.fsf.org/ 24644 22. https://gcc.gnu.org/about.html 24645 ====================================================================== 24646 http://gcc.gnu.org/gcc-2.95/caveats.html 24647 24648 GCC 2.95 Caveats 24649 24650 * GCC 2.95 will issue an error for invalid asm statements that had 24651 been silently accepted by earlier versions of the compiler. This is 24652 particularly noticeable when compiling older versions of the Linux 24653 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 24654 for more information on this issue. 24655 * GCC 2.95 implements type based alias analysis to disambiguate 24656 memory references. Some programs, particularly the Linux kernel 24657 violate ANSI/ISO aliasing rules and therefore may not operate 24658 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 24659 shipped with GCC 2.95) for more information on this issue. 24660 * GCC 2.95 has a known bug in its handling of complex variables for 24661 64bit targets. Instead of silently generating incorrect code, GCC 24662 2.95 will issue a fatal error for situations it can not handle. 24663 This primarily affects the Fortran community as Fortran makes more 24664 use of complex variables than C or C++. 24665 * GCC 2.95 has an integrated libstdc++, but does not have an 24666 integrated libg++. Furthermore old libg++ releases will not work 24667 with GCC 2.95. You can retrieve a recent copy of libg++ from the 24668 [1]GCC ftp server. 24669 Note most C++ programs only need libstdc++. 24670 * Exception handling may not work with shared libraries, particularly 24671 on alphas, hppas, rs6000/powerpc and mips based platforms. 24672 Exception handling is known to work on x86 GNU/Linux platforms with 24673 shared libraries. 24674 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 24675 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 24676 or EGCS 1.1. As a result it may be necessary to fix C++ code before 24677 it will compile with GCC 2.95. 24678 * G++ is also converting toward the ISO C++ standard; as a result 24679 code which was previously valid (and thus accepted by other 24680 compilers and older versions of g++) may no longer be accepted. The 24681 flag -fpermissive may allow some non-conforming code to compile 24682 with GCC 2.95. 24683 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 24684 1.1.x, EGCS 1.0.x or GCC 2.8.x. 24685 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 24686 between Sept 30, 1998 and April 30, 1999 (the official end of the 24687 GCC 2.8 project). Future GCC releases will include all the changes 24688 from the defunct GCC 2.8 sources. 24689 24690 24691 For questions related to the use of GCC, please consult these web 24692 pages and the [2]GCC manuals. If that fails, the 24693 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24694 web pages and the development of GCC are welcome on our developer 24695 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 24696 24697 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 24698 distribution of this entire article is permitted in any medium, 24699 provided this notice is preserved. 24700 24701 These pages are [7]maintained by the GCC team. Last modified 24702 2022-10-26. 24703 24704 References 24705 24706 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 24707 2. https://gcc.gnu.org/onlinedocs/ 24708 3. mailto:gcc-help (a] gcc.gnu.org 24709 4. mailto:gcc (a] gcc.gnu.org 24710 5. https://gcc.gnu.org/lists.html 24711 6. https://www.fsf.org/ 24712 7. https://gcc.gnu.org/about.html 24713 ====================================================================== 24714 http://gcc.gnu.org/egcs-1.1/index.html 24715 24716 EGCS 1.1 24717 24718 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 24719 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 24720 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 24721 24722 EGCS is a free software project to further the development of the GNU 24723 compilers using an open development environment. 24724 24725 EGCS 1.1 is a major new release of the EGCS compiler system. It has 24726 been [1]extensively tested and is believed to be stable and suitable 24727 for widespread use. 24728 24729 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 24730 development sources; it contains all of the new features found in GCC 24731 2.8.1 as well as all new development from GCC up to June 6, 1998. 24732 24733 EGCS 1.1 also contains many improvements and features not found in GCC 24734 or in older versions of EGCS: 24735 * Global common subexpression elimination and global constant/copy 24736 propagation (aka [2]gcse) 24737 * Ongoing improvements to the [3]alias analysis support to allow for 24738 better optimizations throughout the compiler. 24739 * Vastly improved [4]C++ compiler and integrated C++ runtime 24740 libraries. 24741 * Fixes for the /tmp symlink race security problems. 24742 * New targets including mips16, arm-thumb and 64-bit PowerPC. 24743 * Improvements to GNU Fortran (g77) compiler and runtime library made 24744 since g77 version 0.5.23. 24745 24746 See the [5]new features page for a more complete list of new features 24747 found in EGCS 1.1 releases. 24748 24749 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 24750 1.1: 24751 * General improvements and fixes 24752 + Avoid some stack overflows when compiling large functions. 24753 + Avoid incorrect loop invariant code motions. 24754 + Fix some core dumps on Linux kernel code. 24755 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 24756 + Fix code generation problem in gcse. 24757 + Various documentation related fixes. 24758 * g++/libstdc++ improvements and fixes 24759 + MT safe EH fix for setjmp/longjmp based exception handling. 24760 + Fix a few bad interactions between optimization and exception 24761 handling. 24762 + Fixes for demangling of template names starting with "__". 24763 + Fix a bug that would fail to run destructors in some cases 24764 with -O2. 24765 + Fix 'new' of classes with virtual bases. 24766 + Fix crash building Qt on the Alpha. 24767 + Fix failure compiling WIFEXITED macro on GNU/Linux. 24768 + Fix some -frepo failures. 24769 * g77 and libf2c improvements and fixes 24770 + Various documentation fixes. 24771 + Avoid compiler crash on RAND intrinsic. 24772 + Fix minor bugs in makefiles exposed by BSD make programs. 24773 + Define _XOPEN_SOURCE for libI77 build to avoid potential 24774 problems on some 64-bit systems. 24775 + Fix problem with implicit endfile on rewind. 24776 + Fix spurious recursive I/O errors. 24777 * platform specific improvements and fixes 24778 + Match all versions of UnixWare7. 24779 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 24780 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 24781 from unsigned ints to double precision floats. 24782 + Fix ARM ABI issue with NetBSD. 24783 + Fix a few arm code generation bugs. 24784 + Fixincludes will fix additional broken SCO OpenServer header 24785 files. 24786 + Fix a m68k back end bug which caused invalid offsets in reg+d 24787 addresses. 24788 + Fix problems with 64bit AIX 4.3 support. 24789 + Fix handling of long longs for varargs/stdarg functions on the 24790 ppc. 24791 + Minor fixes to CPP predefines for Windows. 24792 + Fix code generation problems with gpr<->fpr copies for 64bit 24793 ppc. 24794 + Fix a few coldfire code generation bugs. 24795 + Fix some more header file problems on SunOS 4.x. 24796 + Fix assert.h handling for RTEMS. 24797 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 24798 + Fix x86 compiler abort in reg-stack pass. 24799 + Fix cygwin/windows problem with section attributes. 24800 + Fix Alpha code generation problem exposed by SMP Linux 24801 kernels. 24802 + Fix typo in m68k 32->64bit integer conversion. 24803 + Make sure target libraries build with -fPIC for PPC & Alpha 24804 targets. 24805 24806 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 24807 1.1.1: 24808 * General improvements and fixes 24809 + Fix bug in loop optimizer which caused the SPARC (and 24810 potentially other) ports to segfault. 24811 + Fix infinite recursion in alias analysis and combiner code. 24812 + Fix bug in regclass preferencing. 24813 + Fix incorrect loop reversal which caused incorrect code to be 24814 generated for several targets. 24815 + Fix return value for builtin memcpy. 24816 + Reduce compile time for certain loops which exposed quadratic 24817 behavior in the loop optimizer. 24818 + Fix bug which caused volatile memory to be written multiple 24819 times when only one write was needed/desired. 24820 + Fix compiler abort in caller-save.c 24821 + Fix combiner bug which caused incorrect code generation for 24822 certain division by constant operations. 24823 + Fix incorrect code generation due to a bug in range check 24824 optimizations. 24825 + Fix incorrect code generation due to mis-handling of clobbered 24826 values in CSE. 24827 + Fix compiler abort/segfault due to incorrect register 24828 splitting when unrolling loops. 24829 + Fix code generation involving autoincremented addresses with 24830 ternary operators. 24831 + Work around bug in the scheduler which caused qt to be 24832 mis-compiled on some platforms. 24833 + Fix code generation problems with -fshort-enums. 24834 + Tighten security for temporary files. 24835 + Improve compile time for codes which make heavy use of 24836 overloaded functions. 24837 + Fix multiply defined constructor/destructor symbol problems. 24838 + Avoid setting bogus RPATH environment variable during 24839 bootstrap. 24840 + Avoid GNU-make dependencies in the texinfo subdir. 24841 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 24842 --enable-cpp=<dirname> can be used to specify an additional 24843 install directory for the cpp wrapper script. 24844 + Fix CSE bug which caused incorrect label-label refs to appear 24845 on some platforms. 24846 + Avoid linking in EH routines from libgcc if they are not 24847 needed. 24848 + Avoid obscure bug in aliasing code. 24849 + Fix bug in weak symbol handling. 24850 * Platform-specific improvements and fixes 24851 + Fix detection of PPro/PII on Unixware 7. 24852 + Fix compiler segfault when building spec99 and other programs 24853 for SPARC targets. 24854 + Fix code-generation bugs for integer and floating point 24855 conditional move instructions on the PPro/PII. 24856 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 24857 + Fix build failure for the arc port. 24858 + Fix floating point format configuration for i?86-gnu port. 24859 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 24860 threads are enabled. 24861 + Fix coldfire code generation bugs. 24862 + Fix "unrecognized insn" problems for Alpha and PPC ports. 24863 + Fix h8/300 code generation problem with floating point values 24864 in memory. 24865 + Fix unrecognized insn problems for the m68k port. 24866 + Fix namespace-pollution problem for the x86 port. 24867 + Fix problems with old assembler on x86 NeXT systems. 24868 + Fix PIC code-generation problems for the SPARC port. 24869 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 24870 + Fix minor ISO namespace violation in Alpha varargs/stdarg 24871 support. 24872 + Fix incorrect "braf" instruction usage for the SH port. 24873 + Fix minor bug in va-sh which prevented its use with -ansi. 24874 + Fix problems recognizing and supporting FreeBSD. 24875 + Handle OpenBSD systems correctly. 24876 + Minor fixincludes fix for Digital UNIX 4.0B. 24877 + Fix problems with ctors/dtors in SCO shared libraries. 24878 + Abort instead of generating incorrect code for PPro/PII 24879 floating point conditional moves. 24880 + Avoid multiply defined symbols on GNU/Linux systems using 24881 libc-5.4.xx. 24882 + Fix abort in alpha compiler. 24883 * Fortran-specific fixes 24884 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 24885 is in the documented, non-Y2K-compliant range of 0-99, instead 24886 of being returned as 100 in the year 2000. 24887 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 24888 milliseconds value properly in Values(8). 24889 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 24890 information properly in SArray(7). 24891 24892 Each release includes installation instructions in both HTML and 24893 plaintext forms (see the INSTALL directory in the toplevel directory of 24894 the distribution). However, we also keep the most up to date 24895 installation instructions and [6]build/test status on our web page. We 24896 will update those pages as new information becomes available. 24897 24898 The EGCS project would like to thank the numerous people that have 24899 contributed new features, test results, bugfixes, etc. This [7]amazing 24900 group of volunteers is what makes EGCS successful. 24901 24902 And finally, we can't in good conscience fail to mention some 24903 [8]caveats to using EGCS 1.1. 24904 24905 Download EGCS from egcs.cygnus.com (USA California). 24906 24907 The EGCS 1.1 release is also available on many mirror sites. 24908 [9]Goto mirror list to find a closer site. 24909 24910 24911 For questions related to the use of GCC, please consult these web 24912 pages and the [10]GCC manuals. If that fails, the 24913 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 24914 web pages and the development of GCC are welcome on our developer 24915 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 24916 archives. 24917 24918 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 24919 distribution of this entire article is permitted in any medium, 24920 provided this notice is preserved. 24921 24922 These pages are [15]maintained by the GCC team. Last modified 24923 2024-05-18. 24924 24925 References 24926 24927 1. https://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 24928 2. https://gcc.gnu.org/news/gcse.html 24929 3. https://gcc.gnu.org/news/alias.html 24930 4. https://gcc.gnu.org/egcs-1.1/c++features.html 24931 5. https://gcc.gnu.org/egcs-1.1/features.html 24932 6. https://gcc.gnu.org/egcs-1.1/buildstat.html 24933 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 24934 8. https://gcc.gnu.org/egcs-1.1/caveats.html 24935 9. https://gcc.gnu.org/mirrors.html 24936 10. https://gcc.gnu.org/onlinedocs/ 24937 11. mailto:gcc-help (a] gcc.gnu.org 24938 12. mailto:gcc (a] gcc.gnu.org 24939 13. https://gcc.gnu.org/lists.html 24940 14. https://www.fsf.org/ 24941 15. https://gcc.gnu.org/about.html 24942 ====================================================================== 24943 http://gcc.gnu.org/egcs-1.1/features.html 24944 24945 EGCS 1.1 new features 24946 24947 * Integrated GNU Fortran (g77) compiler and runtime library with 24948 improvements, based on g77 version 0.5.23. 24949 * Vast improvements in the C++ compiler; so many they have [1]page of 24950 their own! 24951 * Compiler implements [2]global common subexpression elimination and 24952 global copy/constant propagation. 24953 * More major improvements in the [3]alias analysis code. 24954 * More major improvements in the exception handling code to improve 24955 performance, lower static overhead and provide the infrastructure 24956 for future improvements. 24957 * The infamous /tmp symlink race security problems have been fixed. 24958 * The regmove optimization pass has been nearly completely rewritten 24959 to improve performance of generated code. 24960 * The compiler now recomputes register usage information before local 24961 register allocation. By providing more accurate information to the 24962 priority based allocator, we get better register allocation. 24963 * The register reloading phase of the compiler optimizes spill code 24964 much better than in previous releases. 24965 * Some bad interactions between the register allocator and 24966 instruction scheduler have been fixed, resulting in much better 24967 code for certain programs. Additionally, we have tuned the 24968 scheduler in various ways to improve performance of generated code 24969 for some architectures. 24970 * The compiler's branch shortening algorithms have been significantly 24971 improved to work better on targets which align jump targets. 24972 * The compiler now supports -Os to prefer optimizing for code space 24973 over optimizing for code speed. 24974 * The compiler will now totally eliminate library calls which compute 24975 constant values. This primarily helps targets with no integer 24976 div/mul support and targets without floating point support. 24977 * The compiler now supports an extensive "--help" option. 24978 * cpplib has been greatly improved and may be suitable for limited 24979 use. 24980 * Memory footprint for the compiler has been significantly reduced 24981 for some pathological cases. 24982 * The time to build EGCS has been improved for certain targets 24983 (particularly the alpha and mips platforms). 24984 * Many infrastructure improvements throughout the compiler, plus the 24985 usual mountain of bugfixes and minor improvements. 24986 * Target dependent improvements: 24987 + SPARC port now includes V8 plus and V9 support as well as 24988 performance tuning for Ultra class machines. The SPARC port 24989 now uses the Haifa scheduler. 24990 + Alpha port has been tuned for the EV6 processor and has an 24991 optimized expansion of memcpy/bzero. The Alpha port now uses 24992 the Haifa scheduler. 24993 + RS6000/PowerPC: support for the Power64 architecture and AIX 24994 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 24995 + x86: Alignment of static store data and jump targets is per 24996 Intel recommendations now. Various improvements throughout the 24997 x86 port to improve performance on Pentium processors 24998 (including improved epilogue sequences for Pentium chips and 24999 back end improvements which should help register allocation on 25000 all x86 variants. Conditional move support has been fixed and 25001 enabled for PPro processors. The x86 port also better supports 25002 64bit operations now. Unixware 7, a System V Release 5 target, 25003 is now supported and SCO OpenServer targets can support GAS. 25004 + MIPS has improved multiply/multiply-add support and now 25005 includes mips16 ISA support. 25006 + M68k has many micro-optimizations and Coldfire fixes. 25007 * Core compiler is based on the GCC development tree from June 9, 25008 1998, so we have all of the [4]features found in GCC 2.8. 25009 25010 25011 For questions related to the use of GCC, please consult these web 25012 pages and the [5]GCC manuals. If that fails, the 25013 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25014 web pages and the development of GCC are welcome on our developer 25015 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 25016 25017 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 25018 distribution of this entire article is permitted in any medium, 25019 provided this notice is preserved. 25020 25021 These pages are [10]maintained by the GCC team. Last modified 25022 2023-09-02. 25023 25024 References 25025 25026 1. https://gcc.gnu.org/egcs-1.1/c++features.html 25027 2. https://gcc.gnu.org/news/gcse.html 25028 3. https://gcc.gnu.org/news/alias.html 25029 4. https://gcc.gnu.org/egcs-1.0/features-2.8.html 25030 5. https://gcc.gnu.org/onlinedocs/ 25031 6. mailto:gcc-help (a] gcc.gnu.org 25032 7. mailto:gcc (a] gcc.gnu.org 25033 8. https://gcc.gnu.org/lists.html 25034 9. https://www.fsf.org/ 25035 10. https://gcc.gnu.org/about.html 25036 ====================================================================== 25037 http://gcc.gnu.org/egcs-1.1/caveats.html 25038 25039 EGCS 1.1 Caveats 25040 25041 * EGCS has an integrated libstdc++, but does not have an integrated 25042 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 25043 Lu has made a libg++-2.8.1.2 snapshot available which may work with 25044 EGCS. 25045 Note most C++ programs only need libstdc++. 25046 * Exception handling may not work with shared libraries, particularly 25047 on alphas, hppas, rs6000/powerpc and mips based platforms. 25048 Exception handling is known to work on x86-linux platforms with 25049 shared libraries. 25050 * Some versions of the Linux kernel have bugs which prevent them from 25051 being compiled or from running when compiled by EGCS. See the FAQ 25052 (as shipped with EGCS 1.1) for additional information. 25053 * In general, EGCS is more rigorous about rejecting invalid C++ code 25054 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 25055 a result it may be necessary to fix C++ code before it will compile 25056 with EGCS. 25057 * G++ is also converting toward the ISO C++ standard; as a result 25058 code which was previously valid (and thus accepted by other 25059 compilers and older versions of g++) may no longer be accepted. 25060 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 25061 or GCC 2.8.x due to changes necessary to support thread safe 25062 exception handling. 25063 25064 25065 For questions related to the use of GCC, please consult these web 25066 pages and the [1]GCC manuals. If that fails, the 25067 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25068 web pages and the development of GCC are welcome on our developer 25069 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 25070 25071 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 25072 distribution of this entire article is permitted in any medium, 25073 provided this notice is preserved. 25074 25075 These pages are [6]maintained by the GCC team. Last modified 25076 2022-10-26. 25077 25078 References 25079 25080 1. https://gcc.gnu.org/onlinedocs/ 25081 2. mailto:gcc-help (a] gcc.gnu.org 25082 3. mailto:gcc (a] gcc.gnu.org 25083 4. https://gcc.gnu.org/lists.html 25084 5. https://www.fsf.org/ 25085 6. https://gcc.gnu.org/about.html 25086 ====================================================================== 25087 http://gcc.gnu.org/egcs-1.0/index.html 25088 25089 EGCS 1.0 25090 25091 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 25092 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 25093 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 25094 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 25095 25096 EGCS is a collaborative effort involving several groups of hackers 25097 using an open development model to accelerate development and testing 25098 of GNU compilers and runtime libraries. 25099 25100 An important goal of EGCS is to allow wide scale testing of 25101 experimental features and optimizations; therefore, EGCS contains some 25102 features and optimizations which are still under development. However, 25103 EGCS has been carefully tested and should be comparable in quality to 25104 most GCC releases. 25105 25106 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 25107 development sources; it contains nearly all of the new features found 25108 in GCC 2.8. 25109 25110 EGCS 1.0 also contains many improvements and features not found in GCC 25111 2.7 and even the GCC 2.8 series (which was released after the original 25112 EGCS 1.0 release). 25113 * Integrated C++ runtime libraries, including support for most major 25114 GNU/Linux systems! 25115 * The integrated libstdc++ library includes a verbatim copy of SGI's 25116 STL release. 25117 * Integrated GNU Fortran compiler. 25118 * New instruction scheduler. 25119 * New alias analysis code. 25120 25121 See the [1]new features page for a more complete list of new features. 25122 25123 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 25124 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 25125 EGCS 1.0 release: 25126 * Add support for Red Hat 5.0 Linux and better support for Linux 25127 systems using glibc2. 25128 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 25129 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 25130 fix these problems. 25131 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 25132 handling interfaces. 25133 To avoid future compatibility problems, we strongly urge anyone who 25134 is planning on distributing shared libraries that contain C++ code 25135 to upgrade to EGCS 1.0.1 first. 25136 Soon after EGCS 1.0 was released, the GCC developers made some 25137 incompatible changes in libgcc's exception handling interfaces. 25138 These changes were needed to solve problems on some platforms. This 25139 means that GCC 2.8.0, when released, will not be seamlessly 25140 compatible with shared libraries built by EGCS 1.0. The reason is 25141 that the libgcc.a in GCC 2.8.0 will not contain a function needed 25142 by the old interface. 25143 The result of this is that there may be compatibility problems with 25144 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 25145 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 25146 and libgcc.a has the support routines for both the old and the new 25147 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 25148 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 25149 The maintainers of GCC 2.x have decided against including seamless 25150 support for the old interface in 2.8.0, since it was never 25151 "official", so to avoid future compatibility problems we recommend 25152 against distributing any shared libraries built by EGCS 1.0 that 25153 contain C++ code (upgrade to 1.0.1 and use that). 25154 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 25155 The x86 changes fix code generation errors exposed when building 25156 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 25157 The hppa change fixes a compiler abort when configured for use with 25158 RTEMS. 25159 The MIPS changes fix problems with the definition of LONG_MAX on 25160 newer systems, allow for command line selection of the target ABI, 25161 and fix one code generation problem. 25162 The rs6000/ppc change fixes some problems with passing structures 25163 to varargs/stdarg functions. 25164 * A few machine independent bugfixes, mostly to fix code generation 25165 errors when building Linux kernels or glibc. 25166 * Fix a few critical exception handling and template bugs in the C++ 25167 compiler. 25168 * Fix Fortran namelist bug on alphas. 25169 * Fix build problems on x86-solaris systems. 25170 25171 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 25172 serious problems in EGCS 1.0.1. 25173 * General improvements and fixes 25174 + Memory consumption significantly reduced, especially for 25175 templates and inline functions. 25176 + Fix various problems with glibc2.1. 25177 + Fix loop optimization bug exposed by rs6000/ppc port. 25178 + Fix to avoid potential code generation problems in jump.c. 25179 + Fix some undefined symbol problems in dwarf1 debug support. 25180 * g++/libstdc++ improvements and fixes 25181 + libstdc++ in the EGCS release has been updated and should be 25182 link compatible with libstdc++-2.8. 25183 + Various fixes in libio/libstdc++ to work better on GNU/Linux 25184 systems. 25185 + Fix problems with duplicate symbols on systems that do not 25186 support weak symbols. 25187 + Memory corruption bug and undefined symbols in bastring have 25188 been fixed. 25189 + Various exception handling fixes. 25190 + Fix compiler abort for very long thunk names. 25191 * g77 improvements and fixes 25192 + Fix compiler crash for omitted bound in Fortran CASE 25193 statement. 25194 + Add missing entries to g77 lang-options. 25195 + Fix problem with -fpedantic in the g77 compiler. 25196 + Fix "backspace" problem with g77 on alphas. 25197 + Fix x86 back end problem with Fortran literals and -fpic. 25198 + Fix some of the problems with negative subscripts for g77 on 25199 alphas. 25200 + Fixes for Fortran builds on cygwin32/mingw32. 25201 * platform specific improvements and fixes 25202 + Fix long double problems on x86 (exposed by glibc). 25203 + x86 ports define i386 again to keep imake happy. 25204 + Fix exception handling support on NetBSD ports. 25205 + Several changes to collect2 to fix many problems with AIX. 25206 + Define __ELF__ for GNU/Linux on rs6000. 25207 + Fix -mcall-linux problem on GNU/Linux on rs6000. 25208 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 25209 + Allow autoconf to select a proper install problem on AIX 3.1. 25210 + m68k port support includes -mcpu32 option as well as cpu32 25211 multilibs. 25212 + Fix stdarg bug for irix6. 25213 + Allow EGCS to build on irix5 without the gnu assembler. 25214 + Fix problem with static linking on sco5. 25215 + Fix bootstrap on sco5 with native compiler. 25216 + Fix for abort building newlib on H8 target. 25217 + Fix fixincludes handling of math.h on SunOS. 25218 + Minor fix for Motorola 3300 m68k systems. 25219 25220 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 25221 problems reported by Red Hat for builds of Red Hat 5.1. 25222 * Generic bugfixes: 25223 + Fix a typo in the libio library which resulted in incorrect 25224 behavior of istream::get. 25225 + Fix the Fortran negative array index problem. 25226 + Fix a major problem with the ObjC runtime thread support 25227 exposed by glibc2. 25228 + Reduce memory consumption of the Haifa scheduler. 25229 * Target specific bugfixes: 25230 + Fix one x86 floating point code generation bug exposed by 25231 glibc2 builds. 25232 + Fix one x86 internal compiler error exposed by glibc2 builds. 25233 + Fix profiling bugs on the Alpha. 25234 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 25235 + Fix rs6000/ppc bug when converting values from integer types 25236 to floating point types. 25237 25238 The EGCS 1.0 releases include installation instructions in both HTML 25239 and plaintext forms (see the INSTALL directory in the toplevel 25240 directory of the distribution). However, we also keep the most up to 25241 date installation instructions and [2]build/test status on our web 25242 page. We will update those pages as new information becomes available. 25243 25244 And, we can't in good conscience fail to mention some [3]caveats to 25245 using EGCS. 25246 25247 Update: Big thanks to Stanford for providing a high speed link for 25248 downloading EGCS (go.cygnus.com)! 25249 25250 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 25251 (USA California -- High speed link provided by Stanford). 25252 25253 The EGCS 1.0 release is also available many mirror sites. 25254 [4]Goto mirror list to find a closer site 25255 25256 We'd like to thank the numerous people that have contributed new 25257 features, test results, bugfixes, etc. Unfortunately, they're far too 25258 numerous to mention by name. 25259 25260 25261 For questions related to the use of GCC, please consult these web 25262 pages and the [5]GCC manuals. If that fails, the 25263 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25264 web pages and the development of GCC are welcome on our developer 25265 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 25266 25267 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 25268 distribution of this entire article is permitted in any medium, 25269 provided this notice is preserved. 25270 25271 These pages are [10]maintained by the GCC team. Last modified 25272 2023-09-02. 25273 25274 References 25275 25276 1. https://gcc.gnu.org/egcs-1.0/features.html 25277 2. https://gcc.gnu.org/egcs-1.0/buildstat.html 25278 3. https://gcc.gnu.org/egcs-1.0/caveats.html 25279 4. https://gcc.gnu.org/mirrors.html 25280 5. https://gcc.gnu.org/onlinedocs/ 25281 6. mailto:gcc-help (a] gcc.gnu.org 25282 7. mailto:gcc (a] gcc.gnu.org 25283 8. https://gcc.gnu.org/lists.html 25284 9. https://www.fsf.org/ 25285 10. https://gcc.gnu.org/about.html 25286 ====================================================================== 25287 http://gcc.gnu.org/egcs-1.0/features.html 25288 25289 EGCS 1.0 features 25290 25291 * Core compiler is based on the gcc2 development tree from Aug 2, 25292 1997, so we have most of the [1]features found in GCC 2.8. 25293 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 25294 * Vast improvements in the C++ compiler; so many they have [2]page of 25295 their own! 25296 * Integrated C++ runtime libraries, including support for most major 25297 GNU/Linux systems! 25298 * New instruction scheduler from IBM Haifa which includes support for 25299 function wide instruction scheduling as well as superscalar 25300 scheduling. 25301 * Significantly improved alias analysis code. 25302 * Improved register allocation for two address machines. 25303 * Significant code generation improvements for Fortran code on 25304 Alphas. 25305 * Various optimizations from the g77 project as well as improved loop 25306 optimizations. 25307 * Dwarf2 debug format support for some targets. 25308 * egcs libstdc++ includes the SGI STL implementation without changes. 25309 * As a result of these and other changes, egcs libstc++ is not binary 25310 compatible with previous releases of libstdc++. 25311 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 25312 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 25313 1.1), Support for RTEMS on several embedded targets, Support for 25314 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 25315 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 25316 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 25317 * RS6000/PowerPC ports generate code which can run on all 25318 RS6000/PowerPC variants by default. 25319 * -mcpu= and -march= switches for the x86 port to allow better 25320 control over how the x86 port generates code. 25321 * Includes the template repository patch (aka repo patch); note the 25322 new template code makes repo obsolete for ELF systems using gnu-ld 25323 such as GNU/Linux. 25324 * Plus the usual assortment of bugfixes and improvements. 25325 25326 25327 For questions related to the use of GCC, please consult these web 25328 pages and the [3]GCC manuals. If that fails, the 25329 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25330 web pages and the development of GCC are welcome on our developer 25331 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 25332 25333 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 25334 distribution of this entire article is permitted in any medium, 25335 provided this notice is preserved. 25336 25337 These pages are [8]maintained by the GCC team. Last modified 25338 2022-10-26. 25339 25340 References 25341 25342 1. https://gcc.gnu.org/egcs-1.0/features-2.8.html 25343 2. https://gcc.gnu.org/egcs-1.0/c++features.html 25344 3. https://gcc.gnu.org/onlinedocs/ 25345 4. mailto:gcc-help (a] gcc.gnu.org 25346 5. mailto:gcc (a] gcc.gnu.org 25347 6. https://gcc.gnu.org/lists.html 25348 7. https://www.fsf.org/ 25349 8. https://gcc.gnu.org/about.html 25350 ====================================================================== 25351 http://gcc.gnu.org/egcs-1.0/caveats.html 25352 25353 EGCS 1.0 Caveats 25354 25355 * EGCS has an integrated libstdc++, but does not have an integrated 25356 libg++. Furthermore old libg++ releases will not work with egc; HJ 25357 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 25358 Note most C++ programs only need libstdc++. 25359 * Note that using -pedantic or -Wreturn-type can cause an explosion 25360 in the amount of memory needed for template-heavy C++ code, such as 25361 code that uses STL. Also note that -Wall includes -Wreturn-type, so 25362 if you use -Wall you will need to specify -Wno-return-type to turn 25363 it off. 25364 * Exception handling may not work with shared libraries, particularly 25365 on alphas, hppas, and mips based platforms. Exception handling is 25366 known to work on x86-linux platforms with shared libraries. 25367 * Some versions of the Linux kernel have bugs which prevent them from 25368 being compiled or from running when compiled by EGCS. See the FAQ 25369 (as shipped with EGCS 1.0) for additional information. 25370 * In general, EGCS is more rigorous about rejecting invalid C++ code 25371 or deprecated C++ constructs than G++ 2.7. As a result it may be 25372 necessary to fix C++ code before it will compile with EGCS. 25373 * G++ is also aggressively tracking the C++ standard; as a result 25374 code which was previously valid (and thus accepted by other 25375 compilers and older versions of G++) may no longer be accepted. 25376 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 25377 1.0.x and later releases should work with Red Hat Linux 5.0. 25378 25379 25380 For questions related to the use of GCC, please consult these web 25381 pages and the [1]GCC manuals. If that fails, the 25382 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 25383 web pages and the development of GCC are welcome on our developer 25384 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 25385 25386 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 25387 distribution of this entire article is permitted in any medium, 25388 provided this notice is preserved. 25389 25390 These pages are [6]maintained by the GCC team. Last modified 25391 2022-10-26. 25392 25393 References 25394 25395 1. https://gcc.gnu.org/onlinedocs/ 25396 2. mailto:gcc-help (a] gcc.gnu.org 25397 3. mailto:gcc (a] gcc.gnu.org 25398 4. https://gcc.gnu.org/lists.html 25399 5. https://www.fsf.org/ 25400 6. https://gcc.gnu.org/about.html 25401 ====================================================================== 25402
Indexes created Tue Oct 27 00:25:24 UTC 2026