NEWS revision 1.19
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-12/index.html 9 10 GCC 12 Release Series 11 12 July 11, 2025 13 14 The GCC developers are pleased to announce the release of GCC 12.5. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 12.4 relative to previous releases of GCC. 18 19 Release History 20 21 GCC 12.5 22 July 11, 2025 ([1]changes, [2]documentation) 23 24 GCC 12.4 25 Jun 20, 2024 ([3]changes, [4]documentation) 26 27 GCC 12.3 28 May 8, 2023 ([5]changes, [6]documentation) 29 30 GCC 12.2 31 Aug 19, 2022 ([7]changes, [8]documentation) 32 33 GCC 12.1 34 May 6, 2022 ([9]changes, [10]documentation) 35 36 References and Acknowledgements 37 38 GCC used to stand for the GNU C Compiler, but since the compiler 39 supports several other languages aside from C, it now stands for the 40 GNU Compiler Collection. 41 42 The GCC developers would like to thank the numerous people that have 43 contributed new features, improvements, bug fixes, and other changes as 44 well as test results to GCC. This [11]amazing group of volunteers is 45 what makes GCC successful. 46 47 For additional information about GCC please refer to the [12]GCC 48 project web site or contact the [13]GCC development mailing list. 49 50 To obtain GCC please use [14]our mirror sites or [15]our version 51 control system. 52 53 54 For questions related to the use of GCC, please consult these web 55 pages and the [16]GCC manuals. If that fails, the 56 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 57 web pages and the development of GCC are welcome on our developer 58 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 59 archives. 60 61 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 62 distribution of this entire article is permitted in any medium, 63 provided this notice is preserved. 64 65 These pages are [21]maintained by the GCC team. Last modified 66 2025-07-11. 67 68 References 69 70 1. https://gcc.gnu.org/gcc-12/changes.html 71 2. https://gcc.gnu.org/onlinedocs/12.5.0/ 72 3. https://gcc.gnu.org/gcc-12/changes.html 73 4. https://gcc.gnu.org/onlinedocs/12.4.0/ 74 5. https://gcc.gnu.org/gcc-12/changes.html 75 6. https://gcc.gnu.org/onlinedocs/12.3.0/ 76 7. https://gcc.gnu.org/gcc-12/changes.html 77 8. https://gcc.gnu.org/onlinedocs/12.2.0/ 78 9. https://gcc.gnu.org/gcc-12/changes.html 79 10. https://gcc.gnu.org/onlinedocs/12.1.0/ 80 11. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Contributors.html 81 12. https://gcc.gnu.org/index.html 82 13. mailto:gcc (a] gcc.gnu.org 83 14. https://gcc.gnu.org/mirrors.html 84 15. https://gcc.gnu.org/git.html 85 16. https://gcc.gnu.org/onlinedocs/ 86 17. mailto:gcc-help (a] gcc.gnu.org 87 18. mailto:gcc (a] gcc.gnu.org 88 19. https://gcc.gnu.org/lists.html 89 20. https://www.fsf.org/ 90 21. https://gcc.gnu.org/about.html 91 ====================================================================== 92 http://gcc.gnu.org/gcc-12/changes.html 93 94 GCC 12 Release Series 95 Changes, New Features, and Fixes 96 97 This page is a "brief" summary of some of the huge number of 98 improvements in GCC 12. You may also want to check out our [1]Porting 99 to GCC 12 page and the [2]full GCC documentation. 100 101 Caveats 102 103 * An ABI incompatibility between C and C++ when passing or returning 104 by value certain aggregates containing zero width bit-fields has 105 been discovered on various targets. As mentioned in [3]PR102024, 106 since the [4]PR42217 fix in GCC 4.5 the C++ front-end has been 107 removing zero width bit-fields from the internal representation of 108 the aggregates after the layout of those aggregates, but the C 109 front-end kept them, so passing e.g. struct S { float a; int : 0; 110 float b; } or struct T { float c; int : 0; } by value could differ 111 between C and C++. Starting with GCC 12 the C++ front-end no longer 112 removes those bit-fields from the internal representation and per 113 clarified psABI some targets have been changed, so that they either 114 ignore those bit-fields in the argument passing by value decisions 115 in both C and C++, or they always take them into account. x86-64, 116 ARM and AArch64 will always ignore them (so there is a C ABI 117 incompatibility between GCC 11 and earlier with GCC 12 or later), 118 PowerPC64 ELFv2 and S/390 always take them into account (so there 119 is a C++ ABI incompatibility, GCC 4.4 and earlier compatible with 120 GCC 12 or later, incompatible with GCC 4.5 through GCC 11). RISC-V 121 has changed the handling of these already starting with GCC 10. As 122 the ABI requires, MIPS takes them into account handling function 123 return values so there is a C++ ABI incompatibility with GCC 4.5 124 through 11. For function arguments on MIPS, refer to [5]the MIPS 125 specific entry. GCC 12 on the above targets will report such 126 incompatibilities as warnings or other diagnostics unless 127 -Wno-psabi is used. 128 * C: Computed gotos require a pointer type now. 129 * C++: Two non-standard std::pair constructors have been deprecated. 130 These allowed the use of an rvalue and a literal 0 to construct a 131 pair containing a move-only type and a pointer. The nullptr keyword 132 should be used to initialize the pointer member instead of a 133 literal 0, as this is portable to other C++ implementations. 134 * The configuration option --enable-libstdcxx-allocator no longer 135 supports the bitmap, mt, and pool arguments. Those configurations 136 had been broken for some time. 137 * D: Building and bootstrapping GDC, the D compiler, now requires a 138 working GDC (GCC version 9.1 or later) and D runtime library, 139 libphobos, as the D front end is written in D. On some targets, 140 libphobos isn't enabled by default, but compiles and works if 141 --enable-libphobos is used. Other targets may require a more recent 142 version of GCC to bootstrap. Specifics are documented for affected 143 targets in the [6]Installing GCC manual. 144 * Fortran: OpenMP code using the omp_lib.h include file can no longer 145 be compiled with -std=f95 but now requires at least -std=f2003. 146 Alternatively, use the omp_lib module, which still supports 147 -std=f95 and is recommended to be used instead in general. 148 * OpenMP offloading to Intel MIC has been deprecated and will be 149 removed in a future release. 150 * The cr16 target with the cr16-*-* configuration has been obsoleted 151 and will be removed in a future release. 152 * The hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11* configurations 153 targeting 32-bit PA-RISC with HP-UX have been obsoleted and will be 154 removed in a future release. 155 * The m32c*-*-rtems* configuration has been obsoleted and will be 156 removed in a future release. 157 * The support for the m32r-*-linux*, m32rle-*-linux*, 158 m68k*-*-openbsd* and vax-*-openbsd* configurations has been 159 removed. 160 * STABS: Support for emitting the STABS debugging format is 161 deprecated and will be removed in the next release. All ports now 162 default to emit DWARF (version 2 or later) debugging info or are 163 obsoleted. 164 * The optimization level -Ofast now implies 165 -fno-semantic-interposition. 166 167 [7]General Improvements 168 169 * Vectorization is enabled at -O2 which is now equivalent to what 170 would have been -O2 -ftree-vectorize -fvect-cost-model=very-cheap 171 in the past. Note that the default vectorizer cost model has been 172 changed; it used to behave as if -fvect-cost-model=cheap had been 173 specified. 174 * GCC now supports the [8]ShadowCallStack sanitizer, which can be 175 enabled using the command-line option 176 [9]-fsanitize=shadow-call-stack. This sanitizer currently only 177 works on AArch64 targets and it requires an environment in which 178 all code has been compiled with -ffixed-r18. Its primary initial 179 user is the Linux kernel. 180 181 [10]New Languages and Language specific improvements 182 183 * OpenMP 184 + OpenMP 5.0 support has been extended: The close map modifier 185 and the affinity clause are now supported. In addition, 186 Fortran gained the following features which were available in 187 C and C++ before: declare variant is now available, depobj, 188 mutexinoutset and iterator can now also be used with the 189 depend clause, defaultmap has been updated for OpenMP 5.0, and 190 the loop directive and combined directives involving the 191 master directive have been added. 192 + The following OpenMP 5.1 features have been added: support for 193 expressing OpenMP directives as C++ 11 attributes, the masked 194 and scope constructs, the nothing and error directives, and 195 using primary with the proc_bind clause and OMP_PROC_BIND 196 environment variable, the reproducible and unconstrained 197 modifiers to the order clause, and, for C/C++ only, the align 198 and allocator modifiers to the allocate clause and the atomic 199 extensions are now available. The OMP_PLACE environment 200 variable supports the OpenMP 5.1 features. In addition, the 201 OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT environment variables 202 and their associated API routines are now supported as well as 203 the memory-allocation routines added for Fortran and extended 204 for C/C++ in OpenMP 5.1. In Fortran code, strictly structured 205 blocks can be used. 206 + The [11]OpenMP Implementation Status can be found in the 207 libgomp manual. 208 * Version 2.6 of the [12]OpenACC specification continues to be 209 maintained and improved in the C, C++ and Fortran compilers. See 210 the [13]implementation status section on the OpenACC wiki page and 211 the [14]run-time library documentation for further information. In 212 addition to general performance tuning and bug fixing, new features 213 include: 214 + OpenACC worker parallelism for [15]AMD GPUs (already for a 215 long time supported for [16]Nvidia GPUs). 216 + Data privatization/sharing at the OpenACC gang level. 217 + Considerable improvements for the experimental OpenACC 218 'kernels' decomposition ([17]--param 219 openacc-kernels=decompose). 220 + A new warning flag [18]-Wopenacc-parallelism to warn about 221 potentially suboptimal choices related to OpenACC parallelism. 222 * The offload target code generation for OpenMP and OpenACC can now 223 be better adjusted using the new [19]-foffload-options= flag and 224 the pre-existing but now documented [20]-foffload= flag. 225 226 [21]Ada 227 228 * Ada 2022 229 + Added the -gnat2022 flag to indicate strict Ada 2022 230 compliance. The old -gnat2020 flag is now deprecated. 231 + Support for Big Numbers (Annex G) has seen continuous 232 improvements. It is now considered complete. It is compatible 233 with SPARK, i.e. can be used from SPARK code. 234 + Continuous improvements to the Ada 2022 standard since GCC 11. 235 + Greatly improved compile time support. More functions can now 236 have the with Static aspect and can be used in more contexts. 237 * Ada 2022 extensions. The use of the -gnatX flag is necessary to 238 access these features as they are not considered stable or 239 standard. 240 + Fixed lower bound for unconstrained arrays. 241 o type Matrix is array (Natural range 0 .. <>, Natural 242 range 0 .. <>) of Integer; is now valid. 243 o Subtypes can also specify a lower bound: subtype String_1 244 is String (1 .. <>);. Boundaries from slices will "slide" 245 to the correct lower bound of the subtype. 246 + Generalized Object.Operand notation. The following code is now 247 valid V.Add_Element(42);, with V being a vector, for example. 248 + Additional when constructs. Keywords return, goto and raise 249 can now use when in addition to the existing exit when. The 250 following expression is therefore now valid raise 251 Constraint_Error with "Element is null" when Element = null; 252 + Pattern matching 253 o The case statement has been extended to cover records and 254 arrays as well as finer grained casing on scalar types. 255 In the future it is expected to provide more compile time 256 guarantees when accessing discriminated fields. Case 257 exhaustion is supported for pattern matching. An example 258 would be 259 type Sign is (Neg, Zero, Pos); 260 261 function Multiply (S1, S2 : Sign) return Sign is 262 (case (S1, S2) is 263 when (Neg, Neg) | (Pos, Pos) => Pos, 264 when (Zero, <>) | (<>, Zero) => Zero, 265 when (Neg, Pos) | (Pos, Neg) => Neg); 266 267 * gnatfind and gnatxref, which were already deprecated, have been 268 removed. 269 * Greatly expanded code covered by contracts. Thanks to this work, 270 there are now several Ada standard libraries fully proven in SPARK 271 which means they have no runtime nor logical errors. They are 272 mostly numeric and string handling libraries. 273 * Enable return-slot optimization for Pure functions. 274 * General optimizations, improvements and additions to the standard 275 library. Performance, correctness and in some cases stability was 276 improved. Memory pools have also seen some minor enhancements. 277 * Improvements to embedded-RTOS targets such as RTEMS, VxWorks and 278 QNX. Older targets were removed or cleaned. 279 * Added some [22]hardening features. 280 281 [23]C family 282 283 * Support for __builtin_shufflevector compatible with the clang 284 language extension was added. 285 * Support for attribute unavailable was added. 286 * A new built-in function, __builtin_assoc_barrier, was added. It can 287 be used to inhibit re-association of floating-point expressions. 288 * Support for __builtin_dynamic_object_size compatible with the clang 289 language extension was added. 290 * New warnings: 291 + [24]-Wbidi-chars warns about potentially misleading UTF-8 292 bidirectional control characters. The default is 293 -Wbidi-chars=unpaired ([25]PR103026) 294 + [26]-Warray-compare warns about comparisons between two 295 operands of array type ([27]PR97573) 296 * Enhancements to existing warnings: 297 + [28]-Wattributes has been extended so that it's possible to 298 use -Wno-attributes=ns::attr or -Wno-attributes=ns:: to 299 suppress warnings about unknown scoped attributes (in C++11 300 and C2X). Similarly, #pragma GCC diagnostic ignored_attributes 301 "vendor::attr" can be used to achieve the same effect 302 ([29]PR101940) 303 304 [30]C 305 306 * Some new features from the upcoming C2X revision of the ISO C 307 standard are supported with -std=c2x and -std=gnu2x. Some of these 308 features are also supported as extensions when compiling for older 309 language versions. In addition to the features listed, some 310 features previously supported as extensions and now added to the C 311 standard are enabled by default in C2X mode and not diagnosed with 312 -std=c2x -Wpedantic. 313 + Digit separators (as in C++) are supported for C2X. 314 + The #elifdef and #elifndef preprocessing directives are now 315 supported. 316 + The printf and scanf format checking with [31]-Wformat now 317 supports the %b format specified by C2X for binary integers, 318 and the %B format recommended by C2X for printf. 319 320 [32]C++ 321 322 * Several C++23 features have been implemented: 323 + [33]P1938R3, if consteval ([34]PR100974) 324 + [35]P0849R8, auto(x): decay-copy in the language 325 ([36]PR103049) 326 + [37]P2242R3, Non-literal variables (and labels and gotos) in 327 constexpr functions ([38]PR102612) 328 + [39]P2334R1, Support for preprocessing directives elifdef and 329 elifndef ([40]PR102616) 330 + [41]P2360R0, Extend init-statement to allow alias-declaration 331 ([42]PR102617) 332 + [43]P2128R6, Multidimensional subscript operator 333 + [44]DR 2397, auto specifier for pointers and references to 334 arrays ([45]PR100975) 335 * Several C++ Defect Reports have been resolved, e.g.: 336 + [46]DR 960, Covariant functions and lvalue/rvalue references 337 + [47]DR 1227, Mixing immediate and non-immediate contexts in 338 deduction failure 339 + [48]DR 1315, Restrictions on non-type template arguments in 340 partial specializations 341 + [49]DR 2082, Referring to parameters in unevaluated operands 342 of default arguments 343 + [50]DR 2351, void{} 344 + [51]DR 2374, Overly permissive specification of enum 345 direct-list-initialization 346 + [52]DR 2397, auto specifier for pointers and references to 347 arrays 348 + [53]DR 2446, Questionable type-dependency of concept-ids 349 * New command-line option -fimplicit-constexpr can be used to make 350 inline functions implicitly constexpr ([54]git) 351 * New command-line option -ffold-simple-inlines can be used to fold 352 calls to certain trivial inline functions (currently std::move, 353 std::forward, std::addressof and std::as_const). In contrast to 354 inlining such calls, folding means that no intermediate code or 355 debug information will be generated for them; this minimizes the 356 abstraction penalty incurred for using these functions versus using 357 the fundamental operations from which they're defined (e.g. 358 std::move versus static_cast). This flag is enabled by default when 359 -fno-inline is not active. 360 * Deduction guides can be declared at class scope ([55]PR79501) 361 * [56]-Wuninitialized warns about using uninitialized variables in 362 member initializer lists ([57]PR19808) 363 * [58]-Wint-in-bool-context is now disabled when instantiating a 364 template ([59]git) 365 * Stricter checking of attributes on friend declarations: if a friend 366 declaration has an attribute, that declaration must be a 367 definition. Moreover, a C++11 attribute cannot appear in the middle 368 of the decl-specifier-seq. ([60]PR99032) 369 * New warning options for C++ language mismatches: 370 -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions, 371 -Wc++20-extensions, and -Wc++23-extensions. They are enabled by 372 default and can be used to control existing pedwarns about 373 occurrences of new C++ constructs in code using an old C++ standard 374 dialect. 375 * New warning [61]-Wmissing-requires warns about missing requires 376 ([62]git) 377 * The existing std::is_constant_evaluated in if warning was extended 378 to warn in more cases ([63]PR100995) 379 * [64]-Waddress has been enhanced so that it now warns about, for 380 instance, comparing the address of a nonstatic member function to 381 null ([65]PR102103) 382 * Errors about narrowing are no longer hidden if they occur in system 383 headers 384 * Ordered comparison of null pointers is now rejected ([66]PR99701) 385 * Anonymous structs with bases are now rejected ([67]git) 386 * The compiler rejects taking the address of an immediate member 387 function ([68]PR102753) 388 * The compiler has support for C++20 389 __cpp_lib_is_pointer_interconvertible and 390 __cpp_lib_is_layout_compatible to help the C++ library implement 391 [69]P0466, Layout-compatibility and Pointer-interconvertibility 392 Traits ([70]PR101539) 393 * Memory usage of constraint subsumption has been improved 394 ([71]PR100828) 395 * constinit thread_local variables are optimized better 396 ([72]PR101786) 397 * Support for C++17 std::hardware_destructive_interference_size was 398 added, along with the [73]-Winterference-size warning ([74]git) 399 * Many bugs in the CTAD handling have been fixed ([75]PR101344, 400 [76]PR101883, [77]PR89062, [78]PR101233, [79]PR88252, [80]PR86439, 401 [81]PR98832, [82]PR102933 ...) 402 * Two-stage name lookup for dependent operator expressions has been 403 corrected ([83]PR51577) 404 * Several issues with constrained variable templates have been fixed 405 ([84]PR98486) 406 * The compiler performs less instantiating when doing speculative 407 constant evaluation ([85]git) 408 * Various diagnostic improvements; e.g., a more precise caret 409 location for pointer-to-member expressions 410 * The new -fconstexpr-fp-except flag allows IEC559 floating point 411 exceptions in constant-expressions. 412 413 [86]Runtime Library (libstdc++) 414 415 * Improved experimental C++20 support, including: 416 + std::vector, std::basic_string, std::optional, and 417 std::variant can be used in constexpr functions. 418 + std::make_shared for arrays with default initialization, and 419 std::atomic<std::shared_ptr<T>>. 420 + Layout-compatibility and pointer-interconvertibility traits. 421 * Improved experimental C++23 support, including: 422 + Monadic operations for std::optional. 423 + std::expected 424 + std::move_only_function 425 + <spanstream> 426 + std::basic_string::resize_and_overwrite 427 + std::unique_ptr can be used in constexpr functions. 428 + <stacktrace> (not built by default, requires linking to an 429 extra library). 430 + <stdatomic.h> 431 + std::invoke_r 432 + constexpr std::type_info::operator== 433 434 [87]D 435 436 * New features: 437 + Support for the D programming language has been updated to 438 version 2.100.1 of the language and run-time library. Full 439 changelog for this release and previous releases can be found 440 on the [88]dlang.org website. 441 + On supported targets, the __traits(compiles) expression can 442 now be used to determine whether a target-specific built-in is 443 available without error during CTFE ([89]PR101127). 444 + Functions annotated with pragma(inline, true) are now compiled 445 into every module where they are used from ([90]PR106563). 446 + Partial support for directly importing C99 sources into a D 447 compilation ([91]ImportC) has been added to the language. A 448 notable missing feature is support for preprocessing C 449 imports, which can be worked around by preprocessing all C 450 sources used for importing ahead of time. 451 * New language options: 452 + -fcheck=, enables or disables the code generation of specific 453 run-time contract checks. 454 + -fcheckaction=, controls the run-time behavior on an assert, 455 array bounds check, or final switch contract failure. The 456 default is -fcheckaction=throw. 457 + -fdump-c++-spec=, dumps all compiled extern(C++) declarations 458 as C++ code to the given file. The supplementary option 459 -fdump-c++-spec-verbose turns on emission of comments for 460 ignored declarations in the generated spec. 461 + -fextern-std=, controls which C++ standard extern(C++) 462 declarations are compiled to be compatible with. The default 463 is -fextern-std=c++17. 464 + -fpreview=, added to enable upcoming D language features in 465 the compiler. 466 + -frevert=, added to revert D language changes to support older 467 D codebases that need more time to transition. 468 + -fsave-mixins=, saves mixins expanded at compile-time to a 469 file. 470 * Deprecated and removed features: 471 + The -Wtemplates compiler switch has been removed, as it had 472 been superceded by -ftransition=templates, which more 473 accurately reports on which templates have been instantiated. 474 + The -ftransition=dip25 and -ftransition=dip1000 compiler 475 switches have been renamed to -fpreview=dip25 and 476 -fpreview=dip1000. 477 478 [92]Fortran 479 480 * WG5/N1942, "TS 29113 Further Interoperability of Fortran with C", 481 is now fully supported. In addition to implementing previously 482 missing functionality, such as support for character arguments of 483 length greater than one in functions marked bind(c) and gaps in the 484 handling for assumed-rank arrays, numerous other bugs have been 485 fixed, and an extensive set of new conformance test cases has been 486 added. 487 * GCC 12 now uses OPERATION as the name of the function to the 488 CO_REDUCE intrinsic for the pairwise reduction, thus conforming to 489 the Fortran 2018 standard. Previous versions used OPERATOR which 490 conforms to TS 18508. 491 * On POWER systems which support it, the -mabi=ieeelongdouble option 492 now selects the IEEE 128-bit floating point format for 493 REAL(KIND=16). R16_IBM and R16_IEEE have been added to the 494 -fconvert option, the CONVERT specifier of the OPEN statement and 495 the GFORTRAN_CONVERT_UNIT environment variable. 496 497 [93]Go 498 499 * GCC 12 provides a complete implementation of the Go 1.18 user 500 packages. 501 * Although Go 1.18 includes support for generic programming, that 502 support is not yet available in GCC. 503 504 [94]libgccjit 505 506 * The libgccjit API gained 30 new entry points: 507 + 17 new "reflection" entry points for querying functions and 508 types ([95]LIBGCCJIT_ABI_16) 509 + [96]gcc_jit_lvalue_set_tls_model for supporting thread-local 510 variables ([97]LIBGCCJIT_ABI_17) 511 + [98]gcc_jit_lvalue_set_link_section for setting the link 512 section of global variables, analogous to 513 [99]__attribute__((section(".section"))) 514 ([100]LIBGCCJIT_ABI_18) 515 + 4 new entry points for initializing global variables and 516 creating constructors for rvalues ([101]LIBGCCJIT_ABI_19) 517 + Support for sized integer types, including 128-bit integers 518 and helper functions for such types ([102]LIBGCCJIT_ABI_20) 519 + [103]gcc_jit_context_new_bitcast for reinterpreting the bits 520 of an rvalue as a different type ([104]LIBGCCJIT_ABI_21) 521 + [105]gcc_jit_lvalue_set_register_name for setting a specific 522 register for a variable ([106]LIBGCCJIT_ABI_22) 523 + [107]gcc_jit_context_set_bool_print_errors_to_stderr 524 ([108]LIBGCCJIT_ABI_23) 525 + 2 new entry points for setting the alignment of a variable 526 ([109]LIBGCCJIT_ABI_24) 527 * libgccjit has gained support for the use of various atomic builtins 528 ([110]PR96066, [111]PR96067) 529 * [112]gcc_jit_context_new_cast is now able to handle truncation and 530 extension between different integer types ([113]PR95498) 531 532 [114]New Targets and Target Specific Improvements 533 534 [115]AArch64 & arm 535 536 * Newer revisions of the Arm Architecture are supported as arguments 537 to the -march option: armv8.7-a, armv8.8-a, armv9-a. 538 * The Arm Cortex-A510 CPU is now supported through the cortex-a510 539 argument to the -mcpu and -mtune options. 540 * GCC can now auto-vectorize operations performing sign-differing 541 dot-product operations, taking advantage of instructions in the 542 Advanced SIMD (AArch64/AArch32) and SVE (AArch64) instruction sets. 543 544 [116]AArch64 545 546 * A number of new CPUs are supported through the -mcpu and -mtune 547 options (GCC identifiers in parentheses). 548 + Ampere-1 (ampere1). 549 + Arm Cortex-A710 (cortex-a710). 550 + Arm Cortex-X2 (cortex-x2). 551 * The 64-byte atomic load/store intrinsics to accelerator memory from 552 the [117]2020 Arm Architecture extensions are supported through the 553 +ls64 option extension. 554 * Initial code generation support is supported for hardware 555 instructions used to accelerate the memcpy,memmove and memset 556 standard functions. These instructions can be generated when 557 compiling with the +mopsoption extension. 558 * The ACLE Advanced SIMD intrinsics accessible through the arm_neon.h 559 header have been significantly reimplemented and generate 560 higher-performing code than previous GCC versions. 561 * The option -mtune=neoverse-512tvb is added to tune for Arm Neoverse 562 cores that have a total vector bandwidth of 512 bits. Please refer 563 to the documentation for more details. 564 565 [118]AMD Radeon (GCN) 566 567 * Debug experience with ROCGDB has been improved. 568 * Support for the type __int128_t/integer(kind=16) was added. 569 * For offloading, the limitation of using only one wavefront per 570 compute unit (CU) has been lifted. Up to 40 workgroups per CU and 571 16 wavefronts per workgroup are supported (up to a limit of 40 572 wavefronts in total, per CU). Additionally, the number of used 573 wavefronts and workgroups was tuned for performance. 574 575 [119]arm 576 577 * Support is added for accessing the stack canary value via the TLS 578 register through the -fstack-protector-guard=tls and 579 -mstack-protector-guard-offset= options. This intended for use in 580 Linux kernel development. Please refer to the documentation for 581 more details. 582 583 [120]BPF 584 585 * Support for CO-RE (compile-once, run-everywhere) has been added to 586 the BPF back end. CO-RE allows to compile portable BPF programs 587 that are able to run among different versions of the Linux kernel. 588 589 [121]IA-32/x86-64 590 591 * New ISA extension support for Intel AVX512-FP16 was added. 592 AVX512-FP16 intrinsics are available via the -mavx512fp16 compiler 593 switch. 594 * For both C and C++ the _Float16 type is supported on x86 systems 595 with SSE2 enabled. Without {-mavx512fp16}, all operations will be 596 emulated in software and float instructions. 597 * Mitigation against straight line speculation (SLS) for function 598 return and indirect jump is supported via 599 -mharden-sls=[none|all|return|indirect-jmp]. 600 * Add CS prefix to call and jmp to indirect thunk with branch target 601 in r8-r15 registers via -mindirect-branch-cs-prefix. 602 * Always use global offset table (GOT) to access external data and 603 function symbols when the new -mno-direct-extern-access 604 command-line option is specified. 605 606 [122]LoongArch 607 608 * Support for the LoongArch architecture instruction set has been 609 added. 610 * The Loongson CPU codename LA464 and LoongArch 64-bit generic CPU 611 codename loongarch64 are supported through the -march= and -mtune= 612 options (GCC identifiers in parentheses). 613 + Loongson LA464 core (la464). 614 + LoongArch 64-bit generic core (loongarch64). 615 616 [123]MIPS 617 618 * The ABI passing arguments containing zero-width fields (for 619 example, C/C++ zero-width bit-fields, GNU C/C++ zero-length arrays, 620 and GNU C empty structs) has changed. Now a zero-width field will 621 not prevent an aligned 64-bit floating-point field next to it from 622 being passed through FPR. This is compatible with LLVM, but 623 incompatible with previous GCC releases. GCC 12 on MIPS will report 624 such incompatibilities as an inform unless -Wno-psabi is used. 625 * The ABI returning values containing C++17 empty bases has changed. 626 Now an empty base will not prevent an aggregate containing only one 627 or two floating-point fields from being returned through FPR. This 628 is compatible with GCC 6 and earlier, but incompatible with GCC 7 629 through 11. GCC 12 on MIPS will report such incompatibilities as an 630 inform unless -Wno-psabi is used. 631 632 [124]NVPTX 633 634 * The -march flag has been added. The -misa flag is now considered an 635 alias of the -march flag. 636 * Support for PTX ISA target architectures sm_53, sm_70, sm_75 and 637 sm_80 has been added. These can be specified using the -march flag. 638 * The default PTX ISA target architecture has been set back to sm_30, 639 to fix support for sm_30 boards. 640 * The -march-map flag has been added. The -march-map value will be 641 mapped to an valid -march flag value. For instance, 642 -march-map=sm_50 maps to -march=sm_35. This can be used to specify 643 that generated code is to be executed on a board with at least some 644 specific compute capability, without having to know the valid 645 values for the -march flag. 646 * The -mptx flag has been added to specify the PTX ISA version for 647 the generated code; permitted values are 3.1 (matches previous GCC 648 versions), 6.0, 6.3, and 7.0. If not specified, the used version is 649 the minimal version required for -march but at least 6.0. 650 * An mptx-3.1 multilib was added. This allows using older drivers 651 which do not support PTX ISA version 6.0. 652 * The new __PTX_SM__ predefined macro allows code to check the PTX 653 ISA target architecture being targeted by the compiler. 654 * The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__ 655 predefined macros allows code to check the PTX ISA version being 656 targeted by the compiler. 657 658 [125]PowerPC / PowerPC64 / RS6000 659 660 * The internal implementation of Power's target-specific built-in 661 functions has been rewritten to be easier and less error-prone to 662 maintain. Every attempt has been made to ensure that the new 663 behavior matches the old behavior, but inevitably some bugs can be 664 expected. Please [126]report any problems. 665 * The built-in functions __builtin_get_texasr, __builtin_get_texasru, 666 __builtin_get_tfhar, __builtin_get_tfiar, __builtin_set_texasr, 667 __builtin_set_texasru, __builtin_set_tfhar, and __builtin_set_tfiar 668 now behave as documented in all supported configurations. On prior 669 releases, the arguments and return values of these functions were 670 treated as unsigned long long instead of as unsigned long, when the 671 options -m32 -mpowerpc64 were in effect. 672 * The overloaded built-in functions vec_cntlz_lsbb and vec_cnttz_lsbb 673 now behave as documented. On prior releases, these built-in 674 functions had incorrect semantics on little-endian targets. 675 676 [127]PRU 677 678 * The [128]__regio_symbol variable qualifier has been added. It 679 allows easier access in C programs to the __R30 and __R31 CPU I/O 680 registers. 681 682 [129]RISC-V 683 684 * The default ISA spec version was bumped to 20191213; see [130]this 685 announcement for details. 686 * New ISA extension support for zba, zbb, zbc, zbs was added. 687 * New ISA extension support for vector and scalar crypto was added, 688 only support architecture testing macro and -march= parsing. 689 * The option -mtune=thead-c906 is added to tune for T-HEAD c906 690 cores. 691 * libstdc++ no longer attempts to detect built-in atomics. 692 Distributions that have out-of-tree workarounds for -latomic should 693 check their ABIs again. 694 695 [131]Operating Systems 696 697 [132]Improvements to Static Analyzer 698 699 * The analyzer has gained a 700 [133]-Wanalyzer-use-of-uninitialized-value warning, similar to 701 [134]-Wuninitialized and [135]-Wmaybe-uninitialized, but based on 702 an interprocedural path-sensitive analysis ([136]PR95006). 703 Such warnings are not disabled by the new 704 [137]-ftrivial-auto-var-init (see below), as the latter is 705 considered a mitigation option. 706 * [138]-Wanalyzer-write-to-const and 707 [139]-Wanalyzer-write-to-string-literal will now check for 708 [140]__attribute__ ((access, ....)) on calls to externally-defined 709 functions, and complain about read-only regions pointed to by 710 arguments marked with a write_only or read_write attribute 711 ([141]PR104793). 712 * The analyzer's "taint" mode, activated by 713 [142]-fanalyzer-checker=taint (in addition to [143]-fanalyzer), has 714 gained four new taint-based warnings: 715 + [144]-Wanalyzer-tainted-allocation-size for e.g. 716 attacker-controlled malloc and alloca, 717 + [145]-Wanalyzer-tainted-divisor for detecting where an 718 attacker can inject a divide-by-zero, 719 + [146]-Wanalyzer-tainted-offset for attacker-controlled pointer 720 offsets, 721 + [147]-Wanalyzer-tainted-size for attacker-controlled values 722 being used as a size parameter to calls to memset or to 723 functions marked with [148]__attribute__ ((access, ....)). 724 The existing [149]-Wanalyzer-tainted-array-index has been reworded 725 to talk about "attacker-controlled" rather than "tainted" values, 726 for consistency with the new warnings. 727 A new [150]__attribute__ ((tainted_args)) has been added to the C 728 and C++ front ends, usable on functions, and on function pointer 729 callback fields in structs. The analyzer's taint mode will treat 730 all parameters and buffers pointed to by parameters of such 731 functions as being attacker-controlled, such as for annotating 732 system calls in an operating system kernel as being an "attack 733 surface". 734 * The analyzer now respects [151]__attribute__((const)): it will 735 treat such functions as returning the same value when given the 736 same inputs ([152]PR104434), and as having no side effects 737 ([153]PR104576). 738 * The analyzer is now able to split its analysis into multiple 739 execution paths in places where there isn't a split in the control 740 flow graph. For example, it now handles realloc calls by splitting 741 the execution path into three possible outcomes for the call: 742 + failure, returning NULL 743 + success, growing the buffer in-place without moving it 744 + success, allocating a new buffer, copying the content of the 745 old buffer to it, and freeing the old buffer 746 * The analyzer's interprocedural path exploration logic is now able 747 to track calls through function pointers. 748 * The analyzer now makes the assumption that if we know PTR is 749 non-NULL, then (PTR + OFFSET) is also non-NULL. This isn't strictly 750 true, but eliminates false positives in practice ([154]PR101962). 751 * The analyzer has gained some initial support for inline assembler 752 code. This is extremely limited, and is purely to help suppress 753 false positives when analyzing the Linux kernel, which makes heavy 754 use of inline assembler ([155]PR101570). 755 * The way the analyzer tracks the state of memory along an execution 756 path has been improved in various ways for GCC 12: 757 + An optimization for representing bulk updates to memory (e.g. 758 zero fills) has been removed as it never worked well. In GCC 759 12 it has been replaced with a simpler and more accurate 760 approach, eliminating many false positives ([156]PR95006). 761 + Various optimizations have been added, speeding up the 762 analysis on a particularly problematic source file from 4 763 minutes down to 17 seconds ([157]PR104943, [158]PR104954, and 764 [159]PR104955). 765 + The analyzer now tracks the sizes of dynamically-allocated 766 regions, both on the heap (via malloc etc) and stack (via 767 alloca), though none of the analyzer warnings make use of this 768 yet in GCC 12. 769 * The analyzer's handling of switch statements has been rewritten, 770 fixing various bugs. 771 772 Other significant improvements 773 774 [160]Eliminating uninitialized variables 775 776 * GCC can now [161]initialize all stack variables implicitly, 777 including padding. This is intended to eliminate all classes of 778 uninitialized stack variable flaws. Lack of explicit initialization 779 will still warn when [162]-Wuninitialized is active. For best 780 debugging, use of the new command-line option 781 [163]-ftrivial-auto-var-init=pattern can be used to fill variables 782 with a repeated 0xFE pattern, which tends to illuminate many bugs 783 (e.g. pointers receive invalid addresses, sizes and indices are 784 very large). For best production results, the new command-line 785 option [164]-ftrivial-auto-var-init=zero can be used to fill 786 variables with 0x00, which tends to provide a safer state for bugs 787 (e.g. pointers are NULL, strings are NUL filled, and sizes and 788 indices are 0). 789 790 [165]Debugging formats 791 792 * GCC can now generate debugging information in [166]CTF, a 793 lightweight debugging format that provides information about C 794 types and the association between functions and data symbols and 795 types. This format is designed to be embedded in ELF files and to 796 be very compact and simple. A new command-line option -gctf enables 797 the generation of CTF. 798 * GCC can now generate debugging information in BTF. This is a 799 debugging format mainly used in BPF programs and the Linux kernel. 800 The compiler can generate BTF for any target, when enabled with the 801 command-line option -gbtf 802 803 [167]GCC 12.1 804 805 This is the [168]list of problem reports (PRs) from GCC's bug tracking 806 system that are known to be fixed in the 12.1 release. This list might 807 not be complete (that is, it is possible that some PRs that have been 808 fixed are not listed here). 809 810 [169]GCC 12.2 811 812 This is the [170]list of problem reports (PRs) from GCC's bug tracking 813 system that are known to be fixed in the 12.2 release. This list might 814 not be complete (that is, it is possible that some PRs that have been 815 fixed are not listed here). 816 817 Target Specific Changes 818 819 LoongArch 820 821 * The default setting of -m[check|no-check]-zero-division is changed 822 for optimized code. Now -mno-check-zero-division is the default for 823 all optimization levels but -O0 and -Og. The old behavior can be 824 obtained by explicitly passing -mcheck-zero-division to GCC. 825 826 [171]GCC 12.3 827 828 Target Specific Changes 829 830 x86-64 831 832 * GCC now supports AMD CPUs based on the znver4 core via 833 -march=znver4. The switch makes GCC consider using 512-bit vectors 834 when auto-vectorizing. 835 836 This is the [172]list of problem reports (PRs) from GCC's bug tracking 837 system that are known to be fixed in the 12.3 release. This list might 838 not be complete (that is, it is possible that some PRs that have been 839 fixed are not listed here). 840 841 [173]GCC 12.4 842 843 This is the [174]list of problem reports (PRs) from GCC's bug tracking 844 system that are known to be fixed in the 12.4 release. This list might 845 not be complete (that is, it is possible that some PRs that have been 846 fixed are not listed here). 847 848 [175]GCC 12.5 849 850 This is the [176]list of problem reports (PRs) from GCC's bug tracking 851 system that are known to be fixed in the 12.5 release. This list might 852 not be complete (that is, it is possible that some PRs that have been 853 fixed are not listed here). 854 855 856 For questions related to the use of GCC, please consult these web 857 pages and the [177]GCC manuals. If that fails, the 858 [178]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 859 web pages and the development of GCC are welcome on our developer 860 list at [179]gcc (a] gcc.gnu.org. All of [180]our lists have public 861 archives. 862 863 Copyright (C) [181]Free Software Foundation, Inc. Verbatim copying and 864 distribution of this entire article is permitted in any medium, 865 provided this notice is preserved. 866 867 These pages are [182]maintained by the GCC team. Last modified 868 2025-07-11. 869 870 References 871 872 1. https://gcc.gnu.org/gcc-12/porting_to.html 873 2. https://gcc.gnu.org/onlinedocs/index.html#current 874 3. https://gcc.gnu.org/PR102024 875 4. https://gcc.gnu.org/PR42217 876 5. https://gcc.gnu.org/gcc-12/changes.html#mips_zero_width_fields 877 6. https://gcc.gnu.org/install/specific.html 878 7. https://gcc.gnu.org/gcc-12/changes.html#general 879 8. https://clang.llvm.org/docs/ShadowCallStack.html 880 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Instrumentation-Options.html#index-fsanitize_003dshadow-call-stack 881 10. https://gcc.gnu.org/gcc-12/changes.html#languages 882 11. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/OpenMP-Implementation-Status.html 883 12. https://www.openacc.org/ 884 13. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-12 885 14. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/Enabling-OpenACC.html 886 15. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 887 16. https://gcc.gnu.org/gcc-12/changes.html#nvptx 888 17. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-param 889 18. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wopenacc-parallelism 890 19. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload-options 891 20. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload 892 21. https://gcc.gnu.org/gcc-12/changes.html#ada 893 22. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gnat_rm/Security-Hardening-Features.html#Security-Hardening-Features 894 23. https://gcc.gnu.org/gcc-12/changes.html#c-family 895 24. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wbidi-chars 896 25. https://gcc.gnu.org/PR103026 897 26. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Warray-compare 898 27. https://gcc.gnu.org/PR97573 899 28. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wattributes 900 29. https://gcc.gnu.org/PR101940 901 30. https://gcc.gnu.org/gcc-12/changes.html#c 902 31. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wformat 903 32. https://gcc.gnu.org/gcc-12/changes.html#cxx 904 33. https://wg21.link/p1938 905 34. https://gcc.gnu.org/PR100974 906 35. https://wg21.link/p0849 907 36. https://gcc.gnu.org/PR103049 908 37. https://wg21.link/p2242 909 38. https://gcc.gnu.org/PR102612 910 39. https://wg21.link/p2334 911 40. https://gcc.gnu.org/PR102616 912 41. https://wg21.link/p2360 913 42. https://gcc.gnu.org/PR102617 914 43. https://wg21.link/p2128 915 44. https://wg21.link/cwg2397 916 45. https://gcc.gnu.org/PR100975 917 46. https://wg21.link/cwg960 918 47. https://wg21.link/cwg1227 919 48. https://wg21.link/cwg1315 920 49. https://wg21.link/cwg2082 921 50. https://wg21.link/cwg2351 922 51. https://wg21.link/cwg2374 923 52. https://wg21.link/cwg2397 924 53. https://wg21.link/cwg2446 925 54. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=87c2080b 926 55. https://gcc.gnu.org/PR79501 927 56. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 928 57. https://gcc.gnu.org/PR19808 929 58. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wint-in-bool-context 930 59. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a2b12bc 931 60. https://gcc.gnu.org/PR99032 932 61. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmissing-requires 933 62. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=e18e56c7 934 63. https://gcc.gnu.org/PR100995 935 64. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Waddress 936 65. https://gcc.gnu.org/PR102103 937 66. https://gcc.gnu.org/PR99701 938 67. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3ead06c1 939 68. https://gcc.gnu.org/PR102753 940 69. https://wg21.link/p0466 941 70. https://gcc.gnu.org/PR101539 942 71. https://gcc.gnu.org/PR100828 943 72. https://gcc.gnu.org/PR101786 944 73. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Winterference-size 945 74. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=76b75018 946 75. https://gcc.gnu.org/PR101344 947 76. https://gcc.gnu.org/PR101883 948 77. https://gcc.gnu.org/PR89062 949 78. https://gcc.gnu.org/PR101233 950 79. https://gcc.gnu.org/PR88252 951 80. https://gcc.gnu.org/PR86439 952 81. https://gcc.gnu.org/PR98832 953 82. https://gcc.gnu.org/PR102933 954 83. https://gcc.gnu.org/PR51577 955 84. https://gcc.gnu.org/PR98486 956 85. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=1595fe44 957 86. https://gcc.gnu.org/gcc-12/changes.html#libstdcxx 958 87. https://gcc.gnu.org/gcc-12/changes.html#d 959 88. https://dlang.org/changelog/2.100.1.html 960 89. https://gcc.gnu.org/PR101127 961 90. https://gcc.gnu.org/PR106563 962 91. https://dlang.org/spec/importc.html 963 92. https://gcc.gnu.org/gcc-12/changes.html#fortran 964 93. https://gcc.gnu.org/gcc-12/changes.html#go 965 94. https://gcc.gnu.org/gcc-12/changes.html#jit 966 95. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-16 967 96. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_tls_model 968 97. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-17 969 98. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_link_section 970 99. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Variable-Attributes.html#index-section-variable-attribute 971 100. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-18 972 101. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-19 973 102. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-20 974 103. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_bitcast 975 104. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-21 976 105. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_register_name 977 106. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-22 978 107. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/contexts.html#c.gcc_jit_context_set_bool_print_errors_to_stderr 979 108. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-23 980 109. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-24 981 110. https://gcc.gnu.org/PR96066 982 111. https://gcc.gnu.org/PR96067 983 112. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_cast 984 113. https://gcc.gnu.org/PR95498 985 114. https://gcc.gnu.org/gcc-12/changes.html#targets 986 115. https://gcc.gnu.org/gcc-12/changes.html#arm-targets 987 116. https://gcc.gnu.org/gcc-12/changes.html#aarch64 988 117. https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/arm-a-profile-architecture-developments-2020 989 118. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 990 119. https://gcc.gnu.org/gcc-12/changes.html#arm 991 120. https://gcc.gnu.org/gcc-12/changes.html#bpf 992 121. https://gcc.gnu.org/gcc-12/changes.html#x86 993 122. https://gcc.gnu.org/gcc-12/changes.html#loongarch 994 123. https://gcc.gnu.org/gcc-12/changes.html#mips 995 124. https://gcc.gnu.org/gcc-12/changes.html#nvptx 996 125. https://gcc.gnu.org/gcc-12/changes.html#powerpc 997 126. https://gcc.gnu.org/bugs/ 998 127. https://gcc.gnu.org/gcc-12/changes.html#pru 999 128. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Named-Address-Spaces.html#PRU-Named-Address-Spaces 1000 129. https://gcc.gnu.org/gcc-12/changes.html#riscv 1001 130. https://groups.google.com/a/groups.riscv.org/g/sw-dev/c/aE1ZeHHCYf4 1002 131. https://gcc.gnu.org/gcc-12/changes.html#os 1003 132. https://gcc.gnu.org/gcc-12/changes.html#analyzer 1004 133. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-use-of-uninitialized-value 1005 134. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 1006 135. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 1007 136. https://gcc.gnu.org/PR95006 1008 137. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 1009 138. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 1010 139. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 1011 140. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 1012 141. https://gcc.gnu.org/PR104793 1013 142. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 1014 143. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 1015 144. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 1016 145. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 1017 146. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 1018 147. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 1019 148. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 1020 149. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 1021 150. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-tainted_005fargs-function-attribute 1022 151. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-const-function-attribute 1023 152. https://gcc.gnu.org/PR104434 1024 153. https://gcc.gnu.org/PR104576 1025 154. https://gcc.gnu.org/PR101962 1026 155. https://gcc.gnu.org/PR101570 1027 156. https://gcc.gnu.org/PR95006 1028 157. https://gcc.gnu.org/PR104943 1029 158. https://gcc.gnu.org/PR104954 1030 159. https://gcc.gnu.org/PR104955 1031 160. https://gcc.gnu.org/gcc-12/changes.html#uninitialized 1032 161. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 1033 162. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 1034 163. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 1035 164. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 1036 165. https://gcc.gnu.org/gcc-12/changes.html#debug 1037 166. https://ctfstd.org/ 1038 167. https://gcc.gnu.org/gcc-12/changes.html#12.1 1039 168. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.0 1040 169. https://gcc.gnu.org/gcc-12/changes.html#12.2 1041 170. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.2 1042 171. https://gcc.gnu.org/gcc-12/changes.html#12.3 1043 172. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.3 1044 173. https://gcc.gnu.org/gcc-12/changes.html#12.4 1045 174. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.4 1046 175. https://gcc.gnu.org/gcc-12/changes.html#12.5 1047 176. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.5 1048 177. https://gcc.gnu.org/onlinedocs/ 1049 178. mailto:gcc-help (a] gcc.gnu.org 1050 179. mailto:gcc (a] gcc.gnu.org 1051 180. https://gcc.gnu.org/lists.html 1052 181. https://www.fsf.org/ 1053 182. https://gcc.gnu.org/about.html 1054 ====================================================================== 1055 http://gcc.gnu.org/gcc-11/index.html 1056 1057 GCC 11 Release Series 1058 1059 July 19, 2024 1060 1061 The GCC developers are pleased to announce the release of GCC 11.5. 1062 1063 This release is a bug-fix release, containing fixes for regressions in 1064 GCC 11.4 relative to previous releases of GCC. 1065 1066 Release History 1067 1068 GCC 11.5 1069 July 19, 2024 ([1]changes, [2]documentation) 1070 1071 GCC 11.4 1072 May 29, 2023 ([3]changes, [4]documentation) 1073 1074 GCC 11.3 1075 April 21, 2022 ([5]changes, [6]documentation) 1076 1077 GCC 11.2 1078 July 28, 2021 ([7]changes, [8]documentation) 1079 1080 GCC 11.1 1081 April 27, 2021 ([9]changes, [10]documentation) 1082 1083 References and Acknowledgements 1084 1085 GCC used to stand for the GNU C Compiler, but since the compiler 1086 supports several other languages aside from C, it now stands for the 1087 GNU Compiler Collection. 1088 1089 The GCC developers would like to thank the numerous people that have 1090 contributed new features, improvements, bug fixes, and other changes as 1091 well as test results to GCC. This [11]amazing group of volunteers is 1092 what makes GCC successful. 1093 1094 For additional information about GCC please refer to the [12]GCC 1095 project web site or contact the [13]GCC development mailing list. 1096 1097 To obtain GCC please use [14]our mirror sites or [15]our version 1098 control system. 1099 1100 1101 For questions related to the use of GCC, please consult these web 1102 pages and the [16]GCC manuals. If that fails, the 1103 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1104 web pages and the development of GCC are welcome on our developer 1105 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 1106 archives. 1107 1108 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 1109 distribution of this entire article is permitted in any medium, 1110 provided this notice is preserved. 1111 1112 These pages are [21]maintained by the GCC team. Last modified 1113 2024-07-19. 1114 1115 References 1116 1117 1. https://gcc.gnu.org/gcc-11/changes.html 1118 2. https://gcc.gnu.org/onlinedocs/11.5.0/ 1119 3. https://gcc.gnu.org/gcc-11/changes.html 1120 4. https://gcc.gnu.org/onlinedocs/11.4.0/ 1121 5. https://gcc.gnu.org/gcc-11/changes.html 1122 6. https://gcc.gnu.org/onlinedocs/11.3.0/ 1123 7. https://gcc.gnu.org/gcc-11/changes.html 1124 8. https://gcc.gnu.org/onlinedocs/11.2.0/ 1125 9. https://gcc.gnu.org/gcc-11/changes.html 1126 10. https://gcc.gnu.org/onlinedocs/11.1.0/ 1127 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html 1128 12. https://gcc.gnu.org/index.html 1129 13. mailto:gcc (a] gcc.gnu.org 1130 14. https://gcc.gnu.org/mirrors.html 1131 15. https://gcc.gnu.org/git.html 1132 16. https://gcc.gnu.org/onlinedocs/ 1133 17. mailto:gcc-help (a] gcc.gnu.org 1134 18. mailto:gcc (a] gcc.gnu.org 1135 19. https://gcc.gnu.org/lists.html 1136 20. https://www.fsf.org/ 1137 21. https://gcc.gnu.org/about.html 1138 ====================================================================== 1139 http://gcc.gnu.org/gcc-11/changes.html 1140 1141 GCC 11 Release Series 1142 Changes, New Features, and Fixes 1143 1144 This page is a "brief" summary of some of the huge number of 1145 improvements in GCC 11. You may also want to check out our [1]Porting 1146 to GCC 11 page and the [2]full GCC documentation. 1147 1148 Caveats 1149 1150 * The default mode for C++ is now -std=gnu++17 instead of 1151 -std=gnu++14. Note that [3]C++17 changes to template template 1152 parameter matching can be disabled independently of other features 1153 with -fno-new-ttp-matching. 1154 * When building GCC itself, the host compiler must now support C++11, 1155 rather than C++98. In particular bootstrapping GCC 11 using an 1156 older version of GCC requires a binary of GCC 4.8 or later, rather 1157 than of GCC 3.4 or later as was the case for bootstrapping GCC 10. 1158 * Naming and location of auxiliary and dump output files changed. If 1159 you compile multiple input files in a single command, if you enable 1160 Link Time Optimization, or if you use -dumpbase, -dumpdir, 1161 -save-temps=*, and you expect any file other than the primary 1162 output file(s) to be created as a side effect, watch out for 1163 improvements and a few surprises. See [4]the patch, particularly 1164 its textual description, for more details about the changes. 1165 * -gsplit-dwarf no longer enables debug info generation on its own 1166 but requires a separate -g for this. 1167 * The libstdc++ configure option --enable-cheaders=c_std is 1168 deprecated and will be removed in a future release. It should be 1169 possible to use --enable-cheaders=c_global (the default) with no 1170 change in behaviour. 1171 * The front end for compiling BRIG format of Heterogeneous System 1172 Architecture Intermediate Language (HSAIL) has been deprecated and 1173 will likely be removed in a future release. 1174 * Some short options of the gcov tool have been renamed: -i to -j and 1175 -j to -H. 1176 1177 [5]General Improvements 1178 1179 * [6]ThreadSanitizer improvements to support alternative runtimes and 1180 environments. The [7]Linux Kernel Concurrency Sanitizer (KCSAN) is 1181 now supported. 1182 + Add --param tsan-distinguish-volatile to optionally emit 1183 instrumentation distinguishing volatile accesses. 1184 + Add --param tsan-instrument-func-entry-exit to optionally 1185 control if function entries and exits should be instrumented. 1186 * In previous releases of GCC, the "column numbers" emitted in 1187 diagnostics were actually a count of bytes from the start of the 1188 source line. This could be problematic, both because of: 1189 + multibyte characters (requiring more than one byte to encode), 1190 and 1191 + multicolumn characters (requiring more than one column to 1192 display in a monospace font) 1193 For example, the character p ("GREEK SMALL LETTER PI (U+03C0)") 1194 occupies one column, and its UTF-8 encoding requires two bytes; the 1195 character 🙂 ("SLIGHTLY SMILING FACE (U+1F642)") occupies 1196 two columns, and its UTF-8 encoding requires four bytes. 1197 In GCC 11 the column numbers default to being column numbers, 1198 respecting multi-column characters. The old behavior can be 1199 restored using a new option [8]-fdiagnostics-column-unit=byte. 1200 There is also a new option [9]-fdiagnostics-column-origin=, 1201 allowing the pre-existing default of the left-hand column being 1202 column 1 to be overridden if desired (e.g. for 0-based columns). 1203 The output of [10]-fdiagnostics-format=json has been extended to 1204 supply both byte counts and column numbers for all source 1205 locations. 1206 Additionally, in previous releases of GCC, tab characters in the 1207 source would be emitted verbatim when quoting source code, but be 1208 prefixed with whitespace or line number information, leading to 1209 misalignments in the resulting output when compared with the actual 1210 source. Tab characters are now printed as an appropriate number of 1211 spaces, using the [11]-ftabstop option (which defaults to 8 spaces 1212 per tab stop). 1213 * Introduce [12]Hardware-assisted AddressSanitizer support. This 1214 sanitizer currently only works for the AArch64 target. It helps 1215 debug address problems similarly to [13]AddressSanitizer but is 1216 based on partial hardware assistance and provides probabilistic 1217 protection to use less RAM at run time. [14]Hardware-assisted 1218 AddressSanitizer is not production-ready for user space, and is 1219 provided mainly for use compiling the Linux Kernel. 1220 To use this sanitizer the command line arguments are: 1221 + -fsanitize=hwaddress to instrument userspace code. 1222 + -fsanitize=kernel-hwaddress to instrument kernel code. 1223 * For targets that produce DWARF debugging information GCC now 1224 defaults to [15]DWARF version 5 (with the exception of VxWorks and 1225 Darwin/Mac OS X which default to version 2 and AIX which defaults 1226 to version 4). This can produce up to 25% more compact debug 1227 information compared to earlier versions. 1228 To take full advantage of DWARF version 5 GCC needs to be built 1229 against binutils version 2.35.2 or higher. When GCC is built 1230 against earlier versions of binutils GCC will still emit DWARF 1231 version 5 for most debuginfo data, but will generate version 4 1232 debug line tables (even when explicitly given -gdwarf-5). 1233 The following debug information consumers can process DWARF version 1234 5: 1235 + GDB 8.0, or higher 1236 + valgrind 3.17.0 1237 + elfutils 0.172, or higher (for use with systemtap, 1238 dwarves/pahole, perf and libabigail) 1239 + dwz 0.14 1240 Programs embedding libbacktrace are urged to upgrade to the version 1241 shipping with GCC 11. 1242 To make GCC 11 generate an older DWARF version use -g together with 1243 -gdwarf-2, -gdwarf-3 or -gdwarf-4. 1244 * Vectorizer improvements: 1245 + The straight-line code vectorizer now considers the whole 1246 function when vectorizing and can handle opportunities 1247 crossing CFG merges and backedges. 1248 * A series of conditional expressions that compare the same variable 1249 can be transformed into a switch statement if each of them contains 1250 a comparison expression. Example: 1251 int IsHTMLWhitespace(int aChar) { 1252 return aChar == 0x0009 || aChar == 0x000A || 1253 aChar == 0x000C || aChar == 0x000D || 1254 aChar == 0x0020; 1255 } 1256 1257 This statement can be transformed into a switch statement and then 1258 expanded into a bit-test. 1259 * New command-line options: 1260 + [16]-fbit-tests, enabled by default, can be used to enable or 1261 disable switch expansion using bit-tests. 1262 * Inter-procedural optimization improvements: 1263 + A new IPA-modref pass was added to track side effects of 1264 function calls and improve precision of points-to-analysis. 1265 The pass can be controlled by the [17]-fipa-modref option. 1266 + The identical code folding pass (controlled by [18]-fipa-icf) 1267 was significantly improved to increase the number of unified 1268 functions and to reduce compile-time memory use. 1269 + IPA-CP (Interprocedural constant propagation) heuristics 1270 improved its estimation of potential usefulness of known loop 1271 bounds and strides by taking the estimated frequency of these 1272 loops into account. 1273 * Link-time optimization improvements: 1274 + The LTO bytecode format was optimized for smaller object files 1275 and faster streaming. 1276 + Memory allocation of the linking stage was improved to reduce 1277 peak memory use. 1278 * Profile driven optimization improvements: 1279 + Using [19]-fprofile-values, was improved by tracking more 1280 target values for e.g. indirect calls. 1281 + GCOV data file format outputs smaller files by representing 1282 zero counters in a more compact way. 1283 1284 [20]New Languages and Language specific improvements 1285 1286 * GCC 11 adds support for non-rectangular loop nests in OpenMP 1287 constructs and the allocator routines of [21]OpenMP 5.0, including 1288 initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD 1289 environment variable and the active-levels routines are now 1290 supported. For C/C++, the declare variant and map support has been 1291 extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP 1292 5.0 support has been extended, including the following features 1293 which were before only available in C and C++: order(concurrent), 1294 device_type, memorder-clauses for flush, lastprivate with 1295 conditional modifier, atomic construct and reduction clause 1296 extensions of OpenMP 5.0, if clause with simd and cancel modifiers, 1297 target data without map clause, and limited support for the 1298 requires construct. 1299 * Version 2.6 of the [22]OpenACC specification continues to be 1300 maintained and improved in the C, C++ and Fortran compilers. See 1301 the [23]implementation status section on the OpenACC wiki page and 1302 the [24]run-time library documentation for further information. 1303 1304 [25]C family 1305 1306 * New attributes: 1307 + The [26]no_stack_protector attribute has been added to mark 1308 functions which should not be instrumented with stack 1309 protection (-fstack-protector). 1310 + The existing [27]malloc attribute has been extended so that it 1311 can be used to identify allocator/deallocator API pairs. A 1312 pair of new [28]-Wmismatched-dealloc and 1313 [29]-Wmismatched-new-delete warnings will complain about 1314 mismatched calls, and [30]-Wfree-nonheap-object about 1315 deallocation calls with pointers not obtained from allocation 1316 functions. Additionally, the static analyzer will use these 1317 attributes when checking for leaks, double-frees, 1318 use-after-frees, and similar issues. 1319 * New warnings: 1320 + [31]-Wmismatched-dealloc, enabled by default, warns about 1321 calls to deallocation functions with pointers returned from 1322 mismatched allocation functions. 1323 + [32]-Wsizeof-array-div, enabled by -Wall, warns about 1324 divisions of two sizeof operators when the first one is 1325 applied to an array and the divisor does not equal the size of 1326 the array element. 1327 + [33]-Wstringop-overread, enabled by default, warns about calls 1328 to string functions reading past the end of the arrays passed 1329 to them as arguments. In prior GCC releases most instances of 1330 his warning are diagnosed by -Wstringop-overflow. 1331 + [34]-Wtsan, enabled by default, warns about unsupported 1332 features in ThreadSanitizer (currently 1333 std::atomic_thread_fence). 1334 * Enhancements to existing warnings: 1335 + [35]-Wfree-nonheap-object detects many more instances of calls 1336 to deallocation functions with pointers not returned from a 1337 dynamic memory allocation function. 1338 + [36]-Wmaybe-uninitialized diagnoses passing pointers or 1339 references to uninitialized memory to functions taking 1340 const-qualified arguments. 1341 + [37]-Wuninitialized detects reads from uninitialized 1342 dynamically allocated memory. 1343 * For ELF targets that support the GNU or FreeBSD OSABIs, the used 1344 attribute will now save the symbol declaration it is applied to 1345 from linker garbage collection. 1346 To support this behavior, used symbols that have not been placed in 1347 specific sections (e.g. with the section attribute, or the 1348 -f{function,data}-sections options) will be placed in new, unique 1349 sections. 1350 This functionality requires Binutils version 2.36 or later. 1351 1352 [38]C 1353 1354 * Several new features from the upcoming C2X revision of the ISO C 1355 standard are supported with -std=c2x and -std=gnu2x. Some of these 1356 features are also supported as extensions when compiling for older 1357 language versions. In addition to the features listed, some 1358 features previously supported as extensions and now added to the C 1359 standard are enabled by default in C2X mode and not diagnosed with 1360 -std=c2x -Wpedantic. 1361 + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>. 1362 + As in C++, function definitions no longer need to give names 1363 for unused function parameters. 1364 + The expansions of the true and false macros in <stdbool.h> 1365 have changed so that they have type bool. 1366 + The [[nodiscard]] standard attribute is now supported. 1367 + The __has_c_attribute preprocessor operator is now supported. 1368 + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN, 1369 DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros 1370 for _FloatN, _FloatNx and _DecimalN types, are provided in 1371 <float.h>. There are also corresponding built-in functions 1372 __builtin_nansdN for decimal signaling NaNs. 1373 + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and 1374 LDBL_IS_IEC_60559 are provided in <float.h>. 1375 + The feature test macro __STDC_WANT_IEC_60559_EXT__ is 1376 supported by <float.h>. 1377 + Labels may appear before declarations and at the end of a 1378 compound statement. 1379 * New warnings: 1380 + [39]-Warray-parameter, enabled by -Wall, warns about 1381 redeclarations of functions with ordinary array arguments 1382 declared using inconsistent forms. The warning also enables 1383 the detection of the likely out of bounds accesses in calls to 1384 such functions with smaller arrays. 1385 + [40]-Wvla-parameter, enabled by -Wall, warns redeclarations of 1386 functions with variable length array arguments declared using 1387 inconsistent forms or with mismatched bounds. The warning also 1388 enables the detection of the likely out of bounds accesses in 1389 calls to such functions with smaller arrays. 1390 1391 [41]C++ 1392 1393 * The default mode has been changed to -std=gnu++17. 1394 * Several C++20 features have been implemented: 1395 + the compiler now supports consteval virtual functions 1396 + P2082R1, Fixing CTAD for aggregates 1397 + P0593R6, Pseudo-destructors end object lifetimes 1398 + P1907R1, Inconsistencies with non-type template parameters 1399 (complete implementation) 1400 + P1975R0, Fixing the wording of parenthesized 1401 aggregate-initialization 1402 + P1009R2, Array size deduction in new-expressions 1403 + P1099R5, using enum 1404 + Modules, Requires -fmodules-ts and some aspects are 1405 incomplete. Refer to [42]C++ 20 Status 1406 * The C++ front end has experimental support for some of the upcoming 1407 C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b 1408 or -std=gnu++2b flags, including 1409 + P0330R8, Literal Suffix for (signed) size_t. 1410 For a full list of new features, see [43]the C++ status page. 1411 * Several C++ Defect Reports have been resolved, e.g.: 1412 + DR 625, Use of auto as a template-argument 1413 + DR 1512, Pointer comparison vs qualification conversions 1414 + DR 1722, Should lambda to function pointer conversion function 1415 be noexcept? 1416 + DR 1914, Duplicate standard attributes 1417 + DR 2032, Default template-arguments of variable templates 1418 + DR 2289, Uniqueness of decomposition declaration names 1419 + DR 2237, Can a template-id name a constructor? 1420 + DR 2303, Partial ordering and recursive variadic inheritance 1421 + DR 2369, Ordering between constraints and substitution 1422 + DR 2450, braced-init-list as a template-argument 1423 * G++ now performs better access checking in templates ([44]PR41437). 1424 * reinterpret_casts in constexpr evaluation are now checked more 1425 completely ([45]PR95307). 1426 * The diagnostic for static_assert has been improved: the compiler 1427 now shows the expression including its template arguments (if there 1428 were any), and can point to the failing clause if the condition 1429 comprised of any logical AND operators ([46]PR97518). 1430 * New warnings: 1431 + [47]-Wctad-maybe-unsupported, disabled by default, warns about 1432 performing class template argument deduction on a type with no 1433 deduction guides. 1434 + [48]-Wrange-loop-construct, enabled by -Wall, warns when a 1435 range-based for-loop is creating unnecessary and expensive 1436 copies. 1437 + [49]-Wdeprecated-enum-enum-conversion, enabled by default in 1438 C++20, warns about deprecated arithmetic conversions on 1439 operands of enumeration types, as outlined in 1440 [depr.arith.conv.enum]. 1441 + [50]-Wdeprecated-enum-float-conversion, enabled by default in 1442 C++20, warns about deprecated arithmetic conversions on 1443 operands where one is of enumeration type and the other is of 1444 a floating-point type, as outlined in [depr.arith.conv.enum]. 1445 + [51]-Wmismatched-new-delete, enabled by -Wall, warns about 1446 calls to C++ operator delete with pointers returned from 1447 mismatched forms of operator new or from other mismatched 1448 allocation functions. 1449 + [52]-Wvexing-parse, enabled by default, warns about the most 1450 vexing parse rule: the cases when a declaration looks like a 1451 variable definition, but the C++ language requires it to be 1452 interpreted as a function declaration. 1453 * Enhancements to existing warnings: 1454 + [53]-Wnonnull considers the implicit this argument of every 1455 C++ nonstatic member function to have been implicitly declared 1456 with attribute nonnull and triggers warnings for calls where 1457 the pointer is null. 1458 1459 [54]Runtime Library (libstdc++) 1460 1461 * Improved C++17 support, including: 1462 + std::from_chars and std::to_chars for floating-point types. 1463 * Improved experimental C++20 support, including: 1464 + Calendar additions to <chrono>. Thanks to Cassio Neri for 1465 optimizations. 1466 + std::bit_cast 1467 + std::source_location 1468 + Atomic wait and notify operations. 1469 + <barrier>, <latch>, and <semaphore> 1470 + <syncstream> 1471 + Efficient access to basic_stringbuf's buffer. 1472 + Heterogeneous lookup in unordered containers. 1473 * Experimental C++23 support, including: 1474 + contains member functions for strings, thanks to Paul Fee. 1475 + std::to_underlying, std::is_scoped_enum 1476 * Experimental support for Data-Parallel Types (simd) from the 1477 Parallelism 2 TS, thanks to Matthias Kretz. 1478 * Faster std::uniform_int_distribution, thanks to Daniel Lemire. 1479 1480 [55]D 1481 1482 * New features: 1483 + A new bottom type typeof(*null) has been added to represent 1484 run-time errors and non-terminating functions. This also 1485 introduces a new standard alias for the type named noreturn, 1486 and is implicitly imported into every module. 1487 + Printf-like and scanf-like functions are now detected by 1488 prefixing them with pragma(printf) for printf-like functions 1489 or pragma(scanf) for scanf-like functions. 1490 + The __traits() expression now supports the extensions 1491 isDeprecated, isDisabled, isFuture, isModule, isPackage, 1492 child, isReturnOnStack, isZeroInit, getTargetInfo, 1493 getLocation, hasPostblit, isCopyable, getVisibility, and 1494 totype. 1495 + An expression-based contract syntax has been added to the 1496 language. 1497 + Function literals can now return a value by reference with the 1498 ref keyword. 1499 + A new syntax is available to declare aliases to function types 1500 using the alias syntax based on the assignment operator. 1501 + New types __c_complex_float, __c_complex_double, 1502 __c_complex_real, and __c_wchar_t have been added for 1503 interfacing with C and C++ code, and are available from the 1504 core.stdc.config module. 1505 + User-defined attributes can now be used to annotate enum 1506 members, alias declarations, and function parameters. 1507 + Template alias parameters can now be instantiated with basic 1508 types such as int or void function(). 1509 + The mixin construct can now be used as types in the form 1510 mixin(string) var. 1511 + The mixin construct can now take an argument list, same as 1512 pragma(msg). 1513 * New intrinsics: 1514 + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror 1515 have been added. 1516 + Byte swap intrinsic core.bitop.byteswap for swapping bytes in 1517 a 2-byte ushort has been added. 1518 + Math intrinsics available from core.math now have overloads 1519 for float and double types. 1520 + Volatile intrinsics core.volatile.volatileLoad and 1521 core.volatile.volatileStore have been moved from the 1522 core.bitop module. 1523 * New attributes: 1524 + The following GCC attributes are now recognized and available 1525 from the gcc.attributes module with short-hand aliases for 1526 convenience: 1527 o @attribute("alloc_size", arguments) or 1528 @alloc_size(arguments). 1529 o @attribute("always_inline") or @always_inline. 1530 o @attribute("used") or @used. 1531 o @attribute("optimize", arguments) or 1532 @optimize(arguments). 1533 o @attribute("cold") or @cold. 1534 o @attribute("noplt") or @noplt. 1535 o @attribute("target_clones", arguments) or 1536 @target_clones(arguments). 1537 o @attribute("no_icf") or @no_icf. 1538 o @attribute("noipa") or @noipa. 1539 o @attribute("symver", arguments) or @symver(arguments). 1540 + New aliases have been added to gcc.attributes for 1541 compatibility with ldc.attributes. 1542 o The @allocSize(arguments) attribute is the same as 1543 @alloc_size(arguments), but uses a 0-based index for 1544 function arguments. 1545 o The @assumeUsed attribute is an alias for 1546 @attribute("used"). 1547 o The @fastmath attribute is an alias for 1548 @optimize("Ofast"). 1549 o The @naked attribute is an alias for @attribute("naked"). 1550 This attribute may not be available on all targets. 1551 o The @restrict attribute has been added to specify that a 1552 function parameter is to be restrict-qualified in the C99 1553 sense of the term. 1554 o The @optStrategy(strategy) attribute is an alias for 1555 @optimize("O0") when the strategy is "none", otherwise 1556 @optimize("Os") for the "optsize" and "minsize" 1557 strategies. 1558 o The @polly attribute is an alias for 1559 @optimize("loop-parallelize-all"). 1560 o The @section(name) attribute is an alias for 1561 @attribute("section", name). 1562 o The @target(arguments) attribute is an alias for 1563 attribute("target", arguments). 1564 o The @weak attribute is an alias for @attribute("weak"). 1565 * New language options: 1566 + -fweak-templates, added to control whether declarations that 1567 can be defined in multiple objects should be emitted as weak 1568 symbols. The default is to emit all symbols with extern 1569 linkage as weak, unless the target lacks support for weak 1570 symbols. 1571 + -Wdeprecated, this option is now enabled by default. 1572 + -Wextra, this option now turns on all warnings that are not 1573 part of the core D language front-end - -Waddress, 1574 -Wcast-result, -Wunknown-pragmas. 1575 + -Wvarargs, added to turn on warnings about questionable usage 1576 of the va_start intrinsic. 1577 * Deprecated and removed features: 1578 + Compiler-recognized attributes are now made available from the 1579 gcc.attributes module, the former module gcc.attribute has 1580 been deprecated and will be removed in a future release. 1581 + The @attribute("alias") attribute has been removed, as it had 1582 been superseded by pragma(mangle). 1583 + The @attribute("forceinline") attribute has been removed and 1584 renamed to @always_inline. 1585 + __vector types that are not supported in hardware are now 1586 rejected at compile-time. Previously all vector types were 1587 accepted by the compiler and emulated when target support was 1588 absent. 1589 + The extern(Pascal) linkage attribute has been removed. 1590 + The deprecation phase for -ftransition=import and 1591 -ftransition=checkimports is finished. These switches no 1592 longer have an effect and are now removed. Symbols that are 1593 not visible in a particular scope will no longer be found by 1594 the compiler. 1595 + It is now an error to use private variables selectively 1596 imported from other modules. Due to a bug, some imported 1597 private members were visible from other modules, violating the 1598 specification. 1599 + The -fweak compiler switch has been removed, as it existed 1600 only for testing. 1601 1602 [56]Fortran 1603 1604 * Added DEPRECATED to !GCC$'s attributes directive. 1605 1606 [57]Go 1607 1608 * GCC 11 provides a complete implementation of the Go 1.16.3 user 1609 packages. 1610 1611 [58]libgccjit 1612 1613 * libgccjit was marked as merely "Alpha" quality when [59]originally 1614 added in GCC 5. Given that we have maintained [60]API and ABI 1615 compatibility since then and it is in use by various projects, we 1616 have removed that caveat. 1617 * libgccjit can now be built for MinGW 1618 * The libgccjit API gained 10 new entry points: 1619 + [61]gcc_jit_global_set_initializer 1620 + 9 entrypoints for [62]directly embedding asm statements into a 1621 compile, analogous to inline asm in the C front end 1622 1623 [63]New Targets and Target Specific Improvements 1624 1625 [64]AArch64 & arm 1626 1627 * A number of new CPUs are supported through arguments to the -mcpu 1628 and -mtune options in both the arm and aarch64 back ends (GCC 1629 identifiers in parentheses): 1630 + Arm Cortex-A78 (cortex-a78). 1631 + Arm Cortex-A78AE (cortex-a78ae). 1632 + Arm Cortex-A78C (cortex-a78c). 1633 + Arm Cortex-X1 (cortex-x1). 1634 + Arm Neoverse V1 (neoverse-v1). 1635 + Arm Neoverse N2 (neoverse-n2). 1636 * GCC can now auto-vectorize operations performing addition, 1637 subtraction, multiplication and the accumulate/subtract variants on 1638 complex numbers, taking advantage of the Advanced SIMD instructions 1639 in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64) 1640 and MVE (AArch32 M-profile) instruction sets. 1641 1642 [65]AArch64 1643 1644 * In addition to the above, the following AArch64-only CPUs are now 1645 supported: 1646 + Fujitsu A64FX (a64fx). 1647 + Arm Cortex-R82 (cortex-r82). 1648 * The AArch64 Armv8-R architecture is now supported through the 1649 -march=armv8-r option. 1650 * Mitigation against the [66]Straight-line Speculation vulnerability 1651 is supported with the -mharden-sls= option. Please refer to the 1652 documentation for usage instructions. 1653 * The availability of Advanced SIMD intrinsics available through the 1654 arm_neon.h header is improved and GCC 11 supports the full set of 1655 intrinsics defined by ACLE Q3 2020. 1656 1657 [67]AMD Radeon (GCN) 1658 1659 * Initial support for gfx908 GPUs has been added. 1660 1661 [68]arm 1662 1663 * Initial auto-vectorization is now available when targeting the MVE 1664 instruction set. 1665 * GCC can now make use of the Low Overhead Branch instruction in 1666 Armv8.1-M to optimize loop counters and checks. 1667 * The -mcpu=cortex-m55 option now supports the extensions +nomve and 1668 +nomve.fp to control generation of MVE and MVE floating-point 1669 instructions. 1670 1671 [69]IA-32/x86-64 1672 1673 * New ISA extension support for Intel TSXLDTRK was added to GCC. 1674 TSXLDTRK intrinsics are available via the -mtsxldtrk compiler 1675 switch. 1676 * New ISA extension support for Intel SERIALIZE was added to GCC. 1677 SERIALIZE intrinsics are available via the -mserialize compiler 1678 switch. 1679 * New ISA extension support for Intel HRESET was added to GCC. HRESET 1680 intrinsics are available via the -mhreset compiler switch. 1681 * New ISA extension support for Intel UINTR was added to GCC. UINTR 1682 intrinsics are available via the -muintr compiler switch. 1683 * New ISA extension support for Intel KEYLOCKER was added to GCC. 1684 KEYLOCKER intrinsics are available via the -mkeylocker compiler 1685 switch. 1686 * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16 1687 was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are 1688 available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler 1689 switches. 1690 * New ISA extension support for Intel AVX-VNNI was added to GCC. 1691 AVX-VNNI intrinsics are available via the -mavxvnni compiler 1692 switch. 1693 * GCC now supports the Intel CPU named Sapphire Rapids through 1694 -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B, 1695 AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG, 1696 TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA 1697 extensions. 1698 * GCC now supports the Intel CPU named Alderlake through 1699 -march=alderlake. The switch enables the CLDEMOTE, PTWRITE, 1700 WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions. 1701 * GCC now supports the Intel CPU named Rocketlake through 1702 -march=rocketlake. Rocket Lake is based on Icelake client and minus 1703 SGX. 1704 * GCC now supports AMD CPUs based on the znver3 core via 1705 -march=znver3. 1706 * GCC now supports micro-architecture levels defined in the x86-64 1707 psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4. 1708 1709 [70]Nios II 1710 1711 * The options -mcustom-insn=N no longer produce compiler warnings if 1712 the custom instruction is not generated due to missing optimization 1713 options such as -fno-math-errno, -ffinite-math-only, or 1714 -funsafe-math-optimizations. These warnings were not consistently 1715 emitted for all custom instructions. 1716 * The -mcustom-fpu-cfg=fph2 option has been added to enable the 1717 custom instructions supported by the Nios II Floating Point 1718 Hardware 2 Component. 1719 1720 [71]NVPTX 1721 1722 * The -misa default has changed from sm_30 to sm_35. 1723 * The -m32 compiler switch has been removed. 1724 * The -msoft-stack-reserve-local format has been fixed. Previously, 1725 it accepted -msoft-stack-reserve-local<n>. It now accepts 1726 -msoft-stack-reserve-local=<n>. 1727 1728 [72]S/390, System z, IBM Z Systems 1729 1730 * The behavior when compiling with -fexcess-precision=standard (e.g., 1731 implied by -std=c99) on s390(x) targets can now be controlled at 1732 configure time with the flag --enable-s390-excess-float-precision. 1733 When enabled, GCC will maintain previous behavior and evaluate 1734 float expressions in double precision, which aligns with the 1735 definition of float_t as double. With the flag disabled, GCC will 1736 always evaluate float expressions in single precision. In native 1737 builds and cross compiles that have target libc headers, GCC will 1738 by default match the definition of float_t in the installed glibc. 1739 1740 [73]RISC-V 1741 1742 * Support address sanitizer for RISC-V. 1743 * Support big-endian for RISC-V, thanks to Marcus Comstedt. 1744 * Implement new style of architecture extension test macros: each 1745 architecture extension has a corresponding feature test macro, 1746 which can be used to test its existence and version information. 1747 * Legacy architecture extension test macros like __riscv_atomic are 1748 deprecated, but will still be supported for at least 2 release 1749 cycles. 1750 * Support IFUNC for riscv*-*-linux*. 1751 * Add new option -misa-spec=* to control ISA spec version. This 1752 controls the default version of each extensions. It defaults to 1753 2.2. 1754 * Introduce the --with-multilib-generator configure time option. This 1755 allows for flexible config multi-lib settings. Its syntax is the 1756 same as RISC-V's multilib-generator. 1757 * Extend the sytax for multilib-generator, support expansion operator 1758 * to reduce the complexity of complicated multi-lib re-use rules. 1759 * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the 1760 pipeline model and architecture extensions, like -mtune=* plus 1761 -march=*. 1762 * Support for TLS stack protector canary access, thanks to Cooper Qu. 1763 * Support __builtin_thread_pointer for RISC-V. 1764 * Introduce shorten_memrefs optimization, which reduces the code size 1765 for memory access, thanks to Craig Blackmore. 1766 1767 [74]Operating Systems 1768 1769 [75]AIX 1770 1771 * GCC for AIX can be built as a 64-bit application and the runtime is 1772 built as FAT libraries containing both 32-bit and 64-bit objects. 1773 * Support AIX Vector Extended ABI with -mabi=vec-extabi. 1774 * Thread-Local uninitiated data placed in local common section. 1775 * Use thread-safe access in ctype. 1776 * Link with libc128.a when long-double-128 enabled. 1777 1778 [76]Improvements to Static Analyzer 1779 1780 * The implementation of how program state is tracked within 1781 [77]-fanalyzer has been completely rewritten for GCC 11, fixing 1782 numerous bugs, and allowing for the analyzer to scale to larger C 1783 source files. 1784 * The analysis of allocations and deallocations has been generalized 1785 beyond malloc and free. 1786 + As preliminary work towards eventually supporting C++, the 1787 malloc/free checking will also check new/delete and 1788 new[]/delete[]. However, C++ is not yet properly supported by 1789 [78]-fanalyzer (for example, exception-handling is 1790 unimplemented). 1791 + As noted above, the existing [79]malloc attribute has been 1792 extended so that it can be used to identify 1793 allocator/deallocator API pairs. The analyzer will use these 1794 attributes when checking for leaks, double-frees, 1795 use-after-frees, and similar issues. 1796 + A new [80]-Wanalyzer-mismatching-deallocation warning has been 1797 added, covering such mismatches as using scalar delete rather 1798 vector delete[]. 1799 * The analyzer has gained warnings 1800 [81]-Wanalyzer-shift-count-negative, 1801 [82]-Wanalyzer-shift-count-overflow, [83]-Wanalyzer-write-to-const, 1802 and [84]-Wanalyzer-write-to-string-literal, all enabled by default 1803 when [85]-fanalyzer is enabled. 1804 * The analyzer can now be extended by GCC plugins, allowing for 1805 domain-specific path-sensitive warnings. An example of using a 1806 [86]GCC plugin to check for misuses of CPython's global interpreter 1807 lock can be seen in the test suite 1808 * The analyzer has gained new debugging options 1809 [87]-fdump-analyzer-json and [88]-fno-analyzer-feasibility. 1810 1811 Other significant improvements 1812 1813 * GCC has gained a new environment variable 1814 [89]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to 1815 request machine-readable fix-it hints without needing to adjust 1816 build flags. 1817 1818 [90]GCC 11.1 1819 1820 This is the [91]list of problem reports (PRs) from GCC's bug tracking 1821 system that are known to be fixed in the 11.1 release. This list might 1822 not be complete (that is, it is possible that some PRs that have been 1823 fixed are not listed here). 1824 1825 [92]GCC 11.2 1826 1827 This is the [93]list of problem reports (PRs) from GCC's bug tracking 1828 system that are known to be fixed in the 11.2 release. This list might 1829 not be complete (that is, it is possible that some PRs that have been 1830 fixed are not listed here). 1831 1832 [94]GCC 11.3 1833 1834 This is the [95]list of problem reports (PRs) from GCC's bug tracking 1835 system that are known to be fixed in the 11.3 release. This list might 1836 not be complete (that is, it is possible that some PRs that have been 1837 fixed are not listed here). 1838 1839 [96]GCC 11.4 1840 1841 This is the [97]list of problem reports (PRs) from GCC's bug tracking 1842 system that are known to be fixed in the 11.4 release. This list might 1843 not be complete (that is, it is possible that some PRs that have been 1844 fixed are not listed here). 1845 1846 Target Specific Changes 1847 1848 x86-64 1849 1850 * The x86-64 ABI of passing and returning structure with a 64-bit 1851 single precision vector changed in GCC 11.1 when 3DNOW is disabled. 1852 Disabling 3DNOW no longer changes how they are passed nor returned. 1853 This ABI change is now diagnosed with -Wpsabi. 1854 * Mitigation against straight line speculation (SLS) for function 1855 return and indirect jump is supported via 1856 -mharden-sls=[none|all|return|indirect-jmp]. 1857 * Add CS prefix to call and jmp to indirect thunk with branch target 1858 in r8-r15 registers via -mindirect-branch-cs-prefix. 1859 1860 [98]GCC 11.5 1861 1862 This is the [99]list of problem reports (PRs) from GCC's bug tracking 1863 system that are known to be fixed in the 11.5 release. This list might 1864 not be complete (that is, it is possible that some PRs that have been 1865 fixed are not listed here). 1866 1867 Caveats 1868 1869 aarch64 1870 1871 * Due to a bug introduced a few weeks before the 11.5.0 release if 1872 the compiler is configured without explicit --with-arch=, 1873 --with=cpu= and/or --with-tune= configure options the compiler 1874 without explicit -march= etc. options might act as if asked for 1875 cortex-a34. This can be fixed by applying manually the 1876 [100]r12-8060 commit patch on top of GCC 11.5.0. See [101]PR116029 1877 for more details. GCC 11.4.0 or earlier releases are not affected. 1878 1879 1880 For questions related to the use of GCC, please consult these web 1881 pages and the [102]GCC manuals. If that fails, the 1882 [103]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1883 web pages and the development of GCC are welcome on our developer 1884 list at [104]gcc (a] gcc.gnu.org. All of [105]our lists have public 1885 archives. 1886 1887 Copyright (C) [106]Free Software Foundation, Inc. Verbatim copying and 1888 distribution of this entire article is permitted in any medium, 1889 provided this notice is preserved. 1890 1891 These pages are [107]maintained by the GCC team. Last modified 1892 2025-01-31. 1893 1894 References 1895 1896 1. https://gcc.gnu.org/gcc-11/porting_to.html 1897 2. https://gcc.gnu.org/onlinedocs/index.html#current 1898 3. https://wg21.link/p0522r0 1899 4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html 1900 5. https://gcc.gnu.org/gcc-11/changes.html#general 1901 6. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual 1902 7. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html 1903 8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit 1904 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin 1905 10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1906 11. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop 1907 12. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 1908 13. https://github.com/google/sanitizers/wiki/AddressSanitizer 1909 14. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 1910 15. https://dwarfstd.org/doc/DWARF5.pdf 1911 16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests 1912 17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref 1913 18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf 1914 19. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values 1915 20. https://gcc.gnu.org/gcc-11/changes.html#languages 1916 21. https://www.openmp.org/specifications/ 1917 22. https://www.openacc.org/ 1918 23. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-11 1919 24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html 1920 25. https://gcc.gnu.org/gcc-11/changes.html#c-family 1921 26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute 1922 27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 1923 28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 1924 29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 1925 30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 1926 31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 1927 32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div 1928 33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread 1929 34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan 1930 35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 1931 36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 1932 37. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized 1933 38. https://gcc.gnu.org/gcc-11/changes.html#c 1934 39. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter 1935 40. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter 1936 41. https://gcc.gnu.org/gcc-11/changes.html#cxx 1937 42. https://gcc.gnu.org/projects/cxx-status.html#cxx20 1938 43. https://gcc.gnu.org/projects/cxx-status.html#cxx23 1939 44. https://gcc.gnu.org/PR41437 1940 45. https://gcc.gnu.org/PR95307 1941 46. https://gcc.gnu.org/PR97518 1942 47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported 1943 48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct 1944 49. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion 1945 50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion 1946 51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 1947 52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse 1948 53. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull 1949 54. https://gcc.gnu.org/gcc-11/changes.html#libstdcxx 1950 55. https://gcc.gnu.org/gcc-11/changes.html#d 1951 56. https://gcc.gnu.org/gcc-11/changes.html#fortran 1952 57. https://gcc.gnu.org/gcc-11/changes.html#go 1953 58. https://gcc.gnu.org/gcc-11/changes.html#jit 1954 59. https://gcc.gnu.org/gcc-5/changes.html#jit 1955 60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html 1956 61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer 1957 62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html 1958 63. https://gcc.gnu.org/gcc-11/changes.html#targets 1959 64. https://gcc.gnu.org/gcc-11/changes.html#arm-targets 1960 65. https://gcc.gnu.org/gcc-11/changes.html#aarch64 1961 66. https://developer.arm.com/documentation/102587/0102/Straight-line-speculation-frequently-asked-questions 1962 67. https://gcc.gnu.org/gcc-11/changes.html#amdgcn 1963 68. https://gcc.gnu.org/gcc-11/changes.html#arm 1964 69. https://gcc.gnu.org/gcc-11/changes.html#x86 1965 70. https://gcc.gnu.org/gcc-11/changes.html#nios2 1966 71. https://gcc.gnu.org/gcc-11/changes.html#nvptx 1967 72. https://gcc.gnu.org/gcc-11/changes.html#s390 1968 73. https://gcc.gnu.org/gcc-11/changes.html#riscv 1969 74. https://gcc.gnu.org/gcc-11/changes.html#os 1970 75. https://gcc.gnu.org/gcc-11/changes.html#aix 1971 76. https://gcc.gnu.org/gcc-11/changes.html#analyzer 1972 77. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1973 78. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1974 79. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 1975 80. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation 1976 81. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative 1977 82. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow 1978 83. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 1979 84. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 1980 85. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1981 86. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825 1982 87. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json 1983 88. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility 1984 89. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT 1985 90. https://gcc.gnu.org/gcc-11/changes.html#11.1 1986 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0 1987 92. https://gcc.gnu.org/gcc-11/changes.html#11.2 1988 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2 1989 94. https://gcc.gnu.org/gcc-11/changes.html#11.3 1990 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.3 1991 96. https://gcc.gnu.org/gcc-11/changes.html#11.4 1992 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.4 1993 98. https://gcc.gnu.org/gcc-11/changes.html#11.5 1994 99. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.5 1995 100. https://gcc.gnu.org/r12-8060 1996 101. https://gcc.gnu.org/PR116029 1997 102. https://gcc.gnu.org/onlinedocs/ 1998 103. mailto:gcc-help (a] gcc.gnu.org 1999 104. mailto:gcc (a] gcc.gnu.org 2000 105. https://gcc.gnu.org/lists.html 2001 106. https://www.fsf.org/ 2002 107. https://gcc.gnu.org/about.html 2003 ====================================================================== 2004 http://gcc.gnu.org/gcc-10/index.html 2005 2006 GCC 10 Release Series 2007 2008 (This release series is no longer supported.) 2009 2010 July, 7, 2023 2011 2012 The [1]GNU project and the GCC developers are pleased to announce the 2013 release of GCC 10.5. 2014 2015 This release is a bug-fix release, containing fixes for regressions in 2016 GCC 10.4 relative to previous releases of GCC. 2017 2018 Release History 2019 2020 GCC 10.5 2021 July 7, 2023 ([2]changes, [3]documentation) 2022 2023 GCC 10.4 2024 June 28, 2022 ([4]changes, [5]documentation) 2025 2026 GCC 10.3 2027 April 8, 2021 ([6]changes, [7]documentation) 2028 2029 GCC 10.2 2030 July 23, 2020 ([8]changes, [9]documentation) 2031 2032 GCC 10.1 2033 May 7, 2020 ([10]changes, [11]documentation) 2034 2035 References and Acknowledgements 2036 2037 GCC used to stand for the GNU C Compiler, but since the compiler 2038 supports several other languages aside from C, it now stands for the 2039 GNU Compiler Collection. 2040 2041 The GCC developers would like to thank the numerous people that have 2042 contributed new features, improvements, bug fixes, and other changes as 2043 well as test results to GCC. This [12]amazing group of volunteers is 2044 what makes GCC successful. 2045 2046 For additional information about GCC please refer to the [13]GCC 2047 project web site or contact the [14]GCC development mailing list. 2048 2049 To obtain GCC please use [15]our mirror sites or [16]our version 2050 control system. 2051 2052 2053 For questions related to the use of GCC, please consult these web 2054 pages and the [17]GCC manuals. If that fails, the 2055 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2056 web pages and the development of GCC are welcome on our developer 2057 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 2058 archives. 2059 2060 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 2061 distribution of this entire article is permitted in any medium, 2062 provided this notice is preserved. 2063 2064 These pages are [22]maintained by the GCC team. Last modified 2065 2024-05-30. 2066 2067 References 2068 2069 1. http://www.gnu.org/ 2070 2. https://gcc.gnu.org/gcc-10/changes.html 2071 3. https://gcc.gnu.org/onlinedocs/10.5.0/ 2072 4. https://gcc.gnu.org/gcc-10/changes.html 2073 5. https://gcc.gnu.org/onlinedocs/10.4.0/ 2074 6. https://gcc.gnu.org/gcc-10/changes.html 2075 7. https://gcc.gnu.org/onlinedocs/10.3.0/ 2076 8. https://gcc.gnu.org/gcc-10/changes.html 2077 9. https://gcc.gnu.org/onlinedocs/10.2.0/ 2078 10. https://gcc.gnu.org/gcc-10/changes.html 2079 11. https://gcc.gnu.org/onlinedocs/10.1.0/ 2080 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html 2081 13. https://gcc.gnu.org/index.html 2082 14. mailto:gcc (a] gcc.gnu.org 2083 15. https://gcc.gnu.org/mirrors.html 2084 16. https://gcc.gnu.org/git.html 2085 17. https://gcc.gnu.org/onlinedocs/ 2086 18. mailto:gcc-help (a] gcc.gnu.org 2087 19. mailto:gcc (a] gcc.gnu.org 2088 20. https://gcc.gnu.org/lists.html 2089 21. https://www.fsf.org/ 2090 22. https://gcc.gnu.org/about.html 2091 ====================================================================== 2092 http://gcc.gnu.org/gcc-10/changes.html 2093 2094 GCC 10 Release Series 2095 Changes, New Features, and Fixes 2096 2097 This page is a "brief" summary of some of the huge number of 2098 improvements in GCC 10. You may also want to check out our [1]Porting 2099 to GCC 10 page and the [2]full GCC documentation. 2100 2101 Caveats 2102 2103 * An ABI incompatibility between C++14 and C++17 has been fixed. On 2104 some targets a class with a zero-sized subobject would be passed 2105 incorrectly when compiled as C++17 or C++20. See the [3]C++ notes 2106 below for more details. 2107 * The deprecated Profile Mode and array_allocator extensions have 2108 been removed from libstdc++. 2109 * The non-standard std::__is_nullptr_t type trait is deprecated and 2110 will be removed from libstdc++ in a future release. The standard 2111 trait std::is_null_pointer should be instead. 2112 * The minimum version of the [4]MPFR library required for building 2113 GCC has been increased to version 3.1.0 (released 2011-10-03). 2114 * The automatic template instantiation at link time (-frepo) has been 2115 removed. 2116 * The --param allow-store-data-races internal parameter has been 2117 removed in favor of a new official option -fallow-store-data-races. 2118 While default behavior is unchanged and the new option allows to 2119 correctly maintain a per compilation unit setting across link-time 2120 optimization, alteration of the default via --param 2121 allow-store-data-races will now be diagnosed and build systems have 2122 to be adjusted accordingly. 2123 * Offloading to Heterogeneous System Architecture Intermediate 2124 Language (HSAIL) has been deprecated and will likely be removed in 2125 a future release. 2126 * The type of the std::iterator base class of 2127 std::istreambuf_iterator was changed in C++98 mode to be consistent 2128 with C++11 and later standards. See the [5]libstdc++ notes below 2129 for more details. 2130 * GCC 10.5 does not bootstrap with a C++98 compiler; if you need to 2131 start from C++98, you should build 10.4 or 9.5 instead. 2132 2133 [6]General Improvements 2134 2135 * New built-in functions: 2136 + The [7]__has_builtin built-in preprocessor operator can be 2137 used to query support for built-in functions provided by GCC 2138 and other compilers that support it. 2139 + __builtin_roundeven for the corresponding function from 2140 ISO/IEC TS 18661. 2141 * New command-line options: 2142 + [8]-fallocation-dce removes unneeded pairs of new and delete 2143 operators. 2144 + [9]-fprofile-partial-training can now be used to inform the 2145 compiler that code paths not covered by the training run 2146 should not be optimized for size. 2147 + [10]-fprofile-reproducible controls level of reproducibility 2148 of profile gathered by [11]-fprofile-generate. This makes it 2149 possible to rebuild program with same outcome which is useful, 2150 for example, for distribution packages. 2151 + [12]-fprofile-prefix-path can be used in combination with 2152 -fprofile-generate=profile_dir and -fprofile-use=profile_dir 2153 to inform GCC where the base directory of build source tree is 2154 in case it differs between instrumentation and optimized 2155 builds. 2156 + [13]-fanalyzer enables a new static analysis pass and 2157 associated warnings. This pass performs a time-consuming 2158 exploration of paths through the code in the hope of detecting 2159 various common errors, such as double-free bugs. This option 2160 should be regarded as experimental in this release. In 2161 particular, analysis of non-C code is unlikely to work. 2162 * Inter-procedural optimization improvements: 2163 + The inter-procedural scalar replacement of aggregates 2164 (IPA-SRA) pass was re-implemented to work at link-time and can 2165 now also remove computing and returning unused return values. 2166 + [14]-finline-functions is now enabled at -O2 and was retuned 2167 for better code size versus runtime performance trade-offs. 2168 Inliner heuristics was also significantly sped up to avoid 2169 negative impact to -flto -O2 compile times. 2170 + Inliner heuristics and function cloning can now use 2171 value-range information to predict effectivity of individual 2172 transformations. 2173 + During link-time optimization the C++ One Definition Rule is 2174 used to increase precision of type based alias analysis. 2175 * Link-time optimization improvements: 2176 + A new binary [15]lto-dump has been added. It dumps various 2177 information about LTO bytecode object files. 2178 + The parallel phase of the LTO can automatically detect a 2179 running make's jobserver or fall back to number of available 2180 cores. 2181 + The LTO bytecode can be compressed with the [16]zstd 2182 algorithm. The configure script automatically detects zstd 2183 support. 2184 + Most --param values can now be specified at translation unit 2185 granularity. This includes all parameters controlling the 2186 inliner and other inter-procedural optimizations. Unlike 2187 earlier releases, GCC 10 will ignore parameters controlling 2188 optimizations specified at link-time and apply parameters 2189 specified at compile-time in the same manner as done for 2190 optimization flags. 2191 * Profile driven optimization improvements: 2192 + Profile maintenance during compilation and hot/cold code 2193 partitioning have been improved. 2194 + Using [17]-fprofile-values, an instrumented binary can track 2195 multiple values (up to 4) for e.g. indirect calls and provide 2196 more precise profile information. 2197 2198 [18]New Languages and Language-Specific Improvements 2199 2200 * Version 2.6 of the [19]OpenACC specification is now supported by 2201 the C, C++ and Fortran compilers. See the [20]implementation status 2202 section on the OpenACC wiki page and the [21]run-time library 2203 documentation for further information. 2204 * GCC 10 adds a number of newly implemented [22]OpenMP 5.0 features 2205 such as conditional lastprivate clause, scan and loop directives, 2206 order(concurrent) and use_device_addr clauses support, if clause on 2207 simd construct, and partial support for the declare variant 2208 directive, getting closer to full support of the OpenMP 5.0 2209 standard. 2210 * OpenMP and OpenACC now support [23]offloading to AMD Radeon (GCN) 2211 GPUs; supported are the third-generation Fiji (fiji) and the 2212 fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906). 2213 2214 [24]C family 2215 2216 * New attributes: 2217 + The access function and type attribute has been added to 2218 describe how a function accesses objects passed to it by 2219 pointer or reference, and to associate such arguments with 2220 integer arguments denoting the objects' sizes. The attribute 2221 is used to enable the detection of invalid accesses by 2222 user-defined functions, such as those diagnosed by 2223 -Wstringop-overflow. 2224 + The symver attribute can be used to bind symbols to specific 2225 version nodes on ELF platforms. This is preferred to using 2226 inline assembly with GNU as symver directive because the 2227 latter is not compatible with link-time optimizations. 2228 * New warnings: 2229 + [25]-Wstring-compare, enabled by -Wextra, warns about equality 2230 and inequality expressions between zero and the result of a 2231 call to either strcmp and strncmp that evaluate to a constant 2232 as a result of the length of one argument being greater than 2233 the size of the array pointed to by the other. 2234 + [26]-Wzero-length-bounds, enabled by -Warray-bounds, warns 2235 about accesses to elements of zero-length arrays that might 2236 overlap other members of the same object. 2237 * Enhancements to existing warnings: 2238 + [27]-Warray-bounds detects more out-of-bounds accesses to 2239 member arrays as well as accesses to elements of zero-length 2240 arrays. 2241 + [28]-Wformat-overflow makes full use of string length 2242 information computed by the strlen optimization pass. 2243 + [29]-Wrestrict detects overlapping accesses to dynamically 2244 allocated objects. 2245 + [30]-Wreturn-local-addr diagnoses more instances of return 2246 statements returning addresses of automatic variables. 2247 + [31]-Wstringop-overflow detects more out-of-bounds stores to 2248 member arrays including zero-length arrays, dynamically 2249 allocated objects and variable length arrays, as well as more 2250 instances of reads of unterminated character arrays by string 2251 built-in functions. The warning also detects out-of-bounds 2252 accesses by calls to user-defined functions declared with the 2253 new attribute access. 2254 + [32]-Warith-conversion re-enables warnings from -Wconversion, 2255 -Wfloat-conversion, and -Wsign-conversion that are now off by 2256 default for an expression where the result of an arithmetic 2257 operation will not fit in the target type due to promotion, 2258 but the operands of the expression do fit in the target type. 2259 * Extended characters in identifiers may now be specified directly in 2260 the input encoding (UTF-8, by default), in addition to the UCN 2261 syntax (\uNNNN or \UNNNNNNNN) that is already supported: 2262 2263 static const int p = 3; 2264 int get_nave_pi() { 2265 return p; 2266 } 2267 2268 [33]C 2269 2270 * Several new features from the upcoming C2X revision of the ISO C 2271 standard are supported with -std=c2x and -std=gnu2x. Some of these 2272 features are also supported as extensions when compiling for older 2273 language versions. In addition to the features listed, some 2274 features previously supported as extensions and now added to the C 2275 standard are enabled by default in C2X mode and not diagnosed with 2276 -std=c2x -Wpedantic. 2277 + The [[]] attribute syntax is supported, as in C++. Existing 2278 attributes can be used with this syntax in forms such as 2279 [[gnu::const]]. The standard attributes [[deprecated]], 2280 [[fallthrough]] and [[maybe_unused]] are supported. 2281 + UTF-8 character constants using the u8'' syntax are supported. 2282 + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and 2283 LDBL_NORM_MAX. 2284 + When decimal floating-point arithmetic is supported, <float.h> 2285 defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and 2286 DEC128_TRUE_MIN, in addition to the macros that were 2287 previously only defined if __STDC_WANT_DEC_FP__ was defined 2288 before including <float.h>. 2289 + In C2X mode, empty parentheses in a function definition give 2290 that function a type with a prototype for subsequent calls; 2291 other old-style function definitions are diagnosed by default 2292 in C2X mode. 2293 + The strftime format checking supports the %OB and %Ob formats. 2294 + In C2X mode, -fno-fp-int-builtin-inexact is enabled by 2295 default. 2296 * GCC now defaults to -fno-common. As a result, global variable 2297 accesses are more efficient on various targets. In C, global 2298 variables with multiple tentative definitions now result in linker 2299 errors. With -fcommon such definitions are silently merged during 2300 linking. 2301 2302 [34]C++ 2303 2304 * Several C++20 features have been implemented: 2305 + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, 2306 P0848R3, P1616R1, P1452R2 2307 + P1668R1, Permitting Unevaluated inline-assembly in constexpr 2308 Functions 2309 + P1161R3, Deprecate a[b,c] 2310 + P0848R3, Conditionally Trivial Special Member Functions 2311 + P1091R3, Extending structured bindings 2312 + P1143R2, Adding the constinit keyword 2313 + P1152R4, Deprecating volatile 2314 + P0388R4, Permit conversions to arrays of unknown bound 2315 + P0784R7, constexpr new 2316 + P1301R4, [[nodiscard("with reason")]] 2317 + P1814R0, class template argument deduction for alias templates 2318 + P1816R0, class template argument deduction for aggregates 2319 + P0960R3, Parenthesized initialization of aggregates 2320 + P1331R2, Allow trivial default initialization in constexpr 2321 contexts 2322 + P1327R1, Allowing dynamic_cast and polymorphic typeid in 2323 constexpr contexts 2324 + P0912R5, Coroutines (requires -fcoroutines) 2325 * Several C++ Defect Reports have been resolved, e.g.: 2326 + DR 1560, lvalue-to-rvalue conversion in ?: 2327 + DR 1813, __is_standard_layout for a class with repeated bases 2328 + DR 2094, volatile scalars are trivially copyable, 2329 + DR 2096, constraints on literal unions 2330 + DR 2413, typename in conversion-function-ids 2331 + DR 2352, Similar types and reference binding 2332 + DR 1601, Promotion of enumeration with fixed underlying type 2333 + DR 330, Qualification conversions and pointers to arrays of 2334 pointers 2335 + DR 1307, Overload resolution based on size of array 2336 initializer-list 2337 + DR 1710, Missing template keyword in class-or-decltype 2338 * New warnings: 2339 + [35]-Wmismatched-tags, disabled by default, warns about 2340 declarations of structs, classes, and class templates and 2341 their specializations with a class-key that does not match 2342 either the definition or the first declaration if no 2343 definition is provided. The option is provided to ease 2344 portability to Windows-based compilers. 2345 + [36]-Wredundant-tags, disabled by default, warns about 2346 redundant class-key and enum-key in contexts where the key can 2347 be eliminated without causing an syntactic ambiguity. 2348 * G++ can now detect modifying constant objects in constexpr 2349 evaluation (which is undefined behavior). 2350 * G++ no longer emits bogus -Wsign-conversion warnings with explicit 2351 casts. 2352 * Narrowing is now detected in more contexts (e.g., case values). 2353 * Memory consumption of the compiler has been reduced in constexpr 2354 evaluation. 2355 * The noexcept-specifier is now properly treated as a complete-class 2356 context as per [class.mem]. 2357 * The attribute deprecated can now be used on namespaces too. 2358 * The ABI of passing and returning certain C++ classes by value 2359 changed on several targets in GCC 10, including [37]AArch64, 2360 [38]ARM, [39]PowerPC ELFv2, [40]S/390 and [41]Itanium. These 2361 changes affect classes with a zero-sized subobject (an empty base 2362 class, or data member with the [[no_unique_address]] attribute) 2363 where all other non-static data members have the same type (this is 2364 called a "homogeneous aggregate" in some ABI specifications, or if 2365 there is only one such member, a "single element"). In -std=c++17 2366 and -std=c++20 modes, classes with an empty base class were not 2367 considered to have a single element or to be a homogeneous 2368 aggregate, and so could be passed differently (in the wrong 2369 registers or at the wrong stack address). This could make code 2370 compiled with -std=c++17 and -std=c++14 ABI incompatible. This has 2371 been corrected and the empty bases are ignored in those ABI 2372 decisions, so functions compiled with -std=c++14 and -std=c++17 are 2373 now ABI compatible again. Example: struct empty {}; struct S : 2374 empty { float f; }; void f(S);. Similarly, in classes containing 2375 non-static data members with empty class types using the C++20 2376 [[no_unique_address]] attribute, those members weren't ignored in 2377 the ABI argument passing decisions as they should be. Both of these 2378 ABI changes are now diagnosed with -Wpsabi. 2379 2380 [42]Runtime Library (libstdc++) 2381 2382 * Improved experimental C++2a support, including: 2383 + Library concepts in <concepts> and <iterator>. 2384 + Constrained algorithms in <ranges>, <algorithm>, and <memory> 2385 (thanks to Patrick Palka). 2386 + New algorithms shift_left and shift_right (thanks to Patrick 2387 Palka). 2388 + std::span (thanks to JeanHeyd Meneide). 2389 + Three-way comparisons in <compare> and throughout the library. 2390 + Constexpr support in <algorithm> and elsewhere (thanks to 2391 Edward Smith-Rowland). 2392 + <stop_token> and std::jthread (thanks to Thomas Rodgers). 2393 + std::atomic_ref and std::atomic<floating point>. 2394 + Integer comparison functions (cmp_equal, cmp_less etc.). 2395 + std::ssize, std::to_array. 2396 + std::construct_at, std::destroy, constexpr std::allocator. 2397 + Mathematical constants in <numbers>. 2398 * Support for RDSEED in std::random_device. 2399 * Reduced header dependencies, leading to faster compilation for some 2400 code. 2401 * The std::iterator base class of std::istreambuf_iterator was 2402 changed in C++98 mode to be consistent with C++11 and later 2403 standards. This is expected to have no noticeable effect except in 2404 the unlikely case of a class which has potentially overlapping 2405 subobjects of type std::istreambuf_iterator<C> and another iterator 2406 type with a std::iterator<input_iterator_tag, C, ...> base class. 2407 The layout of such a type might change when compiled as C++98. 2408 [43]Bug 92285 has more details and concrete examples. 2409 2410 [44]D 2411 2412 * Support for static foreach has been implemented. 2413 * Aliases can now be created directly from any __traits that return 2414 symbols or tuples. Previously, an AliasSeq was necessary in order 2415 to alias their return. 2416 * It is now possible to detect the language ABI specified for a 2417 struct, class, or interface using __traits(getLinkage, ...). 2418 * Support for core.math.toPrec intrinsics has been added. These 2419 intrinsics guarantee the rounding to specific floating-point 2420 precisions at specified points in the code. 2421 * Support for pragma(inline) has been implemented. Previously the 2422 pragma was recognized, but had no effect on the compilation. 2423 * Optional parentheses in asm operands are deprecated and will be 2424 removed in a future release. 2425 * All content imported files are now included in the make dependency 2426 list when compiling with -M. 2427 * Compiler recognized attributes provided by the gcc.attribute module 2428 will now take effect when applied to function prototypes as well as 2429 when applied to full function declarations. 2430 * Added a --enable-libphobos-checking configure option to control 2431 whether run-time checks are compiled into the D runtime library. 2432 * Added a --with-libphobos-druntime-only configure option to indicate 2433 whether to build only the core D runtime library, or both the core 2434 and standard libraries into libphobos. 2435 2436 [45]Fortran 2437 2438 * use_device_addr of version 5.0 of the [46]OpenMP specification is 2439 now supported. Note that otherwise OpenMP 4.5 is partially 2440 supported in the Fortran compiler; the largest missing item is 2441 structure element mapping. 2442 * The default buffer size for I/O using unformatted files has been 2443 increased to 1048576. The buffer size for can now be set at runtime 2444 via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and 2445 GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted 2446 files, respectively. 2447 * Mismatches between actual and dummy argument lists in a single file 2448 are now rejected with an error. Use the new option 2449 -fallow-argument-mismatch to turn these errors into warnings; this 2450 option is implied with -std=legacy. -Wargument-mismatch has been 2451 removed. 2452 * The handling of a BOZ literal constant has been reworked to provide 2453 better conformance to the Fortran 2008 and 2018 standards. In these 2454 Fortran standards, a BOZ literal constant is a typeless and 2455 kindless entity. As a part of the rework, documented and 2456 undocumented extensions to the Fortran standard now emit errors 2457 during compilation. Some of these extensions are permitted with the 2458 -fallow-invalid-boz option, which degrades the error to a warning 2459 and the code is compiled as with older gfortran. 2460 * At any optimization level except-Os, gfortran now uses inline 2461 packing for arguments instead of calling a library routine. If the 2462 source contains a large number of arguments that need to be 2463 repacked, code size or time for compilation can become excessive. 2464 If that is the case, -fno-inline-arg-packing can be used to disable 2465 inline argument packing. 2466 * Legacy extensions: 2467 + For formatted input/output, if the explicit widths after the 2468 data-edit descriptors I, F and G have been omitted, default 2469 widths are used. 2470 + A blank format item at the end of a format specification, i.e. 2471 nothing following the final comma, is allowed. Use the option 2472 -fdec-blank-format-item; this option is implied with -fdec. 2473 + The existing support for AUTOMATIC and STATIC attributes has 2474 been extended to allow variables with the AUTOMATIC attribute 2475 to be used in EQUIVALENCE statements. Use -fdec-static; this 2476 option is implied by -fdec. 2477 + Allow character literals in assignments and DATA statements 2478 for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. 2479 Use the option -fdec-char-conversions; this option is implied 2480 with -fdec. 2481 + DEC comparisons, i.e. allow Hollerith constants to be used in 2482 comparisons with INTEGER, REAL, COMPLEX and CHARACTER 2483 expressions. Use the option -fdec. 2484 * Character type names in errors and warnings now include len in 2485 addition to kind; * is used for assumed length. The kind is omitted 2486 if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4). 2487 * CO_BROADCAST now supports derived type variables including objects 2488 with allocatable components. In this case, the optional arguments 2489 STAT= and ERRMSG= are currently ignored. 2490 * The handling of module and submodule names has been reworked to 2491 allow the full 63-character length mandated by the standard. 2492 Previously symbol names were truncated if the combined length of 2493 module, submodule, and function name exceeded 126 characters. This 2494 change therefore breaks the ABI, but only for cases where this 126 2495 character limit was exceeded. 2496 2497 [47]Go 2498 2499 * GCC 10 provides a complete implementation of the Go 1.14.6 user 2500 packages. 2501 2502 [48]libgccjit 2503 2504 * The libgccjit API gained four new entry points: 2505 + [49]gcc_jit_version_major, [50]gcc_jit_version_minor, and 2506 [51]gcc_jit_version_patchlevel for programmatically checking 2507 the libgccjit version from client code, and 2508 + [52]gcc_jit_context_new_bitfield 2509 2510 [53]New Targets and Target Specific Improvements 2511 2512 [54]AArch64 & arm 2513 2514 * The AArch64 and arm ports now support condition flag output 2515 constraints in inline assembly, as indicated by the 2516 __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for 2517 A32 and T32 targets. Please refer to the documentation for more 2518 details. 2519 2520 [55]AArch64 2521 2522 * There have been several improvements related to the Scalable Vector 2523 Extension (SVE): 2524 + The SVE ACLE types and intrinsics are now supported. They can 2525 be accessed using the header file arm_sve.h. 2526 + It is now possible to create fixed-length SVE types using the 2527 arm_sve_vector_bits attribute. For example: 2528 #if __ARM_FEATURE_SVE_BITS==512 2529 typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512))); 2530 typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512))); 2531 #endif 2532 + -mlow-precision-div, -mlow-precision-sqrt and 2533 -mlow-precision-recip-sqrt now work for SVE. 2534 + -msve-vector-bits=128 now generates vector-length-specific 2535 code for little-endian targets. It continues to generate 2536 vector-length-agnostic code for big-endian targets, just as 2537 previous releases did for all targets. 2538 + The vectorizer is now able to use extending loads and 2539 truncating stores, including gather loads and scatter stores. 2540 + The vectorizer now compares the cost of vectorizing with SVE 2541 and vectorizing with Advanced SIMD and tries to pick the best 2542 one. Previously it would always use SVE if possible. 2543 + If a vector loop uses Advanced SIMD rather than SVE, the 2544 vectorizer now considers using SVE to vectorize the left-over 2545 elements (the "scalar tail" or "epilog"). 2546 + Besides these specific points, there have been many general 2547 improvements to the way that the vectorizer uses SVE. 2548 * The -mbranch-protection=pac-ret option now accepts the optional 2549 argument +b-key extension to perform return address signing with 2550 the B-key instead of the A-key. 2551 * The option -moutline-atomics has been added to aid deployment of 2552 the Large System Extensions (LSE) on GNU/Linux systems built with a 2553 baseline architecture targeting Armv8-A. When the option is 2554 specified code is emitted to detect the presence of LSE 2555 instructions at runtime and use them for standard atomic 2556 operations. For more information please refer to the documentation. 2557 * The Transactional Memory Extension is now supported through ACLE 2558 intrinsics. It can be enabled through the +tme option extension 2559 (for example, -march=armv8.5-a+tme). 2560 * A number of features from Armv8.5-A are now supported through ACLE 2561 intrinsics. These include: 2562 + The random number instructions that can be enabled through the 2563 (already present in GCC 9.1) +rng option extension. 2564 + Floating-point intrinsics to round to integer instructions 2565 from Armv8.5-A when targeting -march=armv8.5-a or later. 2566 + Memory Tagging Extension intrinsics enabled through the 2567 +memtag option extension. 2568 * Similarly, the following Armv8.6-A features are now supported 2569 through ACLE intrinsics: 2570 + The bfloat16 extension. This extension is enabled 2571 automatically when Armv8.6-A is selected (such as by 2572 -march=armv8.6-a). It can also be enabled for Armv8.2-A and 2573 later using the +bf16 option extension. 2574 + The Matrix Multiply extension. This extension is split into 2575 three parts, one for each supported data type: 2576 o Support for 8-bit integer matrix multiply instructions. 2577 This extension is enabled automatically when Armv8.6-A is 2578 selected. It can also be enabled for Armv8.2-A and later 2579 using the +i8mm option extension. 2580 o Support for 32-bit floating-point matrix multiply 2581 instructions. This extension can be enabled using the 2582 +f32mm option extension, which also has the effect of 2583 enabling SVE. 2584 o Support for 64-bit floating-point matrix multiply 2585 instructions. This extension can be enabled using the 2586 +f64mm option extension, which likewise has the effect of 2587 enabling SVE. 2588 * SVE2 is now supported through ACLE intrinsics and (to a limited 2589 extent) through autovectorization. It can be enabled through the 2590 +sve2 option extension (for example, -march=armv8.5-a+sve2). 2591 Additional extensions can be enabled through +sve2-sm4, +sve2-aes, 2592 +sve2-sha3 and +sve2-bitperm. 2593 * Support has been added for the following processors (GCC 2594 identifiers in parentheses): 2595 + Arm Cortex-A77 (cortex-a77). 2596 + Arm Cortex-A76AE (cortex-a76ae). 2597 + Arm Cortex-A65 (cortex-a65). 2598 + Arm Cortex-A65AE (cortex-a65ae). 2599 + Arm Cortex-A34 (cortex-a34). 2600 + Marvell ThunderX3 (thunderx3t110). 2601 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2602 options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as 2603 arguments to the equivalent target attributes and pragmas. 2604 2605 [56]arm 2606 2607 * Support for the FDPIC ABI has been added. It uses 64-bit function 2608 descriptors to represent pointers to functions, and enables code 2609 sharing on MMU-less systems. The corresponding target triple is 2610 arm-uclinuxfdpiceabi, and the C library is uclibc-ng. 2611 * Support has been added for the Arm EABI on NetBSD through the 2612 arm*-*-netbsdelf-*eabi* triplet. 2613 * The handling of 64-bit integer operations has been significantly 2614 reworked and improved leading to improved performance and reduced 2615 stack usage when using 64-bit integral data types. The option 2616 -mneon-for-64bits is now deprecated and will be removed in a future 2617 release. 2618 * Support has been added for the following processors (GCC 2619 identifiers in parentheses): 2620 + Arm Cortex-A77 (cortex-a77). 2621 + Arm Cortex-A76AE (cortex-a76ae). 2622 + Arm Cortex-M35P (cortex-m35p). 2623 + Arm Cortex-M55 (cortex-m55). 2624 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2625 options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p. 2626 * Support has been extended for the ACLE [57]data-processing 2627 intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit 2628 multiplication and other related intrinsics aimed at DSP algorithm 2629 optimization. 2630 * Support for -mpure-code in Thumb-1 (v6m) has been added: this 2631 M-profile feature is no longer restricted to targets with MOVT. For 2632 example, -mcpu=cortex-m0 now supports this option. 2633 * Support for the [58]Armv8.1-M Mainline Architecture has been added. 2634 + Armv8.1-M Mainline can be enabled by using the 2635 -march=armv8.1-m.main command-line option. 2636 * Support for the [59]MVE beta ACLE intrinsics has been added. These 2637 intrinsics can be enabled by including the arm_mve.h header file 2638 and passing the +mve or +mve.fp option extensions (for example: 2639 -march=armv8.1-m.main+mve). 2640 * Support for the Custom Datapath Extension beta ACLE [60]intrinsics 2641 has been added. 2642 * Support for Armv8.1-M Mainline Security Extensions architecture has 2643 been added. The -mcmse option, when used in combination with an 2644 Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main 2645 -mcmse), now leads to the generation of improved code sequences 2646 when changing security states. 2647 2648 [61]AMD Radeon (GCN) 2649 2650 * Code generation and in particular vectorization support have been 2651 much improved. 2652 2653 [62]ARC 2654 2655 * The interrupt service routine functions save all used registers, 2656 including extension registers and auxiliary registers used by Zero 2657 Overhead Loops. 2658 * Improve code size by using multiple short instructions instead of a 2659 single long mov or ior instruction when its long immediate constant 2660 is known. 2661 * Fix usage of the accumulator register for ARC600. 2662 * Fix issues with uncached attribute. 2663 * Remove -mq-class option. 2664 * Improve 64-bit integer addition and subtraction operations. 2665 2666 [63]AVR 2667 2668 * Support for the XMEGA-like devices 2669 2670 ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, 2671 ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, 2672 ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, 2673 ATmega4808, ATmega4809 2674 has been added. 2675 * A new command-line option -nodevicespecs has been added. It allows 2676 to provide a custom device-specs file by means of 2677 2678 avr-gcc -nodevicespecs -specs=my-spec-file <options> 2679 and without the need to provide options -B and -mmcu=. See [64]AVR 2680 command-line options for details. This feature is also available in 2681 GCC 9.3+ and GCC 8.4+. 2682 * New command-line options -mdouble=[32,64] and -mlong-double=[32,64] 2683 have been added. They allow to choose the size (in bits) of the 2684 double and long double types, respectively. Whether or not the 2685 mentioned layouts are available, whether the options act as a 2686 multilib option, and the default for either option are controlled 2687 by the new [65]AVR configure options --with-double= and 2688 --with-long-double=. 2689 * A new configure option --with-libf7= has been added. It controls to 2690 which level avr-libgcc provides 64-bit floating point support by 2691 means of [66]Libf7. 2692 * A new configure option --with-double-comparison= has been added. 2693 It's unlikely you need to set this option by hand. 2694 2695 [67]IA-32/x86-64 2696 2697 * Support to expand __builtin_roundeven into the appropriate SSE 4.1 2698 instruction has been added. 2699 * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD 2700 intrinsics are available via the -menqcmd compiler switch. 2701 * GCC now supports the Intel CPU named Cooperlake through 2702 -march=cooperlake. The switch enables the AVX512BF16 ISA 2703 extensions. 2704 * GCC now supports the Intel CPU named Tigerlake through 2705 -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B 2706 AVX512VP2INTERSECT ISA extensions. 2707 2708 [68]MIPS 2709 2710 * The mips*-*-linux* targets now mark object files with appropriate 2711 GNU-stack note, facilitating use of non-executable stack hardening 2712 on GNU/Linux. The soft-float targets have this feature enabled by 2713 default, while for hard-float targets it is required for GCC to be 2714 configured with --with-glibc-version=2.31 against glibc 2.31 or 2715 later. 2716 2717 [69]PowerPC / PowerPC64 / RS6000 2718 2719 * Many vector builtins have been listed as deprecated in the 2720 [70]64-Bit ELF V2 ABI Specification for quite a number of years. 2721 The vector builtins listed in Tables A.8 through A.10 are now 2722 deprecated for GCC 10, and will likely be removed from support in 2723 GCC 11. Note that this does not result in any loss of function. 2724 These deprecated builtins generally provide somewhat nonsensical 2725 argument lists (for example, mixing signed, unsigned, and bool 2726 vector arguments arbitrarily), or are duplicate builtins that are 2727 inconsistent with the expected naming scheme. We expect that this 2728 will be unlikely to affect much if any code, and any required code 2729 changes will be trivial. 2730 2731 [71]PRU 2732 2733 * A new back end targeting TI PRU I/O processors has been contributed 2734 to GCC. 2735 2736 [72]RISC-V 2737 2738 * The riscv*-*-* targets now require GNU binutils version 2.30 or 2739 later, to support new assembly instructions produced by GCC. 2740 2741 [73]V850 2742 2743 * The ABI for V850 nested functions has been changed. Previously the 2744 V850 port used %r20 for the static chain pointer, now the port uses 2745 %r19. This corrects a long standing latent bug in the v850 port 2746 where a call to a nested function would unexpectedly change the 2747 value in %r20. 2748 2749 [74]Operating Systems 2750 2751 [75]Improvements for plugin authors 2752 2753 * GCC diagnostics can now have a chain of events associated with 2754 them, describing a path through the code that triggers the problem. 2755 These can be printed by the diagnostics subsystem in various ways, 2756 controlled by the [76]-fdiagnostics-path-format option, or captured 2757 in JSON form via [77]-fdiagnostics-format=json. 2758 * GCC diagnostics can now be associated with [78]CWE weakness 2759 identifiers, which will appear on the standard error stream, and in 2760 the JSON output from [79]-fdiagnostics-format=json. 2761 2762 Other significant improvements 2763 2764 * To allow inline expansion of both memcpy and memmove, the existing 2765 movmem instruction patterns used for non-overlapping memory copies 2766 have been renamed to cpymem. The movmem name is now used for 2767 overlapping memory moves, consistent with the library functions 2768 memcpy and memmove. 2769 * For many releases, when GCC emits a warning it prints the option 2770 controlling that warning. As of GCC 10, that option text is now a 2771 clickable hyperlink for the documentation of that option (assuming 2772 a [80]sufficiently capable terminal). This behavior can be 2773 controlled via a new [81]-fdiagnostics-urls option (along with 2774 various environment variables and heuristics documented with that 2775 option). 2776 2777 GCC 10.1 2778 2779 This is the [82]list of problem reports (PRs) from GCC's bug tracking 2780 system that are known to be fixed in the 10.1 release. This list might 2781 not be complete (that is, it is possible that some PRs that have been 2782 fixed are not listed here). 2783 2784 [83]GCC 10.2 2785 2786 This is the [84]list of problem reports (PRs) from GCC's bug tracking 2787 system that are known to be fixed in the 10.2 release. This list might 2788 not be complete (that is, it is possible that some PRs that have been 2789 fixed are not listed here). 2790 2791 [85]GCC 10.3 2792 2793 This is the [86]list of problem reports (PRs) from GCC's bug tracking 2794 system that are known to be fixed in the 10.3 release. This list might 2795 not be complete (that is, it is possible that some PRs that have been 2796 fixed are not listed here). 2797 2798 Target Specific Changes 2799 2800 AArch64 2801 2802 * A bug with the Random Number intrinsics in the arm_acle.h header 2803 that resulted in an incorrect status result being returned has been 2804 fixed. 2805 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 2806 options are -mcpu=a64fx and -mtune=a64fx respectively. In 2807 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 2808 tunes the code for the A64FX. This includes tuning the SVE code, 2809 although by default the code is still length-agnostic and so works 2810 for all SVE implementations. Adding -msve-vector-bits=512 makes the 2811 code specific to 512-bit SVE. 2812 2813 x86-64 2814 2815 * GCC 10.3 supports AMD CPUs based on the znver3 core via 2816 -march=znver3. 2817 2818 [87]GCC 10.4 2819 2820 This is the [88]list of problem reports (PRs) from GCC's bug tracking 2821 system that are known to be fixed in the 10.4 release. This list might 2822 not be complete (that is, it is possible that some PRs that have been 2823 fixed are not listed here). 2824 2825 Target Specific Changes 2826 2827 x86-64 2828 2829 * The x86-64 ABI of passing and returning structures with a 64-bit 2830 integer vector changed in GCC 10.1 when MMX is disabled. Disabling 2831 MMX no longer changes how they are passed nor returned. This ABI 2832 change is now diagnosed with -Wpsabi. 2833 2834 [89]GCC 10.5 2835 2836 This is the [90]list of problem reports (PRs) from GCC's bug tracking 2837 system that are known to be fixed in the 10.5 release. This list might 2838 not be complete (that is, it is possible that some PRs that have been 2839 fixed are not listed here). 2840 2841 2842 For questions related to the use of GCC, please consult these web 2843 pages and the [91]GCC manuals. If that fails, the 2844 [92]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2845 web pages and the development of GCC are welcome on our developer 2846 list at [93]gcc (a] gcc.gnu.org. All of [94]our lists have public 2847 archives. 2848 2849 Copyright (C) [95]Free Software Foundation, Inc. Verbatim copying and 2850 distribution of this entire article is permitted in any medium, 2851 provided this notice is preserved. 2852 2853 These pages are [96]maintained by the GCC team. Last modified 2854 2025-03-21. 2855 2856 References 2857 2858 1. https://gcc.gnu.org/gcc-10/porting_to.html 2859 2. https://gcc.gnu.org/onlinedocs/index.html#current 2860 3. https://gcc.gnu.org/gcc-10/changes.html#empty_base 2861 4. https://www.mpfr.org/ 2862 5. https://gcc.gnu.org/gcc-10/changes.html#iterator_base 2863 6. https://gcc.gnu.org/gcc-10/changes.html#general 2864 7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin 2865 8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce 2866 9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training 2867 10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible 2868 11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate 2869 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path 2870 13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html 2871 14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions 2872 15. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html 2873 16. https://facebook.github.io/zstd/ 2874 17. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values 2875 18. https://gcc.gnu.org/gcc-10/changes.html#languages 2876 19. https://www.openacc.org/ 2877 20. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-10 2878 21. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1 2879 22. https://www.openmp.org/specifications/ 2880 23. https://gcc.gnu.org/wiki/Offloading 2881 24. https://gcc.gnu.org/gcc-10/changes.html#c-family 2882 25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare 2883 26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds 2884 27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds 2885 28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow 2886 29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict 2887 30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr 2888 31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow 2889 32. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion 2890 33. https://gcc.gnu.org/gcc-10/changes.html#c 2891 34. https://gcc.gnu.org/gcc-10/changes.html#cxx 2892 35. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags 2893 36. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags 2894 37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383 2895 38. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711 2896 39. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707 2897 40. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704 2898 41. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706 2899 42. https://gcc.gnu.org/gcc-10/changes.html#libstdcxx 2900 43. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285 2901 44. https://gcc.gnu.org/gcc-10/changes.html#d 2902 45. https://gcc.gnu.org/gcc-10/changes.html#fortran 2903 46. https://www.openmp.org/specifications/ 2904 47. https://gcc.gnu.org/gcc-10/changes.html#go 2905 48. https://gcc.gnu.org/gcc-10/changes.html#jit 2906 49. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major 2907 50. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor 2908 51. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel 2909 52. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield 2910 53. https://gcc.gnu.org/gcc-10/changes.html#targets 2911 54. https://gcc.gnu.org/gcc-10/changes.html#arm-targets 2912 55. https://gcc.gnu.org/gcc-10/changes.html#aarch64 2913 56. https://gcc.gnu.org/gcc-10/changes.html#arm 2914 57. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics 2915 58. https://developer.arm.com/Architectures/M-Profile%20Architecture 2916 59. https://developer.arm.com/architectures/instruction-sets/intrinsics/ 2917 60. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension 2918 61. https://gcc.gnu.org/gcc-10/changes.html#amdgcn 2919 62. https://gcc.gnu.org/gcc-10/changes.html#arc 2920 63. https://gcc.gnu.org/gcc-10/changes.html#avr 2921 64. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs 2922 65. https://gcc.gnu.org/install/configure.html#avr 2923 66. https://gcc.gnu.org/wiki/avr-gcc#Libf7 2924 67. https://gcc.gnu.org/gcc-10/changes.html#x86 2925 68. https://gcc.gnu.org/gcc-10/changes.html#mips 2926 69. https://gcc.gnu.org/gcc-10/changes.html#powerpc 2927 70. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture 2928 71. https://gcc.gnu.org/gcc-10/changes.html#pru 2929 72. https://gcc.gnu.org/gcc-10/changes.html#riscv 2930 73. https://gcc.gnu.org/gcc-10/changes.html#v850 2931 74. https://gcc.gnu.org/gcc-10/changes.html#os 2932 75. https://gcc.gnu.org/gcc-10/changes.html#plugins 2933 76. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format 2934 77. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2935 78. https://cwe.mitre.org/ 2936 79. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2937 80. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda 2938 81. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls 2939 82. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0 2940 83. https://gcc.gnu.org/gcc-10/changes.html#GCC10.2 2941 84. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2 2942 85. https://gcc.gnu.org/gcc-10/changes.html#GCC10.3 2943 86. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3 2944 87. https://gcc.gnu.org/gcc-10/changes.html#GCC10.4 2945 88. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4 2946 89. https://gcc.gnu.org/gcc-10/changes.html#GCC10.5 2947 90. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.5 2948 91. https://gcc.gnu.org/onlinedocs/ 2949 92. mailto:gcc-help (a] gcc.gnu.org 2950 93. mailto:gcc (a] gcc.gnu.org 2951 94. https://gcc.gnu.org/lists.html 2952 95. https://www.fsf.org/ 2953 96. https://gcc.gnu.org/about.html 2954 ====================================================================== 2955 http://gcc.gnu.org/gcc-9/index.html 2956 2957 GCC 9 Release Series 2958 2959 (This release series is no longer supported.) 2960 2961 May 27, 2022 2962 2963 The [1]GNU project and the GCC developers are pleased to announce the 2964 release of GCC 9.5. 2965 2966 This release is a bug-fix release, containing fixes for regressions in 2967 GCC 9.4 relative to previous releases of GCC. 2968 2969 Release History 2970 2971 GCC 9.5 2972 May 27, 2022 ([2]changes, [3]documentation) 2973 2974 GCC 9.4 2975 June 1, 2021 ([4]changes, [5]documentation) 2976 2977 GCC 9.3 2978 Mar 12, 2020 ([6]changes, [7]documentation) 2979 2980 GCC 9.2 2981 Aug 12, 2019 ([8]changes, [9]documentation) 2982 2983 GCC 9.1 2984 May 3, 2019 ([10]changes, [11]documentation) 2985 2986 References and Acknowledgements 2987 2988 GCC used to stand for the GNU C Compiler, but since the compiler 2989 supports several other languages aside from C, it now stands for the 2990 GNU Compiler Collection. 2991 2992 The GCC developers would like to thank the numerous people that have 2993 contributed new features, improvements, bug fixes, and other changes as 2994 well as test results to GCC. This [12]amazing group of volunteers is 2995 what makes GCC successful. 2996 2997 For additional information about GCC please refer to the [13]GCC 2998 project web site or contact the [14]GCC development mailing list. 2999 3000 To obtain GCC please use [15]our mirror sites or [16]our version 3001 control system. 3002 3003 3004 For questions related to the use of GCC, please consult these web 3005 pages and the [17]GCC manuals. If that fails, the 3006 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3007 web pages and the development of GCC are welcome on our developer 3008 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 3009 archives. 3010 3011 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 3012 distribution of this entire article is permitted in any medium, 3013 provided this notice is preserved. 3014 3015 These pages are [22]maintained by the GCC team. Last modified 3016 2024-05-30. 3017 3018 References 3019 3020 1. http://www.gnu.org/ 3021 2. https://gcc.gnu.org/gcc-9/changes.html 3022 3. https://gcc.gnu.org/onlinedocs/9.5.0/ 3023 4. https://gcc.gnu.org/gcc-9/changes.html 3024 5. https://gcc.gnu.org/onlinedocs/9.4.0/ 3025 6. https://gcc.gnu.org/gcc-9/changes.html 3026 7. https://gcc.gnu.org/onlinedocs/9.3.0/ 3027 8. https://gcc.gnu.org/gcc-9/changes.html 3028 9. https://gcc.gnu.org/onlinedocs/9.2.0/ 3029 10. https://gcc.gnu.org/gcc-9/changes.html 3030 11. https://gcc.gnu.org/onlinedocs/9.1.0/ 3031 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html 3032 13. https://gcc.gnu.org/index.html 3033 14. mailto:gcc (a] gcc.gnu.org 3034 15. https://gcc.gnu.org/mirrors.html 3035 16. https://gcc.gnu.org/git.html 3036 17. https://gcc.gnu.org/onlinedocs/ 3037 18. mailto:gcc-help (a] gcc.gnu.org 3038 19. mailto:gcc (a] gcc.gnu.org 3039 20. https://gcc.gnu.org/lists.html 3040 21. https://www.fsf.org/ 3041 22. https://gcc.gnu.org/about.html 3042 ====================================================================== 3043 http://gcc.gnu.org/gcc-9/changes.html 3044 3045 GCC 9 Release Series 3046 Changes, New Features, and Fixes 3047 3048 This page is a "brief" summary of some of the huge number of 3049 improvements in GCC 9. You may also want to check out our [1]Porting to 3050 GCC 9 page and the [2]full GCC documentation. 3051 3052 Caveats 3053 3054 * On Arm targets (arm*-*-*), [3]a bug in the implementation of the 3055 procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has 3056 been fixed: a structure containing a bit-field based on a 64-bit 3057 integral type and where no other element in a structure required 3058 64-bit alignment could be passed incorrectly to functions. This is 3059 an ABI change. If the option -Wpsabi is enabled (on by default) the 3060 compiler will emit a diagnostic note for code that might be 3061 affected. 3062 * Support for a number of older systems and recently unmaintained or 3063 untested target ports of GCC has been declared obsolete in GCC 9. 3064 Unless there is activity to revive them, the next release of GCC 3065 will have their sources permanently removed. 3066 The following ports for individual systems on particular 3067 architectures have been obsoleted: 3068 + Solaris 10 (*-*-solaris2.10). Details can be found in the 3069 [4]announcement. 3070 + Cell Broadband Engine SPU (spu*-*-*). Details can be found in 3071 the [5]announcement. 3072 * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause 3073 ABI incompatibilities with object files compiled with other 3074 versions of GCC. If the std::rotate algorithm is called with an 3075 empty range then it might cause a divide-by-zero error (as a SIGFPE 3076 signal) and crash. The change has been reverted for GCC 9.2.0 and 3077 future releases. For more details see [6]Bug 90920. The problem can 3078 be avoided by recompiling any objects that might call std::rotate 3079 with an empty range, so that the GCC 9.1.0 definition of 3080 std::rotate is not used. 3081 * The automatic template instantiation at link time ([7]-frepo) has 3082 been deprecated and will be removed in a future release. 3083 * The --with-default-libstdcxx-abi=gcc4-compatible configure option 3084 is broken in the 9.1 and 9.2 releases, producing a shared library 3085 with missing symbols (see [8]Bug 90361). As a workaround, configure 3086 without that option and build GCC as normal, then edit the 3087 installed <bits/c++config.h> headers to define the 3088 _GLIBCXX_USE_CXX11_ABI macro to 0. 3089 3090 [9]General Improvements 3091 3092 The following GCC command line options have been introduced or 3093 improved. 3094 * All command line options that take a byte-size argument accept 3095 64-bit integers as well as standard SI and IEC suffixes such as kb 3096 and KiB, MB and MiB, or GB and GiB denoting the corresponding 3097 multiples of bytes. See [10]Invoking GCC for more. 3098 * A new option [11]-flive-patching=[inline-only-static|inline-clone] 3099 generates code suitable for live patching. At the same time it 3100 provides multiple-level control over IPA optimizations. See the 3101 user guide for more details. 3102 * A new option, --completion, has been added to provide more fine 3103 option completion in a shell. It is intended to be used by 3104 Bash-completion. 3105 * GCC's diagnostics now print source code with a left margin showing 3106 line numbers, configurable with 3107 [12]-fno-diagnostics-show-line-numbers. 3108 GCC's diagnostics can also now label regions of the source code to 3109 show pertinent information, such as the types within an expression. 3110 $ g++ t.cc 3111 t.cc: In function 'int test(const shape&, const shape&)': 3112 t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl 3113 e>' and 'boxed_value<double>') 3114 14 | return (width(s1) * height(s1) 3115 | ~~~~~~~~~~~~~~~~~~~~~~ 3116 | | 3117 | boxed_value<[...]> 3118 15 | + width(s2) * height(s2)); 3119 | ^ ~~~~~~~~~~~~~~~~~~~~~~ 3120 | | 3121 | boxed_value<[...]> 3122 3123 These labels can be disabled via [13]-fno-diagnostics-show-labels. 3124 * A new option [14]-fdiagnostics-format=json has been introduced for 3125 emitting diagnostics in a machine-readable format. 3126 * The alignment-related options [15]-falign-functions, 3127 [16]-falign-labels, [17]-falign-loops, and [18]-falign-jumps 3128 received support for a secondary alignment (e.g. 3129 -falign-loops=n:m:n2:m2). 3130 * New pair of profiling options ([19]-fprofile-filter-files and 3131 [20]-fprofile-exclude-files) has been added. The options help to 3132 filter which source files are instrumented. 3133 * AddressSanitizer generates more compact redzones for automatic 3134 variables. That helps to reduce memory footprint of a sanitized 3135 binary. 3136 * Numerous improvements have been made to the output of 3137 [21]-fopt-info. 3138 Messages are now prefixed with optimized, missed, or note, rather 3139 than the old behavior of all being prefixed with note. 3140 The output from -fopt-info can now contain information on inlining 3141 decisions: 3142 $ g++ -c inline.cc -O2 -fopt-info-inline-all 3143 inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)( 3144 E)) [with T = char**; E = char*]/2. 3145 inline.cc:24:11: optimized: Inlining void foreach(T, T, void (*)(E)) [with T = 3146 char**; E = char*]/2 into int main(int, char**)/1. 3147 inline.cc:19:12: missed: not inlinable: void inline_me(char*)/0 -> int std::pu 3148 ts(const char*)/3, function body not available 3149 inline.cc:13:8: optimized: Inlined void inline_me(char*)/4 into int main(int, c 3150 har**)/1 which now has time 127.363637 and size 11, net change of +0. 3151 Unit growth for small function inlining: 16->16 (0%) 3152 3153 Inlined 2 calls, eliminated 1 functions 3154 3155 3156 The output from the vectorizer has been rationalized so that failed 3157 attempts to vectorize a loop are displayed in the form 3158 [LOOP-LOCATION]: couldn't vectorize this loop 3159 [PROBLEM-LOCATION]: because of [REASON] 3160 3161 rather than an exhaustive log of all decisions made by the 3162 vectorizer. For example: 3163 $ gcc -c v.c -O3 -fopt-info-all-vec 3164 v.c:7:3: missed: couldn't vectorize loop 3165 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 3166 mory"); 3167 v.c:3:6: note: vectorized 0 loops in function. 3168 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 3169 mory"); 3170 3171 The old behavior can be obtained via a new -internals suboption of 3172 -fopt-info. 3173 * A new option, [22]-fsave-optimization-record has been added, which 3174 writes a SRCFILE.opt-record.json.gz file describing the 3175 optimization decisions made by GCC. This is similar to the output 3176 of -fopt-info, but with additional metadata such as the inlining 3177 chain, and profile information (if available). 3178 * Inter-procedural propagation of stack alignment can now be 3179 controlled by [23]-fipa-stack-alignment. 3180 * Propagation of addressability, readonly, and writeonly flags on 3181 static variables can now be controlled by 3182 [24]-fipa-reference-addressable. 3183 3184 The following built-in functions have been introduced. 3185 * [25]__builtin_expect_with_probability to provide branch prediction 3186 probability hints to the optimizer. 3187 * [26]__builtin_has_attribute determines whether a function, type, or 3188 variable has been declared with some attribute. 3189 * [27]__builtin_speculation_safe_value can be used to help mitigate 3190 against unsafe speculative execution. 3191 3192 The following attributes have been introduced. 3193 * The [28]copy function attribute has been added. The attribute can 3194 also be applied to type definitions and to variable declarations. 3195 3196 A large number of improvements to code generation have been made, 3197 including but not limited to the following. 3198 * Switch expansion has been improved by using a different strategy 3199 (jump table, bit test, decision tree) for a subset of switch cases. 3200 * A linear function expression defined as a switch statement can be 3201 transformed by [29]-ftree-switch-conversion. For example: 3202 3203 int 3204 foo (int how) 3205 { 3206 switch (how) { 3207 case 2: how = 205; break; 3208 case 3: how = 305; break; 3209 case 4: how = 405; break; 3210 case 5: how = 505; break; 3211 case 6: how = 605; break; 3212 } 3213 return how; 3214 } 3215 3216 can be transformed into 100 * how + 5 (for values defined in the 3217 switch statement). 3218 * Inter-procedural optimization improvements: 3219 + Inliner defaults were tuned to better suit modern C++ 3220 codebases, especially when built with link time-optimizations. 3221 New parameters max-inline-insns-small, max-inline-insns-size, 3222 uninlined-function-insns, uninlined-function-time, 3223 uninlined-thunk-insns, and uninlined-thunk-time were added. 3224 + Hot/cold partitioning is now more precise and aggressive. 3225 + Improved scalability for very large translation units 3226 (especially when link-time optimizing large programs). 3227 * Profile driven optimization improvements: 3228 + [30]-fprofile-use now enables [31]-fversion-loops-for-strides, 3229 [32]-floop-interchange, [33]-floop-unroll-and-jam, 3230 [34]-ftree-loop-distribution. 3231 + Streaming of counter histograms was removed, which reduces the 3232 size of profile files. Histograms are computed on the fly with 3233 link-time optimization. 3234 + The parameter hot-bb-count-ws-permille was reduced from 999 to 3235 990 to account for more precise histograms. 3236 * Link-time optimization improvements: 3237 + Types are now simplified prior to streaming resulting in 3238 significant reductions of LTO object file sizes and link-time 3239 memory use as well as improvements of link-time parallelism. 3240 + The default number of partitions (--param lto-partitions) was 3241 increased from 32 to 128 enabling effective use of CPUs with 3242 more than 32 hyperthreads. --param 3243 lto-max-streaming-parallelism can now be used to control the 3244 number of streaming processes. 3245 + Warnings on C++ One Decl Rule violations (-Wodr) are now more 3246 informative and produce fewer redundant results. 3247 Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an 3248 8-core machine was reduced by about 5% compared to GCC 8.3, and the 3249 size of LTO object files by 7%. LTO link time improves by 11% on an 3250 8-core machine and scales significantly better for more parallel 3251 build environments. The serial stage of the link-time optimization 3252 is 28% faster consuming 20% less memory. The parallel stage now 3253 scales to up to 128 partitions rather than 32 and reduces memory 3254 use for every worker by 30%. 3255 3256 The following improvements to the gcov command-line utility have been 3257 made. 3258 * The gcov tool received a new option [35]--use-hotness-colors (-q) 3259 that can provide perf-like coloring of hot functions. 3260 * The gcov tool has changed its intermediate format to a new JSON 3261 format. 3262 3263 [36]New Languages and Language specific improvements 3264 3265 [37]OpenACC support in C, C++, and Fortran continues to be maintained 3266 and improved. Most of the OpenACC 2.5 specification is implemented. See 3267 the [38]implementation status section on the OpenACC wiki page for 3268 further information. 3269 3270 [39]C family 3271 3272 * Version 5.0 of the [40]OpenMP specification is now partially 3273 supported in the C and C++ compilers. For details which features of 3274 OpenMP 5.0 are and which are not supported in the GCC 9 release see 3275 [41]this mail. 3276 * New extensions: 3277 + [42]__builtin_convertvector built-in for vector conversions 3278 has been added. 3279 * New warnings: 3280 + [43]-Waddress-of-packed-member, enabled by default, warns 3281 about an unaligned pointer value from the address of a packed 3282 member of a struct or union. 3283 * Enhancements to existing warnings: 3284 + [44]-Warray-bounds detects more instances of out-of-bounds 3285 indices. 3286 + [45]-Wattribute-alias also detects attribute mismatches 3287 between alias declarations and their targets, in addition to 3288 mismatches between their types. 3289 + [46]-Wformat-overflow and [47]-Wformat-truncation have been 3290 extended to all formatted input/output functions (where 3291 applicable) and enhanced to detect a subset of instances of 3292 reading past the end of unterminated constant character arrays 3293 in %s directives. 3294 + [48]-Wmissing-attributes detects instances of missing function 3295 attributes on declarations of aliases and weak references. 3296 + [49]-Wstringop-truncation also detects a subset of instances 3297 of reading past the end of unterminated constant character 3298 arrays, 3299 * If a macro is used with the wrong argument count, the C and C++ 3300 front ends now show the definition of that macro via a note. 3301 * The spelling corrector now considers transposed letters, and the 3302 threshold for similarity has been tightened, to avoid nonsensical 3303 suggestions. 3304 3305 [50]C 3306 3307 * There is now experimental support for -std=c2x, to select support 3308 for the upcoming C2X revision of the ISO C standard. This standard 3309 is in the early stages of development and the only feature 3310 supported in GCC 9 is _Static_assert with a single argument 3311 (support for _Static_assert with two arguments was added in C11 and 3312 GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU 3313 extensions, and -Wc11-c2x-compat, to warn for uses of features 3314 added in C2X (such warnings are also enabled by use of -Wpedantic 3315 if not using -std=c2x or -std=gnu2x). 3316 * New warnings: 3317 + [51]-Wabsolute-value warns for calls to standard functions 3318 that compute the absolute value of an argument when a more 3319 appropriate standard function is available. For example, 3320 calling abs(3.14) triggers the warning because the appropriate 3321 function to call to compute the absolute value of a double 3322 argument is fabs. The option also triggers warnings when the 3323 argument in a call to such a function has an unsigned type. 3324 This warning can be suppressed with an explicit type cast and 3325 it is also enabled by -Wextra. 3326 3327 [52]C++ 3328 3329 * New warnings: 3330 + [53]-Wdeprecated-copy, implied by -Wextra, warns about the 3331 C++11 deprecation of implicitly declared copy constructor and 3332 assignment operator if one of them is user-provided. 3333 -Wdeprecated-copy-dtor also warns if the destructor is 3334 user-provided, as specified in C++11. 3335 + [54]-Winit-list-lifetime, on by default, warns about uses of 3336 std::initializer_list that are likely to result in a dangling 3337 pointer, such as returning or assigning from a temporary list. 3338 + [55]-Wredundant-move, implied by -Wextra, warns about 3339 redundant calls to std::move. 3340 + [56]-Wpessimizing-move, implied by -Wall, warns when a call to 3341 std::move prevents copy elision. 3342 + [57]-Wclass-conversion, on by default, warns when a conversion 3343 function will never be called due to the type it converts to. 3344 * The C++ front end has experimental support for some of the upcoming 3345 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 3346 including range-based for statements with initializer, default 3347 constructible and assignable stateless lambdas, lambdas in 3348 unevaluated contexts, language support for empty data members, 3349 allowing pack expansion in lambda init-capture, likely and unlikely 3350 attributes, class types in non-type template parameters, allowing 3351 virtual function calls in constant expressions, explicit(bool), 3352 std::is_constant_evaluated, nested inline namespaces, etc. For a 3353 full list of new features, see [58]the C++ status page. 3354 * The C++ front end now preserves source locations for literals, 3355 id-expression, and mem-initializer for longer. For example it is 3356 now able to pin-point the pertinent locations for bad 3357 initializations such as these 3358 $ g++ -c bad-inits.cc 3359 bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization 3360 10 | { 3, json::object }, 3361 | ~~~~~~^~~~~~ 3362 | | 3363 | json 3364 bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f 3365 permissive] 3366 14 | char buffers[3][5] = { "red", "green", "blue" }; 3367 | ^~~~~~~ 3368 bad-inits.cc: In constructor 'X::X()': 3369 bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv 3370 e] 3371 17 | X() : one(42), two(42), three(42) 3372 | ^~ 3373 | | 3374 | int 3375 3376 rather than emitting the error at the final closing parenthesis or 3377 brace. 3378 * Error-reporting of overload resolution has been special-cased to 3379 make the case of a single failed candidate easier to read. For 3380 example: 3381 $ g++ param-type-mismatch.cc 3382 param-type-mismatch.cc: In function 'int test(int, const char*, float)': 3383 param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char* 3384 *' 3385 8 | return foo::member_1 (first, second, third); 3386 | ^~~~~~ 3387 | | 3388 | const char* 3389 param-type-mismatch.cc:3:46: note: initializing argument 2 of 'static int foo: 3390 :member_1(int, const char**, float)' 3391 3 | static int member_1 (int one, const char **two, float three); 3392 | ~~~~~~~~~~~~~^~~ 3393 3394 highlights both the problematic argument, and the parameter that it 3395 can't be converted to. 3396 * Diagnostics involving binary operators now use color to distinguish 3397 the two operands, and label them separately (as per the example of 3398 source labelling above). 3399 * Diagnostics involving function calls now highlight the pertinent 3400 parameter of the declaration in more places. 3401 $ g++ bad-conversion.cc 3402 bad-conversion.cc: In function 'void caller()': 3403 bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*' 3404 9 | callee (0, false, 2); 3405 | ^~~~~ 3406 | | 3407 | bool 3408 bad-conversion.cc:3:19: note: initializing argument 2 of 'void callee(int, voi 3409 d*, int)' 3410 3 | void callee (int, void *, int) 3411 | ^~~~~~ 3412 3413 * The C++ front end's implementation of [59]-Wformat now shows 3414 precise locations within string literals, and underlines the 3415 pertinent arguments at bogus call sites (the C front end has been 3416 doing this since GCC 7). For example: 3417 $ g++ -c bad-printf.cc -Wall 3418 bad-printf.cc: In function 'void print_field(const char*, float, long int, long 3419 int)': 3420 bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type 3421 'int', but argument 3 has type 'long int' [-Wformat=] 3422 6 | printf ("%s: %*ld ", fieldname, column - width, value); 3423 | ~^~~ ~~~~~~~~~~~~~~ 3424 | | | 3425 | int long int 3426 bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b 3427 ut argument 4 has type 'double' [-Wformat=] 3428 6 | printf ("%s: %*ld ", fieldname, column - width, value); 3429 | ~~~^ ~~~~~ 3430 | | | 3431 | long int double 3432 | %*f 3433 3434 * The C++ front end has gained new fix-it hints for forgetting the 3435 return *this; needed by various C++ operators: 3436 $ g++ -c operator.cc 3437 operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr 3438 &)': 3439 operator.cc:7:3: warning: no return statement in function returning non-void [-W 3440 return-type] 3441 6 | m_ptr = other.m_ptr; 3442 +++ |+ return *this; 3443 7 | } 3444 | ^ 3445 3446 for when the compiler needs a typename: 3447 $ g++ -c template.cc 3448 template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i 3449 s a dependent scope 3450 3 | Traits::type type; 3451 | ^~~~~~ 3452 | typename 3453 3454 when trying to use an accessor member as if it were a data member: 3455 $ g++ -c fncall.cc 3456 fncall.cc: In function 'void hangman(const mystring&)': 3457 fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length 3458 () const' (did you forget the '()' ?) 3459 12 | if (str.get_length > 0) 3460 | ~~~~^~~~~~~~~~ 3461 | () 3462 3463 for C++11's scoped enums: 3464 $ g++ -c enums.cc 3465 enums.cc: In function 'void json::test(const json::value&)': 3466 enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js 3467 on::kind::STRING'? 3468 12 | if (v.get_kind () == STRING) 3469 | ^~~~~~ 3470 | json::kind::STRING 3471 enums.cc:3:44: note: 'json::kind::STRING' declared here 3472 3 | enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ }; 3473 | ^~~~~~ 3474 3475 and a tweak to integrate the suggestions about misspelled members 3476 with that for accessors: 3477 $ g++ -c accessor-fixit.cc 3478 accessor-fixit.cc: In function 'int test(t*)': 3479 accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m 3480 ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const') 3481 17 | return ptr->ratio; 3482 | ^~~~~ 3483 | get_ratio() 3484 3485 In addition, various diagnostics in the C++ front-end have been 3486 streamlined by consolidating the suggestion into the initial error, 3487 rather than emitting a follow-up note: 3488 $ g++ typo.cc 3489 typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF 3490 _SIZE'? 3491 5 | uint8_t buf[BUFSIZE]; 3492 | ^~~~~~~ 3493 | BUF_SIZE 3494 3495 [60]Runtime Library (libstdc++) 3496 3497 * Improved support for C++17, including: 3498 + The C++17 implementation is no longer experimental. 3499 + Parallel algorithms and <execution> (requires [61]Thread 3500 Building Blocks 2018 or newer). 3501 + <memory_resource>. 3502 + Using the types and functions in <filesystem> does not require 3503 linking with -lstdc++fs now. 3504 * Improved experimental support for C++2a, including: 3505 + Type traits std::remove_cvref, std::unwrap_reference, 3506 std::unwrap_decay_ref, std::is_nothrow_convertible, and 3507 std::type_identity. 3508 + Headers <bit> and <version>. 3509 + Uniform container erasure (std::erase_if). 3510 + contains member of maps and sets. 3511 + String prefix and suffix checking (starts_with, ends_with). 3512 + Functions std::midpoint and std::lerp for interpolation. 3513 + std::bind_front. 3514 + std::visit<R>. 3515 + std::assume_aligned. 3516 + Uses-allocator construction utilities. 3517 + std::pmr::polymorphic_allocator<std::byte>. 3518 + Library support for char8_t type. 3519 + Destroying delete. 3520 + std::is_constant_evaluated() function. 3521 * Support for opening file streams with wide character paths on 3522 Windows 3523 * Incomplete support for the C++17 Filesystem library and the 3524 Filesystem TS on Windows. 3525 * Incomplete, experimental support for the Networking TS. 3526 3527 [62]D 3528 3529 * Support for the D programming language has been added to GCC, 3530 implementing version 2.076 of the language and run-time library. 3531 3532 [63]Fortran 3533 3534 * Asynchronous I/O is now fully supported. The program needs to be 3535 linked against the pthreads library to use it, otherwise the I/O is 3536 done synchronously. For systems which do not support POSIX 3537 condition variables, such as AIX, all I/O is still done 3538 synchronously. 3539 * The BACK argument for MINLOC and MAXLOC has been implemented. 3540 * The FINDLOC intrinsic function has been implemented. 3541 * The IS_CONTIGUOUS intrinsic function has been implemented. 3542 * Direct access to the real and imaginary parts of a complex variable 3543 via c%re and c%im has been implemented. 3544 * Type parameter inquiry via str%len and a%kind has been implemented. 3545 * C descriptors and the ISO_Fortran_binding.h source file have been 3546 implemented. 3547 * The MAX and MIN intrinsics are no longer guaranteed to return any 3548 particular value in case one of the arguments is a NaN. Note that 3549 this conforms to the Fortran standard and to what other Fortran 3550 compilers do. If there is a need to handle that case in some 3551 specific way, one needs to explicitly check for NaN's before 3552 calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the 3553 intrinsic module IEEE_ARITHMETIC. 3554 * A new command-line option [64]-fdec-include, set also by the 3555 [65]-fdec option, has been added to increase compatibility with 3556 legacy code. With this option, an INCLUDE directive is also parsed 3557 as a statement, which allows the directive to be spread across 3558 multiple source lines with line continuations. 3559 * A new [66]BUILTIN directive, has been added. The purpose of the 3560 directive is to provide an API between the GCC compiler and the GNU 3561 C Library which would define vector implementations of math 3562 routines. 3563 3564 [67]Go 3565 3566 * GCC 9 provides a complete implementation of the Go 1.12.2 user 3567 packages. 3568 3569 [68]libgccjit 3570 3571 * The libgccjit API gained a new entry point: 3572 [69]gcc_jit_context_add_driver_option. 3573 3574 [70]New Targets and Target Specific Improvements 3575 3576 [71]AArch64 & Arm 3577 3578 * Support has been added for the following processors (GCC 3579 identifiers in parentheses): 3580 + Arm Cortex-A76 (cortex-a76). 3581 + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE 3582 (cortex-a76.cortex-a55). 3583 + Arm Neoverse N1 (neoverse-n1). 3584 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3585 options, for example: -mcpu=cortex-a76 or 3586 -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent 3587 target attributes and pragmas. 3588 * The Armv8.3-A complex number instructions are now supported via 3589 intrinsics when the option -march=armv8.3-a or equivalent is 3590 specified. For the half-precision floating-point variants of these 3591 instructions use the architecture extension flag +fp16, e.g. 3592 -march=armv8.3-a+fp16. 3593 The intrinsics are defined by the ACLE specification. 3594 * The Armv8.5-A architecture is now supported through the 3595 -march=armv8.5-a option. 3596 * The Armv8.5-A architecture also adds some security features that 3597 are optional to all older architecture versions. These are now 3598 supported and only affect the assembler. 3599 + Speculation Barrier instruction through the -march=armv8-a+sb 3600 option. 3601 + Execution and Data Prediction Restriction instructions through 3602 the -march=armv8-a+predres option. 3603 + Speculative Store Bypass Safe instruction through the 3604 -march=armv8-a+ssbs option. This does not require a compiler 3605 option for Arm and thus -march=armv8-a+ssbs is an 3606 AArch64-specific option. 3607 3608 AArch64 specific 3609 3610 * Support has been added for the Arm Neoverse E1 processor 3611 (-mcpu=neoverse-e1). 3612 * The AArch64 port now has support for stack clash protection using 3613 the [72]-fstack-clash-protection option. The probing interval/guard 3614 size can be set by using --param 3615 stack-clash-protection-guard-size=12|16. The value of this 3616 parameter must be in bytes represented as a power of two. The two 3617 supported values for this parameter are 12 (for a 4KiB size, 2^12) 3618 and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and 3619 can be changed at configure time using the flag 3620 --with-stack-clash-protection-guard-size=12|16. 3621 * The option -msign-return-address= has been deprecated. This has 3622 been replaced by the new -mbranch-protection= option. This new 3623 option can now be used to enable the return address signing as well 3624 as the new Branch Target Identification feature of Armv8.5-A 3625 architecture. For more information on the arguments accepted by 3626 this option, please refer to [73]AArch64-Options. 3627 * The following optional extensions to Armv8.5-A architecture are now 3628 supported and only affect the assembler. 3629 + Random Number Generation instructions through the 3630 -march=armv8.5-a+rng option. 3631 + Memory Tagging Extension through the -march=armv8.5-a+memtag 3632 option. 3633 3634 Arm specific 3635 3636 * Support for the deprecated Armv2 and Armv3 architectures and their 3637 variants has been removed. Their corresponding -march values and 3638 the -mcpu options that used these architectures have been removed. 3639 * Support for the Armv5 and Armv5E architectures (which have no known 3640 implementations) has been removed. Note that Armv5T, Armv5TE and 3641 Armv5TEJ architectures remain supported. 3642 * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using 3643 their respective -mcpu options. 3644 3645 [74]AMD GCN 3646 3647 * A new back end targeting AMD GCN GPUs has been contributed to GCC. 3648 The implementation is currently limited to compiling 3649 single-threaded, stand-alone programs. Future versions will add 3650 support for offloading multi-threaded kernels via OpenMP and 3651 OpenACC. The following devices are supported (GCC identifiers in 3652 parentheses): 3653 + Fiji (fiji). 3654 + Vega 10 (gfx900). 3655 3656 [75]ARC 3657 3658 * LRA is now on by default for the ARC target. This can be controlled 3659 by -mlra. 3660 * Add support for frame code-density and branch-and-index 3661 instructions. 3662 3663 [76]C-SKY 3664 3665 * A new back end targeting C-SKY V2 processors has been contributed 3666 to GCC. 3667 3668 [77]IA-32/x86-64 3669 3670 * Support of Intel MPX (Memory Protection Extensions) has been 3671 removed. 3672 * New ISA extension support for Intel PTWRITE was added to GCC. 3673 PTWRITE intrinsics are available via the -mptwrite compiler switch. 3674 * GCC now supports the Intel CPU named Cascade Lake with AVX512 3675 extensions through -march=cascadelake. The switch enables the 3676 following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW, 3677 AVX512DQ, AVX512VNNI. 3678 3679 [78]MIPS 3680 3681 * The Loongson loongson-mmi and loongson-ext extensions have been 3682 split from loongson3a: 3683 + loongson-mmi contains the Loongson MMI (MultiMedia extensions 3684 Instructions). 3685 + loongson-ext contains the Loongson EXT (EXTensions 3686 instructions). 3687 * The Loongson EXT2 (EXTensions R2 instructions) are now supported. 3688 + loongson-ext2 contains the Loongson EXT2 instructions. 3689 Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and 3690 -m[no-]loongson-ext2 enable or disable those extensions. 3691 * Support has been added for the following processors (GCC 3692 identifiers in parentheses): 3693 + Loongson 3A1000 (gs464) which enables loongson-mmi, 3694 loongson-ext by default. 3695 + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi, 3696 loongson-ext, loongson-ext2 by default. 3697 + Loongson 2K1000 (gs264e) which enables loongson-ext, 3698 loongson-ext2, msa by default. 3699 The GCC identifiers can be used as arguments to the -mcpu and 3700 -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments 3701 to the equivalent target attributes and pragmas. 3702 3703 [79]OpenRISC 3704 3705 * A new back end targeting OpenRISC processors has been contributed 3706 to GCC. 3707 3708 [80]S/390, System z, IBM z Systems 3709 3710 * Support for the arch13 architecture has been added. When using the 3711 -march=arch13 option, the compiler will generate code making use of 3712 the new instructions introduced with the vector enhancement 3713 facility 2 and the miscellaneous instruction extension facility 2. 3714 The -mtune=arch13 option enables arch13 specific instruction 3715 scheduling without making use of new instructions. 3716 * Builtins for the new vector instructions have been added and can be 3717 enabled using the -mzvector option. 3718 * Support for ESA architecture machines g5 and g6 is deprecated since 3719 GCC 6.1.0 and has been removed now. 3720 * When compiling with -march=z14 or higher GCC emits alignments hints 3721 on the vector load/store instructions (8 or 16 byte). 3722 * Functions now have a default alignment of 16 bytes. This helps with 3723 branch prediction effects. 3724 * -mfentry is now supported. As well as the mcount mechanism the 3725 __fentry__ is called before the function prologue. However, since 3726 just a single instruction is required to call __fentry__ the call 3727 sequence imposes a smaller overhead than mcount (4 instructions). 3728 The produced code is compatible only with newer glibc versions, 3729 which provide the __fentry__ symbol and do not clobber r0 when 3730 resolving lazily bound functions. -mfentry is only supported when 3731 generating 64-bit code and does not work with nested C functions. 3732 * The -mnop-mcount option can be used to emit NOP instructions 3733 instead of an mcount or fentry call stub. 3734 * With the -mrecord-mcount option a __mcount_loc section is generated 3735 containing pointers to each profiling call stub. This is useful for 3736 automatically patching in and out calls. 3737 3738 [81]Operating Systems 3739 3740 [82]Solaris 3741 3742 * g++ now unconditionally enables large file support when compiling 3743 32-bit code. 3744 * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has 3745 been merged from LLVM. For the moment, this only works for 32-bit 3746 code on both SPARC and x86. 3747 * An initial port of the D runtime library has been completed on 3748 Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC 3749 support is still work-in-progress. 3750 3751 [83]Windows 3752 3753 * A C++ Microsoft ABI bitfield layout bug, [84]PR87137 has been 3754 fixed. A non-field declaration could cause the current bitfield 3755 allocation unit to be completed, incorrectly placing a following 3756 bitfield into a new allocation unit. The Microsoft ABI is selected 3757 for: 3758 + Mingw targets 3759 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 3760 option is specified, or __attribute__((ms_struct)) is used 3761 + SuperH targets when the -mhitachi option is specified, or 3762 __attribute__((renesas)) is used 3763 3764 [85]Improvements for plugin authors 3765 3766 * GCC's diagnostic subsystem now has a way to logically group 3767 together related diagnostics, auto_diagnostic_group. Such 3768 diagnostics will be nested by the output of 3769 [86]-fdiagnostics-format=json. 3770 * GCC now has a set of [87]user experience guidelines for GCC, with 3771 information and advice on implementing new diagnostics. 3772 3773 Other significant improvements 3774 3775 * GCC's internal "selftest" suite now runs for C++ as well as C (in 3776 debug builds of the compiler). 3777 3778 [88]GCC 9.1 3779 3780 This is the [89]list of problem reports (PRs) from GCC's bug tracking 3781 system that are known to be fixed in the 9.1 release. This list might 3782 not be complete (that is, it is possible that some PRs that have been 3783 fixed are not listed here). 3784 3785 [90]GCC 9.2 3786 3787 This is the [91]list of problem reports (PRs) from GCC's bug tracking 3788 system that are known to be fixed in the 9.2 release. This list might 3789 not be complete (that is, it is possible that some PRs that have been 3790 fixed are not listed here). 3791 3792 [92]GCC 9.3 3793 3794 This is the [93]list of problem reports (PRs) from GCC's bug tracking 3795 system that are known to be fixed in the 9.3 release. This list might 3796 not be complete (that is, it is possible that some PRs that have been 3797 fixed are not listed here). 3798 3799 [94]GCC 9.4 3800 3801 This is the [95]list of problem reports (PRs) from GCC's bug tracking 3802 system that are known to be fixed in the 9.4 release. This list might 3803 not be complete (that is, it is possible that some PRs that have been 3804 fixed are not listed here). 3805 3806 Target Specific Changes 3807 3808 AArch64 3809 3810 * The option -moutline-atomics has been added to aid deployment of 3811 the Large System Extensions (LSE) on GNU/Linux systems built with a 3812 baseline architecture targeting Armv8-A. When the option is 3813 specified code is emitted to detect the presence of LSE 3814 instructions at run time and use them for standard atomic 3815 operations. For more information please refer to the documentation. 3816 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 3817 options are -mcpu=a64fx and -mtune=a64fx respectively. In 3818 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 3819 tunes the code for the A64FX. This includes tuning the SVE code, 3820 although by default the code is still length-agnostic and so works 3821 for all SVE implementations. Adding -msve-vector-bits=512 makes the 3822 code specific to 512-bit SVE. 3823 3824 [96]GCC 9.5 3825 3826 This is the [97]list of problem reports (PRs) from GCC's bug tracking 3827 system that are known to be fixed in the 9.5 release. This list might 3828 not be complete (that is, it is possible that some PRs that have been 3829 fixed are not listed here). 3830 3831 3832 For questions related to the use of GCC, please consult these web 3833 pages and the [98]GCC manuals. If that fails, the 3834 [99]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3835 web pages and the development of GCC are welcome on our developer 3836 list at [100]gcc (a] gcc.gnu.org. All of [101]our lists have public 3837 archives. 3838 3839 Copyright (C) [102]Free Software Foundation, Inc. Verbatim copying and 3840 distribution of this entire article is permitted in any medium, 3841 provided this notice is preserved. 3842 3843 These pages are [103]maintained by the GCC team. Last modified 3844 2025-01-31. 3845 3846 References 3847 3848 1. https://gcc.gnu.org/gcc-9/porting_to.html 3849 2. https://gcc.gnu.org/onlinedocs/index.html#current 3850 3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469 3851 4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html 3852 5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html 3853 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920 3854 7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo 3855 8. https://gcc.gnu.org/PR90361 3856 9. https://gcc.gnu.org/gcc-9/changes.html#general 3857 10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC 3858 11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching 3859 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers 3860 13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels 3861 14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 3862 15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions 3863 16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels 3864 17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops 3865 18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps 3866 19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files 3867 20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files 3868 21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info 3869 22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record 3870 23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment 3871 24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable 3872 25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability 3873 26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1 3874 27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1 3875 28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute 3876 29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion 3877 30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use 3878 31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides 3879 32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange 3880 33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam 3881 34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution 3882 35. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov 3883 36. https://gcc.gnu.org/gcc-9/changes.html#languages 3884 37. https://www.openacc.org/ 3885 38. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-9 3886 39. https://gcc.gnu.org/gcc-9/changes.html#c-family 3887 40. https://www.openmp.org/specifications/ 3888 41. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html 3889 42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector 3890 43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member 3891 44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds 3892 45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias 3893 46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow 3894 47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation 3895 48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 3896 49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 3897 50. https://gcc.gnu.org/gcc-9/changes.html#c 3898 51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value 3899 52. https://gcc.gnu.org/gcc-9/changes.html#cxx 3900 53. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy 3901 54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime 3902 55. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 3903 56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 3904 57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion 3905 58. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 3906 59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat 3907 60. https://gcc.gnu.org/gcc-9/changes.html#libstdcxx 3908 61. https://github.com/uxlfoundation/oneTBB 3909 62. https://gcc.gnu.org/gcc-9/changes.html#d 3910 63. https://gcc.gnu.org/gcc-9/changes.html#fortran 3911 64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include 3912 65. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec 3913 66. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive 3914 67. https://gcc.gnu.org/gcc-9/changes.html#go 3915 68. https://gcc.gnu.org/gcc-9/changes.html#jit 3916 69. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option 3917 70. https://gcc.gnu.org/gcc-9/changes.html#targets 3918 71. https://gcc.gnu.org/gcc-9/changes.html#arm-targets 3919 72. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector 3920 73. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options 3921 74. https://gcc.gnu.org/gcc-9/changes.html#amdgcn 3922 75. https://gcc.gnu.org/gcc-9/changes.html#arc 3923 76. https://gcc.gnu.org/gcc-9/changes.html#csky 3924 77. https://gcc.gnu.org/gcc-9/changes.html#x86 3925 78. https://gcc.gnu.org/gcc-9/changes.html#mips 3926 79. https://gcc.gnu.org/gcc-9/changes.html#or1k 3927 80. https://gcc.gnu.org/gcc-9/changes.html#s390 3928 81. https://gcc.gnu.org/gcc-9/changes.html#os 3929 82. https://gcc.gnu.org/gcc-9/changes.html#solaris 3930 83. https://gcc.gnu.org/gcc-9/changes.html#windows 3931 84. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 3932 85. https://gcc.gnu.org/gcc-9/changes.html#plugins 3933 86. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 3934 87. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html 3935 88. https://gcc.gnu.org/gcc-9/changes.html#GCC9.1 3936 89. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0 3937 90. https://gcc.gnu.org/gcc-9/changes.html#GCC9.2 3938 91. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2 3939 92. https://gcc.gnu.org/gcc-9/changes.html#GCC9.3 3940 93. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3 3941 94. https://gcc.gnu.org/gcc-9/changes.html#GCC9.4 3942 95. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4 3943 96. https://gcc.gnu.org/gcc-9/changes.html#GCC9.5 3944 97. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5 3945 98. https://gcc.gnu.org/onlinedocs/ 3946 99. mailto:gcc-help (a] gcc.gnu.org 3947 100. mailto:gcc (a] gcc.gnu.org 3948 101. https://gcc.gnu.org/lists.html 3949 102. https://www.fsf.org/ 3950 103. https://gcc.gnu.org/about.html 3951 ====================================================================== 3952 http://gcc.gnu.org/gcc-8/index.html 3953 3954 GCC 8 Release Series 3955 3956 (This release series is no longer supported.) 3957 3958 May 14, 2021 3959 3960 The [1]GNU project and the GCC developers are pleased to announce the 3961 release of GCC 8.5. 3962 3963 This release is a bug-fix release, containing fixes for regressions in 3964 GCC 8.4 relative to previous releases of GCC. 3965 3966 Release History 3967 3968 GCC 8.5 3969 May 14, 2021 ([2]changes, [3]documentation) 3970 3971 GCC 8.4 3972 Mar 4, 2020 ([4]changes, [5]documentation) 3973 3974 GCC 8.3 3975 Feb 22, 2019 ([6]changes, [7]documentation) 3976 3977 GCC 8.2 3978 Jul 26, 2018 ([8]changes, [9]documentation) 3979 3980 GCC 8.1 3981 May 2, 2018 ([10]changes, [11]documentation) 3982 3983 References and Acknowledgements 3984 3985 GCC used to stand for the GNU C Compiler, but since the compiler 3986 supports several other languages aside from C, it now stands for the 3987 GNU Compiler Collection. 3988 3989 A list of [12]successful builds is updated as new information becomes 3990 available. 3991 3992 The GCC developers would like to thank the numerous people that have 3993 contributed new features, improvements, bug fixes, and other changes as 3994 well as test results to GCC. This [13]amazing group of volunteers is 3995 what makes GCC successful. 3996 3997 For additional information about GCC please refer to the [14]GCC 3998 project web site or contact the [15]GCC development mailing list. 3999 4000 To obtain GCC please use [16]our mirror sites or [17]our version 4001 control system. 4002 4003 4004 For questions related to the use of GCC, please consult these web 4005 pages and the [18]GCC manuals. If that fails, the 4006 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4007 web pages and the development of GCC are welcome on our developer 4008 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4009 archives. 4010 4011 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4012 distribution of this entire article is permitted in any medium, 4013 provided this notice is preserved. 4014 4015 These pages are [23]maintained by the GCC team. Last modified 4016 2024-05-30. 4017 4018 References 4019 4020 1. http://www.gnu.org/ 4021 2. https://gcc.gnu.org/gcc-8/changes.html 4022 3. https://gcc.gnu.org/onlinedocs/8.5.0/ 4023 4. https://gcc.gnu.org/gcc-8/changes.html 4024 5. https://gcc.gnu.org/onlinedocs/8.4.0/ 4025 6. https://gcc.gnu.org/gcc-8/changes.html 4026 7. https://gcc.gnu.org/onlinedocs/8.3.0/ 4027 8. https://gcc.gnu.org/gcc-8/changes.html 4028 9. https://gcc.gnu.org/onlinedocs/8.2.0/ 4029 10. https://gcc.gnu.org/gcc-8/changes.html 4030 11. https://gcc.gnu.org/onlinedocs/8.1.0/ 4031 12. https://gcc.gnu.org/gcc-8/buildstat.html 4032 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 4033 14. https://gcc.gnu.org/index.html 4034 15. mailto:gcc (a] gcc.gnu.org 4035 16. https://gcc.gnu.org/mirrors.html 4036 17. https://gcc.gnu.org/git.html 4037 18. https://gcc.gnu.org/onlinedocs/ 4038 19. mailto:gcc-help (a] gcc.gnu.org 4039 20. mailto:gcc (a] gcc.gnu.org 4040 21. https://gcc.gnu.org/lists.html 4041 22. https://www.fsf.org/ 4042 23. https://gcc.gnu.org/about.html 4043 ====================================================================== 4044 http://gcc.gnu.org/gcc-8/changes.html 4045 4046 GCC 8 Release Series 4047 Changes, New Features, and Fixes 4048 4049 This page is a "brief" summary of some of the huge number of 4050 improvements in GCC 8. You may also want to check out our [1]Porting to 4051 GCC 8 page and the [2]full GCC documentation. 4052 4053 Caveats 4054 4055 * The default mode for C is now -std=gnu17 instead of -std=gnu11. 4056 * Support for the obsolete SDB/coff debug info format has been 4057 removed. The option -gcoff no longer does anything. 4058 * The Cilk+ extensions to the C and C++ languages have been removed. 4059 * The MPX extensions to the C and C++ languages have been deprecated 4060 and will be removed in a future release. 4061 * The extension allowing arithmetic on std::atomic<void*> and types 4062 like std::atomic<R(*)()> has been deprecated. 4063 * The non-standard C++0x std::copy_exception function was removed. 4064 std::make_exception_ptr should be used instead. 4065 * Support for the powerpc*-*-*spe* target ports which have been 4066 recently unmaintained and untested in GCC has been declared 4067 obsolete in GCC 8 as [3]announced. Unless there is activity to 4068 revive them, the next release of GCC will have their sources 4069 permanently removed. 4070 4071 [4]General Improvements 4072 4073 * Inter-procedural optimization improvements: 4074 + Reworked run-time estimation metrics leading to more realistic 4075 guesses driving inliner and cloning heuristics. 4076 + The ipa-pure-const pass is extended to propagate the malloc 4077 attribute, and the corresponding warning option 4078 -Wsuggest-attribute=malloc emits a diagnostic for functions 4079 which can be annotated with the malloc attribute. 4080 * Profile driven optimization improvements: 4081 + New infrastructure for representing profiles (both statically 4082 guessed and profile feedback) which allows propagation of 4083 additional information about the reliability of the profile. 4084 + A number of improvements in the profile updating code solving 4085 problems found by new verification code. 4086 + Static detection of code which is not executed in a valid run 4087 of the program. This includes paths which trigger undefined 4088 behavior as well as calls to functions declared with the cold 4089 attribute. Newly the noreturn attribute does not imply all 4090 effects of cold to differentiate between exit (which is 4091 noreturn) and abort (which is in addition not executed in 4092 valid runs). 4093 + -freorder-blocks-and-partition, a pass splitting function 4094 bodies into hot and cold regions, is now enabled by default at 4095 -O2 and higher for x86 and x86-64. 4096 * Link-time optimization improvements: 4097 + We have significantly improved debug information on ELF 4098 targets using DWARF by properly preserving language-specific 4099 information. This allows for example the libstdc++ 4100 pretty-printers to work with LTO optimized executables. 4101 * A new option -fcf-protection=[full|branch|return|none] is 4102 introduced to perform code instrumentation to increase program 4103 security by checking that target addresses of control-flow transfer 4104 instructions (such as indirect function call, function return, 4105 indirect jump) are valid. Currently the instrumentation is 4106 supported on x86 GNU/Linux targets only. See the user guide for 4107 further information about the option syntax and section "New 4108 Targets and Target Specific Improvements" for IA-32/x86-64 for more 4109 details. 4110 * The -gcolumn-info option is now enabled by default. It includes 4111 column information in addition to just filenames and line numbers 4112 in DWARF debugging information. 4113 * The polyhedral-based loop nest optimization pass 4114 -floop-nest-optimize has been overhauled. It's still considered 4115 experimental and may not result in any runtime improvements. 4116 * Two new classical loop nest optimization passes have been added. 4117 -floop-unroll-and-jam performs outer loop unrolling and fusing of 4118 the inner loop copies. -floop-interchange exchanges loops in a loop 4119 nest to improve data locality. Both passes are enabled by default 4120 at -O3 and above. 4121 * The classic loop nest optimization pass -ftree-loop-distribution 4122 has been improved and enabled by default at -O3 and above. It 4123 supports loop nest distribution in some restricted scenarios; it 4124 also supports cancellable innermost loop distribution with loop 4125 versioning under run-time alias checks. 4126 * The new option -fstack-clash-protection causes the compiler to 4127 insert probes whenever stack space is allocated statically or 4128 dynamically to reliably detect stack overflows and thus mitigate 4129 the attack vector that relies on jumping over a stack guard page as 4130 provided by the operating system. 4131 * A new pragma GCC unroll has been implemented in the C family of 4132 languages, as well as Fortran and Ada, so as to make it possible 4133 for the user to have a finer-grained control over the loop 4134 unrolling optimization. 4135 * GCC has been enhanced to detect more instances of meaningless or 4136 mutually exclusive attribute specifications and handle such 4137 conflicts more consistently. Mutually exclusive attribute 4138 specifications are ignored with a warning regardless of whether 4139 they appear on the same declaration or on distinct declarations of 4140 the same entity. For example, because the noreturn attribute on the 4141 second declaration below is mutually exclusive with the malloc 4142 attribute on the first, it is ignored and a warning is issued. 4143 > 4144 void* __attribute__ ((malloc)) f (unsigned); 4145 void* __attribute__ ((noreturn)) f (unsigned); 4146 4147 warning: ignoring attribute 'noreturn' because it conflicts with attribute 4148 'malloc' [-Wattributes] 4149 * The gcov tool can distinguish functions that begin on a same line 4150 in a source file. This can be a different template instantiation or 4151 a class constructor: 4152 4153 File 'ins.C' 4154 Lines executed:100.00% of 8 4155 Creating 'ins.C.gcov' 4156 4157 -: 0:Source:ins.C 4158 -: 0:Graph:ins.gcno 4159 -: 0:Data:ins.gcda 4160 -: 0:Runs:1 4161 -: 0:Programs:1 4162 -: 1:template<class T> 4163 -: 2:class Foo 4164 -: 3:{ 4165 -: 4: public: 4166 2: 5: Foo(): b (1000) {} 4167 ------------------ 4168 Foo<char>::Foo(): 4169 1: 5: Foo(): b (1000) {} 4170 ------------------ 4171 Foo<int>::Foo(): 4172 1: 5: Foo(): b (1000) {} 4173 ------------------ 4174 2: 6: void inc () { b++; } 4175 ------------------ 4176 Foo<char>::inc(): 4177 1: 6: void inc () { b++; } 4178 ------------------ 4179 Foo<int>::inc(): 4180 1: 6: void inc () { b++; } 4181 ------------------ 4182 -: 7: 4183 -: 8: private: 4184 -: 9: int b; 4185 -: 10:}; 4186 -: 11: 4187 1: 12:int main(int argc, char **argv) 4188 -: 13:{ 4189 1: 14: Foo<int> a; 4190 1: 15: Foo<char> b; 4191 -: 16: 4192 1: 17: a.inc (); 4193 1: 18: b.inc (); 4194 1: 19:} 4195 4196 * The gcov tool has more accurate numbers for execution of lines in a 4197 source file. 4198 * The gcov tool can use TERM colors to provide more readable output. 4199 * AddressSanitizer gained a new pair of sanitization options, 4200 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 4201 warn about subtraction (or comparison) of pointers that point to a 4202 different memory object: 4203 4204 int 4205 main () 4206 { 4207 /* Heap allocated memory. */ 4208 char *heap1 = (char *)__builtin_malloc (42); 4209 char *heap2 = (char *)__builtin_malloc (42); 4210 if (heap1 > heap2) 4211 return 1; 4212 4213 return 0; 4214 } 4215 4216 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 4217 00050 4218 #0 0x40070f in main /tmp/pointer-compare.c:7 4219 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4220 #2 0x400629 in _start (/tmp/a.out+0x400629) 4221 4222 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 4223 00000003a) 4224 allocated by thread T0 here: 4225 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 4226 _malloc_linux.cc:86 4227 #1 0x4006ea in main /tmp/pointer-compare.c:5 4228 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4229 4230 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 4231 00000007a) 4232 allocated by thread T0 here: 4233 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 4234 _malloc_linux.cc:86 4235 #1 0x4006f8 in main /tmp/pointer-compare.c:6 4236 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4237 4238 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 4239 4240 * The store merging pass has been enhanced to handle bit-fields and 4241 not just constant stores, but also data copying from adjacent 4242 memory locations into other adjacent memory locations, including 4243 bitwise logical operations on the data. The pass can also handle 4244 byte swapping into memory locations. 4245 * The undefined behavior sanitizer gained two new options included in 4246 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 4247 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 4248 builtins, and -fsanitize=pointer-overflow which performs cheap run 4249 time tests for pointer wrapping. 4250 * A new attribute no_sanitize can be applied to functions to instruct 4251 the compiler not to do sanitization of the options provided as 4252 arguments to the attribute. Acceptable values for no_sanitize match 4253 those acceptable by the -fsanitize command-line option. 4254 4255 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 4256 f () { /* Do something. */; } 4257 4258 [5]New Languages and Language specific improvements 4259 4260 [6]Ada 4261 4262 * For its internal exception handling used on the host for error 4263 recovery in the front-end, the compiler now relies on the native 4264 exception handling mechanism of the host platform, which should be 4265 more efficient than the former mechanism. 4266 4267 [7]BRIG (HSAIL) 4268 4269 In this release cycle, the focus for the BRIGFE was on stabilization 4270 and performance improvements. Also a couple of completely new features 4271 were added. 4272 * Improved support for function and module scope group segment 4273 variables. PRM specs define function and module scope group segment 4274 variables as an experimental feature. However, PRM test suite uses 4275 them. Now group segment is handled by separate book keeping of 4276 module scope and function (kernel) offsets. Each function has a 4277 "frame" in the group segment offset to which is given as an 4278 argument, similar to traditional call stack frame handling. 4279 * Reduce the number of type conversions due to the untyped HSAIL 4280 registers. Instead of always representing the HSAIL's untyped 4281 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 4282 code and builds the register variables as a type used the most when 4283 storing or reading data to/from each register. This reduces the 4284 number of total casts which cannot be always optimized away. 4285 * Support for BRIG_KIND_NONE directives. 4286 * Made -O3 the default optimization level for BRIGFE. 4287 * Fixed illegal addresses generated from address expressions which 4288 refer only to offset 0. 4289 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 4290 mode, the offset is treated as 32-bit unless it's in global, 4291 read-only or kernarg address space. 4292 * Fixed a crash caused sometimes by calls with more than 4 arguments. 4293 * Fixed a mis-execution issue with kernels that have both unexpanded 4294 ID functions and calls to subfunctions. 4295 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 4296 avoid illegal optimizations. 4297 * Ensure per WI copies of private variables are aligned correctly. 4298 * libhsail-rt: Assume the host runtime allocates the work group 4299 memory. 4300 4301 [8]C family 4302 4303 * New command-line options have been added for the C and C++ 4304 compilers: 4305 + [9]-Wmultistatement-macros warns about unsafe macros expanding 4306 to multiple statements used as a body of a statement such as 4307 if, else, while, switch, or for. 4308 + [10]-Wstringop-truncation warns for calls to bounded string 4309 manipulation functions such as strncat, strncpy, and stpncpy 4310 that might either truncate the copied string or leave the 4311 destination unchanged. For example, the following call to 4312 strncat is diagnosed because it appends just three of the four 4313 characters from the source string. 4314 void append (char *buf, size_t bufsize) 4315 { 4316 strncat (buf, ".txt", 3); 4317 } 4318 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 4319 Wstringop-truncation] 4320 Similarly, in the following example, the call to strncpy 4321 specifies the size of the destination buffer as the bound. If 4322 the length of the source string is equal to or greater than 4323 this size the result of the copy will not be NUL-terminated. 4324 Therefore, the call is also diagnosed. To avoid the warning, 4325 specify sizeof buf - 1 as the bound and set the last element 4326 of the buffer to NUL. 4327 void copy (const char *s) 4328 { 4329 char buf[80]; 4330 strncpy (buf, s, sizeof buf); 4331 ... 4332 } 4333 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 4334 tion] 4335 The -Wstringop-truncation option is included in -Wall. 4336 Note that due to GCC bug [11]82944, defining strncat, strncpy, 4337 or stpncpy as a macro in a system header as some 4338 implementations do, suppresses the warning. 4339 + [12]-Wif-not-aligned controls warnings issued in response to 4340 invalid uses of objects declared with attribute 4341 [13]warn_if_not_aligned. 4342 The -Wif-not-aligned option is included in -Wall. 4343 + [14]-Wmissing-attributes warns when a declaration of a 4344 function is missing one or more attributes that a related 4345 function is declared with and whose absence may adversely 4346 affect the correctness or efficiency of generated code. For 4347 example, in C++, the warning is issued when an explicit 4348 specialization of a primary template declared with attribute 4349 alloc_align, alloc_size, assume_aligned, format, format_arg, 4350 malloc, or nonnull is declared without it. Attributes 4351 deprecated, error, and warning suppress the warning. 4352 The -Wmissing-attributes option is included in -Wall. 4353 + [15]-Wpacked-not-aligned warns when a struct or union declared 4354 with attribute packed defines a member with an explicitly 4355 specified alignment greater than 1. Such a member will wind up 4356 under-aligned. For example, a warning will be issued for the 4357 definition of struct A in the following: 4358 struct __attribute__ ((aligned (8))) 4359 S8 { char a[8]; }; 4360 4361 struct __attribute__ ((packed)) A 4362 { 4363 struct S8 s8; 4364 }; 4365 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 4366 The -Wpacked-not-aligned option is included in -Wall. 4367 + -Wcast-function-type warns when a function pointer is cast to 4368 an incompatible function pointer. This warning is enabled by 4369 -Wextra. 4370 + -Wsizeof-pointer-div warns for suspicious divisions of the 4371 size of a pointer by the size of the elements it points to, 4372 which looks like the usual way to compute the array size but 4373 won't work out correctly with pointers. This warning is 4374 enabled by -Wall. 4375 + -Wcast-align=strict warns whenever a pointer is cast such that 4376 the required alignment of the target is increased. For 4377 example, warn if a char * is cast to an int * regardless of 4378 the target machine. 4379 + -fprofile-abs-path creates absolute path names in the .gcno 4380 files. This allows gcov to find the correct sources in 4381 projects where compilations occur with different working 4382 directories. 4383 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 4384 signed integer overflow is now undefined by default at all 4385 optimization levels. Using -fsanitize=signed-integer-overflow is 4386 now the preferred way to audit code, -Wstrict-overflow is 4387 deprecated. 4388 * The [16]-Warray-bounds option has been improved to detect more 4389 instances of out-of-bounds array indices and pointer offsets. For 4390 example, negative or excessive indices into flexible array members 4391 and string literals are detected. 4392 * The [17]-Wrestrict option introduced in GCC 7 has been enhanced to 4393 detect many more instances of overlapping accesses to objects via 4394 restrict-qualified arguments to standard memory and string 4395 manipulation functions such as memcpy and strcpy. For example, the 4396 strcpy call in the function below attempts to truncate the string 4397 by replacing its initial characters with the last four. However, 4398 because the function writes the terminating NUL into a[4], the 4399 copies overlap and the call is diagnosed. 4400 void f (void) 4401 { 4402 char a[] = "abcd1234"; 4403 strcpy (a, a + 4); 4404 ... 4405 } 4406 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 4407 4 [-Wrestrict] 4408 The -Wrestrict option is included in -Wall. 4409 * Several optimizer enhancements have enabled improvements to the 4410 [18]-Wformat-overflow and [19]-Wformat-truncation options. The 4411 warnings detect more instances of buffer overflow and truncation 4412 than in GCC 7 and are better at avoiding certain kinds of false 4413 positives. 4414 * When reporting mismatching argument types at a function call, the C 4415 and C++ compilers now underline both the argument and the pertinent 4416 parameter in the declaration. 4417 $ gcc arg-type-mismatch.cc 4418 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 4419 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 4420 [-fpermissive] 4421 return callee(first, second, third); 4422 ^~~~~~ 4423 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 4424 onst char*, float)' 4425 extern int callee(int one, const char *two, float three); 4426 ~~~~~~~~~~~~^~~ 4427 4428 * When reporting on unrecognized identifiers, the C and C++ compilers 4429 will now emit fix-it hints suggesting #include directives for 4430 various headers in the C and C++ standard libraries. 4431 $ gcc incomplete.c 4432 incomplete.c: In function 'test': 4433 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 4434 return NULL; 4435 ^~~~ 4436 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 4437 t to '#include <stddef.h>'? 4438 incomplete.c:1:1: 4439 +#include <stddef.h> 4440 const char *test(void) 4441 incomplete.c:3:10: 4442 return NULL; 4443 ^~~~ 4444 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 4445 ch function it appears in 4446 4447 $ gcc incomplete.cc 4448 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 4449 std::string s("hello world"); 4450 ^~~~~~ 4451 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 4452 forget to '#include <string>'? 4453 +#include <string> 4454 std::string s("hello world"); 4455 ^~~ 4456 4457 * The C and C++ compilers now use more intuitive locations when 4458 reporting on missing semicolons, and offer fix-it hints: 4459 $ gcc t.c 4460 t.c: In function 'test': 4461 t.c:3:12: error: expected ';' before '}' token 4462 return 42 4463 ^ 4464 ; 4465 } 4466 ~ 4467 4468 * When reporting on missing '}' and ')' tokens, the C and C++ 4469 compilers will now highlight the corresponding '{' and '(' token, 4470 issuing a 'note' if it's on a separate line: 4471 $ gcc unclosed.c 4472 unclosed.c: In function 'log_when_out_of_range': 4473 unclosed.c:12:50: error: expected ')' before '{' token 4474 && (temperature < MIN || temperature > MAX) { 4475 ^~ 4476 ) 4477 unclosed.c:11:6: note: to match this '(' 4478 if (logging_enabled && check_range () 4479 ^ 4480 4481 or highlighting it directly if it's on the same line: 4482 $ gcc unclosed-2.c 4483 unclosed-2.c: In function 'test': 4484 unclosed-2.c:8:45: error: expected ')' before '{' token 4485 if (temperature < MIN || temperature > MAX { 4486 ~ ^~ 4487 ) 4488 4489 They will also emit fix-it hints. 4490 4491 [20]C 4492 4493 * New options -std=c17, to select support for the 2018 edition of the 4494 ISO C standard (__STDC_VERSION__ == 201710L), and -std=gnu17, for 4495 C17 with GNU extensions. 4496 * The default mode has been changed to -std=gnu17. 4497 4498 [21]C++ 4499 4500 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 4501 convention, which changes the ABI for some uncommon code: 4502 + Passing an empty class as an argument now takes up no space on 4503 x86_64, as required by the psABI. 4504 + Passing or returning a class with only deleted copy and move 4505 constructors now uses the same calling convention as a class 4506 with a non-trivial copy or move constructor. This only affects 4507 C++17 mode, as in earlier standards passing or returning such 4508 a class was impossible. 4509 + WARNING: In GCC 8.1 the second change mistakenly also affects 4510 classes with a deleted copy constructor and defaulted trivial 4511 move constructor (bug [22]c++/86094). This issue is fixed in 4512 GCC 8.2 (-fabi-version=13). 4513 You can test whether these changes affect your code with -Wabi=11 4514 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 4515 problematic for your project, the GCC 7 ABI can be selected with 4516 -fabi-version=11. 4517 * The value of the C++11 alignof operator has been corrected to match 4518 C _Alignof (minimum alignment) rather than GNU __alignof__ 4519 (preferred alignment); on ia32 targets this means that 4520 alignof(double) is now 4 rather than 8. Code that wants the 4521 preferred alignment should use __alignof__ instead. 4522 * New command-line options have been added for the C++ compiler to 4523 control warnings: 4524 + [23]-Wclass-memaccess warns when objects of non-trivial class 4525 types are manipulated in potentially unsafe ways by raw memory 4526 functions such as memcpy, or realloc. The warning helps detect 4527 calls that bypass user-defined constructors or copy-assignment 4528 operators, corrupt virtual table pointers, data members of 4529 const-qualified types or references, or member pointers. The 4530 warning also detects calls that would bypass access controls 4531 to data members. For example, a call such as: 4532 memcpy (&std::cout, &std::cerr, sizeof std::cout); 4533 results in 4534 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 4535 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 4536 o trivial copy-assignment [-Wclass-memaccess] 4537 The -Wclass-memaccess option is included in -Wall. 4538 * The C++ front end has experimental support for some of the upcoming 4539 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 4540 including designated initializers, default member initializers for 4541 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 4542 lambda [=, this] captures, etc. For a full list of new features, 4543 see [24]the C++ status page. 4544 * When reporting on attempts to access private fields of a class or 4545 struct, the C++ compiler will now offer fix-it hints showing how to 4546 use an accessor function to get at the field in question, if one 4547 exists. 4548 $ gcc accessor.cc 4549 accessor.cc: In function 'void test(foo*)': 4550 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 4551 if (ptr->m_ratio >= 0.5) 4552 ^~~~~~~ 4553 accessor.cc:7:10: note: declared private here 4554 double m_ratio; 4555 ^~~~~~~ 4556 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 4557 foo::get_ratio() const' 4558 if (ptr->m_ratio >= 0.5) 4559 ^~~~~~~ 4560 get_ratio() 4561 4562 * The C++ compiler can now give you a hint if you use a macro before 4563 it was defined (e.g. if you mess up the order of your #include 4564 directives): 4565 $ gcc ordering.cc 4566 ordering.cc:2:24: error: expected ';' at end of member declaration 4567 virtual void clone() const OVERRIDE { } 4568 ^~~~~ 4569 ; 4570 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 4571 virtual void clone() const OVERRIDE { } 4572 ^~~~~~~~ 4573 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 4574 In file included from ordering.cc:5: 4575 c++11-compat.h:2: note: it was later defined here 4576 #define OVERRIDE override 4577 4578 4579 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 4580 you when you can use a static_cast, const_cast, or 4581 reinterpret_cast. 4582 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 4583 old-style-cast-fixits.cc: In function 'void test(void*)': 4584 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 4585 -Wold-style-cast] 4586 foo *f = (foo *)ptr; 4587 ^~~ 4588 ---------- 4589 static_cast<foo *> (ptr) 4590 4591 * When reporting on problems within extern "C" linkage 4592 specifications, the C++ compiler will now display the location of 4593 the start of the extern "C". 4594 $ gcc -c extern-c.cc 4595 extern-c.cc:3:1: error: template with C linkage 4596 template <typename T> void test (void); 4597 ^~~~~~~~ 4598 In file included from extern-c.cc:1: 4599 unclosed.h:1:1: note: 'extern "C"' linkage started here 4600 extern "C" { 4601 ^~~~~~~~~~ 4602 extern-c.cc:3:39: error: expected '}' at end of input 4603 template <typename T> void test (void); 4604 ^ 4605 In file included from extern-c.cc:1: 4606 unclosed.h:1:12: note: to match this '{' 4607 extern "C" { 4608 ^ 4609 4610 * When reporting on mismatching template types, the C++ compiler will 4611 now use color to highlight the mismatching parts of the template, 4612 and will elide the parameters that are common between two 4613 mismatching templates, printing [...] instead: 4614 $ gcc templates.cc 4615 templates.cc: In function 'void test()': 4616 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 4617 e>' to 'vector<int>' 4618 fn_1(vector<double> ()); 4619 ^~~~~~~~~~~~~~~~~ 4620 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 4621 ,double>' to 'map<[...],int>' 4622 fn_2(map<int, double>()); 4623 ^~~~~~~~~~~~~~~~~~ 4624 4625 Those [...] elided parameters can be seen using -fno-elide-type: 4626 $ gcc templates.cc -fno-elide-type 4627 templates.cc: In function 'void test()': 4628 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 4629 e>' to 'vector<int>' 4630 fn_1(vector<double> ()); 4631 ^~~~~~~~~~~~~~~~~ 4632 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 4633 ouble>' to 'map<int,int>' 4634 fn_2(map<int, double>()); 4635 ^~~~~~~~~~~~~~~~~~ 4636 4637 The C++ compiler has also gained an option 4638 -fdiagnostics-show-template-tree which visualizes such mismatching 4639 templates in a hierarchical form: 4640 $ gcc templates-2.cc -fdiagnostics-show-template-tree 4641 templates-2.cc: In function 'void test()': 4642 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 4643 ble>' to 'vector<int>' 4644 vector< 4645 [double != int]> 4646 fn_1(vector<double> ()); 4647 ^~~~~~~~~~~~~~~~~ 4648 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 4649 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 4650 ap<[...],vector<float>>,vector<float>>' 4651 map< 4652 map< 4653 [...], 4654 vector< 4655 [double != float]>>, 4656 vector< 4657 [double != float]>> 4658 fn_2(map<map<int, vector<double>>, vector<double>> ()); 4659 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4660 4661 which again works with -fno-elide-type: 4662 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 4663 templates-2.cc: In function 'void test()': 4664 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 4665 ble>' to 'vector<int>' 4666 vector< 4667 [double != int]> 4668 fn_1(vector<double> ()); 4669 ^~~~~~~~~~~~~~~~~ 4670 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 4671 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 4672 <int,vector<float>>,vector<float>>' 4673 map< 4674 map< 4675 int, 4676 vector< 4677 [double != float]>>, 4678 vector< 4679 [double != float]>> 4680 fn_2(map<map<int, vector<double>>, vector<double>> ()); 4681 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4682 4683 * Flowing off the end of a non-void function is considered 4684 unreachable and may be subject to optimization on that basis. As a 4685 result of this change, -Wreturn-type warnings are enabled by 4686 default for C++. 4687 4688 [25]Runtime Library (libstdc++) 4689 4690 * Improved experimental support for C++17, including the following 4691 features: 4692 + Deduction guides to support class template argument deduction. 4693 + std::filesystem implementation. 4694 + std::char_traits<char> and std::char_traits<wchar_t> are 4695 usable in constant expressions. 4696 + std::to_chars and std::from_chars (for integers only, not for 4697 floating point types). 4698 * Experimental support for C++2a: std::to_address (thanks to Glen 4699 Fernandes) and std::endian. 4700 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 4701 entropy count for the random device, if known (thanks to Xi 4702 Ruoyao). 4703 * Support for std::experimental::source_location. 4704 * AddressSanitizer integration for std::vector, detecting 4705 out-of-range accesses to the unused capacity of a vector. 4706 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 4707 Mathematical Special Functions. 4708 4709 [26]Fortran 4710 4711 * The main version of libfortran has been changed to 5. 4712 * Parameterized derived types, a major feature of Fortran 2003, have 4713 been implemented. 4714 * Partial support is provided for Fortran 2018 teams, which are 4715 hierarchical subsets of images that execute independently of other 4716 image subsets. 4717 * The maximum rank for arrays has been increased to 15, conforming to 4718 the Fortran 2008 standard. 4719 * Transformational intrinsics are now fully supported in 4720 initialization expressions. 4721 * New flag -fc-prototypes to write C prototypes for BIND(C) 4722 procedures and variables. 4723 * If -fmax-stack-var-size is honored if given together with -Ofast, 4724 -fstack-arrays is no longer set in that case. 4725 * New options -fdefault-real-16 and -fdefault-real-10 to control the 4726 default kind of REAL variables. 4727 * A warning is now issued if an array subscript inside a DO loop 4728 could lead to an out-of-bounds-access. The new option 4729 -Wdo-subscript, enabled by -Wextra, warns about this even if the 4730 compiler can not prove that the code will be executed. 4731 * The Fortran front end now attempts to interchange loops if it is 4732 deemed profitable. So far, this is restricted to FORALL and DO 4733 CONCURRENT statements with multiple indices. This behavior be 4734 controlled with the new flag -ffrontend-loop-interchange, which is 4735 enabled with optimization by default. The 4736 -Wfrontend-loop-interchange option warns about such occurrences. 4737 * When an actual argument contains too few elements for a dummy 4738 argument, an error is now issued. The -std=legacy option can be 4739 used to still compile such code. 4740 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 4741 integers, making records larger than 2GiB possible. 4742 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 4743 effect. The record length for preconnected units is now larger than 4744 any practical limit, same as for sequential access units opened 4745 without an explicit RECL= specifier. 4746 * Character variables longer than HUGE(0) elements are now possible 4747 on 64-bit targets. Note that this changes the procedure call ABI 4748 for all procedures with character arguments on 64-bit targets, as 4749 the type of the hidden character length argument has changed. The 4750 hidden character length argument is now of type INTEGER(C_SIZE_T). 4751 * Partial support is provided for Fortran 2018 teams, which are 4752 hierarchical subsets of images that execute independently of other 4753 image subsets. 4754 4755 [27]Go 4756 4757 * GCC 8 provides a complete implementation of the Go 1.10.1 user 4758 packages. 4759 * The garbage collector is now fully concurrent. As before, values 4760 stored on the stack are scanned conservatively, but value stored in 4761 the heap are scanned precisely. 4762 * Escape analysis is fully implemented and enabled by default in the 4763 Go front end. This significantly reduces the number of heap 4764 allocations by allocating values on the stack instead. 4765 4766 [28]libgccjit 4767 4768 The libgccjit API gained four new entry points: 4769 * [29]gcc_jit_type_get_vector and 4770 * [30]gcc_jit_context_new_rvalue_from_vector for working with 4771 vectors, 4772 * [31]gcc_jit_type_get_aligned 4773 * [32]gcc_jit_function_get_address 4774 4775 The C code generated by [33]gcc_jit_context_dump_reproducer_to_file is 4776 now easier-to-read. 4777 4778 [34]New Targets and Target Specific Improvements 4779 4780 [35]AArch64 4781 4782 * The Armv8.4-A architecture is now supported. It can be used by 4783 specifying the -march=armv8.4-a option. 4784 * The Dot Product instructions are now supported as an optional 4785 extension to the Armv8.2-A architecture and newer and are mandatory 4786 on Armv8.4-A. The extension can be used by specifying the +dotprod 4787 architecture extension. E.g. -march=armv8.2-a+dotprod. 4788 * The Armv8-A +crypto extension has now been split into two 4789 extensions for finer grained control: 4790 + +aes which contains the Armv8-A AES crytographic instructions. 4791 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 4792 instructions. 4793 Using +crypto will now enable these two extensions. 4794 * New Armv8.4-A FP16 Floating Point Multiplication Variant 4795 instructions have been added. These instructions are mandatory in 4796 Armv8.4-A but available as an optional extension to Armv8.2-A and 4797 Armv8.3-A. The new extension can be used by specifying the +fp16fml 4798 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 4799 the instructions can be enabled by specifying +fp16. 4800 * New cryptographic instructions have been added as optional 4801 extensions to Armv8.2-A and newer. These instructions can be 4802 enabled with: 4803 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 4804 implies +sha2. 4805 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 4806 * The Scalable Vector Extension (SVE) is now supported as an optional 4807 extension to the Armv8.2-A architecture and newer. This support 4808 includes automatic vectorization with SVE instructions, but it does 4809 not yet include the SVE Arm C Language Extensions (ACLE). It can be 4810 enabled by specifying the +sve architecture extension (for example, 4811 -march=armv8.2-a+sve). By default, the generated code works with 4812 all vector lengths, but it can be made specific to N-bit vectors 4813 using -msve-vector-bits=N. 4814 * Support has been added for the following processors (GCC 4815 identifiers in parentheses): 4816 + Arm Cortex-A75 (cortex-a75). 4817 + Arm Cortex-A55 (cortex-a55). 4818 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 4819 (cortex-a75.cortex-a55). 4820 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4821 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 4822 arguments to the equivalent target attributes and pragmas. 4823 4824 [36]ARC 4825 4826 * Added support for: 4827 + Fast interrupts. 4828 + Naked functions. 4829 + aux variable attributes. 4830 + uncached type qualifier. 4831 + Secure functions via sjli instruction. 4832 * New exception handling implementation. 4833 * Revamped trampoline implementation. 4834 * Refactored small data feature implementation, controlled via the -G 4835 command-line option. 4836 * New support for reduced register set ARC architecture 4837 configurations, controlled via the -mrf16 command-line option. 4838 * Refurbished and improved support for zero overhead loops. 4839 Introduced -mlpc-width command-line option to control the width of 4840 the lp_count register. 4841 4842 [37]ARM 4843 4844 * The -mfpu option now takes a new option setting of -mfpu=auto. When 4845 set to this the floating-point and SIMD settings are derived from 4846 the settings of the -mcpu or -march options. The internal CPU 4847 configurations have been updated with information about the 4848 permitted floating-point configurations supported. See the user 4849 guide for further information about the extended option syntax for 4850 controlling architectural extensions via the -march option. 4851 -mfpu=auto is now the default setting unless the compiler has been 4852 configured with an explicit --with-fpu option. 4853 * The -march and -mcpu options now accept optional extensions to the 4854 architecture or CPU option, allowing the user to enable or disable 4855 any such extensions supported by that architecture or CPU such as 4856 (but not limited to) floating-point and AdvancedSIMD. For example: 4857 the option -mcpu=cortex-a53+nofp will generate code for the 4858 Cortex-A53 processor with no floating-point support. This, in 4859 combination with the new -mfpu=auto option, provides a 4860 straightforward way of specifying a valid build target through a 4861 single -mcpu or -march option. The -mtune option accepts the same 4862 arguments as -mcpu but only the CPU name has an effect on tuning. 4863 The architecture extensions do not have any effect. For details of 4864 what extensions a particular architecture or CPU option supports 4865 please refer to the [38]documentation. 4866 * The -mstructure-size-boundary option has been deprecated and will 4867 be removed in a future release. 4868 * The default link behavior for Armv6 and Armv7-R targets has been 4869 changed to produce BE8 format when generating big-endian images. A 4870 new flag -mbe32 can be used to force the linker to produce legacy 4871 BE32 format images. There is no change of behavior for Armv6-M and 4872 other Armv7 or later targets: these already defaulted to BE8 4873 format. This change brings GCC into alignment with other compilers 4874 for the ARM architecture. 4875 * The Armv8-R architecture is now supported. It can be used by 4876 specifying the -march=armv8-r option. 4877 * The Armv8.3-A architecture is now supported. It can be used by 4878 specifying the -march=armv8.3-a option. 4879 * The Armv8.4-A architecture is now supported. It can be used by 4880 specifying the -march=armv8.4-a option. 4881 * The Dot Product instructions are now supported as an optional 4882 extension to the Armv8.2-A architecture and newer and are mandatory 4883 on Armv8.4-A. The extension can be used by specifying the +dotprod 4884 architecture extension. E.g. -march=armv8.2-a+dotprod. 4885 * Support for setting extensions and architectures using the GCC 4886 target pragma and attribute has been added. It can be used by 4887 specifying #pragma GCC target ("arch=..."), #pragma GCC target 4888 ("+extension"), __attribute__((target("arch=..."))) or 4889 __attribute__((target("+extension"))). 4890 * New Armv8.4-A FP16 Floating Point Multiplication Variant 4891 instructions have been added. These instructions are mandatory in 4892 Armv8.4-A but available as an optional extension to Armv8.2-A and 4893 Armv8.3-A. The new extension can be used by specifying the +fp16fml 4894 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 4895 the instructions can be enabled by specifying +fp16. 4896 * Support has been added for the following processors (GCC 4897 identifiers in parentheses): 4898 + Arm Cortex-A75 (cortex-a75). 4899 + Arm Cortex-A55 (cortex-a55). 4900 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 4901 (cortex-a75.cortex-a55). 4902 + Arm Cortex-R52 for Armv8-R (cortex-r52). 4903 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4904 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 4905 arguments to the equivalent target attributes and pragmas. 4906 4907 [39]AVR 4908 4909 * The AVR port now supports the following XMEGA-like devices: 4910 4911 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 4912 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 4913 ATtiny3214, ATtiny3216, ATtiny3217 4914 The new devices are listed under [40]-mmcu=avrxmega3. 4915 + These devices see flash memory in the RAM address space, so 4916 that features like PROGMEM and __flash are not needed any more 4917 (as opposed to other AVR families for which read-only data 4918 will be located in RAM except special, non-standard features 4919 are used to locate and access such data). This requires that 4920 the compiler is used with Binutils 2.29 or newer so that 4921 [41]read-only data will be located in flash memory. 4922 + A new command-line option -mshort-calls is supported. This 4923 option is used internally for multilib selection of the 4924 avrxmega3 variants. It is not an optimization option. Do not 4925 set it by hand. 4926 * The compiler now generates [42]efficient interrupt service routine 4927 (ISR) prologues and epilogues. This is achieved by using the new 4928 [43]AVR pseudo instruction __gcc_isr which is supported and 4929 resolved by the GNU assembler. 4930 + As the __gcc_isr pseudo-instruction will be resolved by the 4931 assembler, inline assembly is transparent to the process. This 4932 means that when inline assembly uses an instruction like INC 4933 that clobbers the condition code, then the assembler will 4934 detect this and generate an appropriate ISR prologue / 4935 epilogue chunk to save / restore SREG as needed. 4936 + A new command-line option -mno-gas-isr-prologues disables the 4937 generation of the __gcc_isr pseudo instruction. Any non-naked 4938 ISR will save and restore SREG, tmp_reg and zero_reg, no 4939 matter whether the respective register is clobbered or used. 4940 + The feature is turned on per default for all optimization 4941 levels except for -O0 and -Og. It is explicitly enabled by 4942 means of option -mgas-isr-prologues. 4943 + Support has been added for a new [44]AVR function attribute 4944 no_gccisr. It can be used to disable __gcc_isr pseudo 4945 instruction generation for individual ISRs. 4946 + This optimization is only available if GCC is configured with 4947 GNU Binutils 2.29 or newer; or at least with a version of 4948 Binutils that implements feature [45]PR21683. 4949 * The compiler no more saves / restores registers in main; the effect 4950 is the same as if attribute OS_task was specified for main. This 4951 optimization can be switched off by the new command-line option 4952 -mno-main-is-OS_task. 4953 4954 [46]IA-32/x86-64 4955 4956 * The x86 port now supports the naked function attribute. 4957 * Better tuning for znver1 and Intel Core based CPUs. 4958 * Vectorization cost metrics has been reworked leading to significant 4959 improvements on some benchmarks. 4960 * GCC now supports the Intel CPU named Cannonlake through 4961 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 4962 and SHA ISA extensions. 4963 * GCC now supports the Intel CPU named Icelake through 4964 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 4965 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 4966 ISA extensions. 4967 * GCC now supports the Intel Control-flow Enforcement Technology 4968 (CET) extension through -fcf-protection option. 4969 4970 [47]NDS32 4971 4972 * New command-line options -mext-perf, -mext-perf2, and -mext-string 4973 have been added for performance extension instructions. 4974 4975 [48]Nios II 4976 4977 * The Nios II back end has been improved to generate better-optimized 4978 code. Changes include switching to LRA, more accurate cost models, 4979 and more compact code for addressing static variables. 4980 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 4981 added. 4982 * The stack-smashing protection options are now enabled on Nios II. 4983 4984 [49]PA-RISC 4985 4986 * The default call ABI on 32-bit linux has been changed from callee 4987 copies to caller copies. This affects objects larger than eight 4988 bytes passed by value. The goal is to improve compatibility with 4989 x86 and resolve issues with OpenMP. 4990 * Other PA-RISC targets are unchanged. 4991 4992 [50]PowerPC / PowerPC64 / RS6000 4993 4994 * The PowerPC SPE support is split off to a separate powerpcspe port. 4995 The separate port is deprecated and might be removed in a future 4996 release. 4997 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 4998 powerpc*-*-linux*paired*) is deprecated and will be removed in a 4999 future release. 5000 * The Xilinx floating point support (-mxilinx-fpu, 5001 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 5002 release. 5003 * Support for using big-endian AltiVec intrinsics on a little-endian 5004 target (-maltivec=be) is deprecated and will be removed in a future 5005 release. 5006 5007 [51]Tile 5008 5009 * The TILE-Gx port is deprecated and will be removed in a future 5010 release. 5011 5012 [52]Operating Systems 5013 5014 [53]Windows 5015 5016 * GCC on Microsoft Windows can now be configured via 5017 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 5018 specific behavior for GCC itself with regards to supporting the 5019 wildcard character. Prior versions of GCC would follow the 5020 configuration of the MinGW runtime. This behavior can still be 5021 obtained by not using the above options or by using 5022 --enable-mingw-wildcard=platform. 5023 5024 [54]Improvements for plugin authors 5025 5026 * Plugins can now register a callback hook for when comments are 5027 encountered by the C and C++ compilers, e.g. allowing for plugins 5028 to handle documentation markup in code comments. 5029 * The gdbinit support script for debugging GCC now has a 5030 break-on-diagnostic command, providing an easy way to trigger a 5031 breakpoint whenever a diagnostic is emitted. 5032 * The API for creating fix-it hints now supports newlines, and for 5033 emitting mutually incompatible fix-it hints for one diagnostic. 5034 5035 [55]GCC 8.1 5036 5037 This is the [56]list of problem reports (PRs) from GCC's bug tracking 5038 system that are known to be fixed in the 8.1 release. This list might 5039 not be complete (that is, it is possible that some PRs that have been 5040 fixed are not listed here). 5041 5042 [57]GCC 8.2 5043 5044 This is the [58]list of problem reports (PRs) from GCC's bug tracking 5045 system that are known to be fixed in the 8.2 release. This list might 5046 not be complete (that is, it is possible that some PRs that have been 5047 fixed are not listed here). 5048 5049 General Improvements 5050 5051 * Fixed LTO link-time performance problems caused by an overflow in 5052 the partitioning algorithm while building large binaries. 5053 5054 Language Specific Changes 5055 5056 C++ 5057 5058 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 5059 returning of classes with a deleted copy constructor and defaulted 5060 trivial move constructor (bug [59]c++/86094). GCC 8.2 introduces 5061 -fabi-version=13 and makes it the default, ABI incompatibilities 5062 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [60]C++ 5063 changes for more details. 5064 5065 Target Specific Changes 5066 5067 IA-32/x86-64 5068 5069 * -mtune=native performance regression [61]PR84413 on Intel Skylake 5070 processors has been fixed. 5071 5072 [62]GCC 8.3 5073 5074 This is the [63]list of problem reports (PRs) from GCC's bug tracking 5075 system that are known to be fixed in the 8.3 release. This list might 5076 not be complete (that is, it is possible that some PRs that have been 5077 fixed are not listed here). 5078 5079 Windows 5080 5081 * A C++ Microsoft ABI bitfield layout bug, [64]PR87137 has been 5082 fixed. A non-field declaration could cause the current bitfield 5083 allocation unit to be completed, incorrectly placing a following 5084 bitfield into a new allocation unit. The Microsoft ABI is selected 5085 for: 5086 + Mingw targets 5087 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 5088 option is specified, or __attribute__((ms_struct)) is used 5089 + SuperH targets when the -mhitachi option is specified, or 5090 __attribute__((renesas)) is used 5091 GCC 8 introduced additional cases of this defect, but rather than 5092 resolve only those regressions, we decided to resolve all the cases 5093 of this defect in single change. 5094 5095 [65]GCC 8.4 5096 5097 This is the [66]list of problem reports (PRs) from GCC's bug tracking 5098 system that are known to be fixed in the 8.4 release. This list might 5099 not be complete (that is, it is possible that some PRs that have been 5100 fixed are not listed here). 5101 5102 [67]GCC 8.5 5103 5104 This is the [68]list of problem reports (PRs) from GCC's bug tracking 5105 system that are known to be fixed in the 8.5 release. This list might 5106 not be complete (that is, it is possible that some PRs that have been 5107 fixed are not listed here). 5108 5109 Target Specific Changes 5110 5111 AArch64 5112 5113 * The option -moutline-atomics has been added to aid deployment of 5114 the Large System Extensions (LSE) on GNU/Linux systems built with a 5115 baseline architecture targeting Armv8-A. When the option is 5116 specified code is emitted to detect the presence of LSE 5117 instructions at run time and use them for standard atomic 5118 operations. For more information please refer to the documentation. 5119 5120 5121 For questions related to the use of GCC, please consult these web 5122 pages and the [69]GCC manuals. If that fails, the 5123 [70]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5124 web pages and the development of GCC are welcome on our developer 5125 list at [71]gcc (a] gcc.gnu.org. All of [72]our lists have public 5126 archives. 5127 5128 Copyright (C) [73]Free Software Foundation, Inc. Verbatim copying and 5129 distribution of this entire article is permitted in any medium, 5130 provided this notice is preserved. 5131 5132 These pages are [74]maintained by the GCC team. Last modified 5133 2025-01-31. 5134 5135 References 5136 5137 1. https://gcc.gnu.org/gcc-8/porting_to.html 5138 2. https://gcc.gnu.org/onlinedocs/index.html#current 5139 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 5140 4. https://gcc.gnu.org/gcc-8/changes.html#general 5141 5. https://gcc.gnu.org/gcc-8/changes.html#languages 5142 6. https://gcc.gnu.org/gcc-8/changes.html#ada 5143 7. https://gcc.gnu.org/gcc-8/changes.html#brig 5144 8. https://gcc.gnu.org/gcc-8/changes.html#c-family 5145 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 5146 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 5147 11. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 5148 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 5149 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 5150 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 5151 15. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 5152 16. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 5153 17. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 5154 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 5155 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 5156 20. https://gcc.gnu.org/gcc-8/changes.html#c 5157 21. https://gcc.gnu.org/gcc-8/changes.html#cxx 5158 22. https://gcc.gnu.org/PR86094 5159 23. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 5160 24. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 5161 25. https://gcc.gnu.org/gcc-8/changes.html#libstdcxx 5162 26. https://gcc.gnu.org/gcc-8/changes.html#fortran 5163 27. https://gcc.gnu.org/gcc-8/changes.html#go 5164 28. https://gcc.gnu.org/gcc-8/changes.html#jit 5165 29. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 5166 30. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 5167 31. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 5168 32. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 5169 33. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 5170 34. https://gcc.gnu.org/gcc-8/changes.html#targets 5171 35. https://gcc.gnu.org/gcc-8/changes.html#aarch64 5172 36. https://gcc.gnu.org/gcc-8/changes.html#arc 5173 37. https://gcc.gnu.org/gcc-8/changes.html#arm 5174 38. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 5175 39. https://gcc.gnu.org/gcc-8/changes.html#avr 5176 40. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 5177 41. https://sourceware.org/PR21472 5178 42. https://gcc.gnu.org/PR20296 5179 43. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 5180 44. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 5181 45. https://sourceware.org/PR21683 5182 46. https://gcc.gnu.org/gcc-8/changes.html#x86 5183 47. https://gcc.gnu.org/gcc-8/changes.html#nds32 5184 48. https://gcc.gnu.org/gcc-8/changes.html#nios2 5185 49. https://gcc.gnu.org/gcc-8/changes.html#hppa 5186 50. https://gcc.gnu.org/gcc-8/changes.html#powerpc 5187 51. https://gcc.gnu.org/gcc-8/changes.html#Tile 5188 52. https://gcc.gnu.org/gcc-8/changes.html#os 5189 53. https://gcc.gnu.org/gcc-8/changes.html#windows 5190 54. https://gcc.gnu.org/gcc-8/changes.html#plugins 5191 55. https://gcc.gnu.org/gcc-8/changes.html#GCC8.1 5192 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 5193 57. https://gcc.gnu.org/gcc-8/changes.html#GCC8.2 5194 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 5195 59. https://gcc.gnu.org/PR86094 5196 60. https://gcc.gnu.org/gcc-8/changes.html#cxx 5197 61. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 5198 62. https://gcc.gnu.org/gcc-8/changes.html#GCC8.3 5199 63. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 5200 64. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 5201 65. https://gcc.gnu.org/gcc-8/changes.html#GCC8.4 5202 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 5203 67. https://gcc.gnu.org/gcc-8/changes.html#GCC8.5 5204 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5 5205 69. https://gcc.gnu.org/onlinedocs/ 5206 70. mailto:gcc-help (a] gcc.gnu.org 5207 71. mailto:gcc (a] gcc.gnu.org 5208 72. https://gcc.gnu.org/lists.html 5209 73. https://www.fsf.org/ 5210 74. https://gcc.gnu.org/about.html 5211 ====================================================================== 5212 http://gcc.gnu.org/gcc-7/index.html 5213 5214 GCC 7 Release Series 5215 5216 (This release series is no longer supported.) 5217 5218 Nov 14, 2019 5219 5220 The [1]GNU project and the GCC developers are pleased to announce the 5221 release of GCC 7.5. 5222 5223 This release is a bug-fix release, containing fixes for regressions in 5224 GCC 7.4 relative to previous releases of GCC. 5225 5226 Release History 5227 5228 GCC 7.5 5229 Nov 14, 2019 ([2]changes, [3]documentation) 5230 5231 GCC 7.4 5232 Dec 6, 2018 ([4]changes, [5]documentation) 5233 5234 GCC 7.3 5235 Jan 25, 2018 ([6]changes, [7]documentation) 5236 5237 GCC 7.2 5238 Aug 14, 2017 ([8]changes, [9]documentation) 5239 5240 GCC 7.1 5241 May 2, 2017 ([10]changes, [11]documentation) 5242 5243 References and Acknowledgements 5244 5245 GCC used to stand for the GNU C Compiler, but since the compiler 5246 supports several other languages aside from C, it now stands for the 5247 GNU Compiler Collection. 5248 5249 A list of [12]successful builds is updated as new information becomes 5250 available. 5251 5252 The GCC developers would like to thank the numerous people that have 5253 contributed new features, improvements, bug fixes, and other changes as 5254 well as test results to GCC. This [13]amazing group of volunteers is 5255 what makes GCC successful. 5256 5257 For additional information about GCC please refer to the [14]GCC 5258 project web site or contact the [15]GCC development mailing list. 5259 5260 To obtain GCC please use [16]our mirror sites or [17]our version 5261 control system. 5262 5263 5264 For questions related to the use of GCC, please consult these web 5265 pages and the [18]GCC manuals. If that fails, the 5266 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5267 web pages and the development of GCC are welcome on our developer 5268 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 5269 archives. 5270 5271 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5272 distribution of this entire article is permitted in any medium, 5273 provided this notice is preserved. 5274 5275 These pages are [23]maintained by the GCC team. Last modified 5276 2024-05-30. 5277 5278 References 5279 5280 1. http://www.gnu.org/ 5281 2. https://gcc.gnu.org/gcc-7/changes.html 5282 3. https://gcc.gnu.org/onlinedocs/7.5.0/ 5283 4. https://gcc.gnu.org/gcc-7/changes.html 5284 5. https://gcc.gnu.org/onlinedocs/7.4.0/ 5285 6. https://gcc.gnu.org/gcc-7/changes.html 5286 7. https://gcc.gnu.org/onlinedocs/7.3.0/ 5287 8. https://gcc.gnu.org/gcc-7/changes.html 5288 9. https://gcc.gnu.org/onlinedocs/7.2.0/ 5289 10. https://gcc.gnu.org/gcc-7/changes.html 5290 11. https://gcc.gnu.org/onlinedocs/7.1.0/ 5291 12. https://gcc.gnu.org/gcc-7/buildstat.html 5292 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5293 14. https://gcc.gnu.org/index.html 5294 15. mailto:gcc (a] gcc.gnu.org 5295 16. https://gcc.gnu.org/mirrors.html 5296 17. https://gcc.gnu.org/git.html 5297 18. https://gcc.gnu.org/onlinedocs/ 5298 19. mailto:gcc-help (a] gcc.gnu.org 5299 20. mailto:gcc (a] gcc.gnu.org 5300 21. https://gcc.gnu.org/lists.html 5301 22. https://www.fsf.org/ 5302 23. https://gcc.gnu.org/about.html 5303 ====================================================================== 5304 http://gcc.gnu.org/gcc-7/changes.html 5305 5306 GCC 7 Release Series 5307 Changes, New Features, and Fixes 5308 5309 This page is a brief summary of some of the huge number of improvements 5310 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 5311 the [2]full GCC documentation. 5312 5313 Caveats 5314 5315 * GCC now uses [3]LRA (a new local register allocator) by default for 5316 new targets. 5317 * The non-standard C++0x type traits has_trivial_default_constructor, 5318 has_trivial_copy_constructor and has_trivial_copy_assign have been 5319 removed. 5320 * The libstdc++ [4]Profile Mode has been deprecated and will be 5321 removed in a future version. 5322 * The Cilk+ extensions to the C and C++ languages have been 5323 deprecated. 5324 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 5325 affects conformance to the procedure call standard (AAPCS) has been 5326 fixed. The bug affects some C++ code where class objects are passed 5327 by value to functions and could result in incorrect or inconsistent 5328 code being generated. This is an ABI change. If the option -Wpsabi 5329 is enabled (on by default) the compiler will emit a diagnostic note 5330 for code that might be affected. 5331 5332 [6]General Optimizer Improvements 5333 5334 * GCC 7 can determine the return value or range of return values of 5335 some calls to the sprintf family of functions and make it available 5336 to other optimization passes. Some calls to the snprintf function 5337 with a zero size argument can be folded into constants. This 5338 optimization is included in -O1 and can be selectively controlled 5339 by the -fprintf-return-value option. 5340 * A new store merging pass has been added. It merges constant stores 5341 to adjacent memory locations into fewer, wider, stores. It is 5342 enabled by the -fstore-merging option and at the -O2 optimization 5343 level or higher (and -Os). 5344 * A new code hoisting optimization has been added to the partial 5345 redundancy elimination pass. It attempts to move evaluation of 5346 expressions executed on all paths to the function exit as early as 5347 possible. This primarily helps improve code size, but can improve 5348 the speed of the generated code as well. It is enabled by the 5349 -fcode-hoisting option and at the -O2 optimization level or higher 5350 (and -Os). 5351 * A new interprocedural bitwise constant propagation optimization has 5352 been added, which propagates knowledge about which bits of 5353 variables are known to be zero (including pointer alignment 5354 information) across the call graph. It is enabled by the 5355 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 5356 at the -O2 optimization level and higher (and -Os). This 5357 optimization supersedes interprocedural alignment propagation of 5358 GCC 6, and therefore the option -fipa-cp-alignment is now 5359 deprecated and ignored. 5360 * A new interprocedural value range propagation optimization has been 5361 added, which propagates integral range information across the call 5362 graph when variable values can be proven to be within those ranges. 5363 It is enabled by the -fipa-vrp option and at the -O2 optimization 5364 level and higher (and -Os). 5365 * A new loop splitting optimization pass has been added. Certain 5366 loops which contain a condition that is always true on one side of 5367 the iteration space and always false on the other are split into 5368 two loops, such that each of the two new loops iterates on just one 5369 side of the iteration space and the condition does not need to be 5370 checked inside of the loop. It is enabled by the -fsplit-loops 5371 option and at the -O3 optimization level or higher. 5372 * The shrink-wrapping optimization can now separate portions of 5373 prologues and epilogues to improve performance if some of the work 5374 done traditionally by prologues and epilogues is not needed on 5375 certain paths. This is controlled by the -fshrink-wrap-separate 5376 option, enabled by default. It requires target support, which is 5377 currently only implemented in the PowerPC and AArch64 ports. 5378 * AddressSanitizer gained a new sanitization option, 5379 -fsanitize-address-use-after-scope, which enables sanitization of 5380 variables whose address is taken and used after a scope where the 5381 variable is defined: 5382 5383 int 5384 main (int argc, char **argv) 5385 { 5386 char *ptr; 5387 { 5388 char my_char; 5389 ptr = &my_char; 5390 } 5391 5392 *ptr = 123; 5393 return *ptr; 5394 } 5395 5396 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 5397 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 5398 WRITE of size 1 at 0x7fffb8dba990 thread T0 5399 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 5400 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 5401 #2 0x400739 in _start (/tmp/a.out+0x400739) 5402 5403 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 5404 #0 0x40067f in main /tmp/use-after-scope-1.c:3 5405 5406 This frame has 1 object(s): 5407 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 5408 5409 The option is enabled by default with -fsanitize=address and 5410 disabled by default with -fsanitize=kernel-address. Compared to the 5411 LLVM compiler, where the option already exists, the implementation 5412 in the GCC compiler has some improvements and advantages: 5413 + Complex uses of gotos and case labels are properly handled and 5414 should not report any false positive or false negatives. 5415 + C++ temporaries are sanitized. 5416 + Sanitization can handle invalid memory stores that are 5417 optimized out by the LLVM compiler when optimization is 5418 enabled. 5419 * The -fsanitize=signed-integer-overflow suboption of the 5420 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 5421 on arithmetic operations with generic vectors. 5422 * Version 5 of the DWARF debugging information standard is supported 5423 through the -gdwarf-5 option. The DWARF version 4 debugging 5424 information remains the default until consumers of debugging 5425 information are adjusted. 5426 5427 [7]New Languages and Language specific improvements 5428 5429 OpenACC support in C, C++, and Fortran continues to be maintained and 5430 improved. See the [8]OpenACC and [9]Offloading wiki pages for further 5431 information. 5432 5433 [10]Ada 5434 5435 * On mainstream native platforms, Ada programs no longer require the 5436 stack to be made executable in order to run properly. 5437 5438 [11]BRIG (HSAIL) 5439 5440 Support for processing BRIG 1.0 files was added in this release. BRIG 5441 is a binary format for HSAIL (Heterogeneous System Architecture 5442 Intermediate Language). The BRIG front end can be used for implementing 5443 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 5444 GCC-supported targets. An implementation of an HSAIL runtime library, 5445 libhsail-rt is also included. 5446 5447 [12]C family 5448 5449 * New command-line options have been added for the C and C++ 5450 compilers: 5451 + -Wimplicit-fallthrough warns when a switch case falls through. 5452 This warning has five different levels. The compiler is able 5453 to parse a wide range of fallthrough comments, depending on 5454 the level. It also handles control-flow statements, such as 5455 ifs. It's possible to suppress the warning by either adding a 5456 fallthrough comment, or by using a null statement: 5457 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 5458 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 5459 is enabled by -Wextra. 5460 + -Wpointer-compare warns when a pointer is compared with a zero 5461 character constant. Such code is now invalid in C++11 and GCC 5462 rejects it. This warning is enabled by default. 5463 + -Wduplicated-branches warns when an if-else has identical 5464 branches. 5465 + -Wrestrict warns when an argument passed to a 5466 restrict-qualified parameter aliases with another argument. 5467 + -Wmemset-elt-size warns for memset calls, when the first 5468 argument references an array, and the third argument is a 5469 number equal to the number of elements of the array, but not 5470 the size of the array. This warning is enabled by -Wall. 5471 + -Wint-in-bool-context warns about suspicious uses of integer 5472 values where boolean values are expected. This warning is 5473 enabled by -Wall. 5474 + -Wswitch-unreachable warns when a switch statement has 5475 statements between the controlling expression and the first 5476 case label which will never be executed. This warning is 5477 enabled by default. 5478 + -Wexpansion-to-defined warns when defined is used outside #if. 5479 This warning is enabled by -Wextra or -Wpedantic. 5480 + -Wregister warns about uses of the register storage specifier. 5481 In C++17 this keyword has been removed and for C++17 this is a 5482 pedantic warning enabled by default. The warning is not 5483 emitted for the GNU Explicit Register Variables extension. 5484 + -Wvla-larger-than=N warns about unbounded uses of 5485 variable-length arrays, and about bounded uses of 5486 variable-length arrays whose bound can be larger than N bytes. 5487 + -Wduplicate-decl-specifier warns when a declaration has 5488 duplicate const, volatile, restrict or _Atomic specifier. This 5489 warning is enabled by -Wall. 5490 * GCC 6's C and C++ front ends were able to offer suggestions for 5491 misspelled field names: 5492 5493 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 5494 you mean 'color'? 5495 return ptr->colour; 5496 ^~~~~~ 5497 5498 GCC 7 greatly expands the scope of these suggestions. Firstly, it 5499 adds fix-it hints to such suggestions: 5500 5501 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 5502 you mean 'color'? 5503 return ptr->colour; 5504 ^~~~~~ 5505 color 5506 5507 The suggestions now cover many other things, such as misspelled 5508 function names: 5509 5510 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 5511 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 5512 ion] 5513 gtk_widget_showall (w); 5514 ^~~~~~~~~~~~~~~~~~ 5515 gtk_widget_show_all 5516 5517 misspelled macro names and enum values: 5518 5519 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 5520 ction); did you mean 'MAX_ITEMS'? 5521 int array[MAX_ITEM]; 5522 ^~~~~~~~ 5523 MAX_ITEMS 5524 5525 misspelled type names: 5526 5527 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 5528 gned'? 5529 void test (singed char e); 5530 ^~~~~~ 5531 signed 5532 5533 and, in the C front end, named initializers: 5534 5535 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 5536 '? 5537 struct s test = { .colour = 3 }; 5538 ^~~~~~ 5539 color 5540 5541 * The preprocessor can now offer suggestions for misspelled 5542 directives, e.g.: 5543 5544 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 5545 #endfi 5546 ^~~~~ 5547 endif 5548 5549 * Warnings about format strings now underline the pertinent part of 5550 the string, and can offer suggested fixes. In some cases, the 5551 pertinent argument is underlined. 5552 5553 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 5554 nt 3 has type 'int' [-Wformat=] 5555 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 5556 ~^ ~~~~~ 5557 %d 5558 5559 * The new -Wdangling-else command-line option has been split out of 5560 -Wparentheses and warns about dangling else. 5561 * The -Wshadow warning has been split into three variants. 5562 -Wshadow=global warns for any shadowing. This is the default when 5563 using -Wshadow without any argument. -Wshadow=local only warns for 5564 a local variable shadowing another local variable or parameter. 5565 -Wshadow=compatible-local only warns for a local variable shadowing 5566 another local variable or parameter whose type is compatible (in 5567 C++ compatible means that the type of the shadowing variable can be 5568 converted to that of the shadowed variable). 5569 The following example shows the different kinds of shadow warnings: 5570 5571 enum operation { add, count }; 5572 struct container { int nr; }; 5573 5574 int 5575 container_count (struct container c, int count) 5576 { 5577 int r = 0; 5578 for (int count = 0; count > 0; count--) 5579 { 5580 struct container count = c; 5581 r += count.nr; 5582 } 5583 return r; 5584 } 5585 5586 -Wshadow=compatible-local will warn for the parameter being 5587 shadowed with the same type: 5588 5589 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 5590 compatible-local] 5591 for (int count = 0; count > 0; count--) 5592 ^~~~~ 5593 warn-test.c:5:42: note: shadowed declaration is here 5594 container_count (struct container c, int count) 5595 ^~~~~ 5596 5597 -Wshadow=local will warn for the above and for the shadowed 5598 declaration with incompatible type: 5599 5600 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 5601 hadow=local] 5602 struct container count = c; 5603 ^~~~~ 5604 warn-test.c:8:12: note: shadowed declaration is here 5605 for (int count = 0; count > 0; count--) 5606 ^~~~~ 5607 5608 -Wshadow=global will warn for all of the above and the shadowing of 5609 the global declaration: 5610 5611 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 5612 -Wshadow] 5613 container_count (struct container c, int count) 5614 ^~~~~ 5615 warn-test.c:1:23: note: shadowed declaration is here 5616 enum operation { add, count }; 5617 ^~~~~ 5618 5619 * GCC 7 contains a number of enhancements that help detect buffer 5620 overflow and other forms of invalid memory accesses. 5621 + The -Walloc-size-larger-than=size option detects calls to 5622 standard and user-defined memory allocation functions 5623 decorated with attribute alloc_size whose argument exceeds the 5624 specified size (PTRDIFF_MAX by default). The option also 5625 detects arithmetic overflow in the computation of the size in 5626 two-argument allocation functions like calloc where the total 5627 size is the product of the two arguments. Since calls with an 5628 excessive size cannot succeed they are typically the result of 5629 programming errors. Such bugs have been known to be the source 5630 of security vulnerabilities and a target of exploits. 5631 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 5632 For example, the following call to malloc incorrectly tries to 5633 avoid passing a negative argument to the function and instead 5634 ends up unconditionally invoking it with an argument less than 5635 or equal to zero. Since after conversion to the type of the 5636 argument of the function (size_t) a negative argument results 5637 in a value in excess of the maximum PTRDIFF_MAX the call is 5638 diagnosed. 5639 5640 void* f (int n) 5641 { 5642 return malloc (n > 0 ? 0 : n); 5643 } 5644 5645 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 5646 147483647 [-Walloc-size-larger-than=] 5647 5648 + The -Walloc-zero option detects calls to standard and 5649 user-defined memory allocation functions decorated with 5650 attribute alloc_size with a zero argument. -Walloc-zero is not 5651 included in either -Wall or -Wextra and must be explicitly 5652 enabled. 5653 + The -Walloca option detects all calls to the alloca function 5654 in the program. -Walloca is not included in either -Wall or 5655 -Wextra and must be explicitly enabled. 5656 + The -Walloca-larger-than=size option detects calls to the 5657 alloca function whose argument either may exceed the specified 5658 size, or that is not known to be sufficiently constrained to 5659 avoid exceeding it. -Walloca-larger-than is not included in 5660 either -Wall or -Wextra and must be explicitly enabled. 5661 For example, compiling the following snippet with 5662 -Walloca-larger-than=1024 results in a warning because even 5663 though the code appears to call alloca only with sizes of 1kb 5664 and less, since n is signed, a negative value would result in 5665 a call to the function well in excess of the limit. 5666 5667 void f (int n) 5668 { 5669 char *d; 5670 if (n < 1025) 5671 d = alloca (n); 5672 else 5673 d = malloc (n); 5674 ... 5675 } 5676 5677 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 5678 ong unsigned int' [-Walloca-larger-than=] 5679 5680 In contrast, a call to alloca that isn't bounded at all such 5681 as in the following function will elicit the warning below 5682 regardless of the size argument to the option. 5683 5684 void f (size_t n) 5685 { 5686 char *d = alloca (n); 5687 ... 5688 } 5689 5690 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 5691 5692 + The -Wformat-overflow=level option detects certain and likely 5693 buffer overflow in calls to the sprintf family of formatted 5694 output functions. Although the option is enabled even without 5695 optimization it works best with -O2 and higher. 5696 For example, in the following snippet the call to sprintf is 5697 diagnosed because even though its output has been constrained 5698 using the modulo operation it could result in as many as three 5699 bytes if mday were negative. The solution is to either 5700 allocate a larger buffer or make sure the argument is not 5701 negative, for example by changing mday's type to unsigned or 5702 by making the type of the second operand of the modulo 5703 expression unsigned: 100U. 5704 5705 void* f (int mday) 5706 { 5707 char *buf = malloc (3); 5708 sprintf (buf, "%02i", mday % 100); 5709 return buf; 5710 } 5711 5712 warning: 'sprintf may write a terminating nul past the end of the destination [- 5713 Wformat-overflow=] 5714 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 5715 5716 + The -Wformat-truncation=level option detects certain and 5717 likely output truncation in calls to the snprintf family of 5718 formatted output functions. -Wformat-truncation=1 is included 5719 in -Wall and enabled without optimization but works best with 5720 -O2 and higher. 5721 For example, the following function attempts to format an 5722 integer between 0 and 255 in hexadecimal, including the 0x 5723 prefix, into a buffer of four characters. But since the 5724 function must always terminate output by the null character 5725 ('\0') such a buffer is only big enough to fit just one digit 5726 plus the prefix. Therefore the snprintf call is diagnosed. To 5727 avoid the warning either use a bigger buffer or handle the 5728 function's return value which indicates whether or not its 5729 output has been truncated. 5730 5731 void f (unsigned x) 5732 { 5733 char d[4]; 5734 snprintf (d, sizeof d, "%#02x", x & 0xff); 5735 ... 5736 } 5737 5738 warning: 'snprintf' output may be truncated before the last format character [-W 5739 format-truncation=] 5740 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 5741 5742 + The -Wnonnull option has been enhanced to detect a broader set 5743 of cases of passing null pointers to functions that expect a 5744 non-null argument (those decorated with attribute nonnull). By 5745 taking advantage of optimizations the option can detect many 5746 more cases of the problem than in prior GCC versions. 5747 + The -Wstringop-overflow=type option detects buffer overflow in 5748 calls to string handling functions like memcpy and strcpy. The 5749 option relies on [13]Object Size Checking and has an effect 5750 similar to defining the _FORTIFY_SOURCE macro. 5751 -Wstringop-overflow=2 is enabled by default. 5752 For example, in the following snippet, because the call to 5753 strncat specifies a maximum that allows the function to write 5754 past the end of the destination, it is diagnosed. To correct 5755 the problem and avoid the overflow the function should be 5756 called with a size of at most sizeof d - strlen(d) - 1. 5757 5758 void f (const char *fname) 5759 { 5760 char d[8]; 5761 strncpy (d, "/tmp/", sizeof d); 5762 strncat (d, fname, sizeof d); 5763 ... 5764 } 5765 5766 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 5767 ow=] 5768 5769 * The <limits.h> header provided by GCC defines macros such as 5770 INT_WIDTH for the width in bits of integer types, if 5771 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 5772 included. The <stdint.h> header defines such macros as SIZE_WIDTH 5773 and INTMAX_WIDTH for the width of some standard typedef names for 5774 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 5775 before the header is included; note that GCC's implementation of 5776 this header is only used for freestanding compilations, not hosted 5777 compilations, on most systems. These macros come from ISO/IEC TS 5778 18661-1:2014. 5779 * The <float.h> header provided by GCC defines the macro 5780 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 5781 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 5782 included. This represents the number of decimal digits for which 5783 conversions between decimal character strings and binary formats, 5784 in both directions, are correctly rounded, and currently has the 5785 value of UINTMAX_MAX on all systems, reflecting that GCC's 5786 compile-time conversions are correctly rounded for any number of 5787 digits. 5788 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 5789 __builtin_mul_overflow_p built-in functions have been added. These 5790 work similarly to their siblings without the _p suffix, but do not 5791 actually store the result of the arithmetics anywhere, just return 5792 whether the operation would overflow. Calls to these built-ins with 5793 integer constant arguments evaluate to integer constants 5794 expressions. 5795 For example, in the following, c is assigned the result of a * b 5796 only if the multiplication does not overflow, otherwise it is 5797 assigned the value zero. The multiplication is performed at 5798 compile-time and without triggering a -Woverflow warning. 5799 5800 enum { 5801 a = 12345678, 5802 b = 87654321, 5803 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 5804 }; 5805 5806 [14]C 5807 5808 * The C front end now supports type names _FloatN for floating-point 5809 types with IEEE interchange formats and _FloatNx for floating-point 5810 types with IEEE extended formats. These type names come from 5811 ISO/IEC TS 18661-3:2015. 5812 The set of types supported depends on the target for which GCC is 5813 configured. Most targets support _Float32, _Float32x and _Float64. 5814 _Float128 is supported on targets where IEEE binary128 encoding was 5815 already supported as long double or __float128. _Float64x is 5816 supported on targets where a type with either binary128 or Intel 5817 extended precision format is available. 5818 Constants with these types are supported using suffixes fN, FN, fNx 5819 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 5820 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 5821 defined before it is included. 5822 These new types are always distinct from each other and from float, 5823 double and long double, even if they have the same encoding. 5824 Complex types such as _Complex _Float128 are also supported. 5825 Type-generic built-in functions such as __builtin_isinf support the 5826 new types, and the following type-specific built-in functions have 5827 versions (suffixed fN or fNx) for the new types: 5828 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 5829 __builtin_inf, __builtin_nan, __builtin_nans. 5830 * Compilation with -fopenmp is now compatible with the C11 _Atomic 5831 keyword. 5832 5833 [15]C++ 5834 5835 * The C++ front end has experimental support for all of the current 5836 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 5837 constexpr, class template argument deduction, auto template 5838 parameters, and structured bindings. For a full list of new 5839 features, see [16]the C++ status page. 5840 * C++17 support for new of over-aligned types can be enabled in other 5841 modes with the -faligned-new flag. 5842 * The C++17 evaluation order requirements can be selected in other 5843 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 5844 with -fno-strong-eval-order. 5845 * The default semantics of inherited constructors has changed in all 5846 modes, following [17]P0136. Essentially, overload resolution 5847 happens as if calling the inherited constructor directly, and the 5848 compiler fills in construction of the other bases and members as 5849 needed. Most uses should not need any changes. The old behavior can 5850 be restored with -fno-new-inheriting-ctors, or -fabi-version less 5851 than 11. 5852 * The resolution of DR 150 on matching of template template 5853 parameters, allowing default template arguments to make a template 5854 match a parameter, is currently enabled by default in C++17 mode 5855 only. The default can be overridden with -f{no-,}new-ttp-matching. 5856 * The C++ front end will now provide fix-it hints for some missing 5857 semicolons, allowing for automatic fixes by IDEs: 5858 5859 test.cc:4:11: error: expected ';' after class definition 5860 class a {} 5861 ^ 5862 ; 5863 5864 * -Waligned-new has been added to the C++ front end. It warns about 5865 new of type with extended alignment without -faligned-new. 5866 5867 [18]Runtime Library (libstdc++) 5868 5869 * The type of exception thrown by iostreams, std::ios_base::failure, 5870 now uses the [19]cxx11 ABI. 5871 * Experimental support for C++17, including the following new 5872 features: 5873 + std::string_view; 5874 + std::any, std::optional, and std::variant; 5875 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 5876 invoke_result; 5877 + std::is_swappable, and std::is_nothrow_swappable; 5878 + std::apply, and std::make_from_tuple; 5879 + std::void_t, std::bool_constant, std::conjunction, 5880 std::disjunction, and std::negation; 5881 + Variable templates for type traits; 5882 + Mathematical Special Functions; 5883 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 5884 std::chrono::abs; 5885 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 5886 + std::scoped_lock, std::shared_mutex, 5887 std::atomic<T>::is_always_lock_free; 5888 + std::sample, std::default_searcher, std::boyer_moore_searcher 5889 and std::boyer_moore_horspool_searcher; 5890 + Extraction and re-insertion of map and set nodes, try_emplace 5891 members for maps, and functions for accessing containers 5892 std::size, std::empty, and std::data; 5893 + std::shared_ptr support for arrays, 5894 std::shared_ptr<T>::weak_type, 5895 std::enable_shared_from_this<T>::weak_from_this(), and 5896 std::owner_less<void>; 5897 + std::byte; 5898 + std::as_const, std::not_fn, 5899 std::has_unique_object_representations, constexpr 5900 std::addressof. 5901 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 5902 Voutilainen for work on the C++17 support. 5903 * A new power-of-two rehashing policy for use with the _Hashtable 5904 internals, thanks to Franois Dumont. 5905 5906 [20]Fortran 5907 5908 * Support for a number of extensions for compatibility with legacy 5909 code with new flags: 5910 + -fdec-structure Support for DEC STRUCTURE and UNION 5911 + -fdec-intrinsic-ints Support for new integer intrinsics with 5912 B/I/J/K prefixes such as BABS, JIAND... 5913 + -fdec-math Support for additional math intrinsics, including 5914 COTAN and degree-valued trigonometric functions such as TAND, 5915 ASIND... 5916 + -fdec Enable the -fdec-* family of extensions. 5917 * New flag -finit-derived to allow default initialization of 5918 derived-type variables. 5919 * Improved DO loops with step equal to 1 or -1, generates faster code 5920 without a loop preheader. A new warning, -Wundefined-do-loop, warns 5921 when a loop iterates either to HUGE(i) (with step equal to 1), or 5922 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 5923 at run time with -fcheck=do enabled: 5924 5925 program test 5926 implicit none 5927 integer(1) :: i 5928 do i = -HUGE(i)+10, -HUGE(i)-1, -1 5929 print *, i 5930 end do 5931 end program test 5932 5933 At line 8 of file do_check_12.f90 5934 Fortran runtime error: Loop iterates infinitely 5935 5936 * Version 4.5 of the [21]OpenMP specification is now partially 5937 supported in the Fortran compiler; the largest missing item is 5938 structure element mapping. 5939 * User-defined derived-type input/output (UDTIO) is added. 5940 * Derived type coarrays with allocatable and pointer components are 5941 partially supported. 5942 * Non-constant stop codes and error stop codes (Fortran 2015 5943 feature). 5944 * Derived types with allocatable components of recursive type. 5945 * Intrinsic assignment to polymorphic variables. 5946 * Improved submodule support. 5947 * Improved diagnostics (polymorphic results in pure functions). 5948 * Coarray: Support for failed images (Fortan 2015 feature). 5949 5950 [22]Go 5951 5952 * GCC 7 provides a complete implementation of the Go 1.8.1 user 5953 packages. 5954 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 5955 conservative and less concurrent. 5956 * Escape analysis is available for experimental use via the 5957 -fgo-optimize-allocs option. The -fgo-debug-escape prints 5958 information useful for debugging escape analysis choices. 5959 5960 [23]Java (GCJ) 5961 5962 The GCC Java front end and associated libjava runtime library have been 5963 removed from GCC. 5964 5965 [24]libgccjit 5966 5967 The libgccjit API gained support for marking calls as requiring 5968 tail-call optimization via a new entry point: 5969 [25]gcc_jit_rvalue_set_bool_require_tail_call. 5970 5971 libgccjit performs numerous checks at the API boundary, but if these 5972 succeed, it previously ignored errors and other diagnostics emitted 5973 within the core of GCC, and treated the compile of a gcc_jit_context as 5974 having succeeded. As of GCC 7 it now ensures that if any diagnostics 5975 are emitted, they are visible from the libgccjit API, and that the the 5976 context is flagged as having failed. 5977 5978 [26]New Targets and Target Specific Improvements 5979 5980 [27]AArch64 5981 5982 * GCC has been updated to the latest revision of the procedure call 5983 standard (AAPCS64) to provide support for parameter passing when 5984 data types have been over-aligned. 5985 * The ARMv8.3-A architecture is now supported. It can be used by 5986 specifying the -march=armv8.3-a option. 5987 * The option -msign-return-address= is supported to enable return 5988 address protection using ARMv8.3-A Pointer Authentication 5989 Extensions. For more information on the arguments accepted by this 5990 option, please refer to [28]AArch64-Options. 5991 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 5992 Extensions are now supported. They can be used by specifying the 5993 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 5994 Floating-Point Extensions introduce new half-precision data 5995 processing floating-point instructions. 5996 * Support has been added for the following processors (GCC 5997 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 5998 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 5999 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 6000 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 6001 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 6002 The GCC identifiers can be used as arguments to the -mcpu or -mtune 6003 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 6004 arguments to the equivalent target attributes and pragmas. 6005 6006 [29]ARC 6007 6008 * Added support for ARC HS and ARC EM processors. 6009 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 6010 * Added support for NPS400 ARC700 based CPUs. 6011 * Thread Local Storage is now supported by ARC CPUs. 6012 * Fixed errors for ARC600 when using 32x16 multiplier option. 6013 * Fixed PIE for ARC CPUs. 6014 * New CPU templates are supported via multilib. 6015 6016 [30]ARM 6017 6018 * Support for the ARMv5 and ARMv5E architectures has been deprecated 6019 (which have no known implementations) and will be removed in a 6020 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 6021 architectures remain supported. The values armv5 and armv5e of 6022 -march are thus deprecated. 6023 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 6024 Extensions are now supported. They can be used by specifying the 6025 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 6026 Floating-Point Extensions introduce new half-precision data 6027 processing floating-point instructions. 6028 * The ARMv8-M architecture is now supported in its two architecture 6029 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 6030 Floating-Point Extensions. They can be used by specifying the 6031 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 6032 * Support has been added for the following processors (GCC 6033 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 6034 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 6035 identifiers can be used as arguments to the -mcpu or -mtune 6036 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 6037 * A new command-line option -mpure-code has been added. It does not 6038 allow constant data to be placed in code sections. This option is 6039 only available when generating non-PIC code for ARMv7-M targets. 6040 * Support for the ACLE Coprocessor Intrinsics has been added. This 6041 enables the generation of coprocessor instructions through the use 6042 of intrinsics such as cdp, ldc, and others. 6043 * The configure option --with-multilib-list now accepts the value 6044 rmprofile to build multilib libraries for a range of embedded 6045 targets. See our [31]installation instructions for details. 6046 6047 [32]AVR 6048 6049 * On the reduced Tiny cores, the progmem [33]variable attribute is 6050 now properly supported. Respective read-only variables are located 6051 in flash memory in section .progmem.data. No special code is needed 6052 to access such variables; the compiler automatically adds an offset 6053 of 0x4000 to all addresses, which is needed to access variables in 6054 flash memory. As opposed to ordinary cores where it is sufficient 6055 to specify the progmem attribute with definitions, on the reduced 6056 Tiny cores the attribute also has to be specified with (external) 6057 declarations: 6058 6059 extern const int array[] __attribute__((__progmem__)); 6060 6061 int get_value2 (void) 6062 { 6063 /* Access via addresses array + 0x4004 and array + 0x4005. */ 6064 return array[2]; 6065 } 6066 6067 const int* get_address (unsigned idx) 6068 { 6069 /* Returns array + 0x4000 + 2 * idx. */ 6070 return &array[idx]; 6071 } 6072 6073 * A new command-line option -Wmisspelled-isr has been added. It turns 6074 off -- or turns into errors -- warnings that are reported for 6075 interrupt service routines (ISRs) which don't follow AVR-LibC's 6076 naming convention of prefixing ISR names with __vector. 6077 * __builtin_avr_nops(n) is a new [34]built-in function that inserts n 6078 NOP instructions into the instruction stream. n must be a value 6079 known at compile time. 6080 6081 [35]IA-32/x86-64 6082 6083 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 6084 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 6085 variable precision (4VNNIW), AVX-512 Vector Population Count 6086 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 6087 been added. 6088 6089 [36]NVPTX 6090 6091 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 6092 See the [37]Offloading Wiki on how to configure it. 6093 6094 [38]PowerPC / PowerPC64 / RS6000 6095 6096 * The PowerPC port now uses LRA by default. 6097 * GCC now diagnoses inline assembly that clobbers register r2. This 6098 has always been invalid code, and is no longer quietly tolerated. 6099 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 6100 enhanced to generate more of the new instructions by default, and 6101 to provide more built-in functions to generate code for other new 6102 instructions. 6103 * The configuration option --enable-gnu-indirect-function is now 6104 enabled by default on PowerPC GNU/Linux builds. 6105 * The PowerPC port will now allow 64-bit and 32-bit integer types to 6106 be allocated to the VSX vector registers (ISA 2.06 and above). In 6107 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 6108 allocated in the vector registers. Previously, only 64-bit integer 6109 types were allowed in the traditional floating point registers. 6110 * New options -mstack-protector-guard=global, 6111 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 6112 -mstack-protector-guard-offset= change how the stack protector gets 6113 the value to use as canary. 6114 6115 [39]S/390, System z, IBM z Systems, IBM Z 6116 6117 * Support for the IBM z14 processor has been added. When using the 6118 -march=z14 option, the compiler will generate code making use of 6119 the new instructions introduced with the vector enhancement 6120 facility and the miscellaneous instruction extension facility 2. 6121 The -mtune=z14 option enables z14 specific instruction scheduling 6122 without making use of new instructions. 6123 * Builtins for the new vector instructions have been added and can be 6124 enabled using the -mzvector option. 6125 6126 [40]RISC-V 6127 6128 * Support for the RISC-V instruction set has been added. 6129 6130 [41]RX 6131 6132 Basic support for atomic built-in function has been added. It is 6133 currently implemented by flipping interrupts off and on as needed. 6134 6135 [42]SH 6136 6137 * Support for SH5/SH64 has been removed. 6138 * Improved utilization of delay slots on SH2A. 6139 * Improved utilization of zero-displacement conditional branches. 6140 * The following deprecated options have been removed 6141 + -mcbranchdi 6142 + -mcmpeqdi 6143 + -minvalid-symbols 6144 + -msoft-atomic 6145 + -mspace 6146 + -madjust-unroll 6147 * Support for the following SH2A instructions has been added 6148 + mov.b @-Rm,R0 6149 + mov.w @-Rm,R0 6150 + mov.l @-Rm,R0 6151 + mov.b R0,@Rn+ 6152 + mov.w R0,@Rn+ 6153 + mov.l R0,@Rn+ 6154 6155 [43]SPARC 6156 6157 * The SPARC port now uses LRA by default. 6158 * Support for the new Subtract-Extended-with-Carry instruction 6159 available in SPARC M7 (Niagara 7) has been added. 6160 6161 [44]Operating Systems 6162 6163 [45]AIX 6164 6165 * Visibility support has been enabled for AIX 7.1 and above. 6166 6167 [46]Fuchsia 6168 6169 * Support has been added for the [47]Fuchsia OS. 6170 6171 [48]RTEMS 6172 6173 * The ABI changes on ARM so that no short enums are used by default. 6174 6175 Other significant improvements 6176 6177 * -fverbose-asm previously emitted information on the meanings of 6178 assembly expressions. This has been extended so that it now also 6179 prints comments showing the source lines that correspond to the 6180 assembly, making it easier to read the generated assembly 6181 (especially with larger functions). For example, given this C 6182 source file: 6183 6184 int test (int n) 6185 { 6186 int i; 6187 int total = 0; 6188 6189 for (i = 0; i < n; i++) 6190 total += i * i; 6191 return total; 6192 } 6193 6194 -fverbose-asm now gives output similar to this for the function 6195 body (when compiling for x86_64, with -Os): 6196 6197 .text 6198 .globl test 6199 .type test, @@function 6200 test: 6201 .LFB0: 6202 .cfi_startproc 6203 # example.c:4: int total = 0; 6204 xorl %eax, %eax # <retval> 6205 # example.c:6: for (i = 0; i < n; i++) 6206 xorl %edx, %edx # i 6207 .L2: 6208 # example.c:6: for (i = 0; i < n; i++) 6209 cmpl %edi, %edx # n, i 6210 jge .L5 #, 6211 # example.c:7: total += i * i; 6212 movl %edx, %ecx # i, tmp92 6213 imull %edx, %ecx # i, tmp92 6214 # example.c:6: for (i = 0; i < n; i++) 6215 incl %edx # i 6216 # example.c:7: total += i * i; 6217 addl %ecx, %eax # tmp92, <retval> 6218 jmp .L2 # 6219 .L5: 6220 # example.c:10: } 6221 ret 6222 .cfi_endproc 6223 6224 * Two new options have been added for printing fix-it hints: 6225 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 6226 emitted in a machine-readable form, suitable for consumption 6227 by IDEs. For example, given: 6228 6229 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 6230 you mean 'color'? 6231 return ptr->colour; 6232 ^~~~~~ 6233 color 6234 6235 it will emit: 6236 6237 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 6238 6239 + -fdiagnostics-generate-patch will print a patch in "unified" 6240 format after any diagnostics are printed, showing the result 6241 of applying all fix-it hints. For the above example it would 6242 emit: 6243 6244 --- spellcheck-fields.cc 6245 +++ spellcheck-fields.cc 6246 @@ -49,5 +49,5 @@ 6247 6248 color get_color(struct s *ptr) 6249 { 6250 - return ptr->colour; 6251 + return ptr->color; 6252 } 6253 6254 * The gcc and g++ driver programs will now provide suggestions for 6255 misspelled arguments to command-line options. 6256 6257 $ gcc -c test.c -ftls-model=global-dinamic 6258 gcc: error: unknown TLS model 'global-dinamic' 6259 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 6260 cal-dynamic local-exec; did you mean 'global-dynamic'? 6261 6262 * The compiler will now provide suggestions for misspelled 6263 parameters. 6264 6265 $ gcc -c test.c --param max-early-inliner-iteration=3 6266 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 6267 x-early-inliner-iterations'? 6268 6269 * Profile-guided optimization (PGO) instrumentation, as well as test 6270 coverage (GCOV), can newly instrument constructors (functions marks 6271 with __attribute__((constructor))), destructors and C++ 6272 constructors (and destructors) of classes that are used as the type 6273 of a global variable. 6274 * A new option -fprofile-update=atomic prevents creation of corrupted 6275 profiles created during an instrumentation run (-fprofile=generate) 6276 of an application. The downside of the option is a speed penalty. 6277 Providing -pthread on the command line selects atomic profile 6278 updating (when supported by the target). 6279 * GCC's already extensive testsuite has gained some new capabilities, 6280 to further improve the reliability of the compiler: 6281 + GCC now has an internal unit-testing API and a suite of tests 6282 for programmatic self-testing of subsystems. 6283 + GCC's C front end has been extended so that it can parse dumps 6284 of GCC's internal representations, allowing for DejaGnu tests 6285 that more directly exercise specific optimization passes. This 6286 covers both the [49]GIMPLE representation (for testing 6287 higher-level optimizations) and the [50]RTL representation, 6288 allowing for more direct testing of lower-level details, such 6289 as register allocation and instruction selection. 6290 6291 [51]GCC 7.1 6292 6293 This is the [52]list of problem reports (PRs) from GCC's bug tracking 6294 system that are known to be fixed in the 7.1 release. This list might 6295 not be complete (that is, it is possible that some PRs that have been 6296 fixed are not listed here). 6297 6298 [53]GCC 7.2 6299 6300 This is the [54]list of problem reports (PRs) from GCC's bug tracking 6301 system that are known to be fixed in the 7.2 release. This list might 6302 not be complete (that is, it is possible that some PRs that have been 6303 fixed are not listed here). 6304 6305 Target Specific Changes 6306 6307 SPARC 6308 6309 * Support for the SPARC M8 processor has been added. 6310 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 6311 around an erratum in LEON3FT processors. 6312 * Use of the Floating-point Multiply Single to Double (FsMULd) 6313 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 6314 options. 6315 6316 Operating Systems 6317 6318 RTEMS 6319 6320 * The Ada run-time support uses now thread-local storage (TLS). 6321 * Support for RISC-V has been added. 6322 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 6323 double has been added. 6324 6325 [55]GCC 7.3 6326 6327 This is the [56]list of problem reports (PRs) from GCC's bug tracking 6328 system that are known to be fixed in the 7.3 release. This list might 6329 not be complete (that is, it is possible that some PRs that have been 6330 fixed are not listed here). 6331 6332 Target Specific Changes 6333 6334 SPARC 6335 6336 * Workarounds for the four [57]LEON3FT errata GRLIB-TN-0010..0013 6337 have been added. Relevant errata are activated by the target 6338 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 6339 6340 Operating Systems 6341 6342 RTEMS 6343 6344 * Support has been added for Epiphany target. 6345 6346 [58]GCC 7.4 6347 6348 This is the [59]list of problem reports (PRs) from GCC's bug tracking 6349 system that are known to be fixed in the 7.4 release. This list might 6350 not be complete (that is, it is possible that some PRs that have been 6351 fixed are not listed here). 6352 6353 [60]GCC 7.5 6354 6355 This is the [61]list of problem reports (PRs) from GCC's bug tracking 6356 system that are known to be fixed in the 7.5 release. This list might 6357 not be complete (that is, it is possible that some PRs that have been 6358 fixed are not listed here). 6359 6360 6361 For questions related to the use of GCC, please consult these web 6362 pages and the [62]GCC manuals. If that fails, the 6363 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6364 web pages and the development of GCC are welcome on our developer 6365 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 6366 archives. 6367 6368 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 6369 distribution of this entire article is permitted in any medium, 6370 provided this notice is preserved. 6371 6372 These pages are [67]maintained by the GCC team. Last modified 6373 2025-01-31. 6374 6375 References 6376 6377 1. https://gcc.gnu.org/gcc-7/porting_to.html 6378 2. https://gcc.gnu.org/onlinedocs/index.html#current 6379 3. https://gcc.gnu.org/wiki/LRAIsDefault 6380 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 6381 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 6382 6. https://gcc.gnu.org/gcc-7/changes.html#general 6383 7. https://gcc.gnu.org/gcc-7/changes.html#languages 6384 8. https://gcc.gnu.org/wiki/OpenACC 6385 9. https://gcc.gnu.org/wiki/Offloading 6386 10. https://gcc.gnu.org/gcc-7/changes.html#ada 6387 11. https://gcc.gnu.org/gcc-7/changes.html#brig 6388 12. https://gcc.gnu.org/gcc-7/changes.html#c-family 6389 13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 6390 14. https://gcc.gnu.org/gcc-7/changes.html#c 6391 15. https://gcc.gnu.org/gcc-7/changes.html#cxx 6392 16. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 6393 17. https://wg21.link/p0136 6394 18. https://gcc.gnu.org/gcc-7/changes.html#libstdcxx 6395 19. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 6396 20. https://gcc.gnu.org/gcc-7/changes.html#fortran 6397 21. https://www.openmp.org/specifications/ 6398 22. https://gcc.gnu.org/gcc-7/changes.html#go 6399 23. https://gcc.gnu.org/gcc-7/changes.html#java 6400 24. https://gcc.gnu.org/gcc-7/changes.html#jit 6401 25. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 6402 26. https://gcc.gnu.org/gcc-7/changes.html#targets 6403 27. https://gcc.gnu.org/gcc-7/changes.html#aarch64 6404 28. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 6405 29. https://gcc.gnu.org/gcc-7/changes.html#arc 6406 30. https://gcc.gnu.org/gcc-7/changes.html#arm 6407 31. https://gcc.gnu.org/install/configure.html 6408 32. https://gcc.gnu.org/gcc-7/changes.html#avr 6409 33. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 6410 34. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 6411 35. https://gcc.gnu.org/gcc-7/changes.html#x86 6412 36. https://gcc.gnu.org/gcc-7/changes.html#nvptx 6413 37. https://gcc.gnu.org/wiki/Offloading 6414 38. https://gcc.gnu.org/gcc-7/changes.html#powerpc 6415 39. https://gcc.gnu.org/gcc-7/changes.html#s390 6416 40. https://gcc.gnu.org/gcc-7/changes.html#riscv 6417 41. https://gcc.gnu.org/gcc-7/changes.html#rx 6418 42. https://gcc.gnu.org/gcc-7/changes.html#sh 6419 43. https://gcc.gnu.org/gcc-7/changes.html#sparc 6420 44. https://gcc.gnu.org/gcc-7/changes.html#os 6421 45. https://gcc.gnu.org/gcc-7/changes.html#aix 6422 46. https://gcc.gnu.org/gcc-7/changes.html#fuchsia 6423 47. https://fuchsia.googlesource.com/ 6424 48. https://gcc.gnu.org/gcc-7/changes.html#rtems 6425 49. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 6426 50. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 6427 51. https://gcc.gnu.org/gcc-7/changes.html#GCC7.1 6428 52. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 6429 53. https://gcc.gnu.org/gcc-7/changes.html#GCC7.2 6430 54. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 6431 55. https://gcc.gnu.org/gcc-7/changes.html#GCC7.3 6432 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 6433 57. https://www.gaisler.com/app-notes-tech-notes-and-white-papers 6434 58. https://gcc.gnu.org/gcc-7/changes.html#GCC7.4 6435 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 6436 60. https://gcc.gnu.org/gcc-7/changes.html#GCC7.5 6437 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 6438 62. https://gcc.gnu.org/onlinedocs/ 6439 63. mailto:gcc-help (a] gcc.gnu.org 6440 64. mailto:gcc (a] gcc.gnu.org 6441 65. https://gcc.gnu.org/lists.html 6442 66. https://www.fsf.org/ 6443 67. https://gcc.gnu.org/about.html 6444 ====================================================================== 6445 http://gcc.gnu.org/gcc-6/index.html 6446 6447 GCC 6 Release Series 6448 6449 (This release series is no longer supported.) 6450 6451 October 26, 2018 6452 6453 The [1]GNU project and the GCC developers are pleased to announce the 6454 release of GCC 6.5. 6455 6456 This release is a bug-fix release, containing fixes for regressions in 6457 GCC 6.4 relative to previous releases of GCC. 6458 6459 Release History 6460 6461 GCC 6.5 6462 October 26, 2018 ([2]changes, [3]documentation) 6463 6464 GCC 6.4 6465 July 4, 2017 ([4]changes, [5]documentation) 6466 6467 GCC 6.3 6468 December 21, 2016 ([6]changes, [7]documentation) 6469 6470 GCC 6.2 6471 August 22, 2016 ([8]changes, [9]documentation) 6472 6473 GCC 6.1 6474 April 27, 2016 ([10]changes, [11]documentation) 6475 6476 References and Acknowledgements 6477 6478 GCC used to stand for the GNU C Compiler, but since the compiler 6479 supports several other languages aside from C, it now stands for the 6480 GNU Compiler Collection. 6481 6482 A list of [12]successful builds is updated as new information becomes 6483 available. 6484 6485 The GCC developers would like to thank the numerous people that have 6486 contributed new features, improvements, bug fixes, and other changes as 6487 well as test results to GCC. This [13]amazing group of volunteers is 6488 what makes GCC successful. 6489 6490 For additional information about GCC please refer to the [14]GCC 6491 project web site or contact the [15]GCC development mailing list. 6492 6493 To obtain GCC please use [16]our mirror sites or [17]our version 6494 control system. 6495 6496 6497 For questions related to the use of GCC, please consult these web 6498 pages and the [18]GCC manuals. If that fails, the 6499 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6500 web pages and the development of GCC are welcome on our developer 6501 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6502 archives. 6503 6504 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6505 distribution of this entire article is permitted in any medium, 6506 provided this notice is preserved. 6507 6508 These pages are [23]maintained by the GCC team. Last modified 6509 2024-05-30. 6510 6511 References 6512 6513 1. http://www.gnu.org/ 6514 2. https://gcc.gnu.org/gcc-6/changes.html 6515 3. https://gcc.gnu.org/onlinedocs/6.5.0/ 6516 4. https://gcc.gnu.org/gcc-6/changes.html 6517 5. https://gcc.gnu.org/onlinedocs/6.4.0/ 6518 6. https://gcc.gnu.org/gcc-6/changes.html 6519 7. https://gcc.gnu.org/onlinedocs/6.3.0/ 6520 8. https://gcc.gnu.org/gcc-6/changes.html 6521 9. https://gcc.gnu.org/onlinedocs/6.2.0/ 6522 10. https://gcc.gnu.org/gcc-6/changes.html 6523 11. https://gcc.gnu.org/onlinedocs/6.1.0/ 6524 12. https://gcc.gnu.org/gcc-6/buildstat.html 6525 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6526 14. https://gcc.gnu.org/index.html 6527 15. mailto:gcc (a] gcc.gnu.org 6528 16. https://gcc.gnu.org/mirrors.html 6529 17. https://gcc.gnu.org/git.html 6530 18. https://gcc.gnu.org/onlinedocs/ 6531 19. mailto:gcc-help (a] gcc.gnu.org 6532 20. mailto:gcc (a] gcc.gnu.org 6533 21. https://gcc.gnu.org/lists.html 6534 22. https://www.fsf.org/ 6535 23. https://gcc.gnu.org/about.html 6536 ====================================================================== 6537 http://gcc.gnu.org/gcc-6/changes.html 6538 6539 GCC 6 Release Series 6540 Changes, New Features, and Fixes 6541 6542 This page is a brief summary of some of the huge number of improvements 6543 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 6544 the [2]full GCC documentation. 6545 6546 Caveats 6547 6548 * The default mode for C++ is now -std=gnu++14 instead of 6549 -std=gnu++98. 6550 * Support for a number of older systems and recently unmaintained or 6551 untested target ports of GCC has been declared obsolete in GCC 6. 6552 Unless there is activity to revive them, the next release of GCC 6553 will have their sources permanently removed. 6554 The following ports for individual systems on particular 6555 architectures have been obsoleted: 6556 + SH5 / SH64 (sh64-*-*) as announced [3]here. 6557 * The AVR port requires binutils version 2.26.1 or later for the fix 6558 for [4]PR71151 to work. 6559 * The GCC 6.5 release has an accidental ABI incompatibility for 6560 nested std::pair objects, for more details see [5]PR 87822. The bug 6561 causes a layout change for pairs where the first member is also a 6562 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 6563 is closed so the bug in GCC 6.5 will not be fixed upstream, but 6564 there is a patch in the bug report to allow it to be fixed by 6565 anybody packaging GCC 6.5 or installing it themselves. 6566 6567 [6]General Optimizer Improvements 6568 6569 * UndefinedBehaviorSanitizer gained a new sanitization option, 6570 -fsanitize=bounds-strict, which enables strict checking of array 6571 bounds. In particular, it enables -fsanitize=bounds as well as 6572 instrumentation of flexible array member-like arrays. 6573 * Type-based alias analysis now disambiguates accesses to different 6574 pointers. This improves precision of the alias oracle by about 6575 20-30% on higher-level C++ programs. Programs doing invalid type 6576 punning of pointer types may now need -fno-strict-aliasing to work 6577 correctly. 6578 * Alias analysis now correctly supports the weakref and alias 6579 attributes. This allows accessing both a variable and its alias in 6580 one translation unit which is common with link-time optimization. 6581 * Value range propagation now assumes that the this pointer in C++ 6582 member functions is non-null. This eliminates common null pointer 6583 checks but also breaks some non-conforming code-bases (such as 6584 Qt-5, Chromium, KDevelop). As a temporary work-around 6585 -fno-delete-null-pointer-checks can be used. Wrong code can be 6586 identified by using -fsanitize=undefined. 6587 * Link-time optimization improvements: 6588 + warning and error attributes are now correctly preserved by 6589 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 6590 supported with -flto. 6591 + Type merging was fixed to handle C and Fortran 6592 interoperability rules as defined by the Fortran 2008 language 6593 standard. 6594 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 6595 with char in all cases because it is an array while char is 6596 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 6597 general, this inter-operability cannot be implemented, for 6598 example on targets where the argument passing convention for 6599 arrays differs from scalars. 6600 + More type information is now preserved at link time, reducing 6601 the loss of accuracy of the type-based alias analysis compared 6602 to builds without link-time optimization. 6603 + Invalid type punning on global variables and declarations is 6604 now reported with -Wodr-type-mismatch. 6605 + The size of LTO object files was reduced by about 11% 6606 (measured by compiling Firefox 46.0). 6607 + Link-time parallelization (enabled using -flto=n) was 6608 significantly improved by decreasing the size of streamed data 6609 when partitioning programs. The size of streamed IL while 6610 compiling Firefox 46.0 was reduced by 66%. 6611 + The linker plugin was extended to pass information about the 6612 type of binary produced to the GCC back end. (That can also be 6613 controlled manually by -flinker-output.) This makes it 6614 possible to properly configure the code generator and support 6615 incremental linking. Incremental linking of LTO objects by gcc 6616 -r is now supported for plugin-enabled setups. 6617 There are two ways to perform incremental linking: 6618 1. Linking by ld -r will result in an object file with all 6619 sections from individual object files mechanically 6620 merged. This delays the actual link-time optimization to 6621 the final linking step and thus permits whole program 6622 optimization. Linking the final binary with such object 6623 files is however slower. 6624 2. Linking by gcc -r will lead to link-time optimization and 6625 emit the final binary into the object file. Linking such 6626 an object file is fast but avoids any benefits from whole 6627 program optimization. 6628 GCC 7 will support incremental link-time optimization with gcc 6629 -r. 6630 * Inter-procedural optimization improvements: 6631 + Basic jump threading is now performed before profile 6632 construction and inline analysis, resulting in more realistic 6633 size and time estimates that drive the heuristics of the 6634 inliner and function cloning passes. 6635 + Function cloning now more aggressively eliminates unused 6636 function parameters. 6637 6638 [7]New Languages and Language specific improvements 6639 6640 Compared to GCC 5, the GCC 6 release series includes a much improved 6641 implementation of the [8]OpenACC 2.0a specification. Highlights are: 6642 * In addition to single-threaded host-fallback execution, offloading 6643 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 6644 little-endian GNU/Linux host systems. For nvptx offloading, with 6645 the OpenACC parallel construct, the execution model allows for an 6646 arbitrary number of gangs, up to 32 workers, and 32 vectors. 6647 * Initial support for parallelized execution of OpenACC kernels 6648 constructs: 6649 + Parallelization of a kernels region is switched on by 6650 -fopenacc combined with -O2 or higher. 6651 + Code is offloaded onto multiple gangs, but executes with just 6652 one worker, and a vector length of 1. 6653 + Directives inside a kernels region are not supported. 6654 + Loops with reductions can be parallelized. 6655 + Only kernels regions with one loop nest are parallelized. 6656 + Only the outer-most loop of a loop nest can be parallelized. 6657 + Loop nests containing sibling loops are not parallelized. 6658 Typically, using the OpenACC parallel construct gives much better 6659 performance, compared to the initial support of the OpenACC kernels 6660 construct. 6661 * The device_type clause is not supported. The bind and nohost 6662 clauses are not supported. The host_data directive is not supported 6663 in Fortran. 6664 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 6665 * Usage of OpenACC constructs inside multithreaded contexts (such as 6666 created by OpenMP, or pthread programming) is not supported. 6667 * If a call to the acc_on_device function has a compile-time constant 6668 argument, the function call evaluates to a compile-time constant 6669 value only for C and C++ but not for Fortran. 6670 6671 See the [9]OpenACC and [10]Offloading wiki pages for further 6672 information. 6673 6674 [11]C family 6675 6676 * Version 4.5 of the [12]OpenMP specification is now supported in the 6677 C and C++ compilers. 6678 * The C and C++ compilers now support attributes on enumerators. For 6679 instance, it is now possible to mark enumerators as deprecated: 6680 6681 enum { 6682 newval, 6683 oldval __attribute__ ((deprecated ("too old"))) 6684 }; 6685 6686 * Source locations for the C and C++ compilers are now tracked as 6687 ranges, rather than just points, making it easier to identify the 6688 subexpression of interest within a complicated expression. For 6689 example: 6690 6691 test.cc: In function 'int test(int, int, foo, int, int)': 6692 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 6693 ) 6694 return p + q * r * s + t; 6695 ~~^~~ 6696 6697 In addition, there is now initial support for precise diagnostic 6698 locations within strings: 6699 6700 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 6701 t' argument [-Wformat=] 6702 printf("%*d"); 6703 ^ 6704 6705 * Diagnostics can now contain "fix-it hints", which are displayed in 6706 context underneath the relevant source code. For example: 6707 6708 fixits.c: In function 'bad_deref': 6709 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 6710 return ptr.x; 6711 ^ 6712 -> 6713 6714 * The C and C++ compilers now offer suggestions for misspelled field 6715 names: 6716 6717 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 6718 you mean 'color'? 6719 return ptr->colour; 6720 ^~~~~~ 6721 6722 * New command-line options have been added for the C and C++ 6723 compilers: 6724 + -Wshift-negative-value warns about left shifting a negative 6725 value. 6726 + -Wshift-overflow warns about left shift overflows. This 6727 warning is enabled by default. -Wshift-overflow=2 also warns 6728 about left-shifting 1 into the sign bit. 6729 + -Wtautological-compare warns if a self-comparison always 6730 evaluates to true or false. This warning is enabled by -Wall. 6731 + -Wnull-dereference warns if the compiler detects paths that 6732 trigger erroneous or undefined behavior due to dereferencing a 6733 null pointer. This option is only active when 6734 -fdelete-null-pointer-checks is active, which is enabled by 6735 optimizations in most targets. The precision of the warnings 6736 depends on the optimization options used. 6737 + -Wduplicated-cond warns about duplicated conditions in an 6738 if-else-if chain. 6739 + -Wmisleading-indentation warns about places where the 6740 indentation of the code gives a misleading idea of the block 6741 structure of the code to a human reader. For example, given 6742 [13]CVE-2014-1266: 6743 6744 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 6745 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 6746 g-indentation] 6747 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 6748 ^~ 6749 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 6750 indented as if it is guarded by the 'if' 6751 goto fail; 6752 ^~~~ 6753 6754 This warning is enabled by -Wall. 6755 * The C and C++ compilers now emit saner error messages if 6756 merge-conflict markers are present in a source file. 6757 6758 test.c:3:1: error: version control conflict marker in file 6759 <<<<<<< HEAD 6760 ^~~~~~~ 6761 6762 [14]C 6763 6764 * It is possible to disable warnings when an initialized field of a 6765 structure or a union with side effects is being overridden when 6766 using designated initializers via a new warning option 6767 -Woverride-init-side-effects. 6768 * A new type attribute scalar_storage_order applying to structures 6769 and unions has been introduced. It specifies the storage order (aka 6770 endianness) in memory of scalar fields in structures or unions. 6771 6772 [15]C++ 6773 6774 * The default mode has been changed to -std=gnu++14. 6775 * [16]C++ Concepts are now supported when compiling with -fconcepts. 6776 * -flifetime-dse is more aggressive in dead-store elimination in 6777 situations where a memory store to a location precedes a 6778 constructor to that memory location. 6779 * G++ now supports [17]C++17 fold expressions, u8 character literals, 6780 extended static_assert, and nested namespace definitions. 6781 * G++ now allows constant evaluation for all non-type template 6782 arguments. 6783 * G++ now supports C++ Transactional Memory when compiling with 6784 -fgnu-tm. 6785 6786 [18]Runtime Library (libstdc++) 6787 6788 * Extensions to the C++ Library to support mathematical special 6789 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 6790 * Experimental support for C++17, including the following new 6791 features: 6792 + std::uncaught_exceptions function (this is also available for 6793 -std=gnu++NN modes); 6794 + new member functions try_emplace and insert_or_assign for 6795 unique_key maps; 6796 + non-member functions std::size, std::empty, and std::data for 6797 accessing containers and arrays; 6798 + std::invoke; 6799 + std::shared_mutex; 6800 + std::void_t and std::bool_constant metaprogramming utilities. 6801 Thanks to Ville Voutilainen for contributing many of the C++17 6802 features. 6803 * An experimental implementation of the File System TS. 6804 * Experimental support for most features of the second version of the 6805 Library Fundamentals TS. This includes polymorphic memory resources 6806 and array support in shared_ptr, thanks to Fan You. 6807 * Some assertions checked by Debug Mode can now also be enabled by 6808 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 6809 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 6810 don't affect the library ABI, so can be enabled per-translation 6811 unit. 6812 * Timed mutex types are supported on more targets, including Darwin. 6813 * Improved std::locale support for DragonFly and FreeBSD, thanks to 6814 John Marino and Andreas Tobler. 6815 6816 [19]Fortran 6817 6818 * Fortran 2008 SUBMODULE support. 6819 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 6820 support. 6821 * Improved support for Fortran 2003 deferred-length character 6822 variables. 6823 * Improved support for OpenMP and OpenACC. 6824 * The MATMUL intrinsic is now inlined for straightforward cases if 6825 front-end optimization is active. The maximum size for inlining can 6826 be set to n with the -finline-matmul-limit=n option and turned off 6827 with -finline-matmul-limit=0. 6828 * The -Wconversion-extra option will warn about REAL constants which 6829 have excess precision for their kind. 6830 * The -Winteger-division option has been added, which warns about 6831 divisions of integer constants which are truncated. This option is 6832 included in -Wall by default. 6833 6834 [20]libgccjit 6835 6836 * The driver code is now run in-process within libgccjit, providing a 6837 small speed-up of the compilation process. 6838 * The API has gained entrypoints for 6839 + [21]timing how long was spent in different parts of code, 6840 + [22]creating switch statements, 6841 + [23]allowing unreachable basic blocks in a function, and 6842 + [24]adding arbitrary command-line options to a compilation. 6843 6844 [25]New Targets and Target Specific Improvements 6845 6846 [26]AArch64 6847 6848 * A number of AArch64-specific options have been added. The most 6849 important ones are summarised in this section; for more detailed 6850 information please refer to the documentation. 6851 * The command-line options -march=native, -mcpu=native and 6852 -mtune=native are now available on native AArch64 GNU/Linux 6853 systems. Specifying these options causes GCC to auto-detect the 6854 host CPU and choose the optimal setting for that system. 6855 * -fpic is now supported when generating code for the small code 6856 model (-mcmodel=small). The size of the global offset table (GOT) 6857 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 6858 ILP32 SysV ABI. 6859 * The AArch64 port now supports target attributes and pragmas. Please 6860 refer to the [27]documentation for details of available attributes 6861 and pragmas as well as usage instructions. 6862 * Link-time optimization across translation units with different 6863 target-specific options is now supported. 6864 * The option -mtls-size= is now supported. It can be used to specify 6865 the bit size of TLS offsets, allowing GCC to generate better TLS 6866 instruction sequences. 6867 * The option -fno-plt is now fully functional. 6868 * The ARMv8.1-A architecture and the Large System Extensions are now 6869 supported. They can be used by specifying the -march=armv8.1-a 6870 option. Additionally, the +lse option extension can be used in a 6871 similar fashion to other option extensions. The Large System 6872 Extensions introduce new instructions that are used in the 6873 implementation of atomic operations. 6874 * The ACLE half-precision floating-point type __fp16 is now supported 6875 in the C and C++ languages. 6876 * The ARM Cortex-A35 processor is now supported via the 6877 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 6878 equivalent target attributes and pragmas. 6879 * The Qualcomm QDF24xx processor is now supported via the 6880 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 6881 target attributes and pragmas. 6882 * Code generation for the ARM Cortex-A57 processor is improved. Among 6883 general code generation improvements, a better algorithm is added 6884 for allocating registers to floating-point multiply-accumulate 6885 instructions offering increased performance when compiling with 6886 -mcpu=cortex-a57 or -mtune=cortex-a57. 6887 * Code generation for the ARM Cortex-A53 processor is improved. A 6888 more accurate instruction scheduling model for the processor is now 6889 used, and a number of compiler tuning parameters have been set to 6890 offer increased performance when compiling with -mcpu=cortex-a53 or 6891 -mtune=cortex-a53. 6892 * Code generation for the Samsung Exynos M1 processor is improved. A 6893 more accurate instruction scheduling model for the processor is now 6894 used, and a number of compiler tuning parameters have been set to 6895 offer increased performance when compiling with -mcpu=exynos-m1 or 6896 -mtune=exynos-m1. 6897 * Improvements in the generation of conditional branches and literal 6898 pools allow the compiler to compile functions of a large size. 6899 Constant pools are now placed into separate rodata sections. The 6900 new option -mpc-relative-literal-loads generates per-function 6901 literal pools, limiting the maximum size of functions to 1MiB. 6902 * Several correctness issues generating Advanced SIMD instructions 6903 for big-endian targets have been fixed resulting in improved code 6904 generation for ACLE intrinsics with -mbig-endian. 6905 6906 [28]ARM 6907 6908 * Support for revisions of the ARM architecture prior to ARMv4t has 6909 been deprecated and will be removed in a future GCC release. The 6910 -mcpu and -mtune values that are deprecated are: arm2, arm250, 6911 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 6912 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 6913 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 6914 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 6915 arm7tdmi is still supported. The values of -march that are 6916 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 6917 * The ARM port now supports target attributes and pragmas. Please 6918 refer to the [29]documentation for details of available attributes 6919 and pragmas as well as usage instructions. 6920 * Support has been added for the following processors (GCC 6921 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 6922 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 6923 identifiers can be used as arguments to the -mcpu or -mtune 6924 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 6925 6926 [30]Heterogeneous Systems Architecture 6927 6928 * GCC can now generate HSAIL (Heterogeneous System Architecture 6929 Intermediate Language) for simple OpenMP device constructs if 6930 configured with --enable-offload-targets=hsa. A new libgomp plugin 6931 then runs the HSA GPU kernels implementing these constructs on HSA 6932 capable GPUs via a standard HSA run time. 6933 If the HSA compilation back end determines it cannot output HSAIL 6934 for a particular input, it gives a warning by default. These 6935 warnings can be suppressed with -Wno-hsa. To give a few examples, 6936 the HSA back end does not implement compilation of code using 6937 function pointers, automatic allocation of variable sized arrays, 6938 functions with variadic arguments as well as a number of other less 6939 common programming constructs. 6940 When compilation for HSA is enabled, the compiler attempts to 6941 compile composite OpenMP constructs 6942 6943 #pragma omp target teams distribute parallel for 6944 6945 into parallel HSA GPU kernels. 6946 6947 [31]IA-32/x86-64 6948 6949 * GCC now supports the Intel CPU named Skylake with AVX-512 6950 extensions through -march=skylake-avx512. The switch enables the 6951 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 6952 AVX-512DQ. 6953 * Support for new AMD instructions monitorx and mwaitx has been 6954 added. This includes new intrinsic and built-in support. It is 6955 enabled through option -mmwaitx. The instructions monitorx and 6956 mwaitx implement the same functionality as the old monitor and 6957 mwait instructions. In addition, mwaitx adds a configurable timer. 6958 The timer value is received as third argument and stored in 6959 register %ebx. 6960 * x86-64 targets now allow stack realignment from a word-aligned 6961 stack pointer using the command-line option -mstackrealign or 6962 __attribute__ ((force_align_arg_pointer)). This allows functions 6963 compiled with a vector-aligned stack to be invoked from objects 6964 that keep only word-alignment. 6965 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 6966 can be used to access data via the %fs and %gs segments without 6967 having to resort to inline assembly. Please refer to the 6968 [32]documentation for usage instructions. 6969 * Support for AMD Zen (family 17h) processors is now available 6970 through the -march=znver1 and -mtune=znver1 options. 6971 6972 [33]MeP 6973 6974 * Support for the MeP (mep-elf) architecture has been deprecated and 6975 will be removed in a future GCC release. 6976 6977 [34]MSP430 6978 6979 * The MSP430 compiler now has the ability to automatically distribute 6980 code and data between low memory (addresses below 64K) and high 6981 memory. This only applies to parts that actually have both memory 6982 regions and only if the linker script for the part has been 6983 specifically set up to support this feature. 6984 A new attribute of either can be applied to both functions and 6985 data, and this tells the compiler to place the object into low 6986 memory if there is room and into high memory otherwise. Two other 6987 new attributes - lower and upper - can be used to explicitly state 6988 that an object should be placed in the specified memory region. If 6989 there is not enough left in that region the compilation will fail. 6990 Two new command-line options - -mcode-region=[lower|upper|either] 6991 and -mdata-region=[lower|upper|either] - can be used to tell the 6992 compiler what to do with objects that do not have one of these new 6993 attributes. 6994 6995 [35]PowerPC / PowerPC64 / RS6000 6996 6997 * PowerPC64 now supports IEEE 128-bit floating-point using the 6998 __float128 data type. In GCC 6, this is not enabled by default, but 6999 you can enable it with -mfloat128. The IEEE 128-bit floating-point 7000 support requires the use of the VSX instruction set. IEEE 128-bit 7001 floating-point values are passed and returned as a single vector 7002 value. The software emulator for IEEE 128-bit floating-point 7003 support is only built on PowerPC GNU/Linux systems where the 7004 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 7005 and later), you will be able to use the -mfloat128-hardware option 7006 to use the ISA 3.0 instructions that support IEEE 128-bit 7007 floating-point. An additional type (__ibm128) has been added to 7008 refer to the IBM extended double type that normally implements long 7009 double. This will allow for a future transition to implementing 7010 long double with IEEE 128-bit floating-point. 7011 * Basic support has been added for POWER9 hardware that will use the 7012 recently published OpenPOWER ISA 3.0 instructions. The following 7013 new switches are available: 7014 + -mcpu=power9: Implement all of the ISA 3.0 instructions 7015 supported by the compiler. 7016 + -mtune=power9: In the future, apply tuning for POWER9 systems. 7017 Currently, POWER8 tunings are used. 7018 + -mmodulo: Generate code using the ISA 3.0 integer instructions 7019 (modulus, count trailing zeros, array index support, integer 7020 multiply/add). 7021 + -mpower9-fusion: Generate code to suitably fuse instruction 7022 sequences for a POWER9 system. 7023 + -mpower9-dform: Generate code to use the new D-form 7024 (register+offset) memory instructions for the vector 7025 registers. 7026 + -mpower9-vector: Generate code using the new ISA 3.0 vector 7027 (VSX or Altivec) instructions. 7028 + -mpower9-minmax: Reserved for future development. 7029 + -mtoc-fusion: Keep TOC entries together to provide more fusion 7030 opportunities. 7031 * New constraints have been added to support IEEE 128-bit 7032 floating-point and ISA 3.0 instructions: 7033 + wb: Altivec register if -mpower9-dform is enabled. 7034 + we: VSX register if -mpower9-vector is enabled for 64-bit code 7035 generation. 7036 + wo: VSX register if -mpower9-vector is enabled. 7037 + wp: Reserved for future use if long double is implemented with 7038 IEEE 128-bit floating-point instead of IBM extended double. 7039 + wq: VSX register if -mfloat128 is enabled. 7040 + wF: Memory operand suitable for POWER9 fusion load/store. 7041 + wG: Memory operand suitable for TOC fusion memory references. 7042 + wL: Integer constant identifying the element number mfvsrld 7043 accesses within a vector. 7044 * Support has been added for __builtin_cpu_is() and 7045 __builtin_cpu_supports(), allowing for very fast access to 7046 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 7047 glibc 2.23 or later. 7048 * All hardware transactional memory builtins now correctly behave as 7049 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 7050 determine whether their "old" compiler treats the builtins as 7051 barriers. 7052 * Split-stack support has been added for gccgo on PowerPC64 for both 7053 big- and little-endian (but not for 32-bit). The gold linker from 7054 at least binutils 2.25.1 must be available in the PATH when 7055 configuring and building gccgo to enable split stack. (The 7056 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 7057 split-stack feature allows a small initial stack size to be 7058 allocated for each goroutine, which increases as needed. 7059 * GCC on PowerPC now supports the standard lround function. 7060 * A new configuration option ---with-advance-toolchain=at was added 7061 for PowerPC 64-bit GNU/Linux systems to use the header files, 7062 library files, and the dynamic linker from a specific Advance 7063 Toolchain release instead of the default versions that are provided 7064 by the GNU/Linux distribution. In general, this option is intended 7065 for the developers of GCC, and it is not intended for general use. 7066 * The "q", "S", "T", and "t" asm-constraints have been removed. 7067 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 7068 7069 [36]S/390, System z, IBM z Systems 7070 7071 * Support for the IBM z13 processor has been added. When using the 7072 -march=z13 option, the compiler will generate code making use of 7073 the new instructions and registers introduced with the vector 7074 extension facility. The -mtune=z13 option enables z13 specific 7075 instruction scheduling without making use of new instructions. 7076 Compiling code with -march=z13 reduces the default alignment of 7077 vector types bigger than 8 bytes to 8. This is an ABI change and 7078 care must be taken when linking modules compiled with different 7079 arch levels which interchange variables containing vector type 7080 values. For newly compiled code the GNU linker will emit a warning. 7081 * The -mzvector option enables a C/C++ language extension. This 7082 extension provides a new keyword vector which can be used to define 7083 vector type variables. (Note: This is not available when enforcing 7084 strict standard compliance e.g. with -std=c99. Either enable GNU 7085 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 7086 Additionally a set of overloaded builtins is provided which is 7087 partially compatible to the PowerPC Altivec builtins. In order to 7088 make use of these builtins the vecintrin.h header file needs to be 7089 included. 7090 * The new command-line options -march=native, and -mtune=native are 7091 now available on native IBM z Systems. Specifying these options 7092 causes GCC to auto-detect the host CPU and choose the optimal 7093 setting for that system. If GCC is unable to detect the host CPU 7094 these options have no effect. 7095 * The IBM z Systems port now supports target attributes and pragmas. 7096 Please refer to the [37]documentation for details of available 7097 attributes and pragmas as well as usage instructions. 7098 * -fsplit-stack is now supported as part of the IBM z Systems port. 7099 This feature requires a recent gold linker to be used. 7100 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 7101 been deprecated and will be removed in a future GCC release. -m31 7102 from now on defaults to -march=z900 if not specified otherwise. 7103 -march=native on a g5/g6 machine will default to -march=z900. 7104 7105 [38]SH 7106 7107 * Support for SH5 / SH64 has been declared obsolete and will be 7108 removed in future releases. 7109 * Support for the FDPIC ABI has been added. It can be enabled using 7110 the new -mfdpic target option and --enable-fdpic configure option. 7111 7112 [39]SPARC 7113 7114 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 7115 change will break binary compatibility with earlier releases for 7116 code it affects, but this should be pretty rare in practice. The 7117 conditions are: a 16-byte structure containing a double or a 8-byte 7118 vector in the second half is passed to a subprogram in slot #15, 7119 for example as 16th parameter if the first 15 ones have at most 8 7120 bytes. The double or vector was wrongly passed in floating-point 7121 register %d32 in lieu of on the stack as per the SPARC calling 7122 conventions. 7123 7124 [40]Operating Systems 7125 7126 [41]AIX 7127 7128 * DWARF debugging support for AIX 7.1 has been enabled as an optional 7129 debugging format. A more recent Technology Level (TL) and GCC built 7130 with that level are required for full exploitation of DWARF 7131 debugging capabilities. 7132 7133 [42]Linux 7134 7135 * Support for the [43]musl C library was added for the AArch64, ARM, 7136 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 7137 x86_64 targets. It can be selected using the new -mmusl option in 7138 case musl is not the default libc. GCC defaults to musl libc if it 7139 is built with a target triplet matching the *-linux-musl* pattern. 7140 7141 [44]RTEMS 7142 7143 * The RTEMS thread model implementation changed. Mutexes now use 7144 self-contained objects defined in newlib <sys/lock.h> instead of 7145 Classic API semaphores. The keys for thread specific data and the 7146 once function are directly defined via <pthread.h>. Self-contained 7147 condition variables are provided via newlib <sys/lock.h>. The RTEMS 7148 thread model also supports C++11 threads. 7149 * OpenMP support now uses self-contained objects provided by newlib 7150 <sys/lock.h> and offers a significantly better performance compared 7151 to the POSIX configuration of libgomp. It is possible to configure 7152 thread pools for each scheduler instance via the environment 7153 variable GOMP_RTEMS_THREAD_POOLS. 7154 7155 [45]Solaris 7156 7157 * Solaris 12 is now fully supported. Minimal support had already been 7158 present in GCC 5.3. 7159 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 7160 crtn.o), which GCC now prefers over its own ones. 7161 * Position independent executables (PIE) are now supported on Solaris 7162 12. 7163 * Constructor priority is now supported on Solaris 12 with the system 7164 linker. 7165 * libvtv has been ported to Solaris 11 and up. 7166 7167 [46]Windows 7168 7169 * The option -mstackrealign is now automatically activated in 32-bit 7170 mode whenever the use of SSE instructions is requested. 7171 7172 Other significant improvements 7173 7174 * The gcc and g++ driver programs will now provide suggestions for 7175 misspelled command-line options. 7176 7177 $ gcc -static-libfortran test.f95 7178 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 7179 '-static-libgfortran'? 7180 7181 * The --enable-default-pie configure option enables generation of PIE 7182 by default. 7183 7184 [47]GCC 6.2 7185 7186 This is the [48]list of problem reports (PRs) from GCC's bug tracking 7187 system that are known to be fixed in the 6.2 release. This list might 7188 not be complete (that is, it is possible that some PRs that have been 7189 fixed are not listed here). 7190 7191 Target Specific Changes 7192 7193 SPARC 7194 7195 * Support for --with-cpu-32 and --with-cpu-64 configure options has 7196 been added on bi-architecture platforms. 7197 * Support for the SPARC M7 (Niagara 7) processor has been added. 7198 * Support for the VIS 4.0 instruction set has been added. 7199 7200 [49]GCC 6.3 7201 7202 This is the [50]list of problem reports (PRs) from GCC's bug tracking 7203 system that are known to be fixed in the 6.3 release. This list might 7204 not be complete (that is, it is possible that some PRs that have been 7205 fixed are not listed here). 7206 7207 Target Specific Changes 7208 7209 IA-32/x86-64 7210 7211 * Support for the [51]deprecated pcommit instruction has been 7212 removed. 7213 7214 [52]GCC 6.4 7215 7216 This is the [53]list of problem reports (PRs) from GCC's bug tracking 7217 system that are known to be fixed in the 6.4 release. This list might 7218 not be complete (that is, it is possible that some PRs that have been 7219 fixed are not listed here). 7220 7221 [54]Operating Systems 7222 7223 [55]RTEMS 7224 7225 * The ABI changes on ARM so that no short enums are used by default. 7226 7227 [56]GCC 6.5 7228 7229 This is the [57]list of problem reports (PRs) from GCC's bug tracking 7230 system that are known to be fixed in the 6.5 release. This list might 7231 not be complete (that is, it is possible that some PRs that have been 7232 fixed are not listed here). 7233 7234 7235 For questions related to the use of GCC, please consult these web 7236 pages and the [58]GCC manuals. If that fails, the 7237 [59]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7238 web pages and the development of GCC are welcome on our developer 7239 list at [60]gcc (a] gcc.gnu.org. All of [61]our lists have public 7240 archives. 7241 7242 Copyright (C) [62]Free Software Foundation, Inc. Verbatim copying and 7243 distribution of this entire article is permitted in any medium, 7244 provided this notice is preserved. 7245 7246 These pages are [63]maintained by the GCC team. Last modified 7247 2025-01-31. 7248 7249 References 7250 7251 1. https://gcc.gnu.org/gcc-6/porting_to.html 7252 2. https://gcc.gnu.org/onlinedocs/index.html#current 7253 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 7254 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 7255 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 7256 6. https://gcc.gnu.org/gcc-6/changes.html#general 7257 7. https://gcc.gnu.org/gcc-6/changes.html#languages 7258 8. https://www.openacc.org/ 7259 9. https://gcc.gnu.org/wiki/OpenACC 7260 10. https://gcc.gnu.org/wiki/Offloading 7261 11. https://gcc.gnu.org/gcc-6/changes.html#c-family 7262 12. https://www.openmp.org/specifications/ 7263 13. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 7264 14. https://gcc.gnu.org/gcc-6/changes.html#c 7265 15. https://gcc.gnu.org/gcc-6/changes.html#cxx 7266 16. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 7267 17. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 7268 18. https://gcc.gnu.org/gcc-6/changes.html#libstdcxx 7269 19. https://gcc.gnu.org/gcc-6/changes.html#fortran 7270 20. https://gcc.gnu.org/gcc-6/changes.html#jit 7271 21. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 7272 22. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 7273 23. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 7274 24. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 7275 25. https://gcc.gnu.org/gcc-6/changes.html#targets 7276 26. https://gcc.gnu.org/gcc-6/changes.html#aarch64 7277 27. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 7278 28. https://gcc.gnu.org/gcc-6/changes.html#arm 7279 29. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 7280 30. https://gcc.gnu.org/gcc-6/changes.html#hsa 7281 31. https://gcc.gnu.org/gcc-6/changes.html#x86 7282 32. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 7283 33. https://gcc.gnu.org/gcc-6/changes.html#mep 7284 34. https://gcc.gnu.org/gcc-6/changes.html#msp430 7285 35. https://gcc.gnu.org/gcc-6/changes.html#powerpc 7286 36. https://gcc.gnu.org/gcc-6/changes.html#s390 7287 37. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 7288 38. https://gcc.gnu.org/gcc-6/changes.html#sh 7289 39. https://gcc.gnu.org/gcc-6/changes.html#sparc 7290 40. https://gcc.gnu.org/gcc-6/changes.html#os 7291 41. https://gcc.gnu.org/gcc-6/changes.html#aix 7292 42. https://gcc.gnu.org/gcc-6/changes.html#linux 7293 43. http://www.musl-libc.org/ 7294 44. https://gcc.gnu.org/gcc-6/changes.html#rtems 7295 45. https://gcc.gnu.org/gcc-6/changes.html#solaris 7296 46. https://gcc.gnu.org/gcc-6/changes.html#windows 7297 47. https://gcc.gnu.org/gcc-6/changes.html#GCC6.2 7298 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 7299 49. https://gcc.gnu.org/gcc-6/changes.html#GCC6.3 7300 50. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 7301 51. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 7302 52. https://gcc.gnu.org/gcc-6/changes.html#GCC6.4 7303 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 7304 54. https://gcc.gnu.org/gcc-6/changes.html#os64 7305 55. https://gcc.gnu.org/gcc-6/changes.html#rtems64 7306 56. https://gcc.gnu.org/gcc-6/changes.html#GCC6.5 7307 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 7308 58. https://gcc.gnu.org/onlinedocs/ 7309 59. mailto:gcc-help (a] gcc.gnu.org 7310 60. mailto:gcc (a] gcc.gnu.org 7311 61. https://gcc.gnu.org/lists.html 7312 62. https://www.fsf.org/ 7313 63. https://gcc.gnu.org/about.html 7314 ====================================================================== 7315 http://gcc.gnu.org/gcc-5/index.html 7316 7317 GCC 5 Release Series 7318 7319 (This release series is no longer supported.) 7320 7321 October 10, 2017 7322 7323 The [1]GNU project and the GCC developers are pleased to announce the 7324 release of GCC 5.5. 7325 7326 This release is a bug-fix release, containing fixes for regressions in 7327 GCC 5.4 relative to previous releases of GCC. 7328 7329 Release History 7330 7331 GCC 5.5 7332 October 10, 2017 ([2]changes, [3]documentation) 7333 7334 GCC 5.4 7335 June 3, 2016 ([4]changes, [5]documentation) 7336 7337 GCC 5.3 7338 December 4, 2015 ([6]changes, [7]documentation) 7339 7340 GCC 5.2 7341 July 16, 2015 ([8]changes, [9]documentation) 7342 7343 GCC 5.1 7344 April 22, 2015 ([10]changes, [11]documentation) 7345 7346 References and Acknowledgements 7347 7348 GCC used to stand for the GNU C Compiler, but since the compiler 7349 supports several other languages aside from C, it now stands for the 7350 GNU Compiler Collection. 7351 7352 A list of [12]successful builds is updated as new information becomes 7353 available. 7354 7355 The GCC developers would like to thank the numerous people that have 7356 contributed new features, improvements, bug fixes, and other changes as 7357 well as test results to GCC. This [13]amazing group of volunteers is 7358 what makes GCC successful. 7359 7360 For additional information about GCC please refer to the [14]GCC 7361 project web site or contact the [15]GCC development mailing list. 7362 7363 To obtain GCC please use [16]our mirror sites or [17]our version 7364 control system. 7365 7366 7367 For questions related to the use of GCC, please consult these web 7368 pages and the [18]GCC manuals. If that fails, the 7369 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7370 web pages and the development of GCC are welcome on our developer 7371 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 7372 archives. 7373 7374 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 7375 distribution of this entire article is permitted in any medium, 7376 provided this notice is preserved. 7377 7378 These pages are [23]maintained by the GCC team. Last modified 7379 2024-05-30. 7380 7381 References 7382 7383 1. http://www.gnu.org/ 7384 2. https://gcc.gnu.org/gcc-5/changes.html 7385 3. https://gcc.gnu.org/onlinedocs/5.5.0/ 7386 4. https://gcc.gnu.org/gcc-5/changes.html 7387 5. https://gcc.gnu.org/onlinedocs/5.4.0/ 7388 6. https://gcc.gnu.org/gcc-5/changes.html 7389 7. https://gcc.gnu.org/onlinedocs/5.3.0/ 7390 8. https://gcc.gnu.org/gcc-5/changes.html 7391 9. https://gcc.gnu.org/onlinedocs/5.2.0/ 7392 10. https://gcc.gnu.org/gcc-5/changes.html 7393 11. https://gcc.gnu.org/onlinedocs/5.1.0/ 7394 12. https://gcc.gnu.org/gcc-5/buildstat.html 7395 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7396 14. https://gcc.gnu.org/index.html 7397 15. mailto:gcc (a] gcc.gnu.org 7398 16. https://gcc.gnu.org/mirrors.html 7399 17. https://gcc.gnu.org/git.html 7400 18. https://gcc.gnu.org/onlinedocs/ 7401 19. mailto:gcc-help (a] gcc.gnu.org 7402 20. mailto:gcc (a] gcc.gnu.org 7403 21. https://gcc.gnu.org/lists.html 7404 22. https://www.fsf.org/ 7405 23. https://gcc.gnu.org/about.html 7406 ====================================================================== 7407 http://gcc.gnu.org/gcc-5/changes.html 7408 7409 GCC 5 Release Series 7410 Changes, New Features, and Fixes 7411 7412 Caveats 7413 7414 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 7415 * The C++ runtime library (libstdc++) uses a new ABI by default (see 7416 [1]below). 7417 * The Graphite framework for loop optimizations no longer requires 7418 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 7419 The installation manual contains more information about 7420 requirements to build GCC. 7421 * The non-standard C++0x type traits has_trivial_default_constructor, 7422 has_trivial_copy_constructor and has_trivial_copy_assign have been 7423 deprecated and will be removed in a future version. The standard 7424 C++11 traits is_trivially_default_constructible, 7425 is_trivially_copy_constructible and is_trivially_copy_assignable 7426 should be used instead. 7427 7428 [2]General Optimizer Improvements 7429 7430 * Inter-procedural optimization improvements: 7431 + An Identical Code Folding (ICF) pass (controlled via 7432 -fipa-icf) has been added. Compared to the identical code 7433 folding performed by the Gold linker this pass does not 7434 require function sections. It also performs merging before 7435 inlining, so inter-procedural optimizations are aware of the 7436 code re-use. On the other hand not all unifications performed 7437 by a linker are doable by GCC which must honor aliasing 7438 information. During link-time optimization of Firefox, this 7439 pass unifies about 31000 functions, that is 14% overall. 7440 + The devirtualization pass was significantly improved by adding 7441 better support for speculative devirtualization and dynamic 7442 type detection. About 50% of virtual calls in Firefox are now 7443 speculatively devirtualized during link-time optimization. 7444 + A new comdat localization pass allows the linker to eliminate 7445 more dead code in presence of C++ inline functions. 7446 + Virtual tables are now optimized. Local aliases are used to 7447 reduce dynamic linking time of C++ virtual tables on ELF 7448 targets and data alignment has been reduced to limit data 7449 segment bloat. 7450 + A new -fno-semantic-interposition option can be used to 7451 improve code quality of shared libraries where interposition 7452 of exported symbols is not allowed. 7453 + Write-only variables are now detected and optimized out. 7454 + With profile feedback the function inliner can now bypass 7455 --param inline-insns-auto and --param inline-insns-single 7456 limits for hot calls. 7457 + The IPA reference pass was significantly sped up making it 7458 feasible to enable -fipa-reference with -fprofile-generate. 7459 This also solves a bottleneck seen when building Chromium with 7460 link-time optimization. 7461 + The symbol table and call-graph API was reworked to C++ and 7462 simplified. 7463 + The interprocedural propagation of constants now also 7464 propagates alignments of pointer parameters. This for example 7465 means that the vectorizer often does not need to generate loop 7466 prologues and epilogues to make up for potential 7467 misalignments. 7468 * Link-time optimization improvements: 7469 + One Definition Rule based merging of C++ types has been 7470 implemented. Type merging enables better devirtualization and 7471 alias analysis. Streaming extra information needed to merge 7472 types adds about 2-6% of memory size and object size increase. 7473 This can be controlled by -flto-odr-type-merging. 7474 + Command-line optimization and target options are now streamed 7475 on a per-function basis and honored by the link-time 7476 optimizer. This change makes link-time optimization a more 7477 transparent replacement of per-file optimizations. It is now 7478 possible to build projects that require different optimization 7479 settings for different translation units (such as -ffast-math, 7480 -mavx, or -finline). Contrary to earlier GCC releases, the 7481 optimization and target options passed on the link command 7482 line are ignored. 7483 Note that this applies only to those command-line options that 7484 can be passed to optimize and target attributes. Command-line 7485 options affecting global code generation (such as -fpic), 7486 warnings (such as -Wodr), optimizations affecting the way 7487 static variables are optimized (such as -fcommon), debug 7488 output (such as -g), and --param parameters can be applied 7489 only to the whole link-time optimization unit. In these cases, 7490 it is recommended to consistently use the same options at both 7491 compile time and link time. 7492 + GCC bootstrap now uses slim LTO object files. 7493 + Memory usage and link times were improved. Tree merging was 7494 sped up, memory usage of GIMPLE declarations and types was 7495 reduced, and, support for on-demand streaming of variable 7496 constructors was added. 7497 * Feedback directed optimization improvements: 7498 + A new auto-FDO mode uses profiles collected by low overhead 7499 profiling tools (perf) instead of more expensive program 7500 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 7501 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 7502 traditional feedback directed optimization. 7503 + Profile precision was improved in presence of C++ inline and 7504 extern inline functions. 7505 + The new gcov-tool utility allows manipulating profiles. 7506 + Profiles are now more tolerant to source file changes (this 7507 can be controlled by --param profile-func-internal-id). 7508 * Register allocation improvements: 7509 + A new local register allocator (LRA) sub-pass, controlled by 7510 -flra-remat, implements control-flow sensitive global register 7511 rematerialization. Instead of spilling and restoring a 7512 register value, it is recalculated if it is profitable. The 7513 sub-pass improved SPEC2000 generated code by 1% and 0.5% 7514 correspondingly on ARM and x86-64. 7515 + Reuse of the PIC hard register, instead of using a fixed 7516 register, was implemented on x86/x86-64 targets. This improves 7517 generated PIC code performance as more hard registers can be 7518 used. Shared libraries can significantly benefit from this 7519 optimization. Currently it is switched on only for x86/x86-64 7520 targets. As RA infrastructure is already implemented for PIC 7521 register reuse, other targets might follow this in the future. 7522 + A simple form of inter-procedural RA was implemented. When it 7523 is known that a called function does not use caller-saved 7524 registers, save/restore code is not generated around the call 7525 for such registers. This optimization can be controlled by 7526 -fipa-ra 7527 + LRA is now much more effective at generating spills of general 7528 registers into vector registers instead of memory on 7529 architectures (e.g., modern Intel processors) where this is 7530 profitable. 7531 * UndefinedBehaviorSanitizer gained a few new sanitization options: 7532 + -fsanitize=float-divide-by-zero: detect floating-point 7533 division by zero; 7534 + -fsanitize=float-cast-overflow: check that the result of 7535 floating-point type to integer conversions do not overflow; 7536 + -fsanitize=bounds: enable instrumentation of array bounds and 7537 detect out-of-bounds accesses; 7538 + -fsanitize=alignment: enable alignment checking, detect 7539 various misaligned objects; 7540 + -fsanitize=object-size: enable object size checking, detect 7541 various out-of-bounds accesses. 7542 + -fsanitize=vptr: enable checking of C++ member function calls, 7543 member accesses and some conversions between pointers to base 7544 and derived classes, detect if the referenced object does not 7545 have the correct dynamic type. 7546 * Pointer Bounds Checker, a bounds violation detector, has been added 7547 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 7548 instrumented with run-time checks of used pointers against their 7549 bounds to detect pointer bounds violations (overflows). The Pointer 7550 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 7551 new ISA extension Intel MPX support. See the Pointer Bounds Checker 7552 [3]Wiki page for more details. 7553 7554 [4]New Languages and Language specific improvements 7555 7556 * [5]OpenMP 4.0 specification offloading features are now supported 7557 by the C, C++, and Fortran compilers. Generic changes: 7558 + Infrastructure (suitable for any vendor). 7559 + Testsuite which covers offloading from the [6]OpenMP 4.0 7560 Examples document. 7561 Specific for upcoming Intel Xeon Phi products: 7562 + Run-time library. 7563 + Card emulator. 7564 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 7565 specification. OpenACC is intended for programming accelerator 7566 devices such as GPUs. See [7]the OpenACC wiki page for more 7567 information. 7568 7569 [8]C family 7570 7571 * The default setting of the -fdiagnostics-color= command-line option 7572 is now [9]configurable when building GCC using configuration option 7573 --with-diagnostics-color=. The possible values are: never, always, 7574 auto and auto-if-env. The new default auto uses color only when the 7575 standard error is a terminal. The default in GCC 4.9 was 7576 auto-if-env, which is equivalent to auto if there is a non-empty 7577 GCC_COLORS environment variable, and never otherwise. As in GCC 7578 4.9, an empty GCC_COLORS variable in the environment will always 7579 disable colors, no matter what the default is or what command-line 7580 options are used. 7581 * A new command-line option -Wswitch-bool has been added for the C 7582 and C++ compilers, which warns whenever a switch statement has an 7583 index of boolean type. 7584 * A new command-line option -Wlogical-not-parentheses has been added 7585 for the C and C++ compilers, which warns about "logical not" used 7586 on the left hand side operand of a comparison. 7587 * A new command-line option -Wsizeof-array-argument has been added 7588 for the C and C++ compilers, which warns when the sizeof operator 7589 is applied to a parameter that has been declared as an array in a 7590 function definition. 7591 * A new command-line option -Wbool-compare has been added for the C 7592 and C++ compilers, which warns about boolean expressions compared 7593 with an integer value different from true/false. 7594 * Full support for Cilk Plus has been added to the GCC compiler. Cilk 7595 Plus is an extension to the C and C++ languages to support data and 7596 task parallelism. 7597 * A new attribute no_reorder prevents reordering of selected symbols 7598 against other such symbols or inline assembler. This enables to 7599 link-time optimize the Linux kernel without having to resort to 7600 -fno-toplevel-reorder that disables several optimizations. 7601 * New preprocessor constructs, __has_include and __has_include_next, 7602 to test the availability of headers have been added. 7603 This demonstrates a way to include the header <optional> only if it 7604 is available: 7605 7606 #ifdef __has_include 7607 # if __has_include(<optional>) 7608 # include <optional> 7609 # define have_optional 1 7610 # elif __has_include(<experimental/optional>) 7611 # include <experimental/optional> 7612 # define have_optional 1 7613 # define experimental_optional 7614 # else 7615 # define have_optional 0 7616 # endif 7617 #endif 7618 7619 The header search paths for __has_include and __has_include_next 7620 are equivalent to those of the standard directive #include and the 7621 extension #include_next respectively. 7622 * A new built-in function-like macro to determine the existence of an 7623 attribute, __has_attribute, has been added. The equivalent built-in 7624 macro __has_cpp_attribute was added to C++ to support 7625 [10]Feature-testing recommendations for C++. The macro 7626 __has_attribute is added to all C-like languages as an extension: 7627 7628 int 7629 #ifdef __has_attribute 7630 # if __has_attribute(__noinline__) 7631 __attribute__((__noinline__)) 7632 # endif 7633 #endif 7634 foo(int x); 7635 7636 If an attribute exists, a nonzero constant integer is returned. For 7637 standardized C++ attributes a date is returned, otherwise the 7638 constant returned is 1. Both __has_attribute and 7639 __has_cpp_attribute will add underscores to an attribute name if 7640 necessary to resolve the name. For C++11 and onwards the attribute 7641 may be scoped. 7642 * A new set of built-in functions for arithmetics with overflow 7643 checking has been added: __builtin_add_overflow, 7644 __builtin_sub_overflow and __builtin_mul_overflow and for 7645 compatibility with clang also other variants. These builtins have 7646 two integral arguments (which don't need to have the same type), 7647 the arguments are extended to infinite precision signed type, +, - 7648 or * is performed on those, and the result is stored in an integer 7649 variable pointed to by the last argument. If the stored value is 7650 equal to the infinite precision result, the built-in functions 7651 return false, otherwise true. The type of the integer variable that 7652 will hold the result can be different from the types of the first 7653 two arguments. The following snippet demonstrates how this can be 7654 used in computing the size for the calloc function: 7655 7656 void * 7657 calloc (size_t x, size_t y) 7658 { 7659 size_t sz; 7660 if (__builtin_mul_overflow (x, y, &sz)) 7661 return NULL; 7662 void *ret = malloc (sz); 7663 if (ret) memset (res, 0, sz); 7664 return ret; 7665 } 7666 7667 On e.g. i?86 or x86-64 the above will result in a mul instruction 7668 followed by a jump on overflow. 7669 * The option -fextended-identifiers is now enabled by default for 7670 C++, and for C99 and later C versions. Various bugs in the 7671 implementation of extended identifiers have been fixed. 7672 7673 [11]C 7674 7675 * The default mode has been changed to -std=gnu11. 7676 * A new command-line option -Wc90-c99-compat has been added to warn 7677 about features not present in ISO C90, but present in ISO C99. 7678 * A new command-line option -Wc99-c11-compat has been added to warn 7679 about features not present in ISO C99, but present in ISO C11. 7680 * It is possible to disable warnings about conversions between 7681 pointers that have incompatible types via a new warning option 7682 -Wno-incompatible-pointer-types; warnings about implicit 7683 incompatible integer to pointer and pointer to integer conversions 7684 via a new warning option -Wno-int-conversion; and warnings about 7685 qualifiers on pointers being discarded via a new warning option 7686 -Wno-discarded-qualifiers. 7687 * To allow proper use of const qualifiers with multidimensional 7688 arrays, GCC will not warn about incompatible pointer types anymore 7689 for conversions between pointers to arrays with and without const 7690 qualifier (except when using -pedantic). Instead, a new warning is 7691 emitted only if the const qualifier is lost. This can be controlled 7692 with a new warning option -Wno-discarded-array-qualifiers. 7693 * The C front end now generates more precise caret diagnostics. 7694 * The -pg command-line option now only affects the current file in an 7695 LTO build. 7696 7697 [12]C++ 7698 7699 * G++ now supports [13]C++14 variable templates. 7700 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 7701 * Excessive template instantiation depth is now a fatal error. This 7702 prevents excessive diagnostics that usually do not help to identify 7703 the problem. 7704 * G++ and libstdc++ now implement the feature-testing macros from 7705 [14]Feature-testing recommendations for C++. 7706 * G++ now allows typename in a template template parameter. 7707 7708 template<template<typename> typename X> struct D; // OK 7709 7710 * G++ now supports [15]C++14 aggregates with non-static data member 7711 initializers. 7712 7713 struct A { int i, j = i; }; 7714 A a = { 42 }; // a.j is also 42 7715 7716 * G++ now supports [16]C++14 extended constexpr. 7717 7718 constexpr int f (int i) 7719 { 7720 int j = 0; 7721 for (; i > 0; --i) 7722 ++j; 7723 return j; 7724 } 7725 7726 constexpr int i = f(42); // i is 42 7727 7728 * G++ now supports the [17]C++14 sized deallocation functions. 7729 7730 void operator delete (void *, std::size_t) noexcept; 7731 void operator delete[] (void *, std::size_t) noexcept; 7732 7733 * A new One Definition Rule violation warning (controlled by -Wodr) 7734 detects mismatches in type definitions and virtual table contents 7735 during link-time optimization. 7736 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 7737 developers to annotate programs with final specifiers (or anonymous 7738 namespaces) to improve code generation. These warnings can be used 7739 at compile time, but they are more useful in combination with 7740 link-time optimization. 7741 * G++ no longer supports [18]N3639 variable length arrays, as they 7742 were removed from the C++14 working paper prior to ratification. 7743 GNU VLAs are still supported, so VLA support is now the same in 7744 C++14 mode as in C++98 and C++11 modes. 7745 * G++ now allows passing a non-trivially-copyable class via C 7746 varargs, which is conditionally-supported with 7747 implementation-defined semantics in the standard. This uses the 7748 same calling convention as a normal value parameter. 7749 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 7750 various mangling bugs are fixed, but G++ will still emit aliases 7751 with the old, wrong mangling where feasible. -Wabi=2 will warn 7752 about differences between ABI version 2 and the current setting. 7753 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 7754 to be unaffected, but -Wabi=8 will warn about a non-static data 7755 member with type std::nullptr_t which changes position due to this 7756 change. 7757 7758 [19]Runtime Library (libstdc++) 7759 7760 * A [20]Dual ABI is provided by the library. A new ABI is enabled by 7761 default. The old ABI is still supported and can be used by defining 7762 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 7763 standard library headers. 7764 * A new implementation of std::string is enabled by default, using 7765 the small string optimization instead of copy-on-write reference 7766 counting. 7767 * A new implementation of std::list is enabled by default, with an 7768 O(1) size() function; 7769 * [21]Full support for C++11, including the following new features: 7770 + std::deque and std::vector<bool> meet the allocator-aware 7771 container requirements; 7772 + movable and swappable iostream classes; 7773 + support for std::align and std::aligned_union; 7774 + type traits std::is_trivially_copyable, 7775 std::is_trivially_constructible, std::is_trivially_assignable 7776 etc.; 7777 + I/O manipulators std::put_time, std::get_time, std::hexfloat 7778 and std::defaultfloat; 7779 + generic locale-aware std::isblank; 7780 + locale facets for Unicode conversion; 7781 + atomic operations for std::shared_ptr; 7782 + std::notify_all_at_thread_exit() and functions for making 7783 futures ready at thread exit. 7784 * Support for the C++11 hexfloat manipulator changes how the num_put 7785 facet formats floating point types when 7786 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 7787 This change affects all language modes, even though the C++98 7788 standard gave no special meaning to that combination of flags. To 7789 prevent the use of hexadecimal notation for floating point types 7790 use str.unsetf(std::ios_base::floatfield) to clear the relevant 7791 bits in str.flags(). 7792 * [22]Full experimental support for C++14, including the following 7793 new features: 7794 + std::is_final type trait; 7795 + heterogeneous comparison lookup in associative containers. 7796 + global functions cbegin, cend, rbegin, rend, crbegin, and 7797 crend for range access to containers, arrays and initializer 7798 lists. 7799 * [23]Improved experimental support for the Library Fundamentals TS, 7800 including: 7801 + class std::experimental::any; 7802 + function template std::experimental::apply; 7803 + function template std::experimental::sample; 7804 + function template std::experimental::search and related 7805 searcher types; 7806 + variable templates for type traits; 7807 + function template std::experimental::not_fn. 7808 * New random number distributions logistic_distribution and 7809 uniform_on_sphere_distribution as extensions. 7810 * [24]GDB Xmethods for containers and std::unique_ptr. 7811 7812 [25]Fortran 7813 7814 * Compatibility notice: 7815 + The version of the module files (.mod) has been incremented. 7816 + For free-form source files [26]-Werror=line-truncation is now 7817 enabled by default. Note that comments exceeding the line 7818 length are not diagnosed. (For fixed-form source code, the 7819 same warning is available but turned off by default, such that 7820 excess characters are ignored. -ffree-line-length-n and 7821 -ffixed-line-length-n can be used to modify the default line 7822 lengths of 132 and 72 columns, respectively.) 7823 + The -Wtabs option is now more sensible: with -Wtabs the 7824 compiler warns if it encounters tabs and with -Wno-tabs this 7825 warning is turned off. Before, -Wno-tabs warned and -Wtabs 7826 disabled the warning. As before, this warning is also enabled 7827 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 7828 options of -std=. 7829 * Incomplete support for colorizing diagnostics emitted by gfortran 7830 has been added. The option [27]-fdiagnostics-color controls when 7831 color is used in diagnostics. The default value of this option can 7832 be [28]configured when building GCC. The GCC_COLORS environment 7833 variable can be used to customize the colors or disable coloring 7834 completely. Sample diagnostics output: 7835 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 7836 test.f90:6:1: 7837 7838 0 continue 7839 1 7840 Error: Zero is not a valid statement label at (1) 7841 test.f90:9:6: 7842 7843 USE foo 7844 1 7845 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 7846 7847 * The -Wuse-without-only option has been added to warn when a USE 7848 statement has no ONLY qualifier and thus implicitly imports all 7849 public entities of the used module. 7850 * Formatted READ and WRITE statements now work correctly in 7851 locale-aware programs. For more information and potential caveats, 7852 see [29]Section 5.3 Thread-safety of the runtime library in the 7853 manual. 7854 * [30]Fortran 2003: 7855 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 7856 IEEE_ARITHMETIC) are now supported. 7857 * [31]Fortran 2008: 7858 + [32]Coarrays: Full experimental support of Fortran 2008's 7859 coarrays with -fcoarray=lib except for allocatable/pointer 7860 components of derived-type coarrays. GCC currently only ships 7861 with a single-image library (libcaf_single), but multi-image 7862 support based on MPI and GASNet is provided by the libraries 7863 of the [33]OpenCoarrays project. 7864 * TS18508 Additional Parallel Features in Fortran: 7865 + Support for the collective intrinsic subroutines CO_MAX, 7866 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 7867 including -fcoarray=lib support. 7868 + Support for the new atomic intrinsics has been added, 7869 including -fcoarray=lib support. 7870 * Fortran 2015: 7871 + Support for IMPLICIT NONE (external, type). 7872 + ERROR STOP is now permitted in pure procedures. 7873 7874 [34]Go 7875 7876 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 7877 * Building GCC 5 with Go enabled will install two new programs: 7878 [35]go and [36]gofmt. 7879 7880 [37]libgccjit 7881 7882 New in GCC 5 is the ability to build GCC as a shared library for 7883 embedding in other processes (such as interpreters), suitable for 7884 Just-In-Time compilation to machine code. 7885 7886 The shared library has a [38]C API and a [39]C++ wrapper API providing 7887 some "syntactic sugar". There are also bindings available from 3rd 7888 parties for [40]Python and for [41]D. 7889 7890 For example, this library can be used by interpreters for [42]compiling 7891 functions from bytecode to machine code. 7892 7893 The library can also be used for ahead-of-time compilation, enabling 7894 GCC to be plugged into a pre-existing front end. An example of using 7895 this to build a compiler for an esoteric language we'll refer to as 7896 "brainf" can be seen [43]here. 7897 7898 libgccjit is licensed under the GPLv3 (or at your option, any later 7899 version) 7900 7901 It should be regarded as experimental at this time. 7902 7903 [44]New Targets and Target Specific Improvements 7904 7905 [45]Reporting stack usage 7906 7907 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 7908 -fstack-usage option. 7909 7910 [46]AArch64 7911 7912 * Code generation for the ARM Cortex-A57 processor has been improved. 7913 A more accurate instruction scheduling model for the processor is 7914 now used, and a number of compiler tuning parameters have been set 7915 to offer increased performance when compiling with -mcpu=cortex-a57 7916 or -mtune=cortex-a57. 7917 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 7918 and can be enabled by giving the -mfix-cortex-a53-835769 option. 7919 Alternatively it can be enabled by default by configuring GCC with 7920 the --enable-fix-cortex-a53-835769 option. 7921 * The optional cryptographic extensions to the ARMv8-A architecture 7922 are no longer enabled by default when specifying the 7923 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 7924 options. To enable these extensions add +crypto to the value of 7925 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 7926 * Support has been added for the following processors (GCC 7927 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 7928 initial support for its big.LITTLE combination with the ARM 7929 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 7930 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 7931 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7932 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 7933 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 7934 support for the Cortex-A72. 7935 * The transitional options -mlra and -mno-lra have been removed. The 7936 AArch64 back end now uses the local register allocator (LRA) only. 7937 7938 [47]ARM 7939 7940 * Thumb-1 assembly code is now generated in unified syntax. The new 7941 option -masm-syntax-unified specifies whether inline assembly code 7942 is using unified syntax. By default the option is off which means 7943 non-unified syntax is used. However this is subject to change in 7944 future releases. Eventually the non-unified syntax will be 7945 deprecated. 7946 * It is now a configure-time error to use the --with-cpu configure 7947 option with either of --with-tune or --with-arch. 7948 * Code generation for the ARM Cortex-A57 processor has been improved. 7949 A more accurate instruction scheduling model for the processor is 7950 now used, and a number of compiler tuning parameters have been set 7951 to offer increased performance when compiling with -mcpu=cortex-a57 7952 or -mtune=cortex-a57. 7953 * Support has been added for the following processors (GCC 7954 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 7955 initial support for its big.LITTLE combination with the ARM 7956 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 7957 initial support for its big.LITTLE combination with the ARM 7958 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 7959 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 7960 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7961 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 7962 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 7963 support for the Cortex-A72. 7964 * The deprecated option -mwords-little-endian has been removed. 7965 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 7966 -mtpcs-leaf-frame which are only applicable to the old ABI have 7967 been deprecated. 7968 * The transitional options -mlra and -mno-lra have been removed. The 7969 ARM back end now uses the local register allocator (LRA) only. 7970 7971 [48]AVR 7972 7973 * Support has been added for the devices ATtiny4/5/9/10/20/40. This 7974 requires Binutils 2.25 or newer. 7975 * The port uses a new scheme to describe supported devices: For each 7976 supported device, the compiler provides a device-specific [49]spec 7977 file. If the compiler is used together with AVR-LibC, this requires 7978 at least GCC 5.2 and a version of AVR-LibC which implements 7979 [50]feature #44574. 7980 As a consequence, the compiler no more supports individual devices 7981 like ATmega8. Specifying, say, -mmcu=atmega8 triggers the usage of 7982 the device-specific [51]spec file specs-atmega8 which is part of 7983 the installation and describes options for the sub-processes like 7984 compiler proper, assembler and linker. You can add support for a 7985 new device -mmcu=mydevice as follows: 7986 1. In an empty directory /someplace, create a new directory 7987 device-specs. 7988 2. Copy a device spec file from the installed device-specs 7989 folder, follow the comments in that file and then save it as 7990 /someplace/device-specs/specs-mydevice. 7991 3. Add -B /someplace -mmcu=mydevice to the compiler's 7992 command-line options. Notice that /someplace must specify an 7993 absolute path and that mydevice must not start with "avr". 7994 4. Provided you have a device-specific library libmydevice.a 7995 available, you can put it at /someplace, dito for a 7996 device-specific startup file crtmydevice.o. 7997 The contents of the device spec files depend on the compiler's 7998 configuration, in particular on --with-avrlibc=no and whether or 7999 not it is configured for RTEMS. 8000 * A new command-line option -nodevicelib has been added. It prevents 8001 the compiler from linking against AVR-LibC's device-specific 8002 library libdevice.a. 8003 * The following three command-line options have been added: 8004 8005 -mrmw 8006 Set if the device supports the read-modify-write 8007 instructions LAC, LAS, LAT and XCH. 8008 8009 -mn-flash=size 8010 Specify the flash size of the device in units of 64 KiB, 8011 rounded up to the next integer as needed. This option 8012 affects the availability of the [52]AVR address-spaces. 8013 8014 -mskip-bug 8015 Set if the device is affected by the respective silicon 8016 bug. 8017 8018 These options are used internally in order to communicate between 8019 the compiler and the device-spces file. They are set in the 8020 device-specs file as needed. Don't set them by hand. 8021 8022 [53]IA-32/x86-64 8023 8024 * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's 8025 CPU codenamed Skylake Server was added to GCC. That includes inline 8026 assembly support, new intrinsics, and basic autovectorization. 8027 These new AVX-512 extensions are available via the following GCC 8028 switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512 8029 Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword 8030 instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma 8031 and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi. 8032 * New ISA extensions support Intel MPX was added to GCC. This new 8033 extension is available via the -mmpx compiler switch. Intel MPX is 8034 a set of processor features which, with compiler, run-time library 8035 and OS support, brings increased robustness to software by run-time 8036 checking pointer references against their bounds. In GCC Intel MPX 8037 is supported by Pointer Bounds Checker and libmpx run-time 8038 libraries. 8039 * The new -mrecord-mcount option for -pg generates a Linux kernel 8040 style table of pointers to mcount or __fentry__ calls at the 8041 beginning of functions. The new -mnop-mcount option in addition 8042 also generates nops in place of the __fentry__ or mcount call, so 8043 that a call per function can be later patched in. This can be used 8044 for low overhead tracing or hot code patching. 8045 * The new -malign-data option controls how GCC aligns variables. 8046 -malign-data=compat uses increased alignment compatible with GCC 8047 4.8 and earlier, -malign-data=abi uses alignment as specified by 8048 the psABI, and -malign-data=cacheline uses increased alignment to 8049 match the cache line size. -malign-data=compat is the default. 8050 * The new -mskip-rax-setup option skips setting up the RAX register 8051 when SSE is disabled and there are no variable arguments passed in 8052 vector registers. This can be used to optimize the Linux kernel. 8053 8054 [54]MIPS 8055 8056 * MIPS Releases 3 and 5 are now directly supported. Use the 8057 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 8058 to enable code-generation for these processors. 8059 * The Imagination P5600 processor is now supported using the 8060 -march=p5600 command-line option. 8061 * The Cavium Octeon3 processor is now supported using the 8062 -march=octeon3 command-line option. 8063 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 8064 command-line options. 8065 * The o32 ABI has been modified and extended. The o32 64-bit 8066 floating-point register support is now obsolete and has been 8067 removed. It has been replaced by three ABI extensions FPXX, FP64A, 8068 and FP64. The meaning of the -mfp64 command-line option has 8069 changed. It is now used to enable the FP64A and FP64 ABI 8070 extensions. 8071 + The FPXX extension requires that code generated to access 8072 double-precision values use even-numbered registers. Code that 8073 adheres to this extension is link-compatible with all other 8074 o32 double-precision ABI variants and will execute correctly 8075 in all hardware FPU modes. The command-line options -mabi=32 8076 -mfpxx can be used to enable this extension. MIPS II is the 8077 minimum processor required. 8078 + The o32 FP64A extension requires that floating-point registers 8079 be 64-bit and odd-numbered single-precision registers are not 8080 allowed. Code that adheres to the o32 FP64A variant is 8081 link-compatible with all other o32 double-precision ABI 8082 variants. The command-line options -mabi=32 -mfp64 8083 -mno-odd-spreg can be used to enable this extension. MIPS32R2 8084 is the minimum processor required. 8085 + The o32 FP64 extension also requires that floating-point 8086 registers be 64-bit, but permits the use of single-precision 8087 registers. Code that adheres to the o32 FP64 variant is 8088 link-compatible with o32 FPXX and o32 FP64A variants only, 8089 i.e. it is not compatible with the original o32 8090 double-precision ABI. The command-line options -mabi=32 -mfp64 8091 -modd-spreg can be used to enable this extension. MIPS32R2 is 8092 the minimum processor required. 8093 The new ABI variants can be enabled by default using the configure 8094 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 8095 It is strongly recommended that all vendors begin to set o32 FPXX 8096 as the default ABI. This will be required to run the generated code 8097 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 8098 MIPSR6 cores. 8099 * GCC will now pass all floating-point options to the assembler if 8100 GNU binutils 2.25 is used. As a result, any inline assembly code 8101 that uses hard-float instructions should be amended to include a 8102 .set directive to override the global assembler options when 8103 compiling for soft-float targets. 8104 8105 [55]NDS32 8106 8107 * The variadic function ABI implementation is now compatible with 8108 past Andes toolchains where the caller uses registers to pass 8109 arguments and the callee is in charge of pushing them on stack. 8110 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 8111 been removed since they are not yet available in the nds32 port of 8112 GNU binutils. 8113 * A new option -mcmodel=[small|medium|large] supports varied code 8114 models on code generation. The -mgp-direct option became 8115 meaningless and can be discarded. 8116 8117 [56]RX 8118 8119 * A new command line option -mno-allow-string-insns can be used to 8120 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 8121 SWHILE and RMPA instructions. An erratum released by Renesas shows 8122 that it is unsafe to use these instructions on addresses within the 8123 I/O space of the processor. The new option can be used when the 8124 programmer is concerned that the I/O space might be accessed. The 8125 default is still to enable these instructions. 8126 8127 [57]SH 8128 8129 * The compiler will now pass the appropriate --isa= option to the 8130 assembler. 8131 * The default handling for the GBR has been changed from call 8132 clobbered to call preserved. The old behavior can be reinstated by 8133 specifying the option -fcall-used-gbr. 8134 * Support for the SH4A fpchg instruction has been added which will be 8135 utilized when switching between single and double precision FPU 8136 modes. 8137 * The compiler no longer uses the __fpscr_values array for switching 8138 between single and double FPU precision modes on non-SH4A targets. 8139 Instead mode switching will now be performed by storing, modifying 8140 and reloading the FPSCR, so that other FPSCR bits are preserved 8141 across mode switches. The __fpscr_values array that is defined in 8142 libgcc is still present for backwards compatibility, but it will 8143 not be referenced by compiler generated code anymore. 8144 * New builtin functions __builtin_sh_get_fpscr and 8145 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 8146 function will mask the specified bits in such a way that the SZ, PR 8147 and FR mode bits will be preserved, while changing the other bits. 8148 These new functions do not reference the __fpscr_values array. The 8149 old functions __set_fpscr and __get_fpscr in libgcc which access 8150 the __fpscr_values array are still present for backwards 8151 compatibility, but their usage is highly discouraged. 8152 * Some improvements to code generated for __atomic built-in 8153 functions. 8154 * When compiling for SH2E the compiler will no longer force the usage 8155 of delay slots for conditional branch instructions bt and bf. The 8156 old behavior can be reinstated (e.g. to work around a hardware bug 8157 in the original SH7055) by specifying the new option 8158 -mcbranch-force-delay-slot. 8159 8160 [58]Operating Systems 8161 8162 [59]AIX 8163 8164 * GCC now supports stabs debugging continuation lines to allow long 8165 stabs debug information without overflow that generates AIX linker 8166 errors. 8167 8168 [60]DragonFly BSD 8169 8170 * GCC now supports the DragonFly BSD operating system. 8171 8172 [61]FreeBSD 8173 8174 * GCC now supports the FreeBSD operating system for the arm port 8175 through the arm*-*-freebsd* target triplets. 8176 8177 [62]VxWorks MILS 8178 8179 * GCC now supports the MILS (Multiple Independent Levels of Security) 8180 variant of WindRiver's VxWorks operating system for PowerPC 8181 targets. 8182 8183 Other significant improvements 8184 8185 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 8186 to set the compiler to use. 8187 8188 * When the new command-line option -freport-bug is used, GCC 8189 automatically generates a developer-friendly reproducer whenever an 8190 internal compiler error is encountered. 8191 8192 [63]GCC 5.2 8193 8194 This is the [64]list of problem reports (PRs) from GCC's bug tracking 8195 system that are known to be fixed in the 5.2 release. This list might 8196 not be complete (that is, it is possible that some PRs that have been 8197 fixed are not listed here). 8198 8199 Target Specific Changes 8200 8201 IA-32/x86-64 8202 8203 * Support for new AMD instructions monitorx and mwaitx has been 8204 added. This includes new intrinsic and built-in support. It is 8205 enabled through option -mmwaitx. The instructions monitorx and 8206 mwaitx implement the same functionality as the old monitor and 8207 mwait instructions. In addition, mwaitx adds a configurable timer. 8208 The timer value is received as third argument and stored in 8209 register %ebx. 8210 8211 S/390, System z, IBM z Systems 8212 8213 * Support for the IBM z13 processor has been added. When using the 8214 -march=z13 option, the compiler will generate code making use of 8215 the new instructions and registers introduced with the vector 8216 extension facility. The -mtune=z13 option enables z13 specific 8217 instruction scheduling without making use of new instructions. 8218 Compiling code with -march=z13 reduces the default alignment of 8219 vector types bigger than 8 bytes to 8. This is an ABI change and 8220 care must be taken when linking modules compiled with different 8221 arch levels which interchange variables containing vector type 8222 values. For newly compiled code the GNU linker will emit a warning. 8223 * The -mzvector option enables a C/C++ language extension. This 8224 extension provides a new keyword vector which can be used to define 8225 vector type variables. (Note: This is not available when enforcing 8226 strict standard compliance e.g. with -std=c99. Either enable GNU 8227 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 8228 Additionally a set of overloaded builtins is provided which is 8229 partially compatible to the PowerPC Altivec builtins. In order to 8230 make use of these builtins the vecintrin.h header file needs to be 8231 included. 8232 8233 [65]GCC 5.3 8234 8235 This is the [66]list of problem reports (PRs) from GCC's bug tracking 8236 system that are known to be fixed in the 5.3 release. This list might 8237 not be complete (that is, it is possible that some PRs that have been 8238 fixed are not listed here). 8239 8240 Target Specific Changes 8241 8242 IA-32/x86-64 8243 8244 * GCC now supports the Intel CPU named Skylake with AVX-512 8245 extensions through -march=skylake-avx512. The switch enables the 8246 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 8247 AVX-512DQ. 8248 8249 S/390, System z, IBM z Systems 8250 8251 * With this version of GCC IBM z Systems support has been added to 8252 the GO runtime environment. GCC 5.3 has proven to be able to 8253 compile larger GO applications on IBM z Systems. 8254 8255 [67]GCC 5.4 8256 8257 This is the [68]list of problem reports (PRs) from GCC's bug tracking 8258 system that are known to be fixed in the 5.4 release. This list might 8259 not be complete (that is, it is possible that some PRs that have been 8260 fixed are not listed here). 8261 8262 [69]GCC 5.5 8263 8264 This is the [70]list of problem reports (PRs) from GCC's bug tracking 8265 system that are known to be fixed in the 5.5 release. This list might 8266 not be complete (that is, it is possible that some PRs that have been 8267 fixed are not listed here). 8268 8269 Target Specific Changes 8270 8271 IA-32/x86-64 8272 8273 * Support for the [71]deprecated pcommit instruction has been 8274 removed. 8275 8276 8277 For questions related to the use of GCC, please consult these web 8278 pages and the [72]GCC manuals. If that fails, the 8279 [73]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8280 web pages and the development of GCC are welcome on our developer 8281 list at [74]gcc (a] gcc.gnu.org. All of [75]our lists have public 8282 archives. 8283 8284 Copyright (C) [76]Free Software Foundation, Inc. Verbatim copying and 8285 distribution of this entire article is permitted in any medium, 8286 provided this notice is preserved. 8287 8288 These pages are [77]maintained by the GCC team. Last modified 8289 2025-02-28. 8290 8291 References 8292 8293 1. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 8294 2. https://gcc.gnu.org/gcc-5/changes.html#general 8295 3. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 8296 4. https://gcc.gnu.org/gcc-5/changes.html#languages 8297 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 8298 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 8299 7. https://gcc.gnu.org/wiki/OpenACC 8300 8. https://gcc.gnu.org/gcc-5/changes.html#c-family 8301 9. https://gcc.gnu.org/install/configure.html 8302 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 8303 11. https://gcc.gnu.org/gcc-5/changes.html#c 8304 12. https://gcc.gnu.org/gcc-5/changes.html#cxx 8305 13. https://gcc.gnu.org/projects/cxx1y.html 8306 14. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 8307 15. https://gcc.gnu.org/projects/cxx1y.html 8308 16. https://gcc.gnu.org/projects/cxx1y.html 8309 17. https://gcc.gnu.org/projects/cxx1y.html 8310 18. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 8311 19. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 8312 20. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 8313 21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 8314 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 8315 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 8316 24. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python 8317 25. https://gcc.gnu.org/gcc-5/changes.html#fortran 8318 26. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 8319 27. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 8320 28. https://gcc.gnu.org/install/configure.html 8321 29. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 8322 30. https://gcc.gnu.org/wiki/Fortran2003Status 8323 31. https://gcc.gnu.org/wiki/Fortran2008Status 8324 32. https://gcc.gnu.org/wiki/Coarray 8325 33. http://www.opencoarrays.org/ 8326 34. https://gcc.gnu.org/gcc-5/changes.html#go 8327 35. https://pkg.go.dev/cmd/go 8328 36. https://pkg.go.dev/cmd/gofmt 8329 37. https://gcc.gnu.org/gcc-5/changes.html#jit 8330 38. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 8331 39. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 8332 40. https://github.com/davidmalcolm/pygccjit 8333 41. https://github.com/ibuclaw/gccjitd 8334 42. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 8335 43. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 8336 44. https://gcc.gnu.org/gcc-5/changes.html#targets 8337 45. https://gcc.gnu.org/gcc-5/changes.html#stack-usage 8338 46. https://gcc.gnu.org/gcc-5/changes.html#aarch64 8339 47. https://gcc.gnu.org/gcc-5/changes.html#arm 8340 48. https://gcc.gnu.org/gcc-5/changes.html#avr 8341 49. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 8342 50. https://savannah.nongnu.org/bugs/?44574 8343 51. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 8344 52. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 8345 53. https://gcc.gnu.org/gcc-5/changes.html#x86 8346 54. https://gcc.gnu.org/gcc-5/changes.html#mips 8347 55. https://gcc.gnu.org/gcc-5/changes.html#nds32 8348 56. https://gcc.gnu.org/gcc-5/changes.html#rx 8349 57. https://gcc.gnu.org/gcc-5/changes.html#sh 8350 58. https://gcc.gnu.org/gcc-5/changes.html#os 8351 59. https://gcc.gnu.org/gcc-5/changes.html#aix 8352 60. https://gcc.gnu.org/gcc-5/changes.html#dragonfly 8353 61. https://gcc.gnu.org/gcc-5/changes.html#freebsd 8354 62. https://gcc.gnu.org/gcc-5/changes.html#vxmils 8355 63. https://gcc.gnu.org/gcc-5/changes.html#GCC5.2 8356 64. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 8357 65. https://gcc.gnu.org/gcc-5/changes.html#GCC5.3 8358 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 8359 67. https://gcc.gnu.org/gcc-5/changes.html#GCC5.4 8360 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 8361 69. https://gcc.gnu.org/gcc-5/changes.html#GCC5.5 8362 70. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 8363 71. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 8364 72. https://gcc.gnu.org/onlinedocs/ 8365 73. mailto:gcc-help (a] gcc.gnu.org 8366 74. mailto:gcc (a] gcc.gnu.org 8367 75. https://gcc.gnu.org/lists.html 8368 76. https://www.fsf.org/ 8369 77. https://gcc.gnu.org/about.html 8370 ====================================================================== 8371 http://gcc.gnu.org/gcc-4.9/index.html 8372 8373 GCC 4.9 Release Series 8374 8375 (This release series is no longer supported.) 8376 8377 Aug 3, 2016 8378 8379 The [1]GNU project and the GCC developers are pleased to announce the 8380 release of GCC 4.9.4. 8381 8382 This release is a bug-fix release, containing fixes for regressions in 8383 GCC 4.9.3 relative to previous releases of GCC. 8384 8385 Release History 8386 8387 GCC 4.9.4 8388 Aug 3, 2016 ([2]changes, [3]documentation) 8389 8390 GCC 4.9.3 8391 June 26, 2015 ([4]changes, [5]documentation) 8392 8393 GCC 4.9.2 8394 October 30, 2014 ([6]changes, [7]documentation) 8395 8396 GCC 4.9.1 8397 July 16, 2014 ([8]changes, [9]documentation) 8398 8399 GCC 4.9.0 8400 April 22, 2014 ([10]changes, [11]documentation) 8401 8402 References and Acknowledgements 8403 8404 GCC used to stand for the GNU C Compiler, but since the compiler 8405 supports several other languages aside from C, it now stands for the 8406 GNU Compiler Collection. 8407 8408 A list of [12]successful builds is updated as new information becomes 8409 available. 8410 8411 The GCC developers would like to thank the numerous people that have 8412 contributed new features, improvements, bug fixes, and other changes as 8413 well as test results to GCC. This [13]amazing group of volunteers is 8414 what makes GCC successful. 8415 8416 For additional information about GCC please refer to the [14]GCC 8417 project web site or contact the [15]GCC development mailing list. 8418 8419 To obtain GCC please use [16]our mirror sites or [17]our version 8420 control system. 8421 8422 8423 For questions related to the use of GCC, please consult these web 8424 pages and the [18]GCC manuals. If that fails, the 8425 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8426 web pages and the development of GCC are welcome on our developer 8427 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 8428 archives. 8429 8430 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8431 distribution of this entire article is permitted in any medium, 8432 provided this notice is preserved. 8433 8434 These pages are [23]maintained by the GCC team. Last modified 8435 2024-05-30. 8436 8437 References 8438 8439 1. http://www.gnu.org/ 8440 2. https://gcc.gnu.org/gcc-4.9/changes.html 8441 3. https://gcc.gnu.org/onlinedocs/4.9.4/ 8442 4. https://gcc.gnu.org/gcc-4.9/changes.html 8443 5. https://gcc.gnu.org/onlinedocs/4.9.3/ 8444 6. https://gcc.gnu.org/gcc-4.9/changes.html 8445 7. https://gcc.gnu.org/onlinedocs/4.9.2/ 8446 8. https://gcc.gnu.org/gcc-4.9/changes.html 8447 9. https://gcc.gnu.org/onlinedocs/4.9.1/ 8448 10. https://gcc.gnu.org/gcc-4.9/changes.html 8449 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 8450 12. https://gcc.gnu.org/gcc-4.9/buildstat.html 8451 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8452 14. https://gcc.gnu.org/index.html 8453 15. mailto:gcc (a] gcc.gnu.org 8454 16. https://gcc.gnu.org/mirrors.html 8455 17. https://gcc.gnu.org/git.html 8456 18. https://gcc.gnu.org/onlinedocs/ 8457 19. mailto:gcc-help (a] gcc.gnu.org 8458 20. mailto:gcc (a] gcc.gnu.org 8459 21. https://gcc.gnu.org/lists.html 8460 22. https://www.fsf.org/ 8461 23. https://gcc.gnu.org/about.html 8462 ====================================================================== 8463 http://gcc.gnu.org/gcc-4.9/changes.html 8464 8465 GCC 4.9 Release Series 8466 Changes, New Features, and Fixes 8467 8468 Caveats 8469 8470 * The mudflap run time checker has been removed. The mudflap options 8471 remain, but do nothing. 8472 * Support for a number of older systems and recently unmaintained or 8473 untested target ports of GCC has been declared obsolete in GCC 4.9. 8474 Unless there is activity to revive them, the next release of GCC 8475 will have their sources permanently removed. 8476 The following ports for individual systems on particular 8477 architectures have been obsoleted: 8478 + Solaris 9 (*-*-solaris2.9). Details can be found in the 8479 [1]announcement. 8480 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 8481 float64x1_t exported by arm_neon.h are defined to be the same as 8482 their base types. This results in incorrect application of 8483 parameter passing rules to arguments of types int64x1_t and 8484 uint64x1_t, with respect to the AAPCS64 ABI specification. In 8485 addition, names of C++ functions with parameters of these types 8486 (including float64x1_t) are not mangled correctly. The current 8487 typedef declarations also unintentionally allow implicit casting 8488 between singleton vector types and their base types. These issues 8489 will be resolved in a near future release. See [2]PR60825 for more 8490 information. 8491 8492 More information on porting to GCC 4.9 from previous versions of GCC 8493 can be found in the [3]porting guide for this release. 8494 8495 General Optimizer Improvements 8496 8497 * AddressSanitizer, a fast memory error detector, is now available on 8498 ARM. 8499 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 8500 detector, has been added and can be enabled via 8501 -fsanitize=undefined. Various computations will be instrumented to 8502 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 8503 currently available for the C and C++ languages. 8504 * Link-time optimization (LTO) improvements: 8505 + Type merging was rewritten. The new implementation is 8506 significantly faster and uses less memory. 8507 + Better partitioning algorithm resulting in less streaming 8508 during link time. 8509 + Early removal of virtual methods reduces the size of object 8510 files and improves link-time memory usage and compile time. 8511 + Function bodies are now loaded on-demand and released early 8512 improving overall memory usage at link time. 8513 + C++ hidden keyed methods can now be optimized out. 8514 + When using a linker plugin, compiling with the -flto option 8515 now generates slim object files (.o) which only contain 8516 intermediate language representation for LTO. Use 8517 -ffat-lto-objects to create files which contain additionally 8518 the object code. To generate static libraries suitable for LTO 8519 processing, use gcc-ar and gcc-ranlib; to list symbols from a 8520 slim object file use gcc-nm. (This requires that ar, ranlib 8521 and nm have been compiled with plugin support.) 8522 Memory usage building Firefox with debug enabled was reduced from 8523 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 8524 * Inter-procedural optimization improvements: 8525 + New type inheritance analysis module improving 8526 devirtualization. Devirtualization now takes into account 8527 anonymous name-spaces and the C++11 final keyword. 8528 + New speculative devirtualization pass (controlled by 8529 -fdevirtualize-speculatively. 8530 + Calls that were speculatively made direct are turned back to 8531 indirect where direct call is not cheaper. 8532 + Local aliases are introduced for symbols that are known to be 8533 semantically equivalent across shared libraries improving 8534 dynamic linking times. 8535 * Feedback directed optimization improvements: 8536 + Profiling of programs using C++ inline functions is now more 8537 reliable. 8538 + New time profiling determines typical order in which functions 8539 are executed. 8540 + A new function reordering pass (controlled by 8541 -freorder-functions) significantly reduces startup time of 8542 large applications. Until binutils support is completed, it is 8543 effective only with link-time optimization. 8544 + Feedback driven indirect call removal and devirtualization now 8545 handle cross-module calls when link-time optimization is 8546 enabled. 8547 8548 [4]New Languages and Language specific improvements 8549 8550 * Version 4.0 of the [5]OpenMP specification is now supported in the 8551 C and C++ compilers and starting with the 4.9.1 release also in the 8552 Fortran compiler. The new -fopenmp-simd option can be used to 8553 enable OpenMP's SIMD directives while ignoring other OpenMP 8554 directives. The new [6]-fsimd-cost-model= option permits to tune 8555 the vectorization cost model for loops annotated with OpenMP and 8556 Cilk Plus simd directives. -Wopenmp-simd warns when the current 8557 cost model overrides simd directives set by the user. 8558 * The -Wdate-time option has been added for the C, C++ and Fortran 8559 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 8560 macros are used. Those macros might prevent bit-wise-identical 8561 reproducible compilations. 8562 8563 [7]Ada 8564 8565 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 8566 8567 [8]C family 8568 8569 * Support for colorizing diagnostics emitted by GCC has been added. 8570 The [9]-fdiagnostics-color=auto will enable it when outputting to 8571 terminals, -fdiagnostics-color=always unconditionally. The 8572 GCC_COLORS environment variable can be used to customize the colors 8573 or disable coloring. If GCC_COLORS variable is present in the 8574 environment, the default is -fdiagnostics-color=auto, otherwise 8575 -fdiagnostics-color=never. 8576 Sample diagnostics output: 8577 $ g++ -fdiagnostics-color=always -S -Wall test.C 8578 test.C: In function `int foo()': 8579 test.C:1:14: warning: no return statement in function returning non-void [-W 8580 return-type] 8581 int foo () { } 8582 ^ 8583 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 8584 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 8585 template <int N> struct X { static const int value = X<N-1>::value; }; temp 8586 late struct X<1000>; 8587 ^ 8588 test.C:2:46: recursively required from `const int X<999>::value' 8589 test.C:2:46: required from `const int X<1000>::value' 8590 test.C:2:88: required from here 8591 8592 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 8593 8594 * With the new [10]#pragma GCC ivdep, the user can assert that there 8595 are no loop-carried dependencies which would prevent concurrent 8596 execution of consecutive iterations using SIMD (single instruction 8597 multiple data) instructions. 8598 * Support for Cilk Plus has been added and can be enabled with the 8599 -fcilkplus option. Cilk Plus is an extension to the C and C++ 8600 languages to support data and task parallelism. The present 8601 implementation follows ABI version 1.2; all features but _Cilk_for 8602 have been implemented. 8603 8604 [11]C 8605 8606 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 8607 <stdatomic.h> header) are now supported. 8608 * ISO C11 generic selections (_Generic keyword) are now supported. 8609 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 8610 __thread) is now supported. 8611 * ISO C11 support is now at a similar level of completeness to ISO 8612 C99 support: substantially complete modulo bugs, extended 8613 identifiers (supported except for corner cases when 8614 -fextended-identifiers is used), floating-point issues (mainly but 8615 not entirely relating to optional C99 features from Annexes F and 8616 G) and the optional Annexes K (Bounds-checking interfaces) and L 8617 (Analyzability). 8618 * A new C extension __auto_type provides a subset of the 8619 functionality of C++11 auto in GNU C. 8620 8621 [12]C++ 8622 8623 * The G++ implementation of [13]C++1y return type deduction for 8624 normal functions has been updated to conform to [14]N3638, the 8625 proposal accepted into the working paper. Most notably, it adds 8626 decltype(auto) for getting decltype semantics rather than the 8627 template argument deduction semantics of plain auto: 8628 8629 int& f(); 8630 auto i1 = f(); // int 8631 decltype(auto) i2 = f(); // int& 8632 8633 * G++ supports [15]C++1y lambda capture initializers: 8634 8635 [x = 42]{ ... }; 8636 8637 Actually, they have been accepted since GCC 4.5, but now the 8638 compiler doesn't warn about them with -std=c++1y, and supports 8639 parenthesized and brace-enclosed initializers as well. 8640 * G++ supports [16]C++1y variable length arrays. G++ has supported 8641 GNU/C99-style VLAs for a long time, but now additionally supports 8642 initializers and lambda capture by reference. In C++1y mode G++ 8643 will complain about VLA uses that are not permitted by the draft 8644 standard, such as forming a pointer to VLA type or applying sizeof 8645 to a VLA variable. Note that it now appears that VLAs will not be 8646 part of C++14, but will be part of a separate document and then 8647 perhaps C++17. 8648 8649 void f(int n) { 8650 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 8651 [&a]{ for (int i : a) { cout << i << endl; } }(); 8652 &a; // error, taking address of VLA 8653 } 8654 8655 * G++ supports the [17]C++1y [[deprecated]] attribute modulo bugs in 8656 the underlying [[gnu::deprecated]] attribute. Classes and functions 8657 can be marked deprecated and a diagnostic message added: 8658 8659 class A; 8660 int bar(int n); 8661 #if __cplusplus > 201103 8662 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 8663 [[deprecated("bar is unsafe; use foo() instead")]] 8664 int bar(int n); 8665 8666 int foo(int n); 8667 class B; 8668 #endif 8669 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 8670 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 8671 o() instead 8672 8673 * G++ supports [18]C++1y digit separators. Long numeric literals can 8674 be subdivided with a single quote ' to enhance readability: 8675 8676 int i = 1048576; 8677 int j = 1'048'576; 8678 int k = 0x10'0000; 8679 int m = 0'004'000'000; 8680 int n = 0b0001'0000'0000'0000'0000'0000; 8681 8682 double x = 1.602'176'565e-19; 8683 double y = 1.602'176'565e-1'9; 8684 8685 * G++ supports [19]C++1y generic (polymorphic) lambdas. 8686 8687 // a functional object that will increment any type 8688 auto incr = [](auto x) { return x++; }; 8689 8690 * As a GNU extension, G++ supports explicit template parameter syntax 8691 for generic lambdas. This can be combined in the expected way with 8692 the standard auto syntax. 8693 8694 // a functional object that will add two like-type objects 8695 auto add = [] <typename T> (T a, T b) { return a + b; }; 8696 8697 * G++ supports unconstrained generic functions as specified by 4.1.2 8698 and 5.1.1 of [20]N3889: Concepts Lite Specification. Briefly, auto 8699 may be used as a type-specifier in a parameter declaration of any 8700 function declarator in order to introduce an implicit function 8701 template parameter, akin to generic lambdas. 8702 8703 // the following two function declarations are equivalent 8704 auto incr(auto x) { return x++; } 8705 template <typename T> 8706 auto incr(T x) { return x++; } 8707 8708 Runtime Library (libstdc++) 8709 8710 * [21]Improved support for C++11, including: 8711 + support for <regex>; 8712 + The associative containers in <map> and <set> and the 8713 unordered associative containers in <unordered_map> and 8714 <unordered_set> meet the allocator-aware container 8715 requirements; 8716 * [22]Improved experimental support for the upcoming ISO C++ 8717 standard, C++14, including: 8718 + fixing constexpr member functions without const; 8719 + implementation of the std::exchange() utility function; 8720 + addressing tuples by type; 8721 + implemention of std::make_unique; 8722 + implemention of std::shared_lock; 8723 + making std::result_of SFINAE-friendly; 8724 + adding operator() to std::integral_constant; 8725 + adding user-defined literals for standard library types 8726 std::basic_string, std::chrono::duration, and std::complex; 8727 + adding two range overloads to non-modifying sequence oprations 8728 std::equal and std::mismatch; 8729 + adding IO manipulators for quoted strings; 8730 + adding constexpr members to <utility>, <complex>, <chrono>, 8731 and some containers; 8732 + adding compile-time std::integer_sequence; 8733 + adding cleaner transformation traits; 8734 + making <functional>s operator functors easier to use and more 8735 generic; 8736 * An implementation of std::experimental::optional. 8737 * An implementation of std::experimental::string_view. 8738 * The non-standard function std::copy_exception has been deprecated 8739 and will be removed in a future version. std::make_exception_ptr 8740 should be used instead. 8741 8742 [23]Fortran 8743 8744 * Compatibility notice: 8745 + Module files: The version of the module files (.mod) has been 8746 incremented; additionally, module files are now compressed. 8747 Fortran MODULEs compiled by earlier GCC versions have to be 8748 recompiled, when they are USEd by files compiled with GCC 4.9. 8749 GCC 4.9 is not able to read .mod files of earlier GCC 8750 versions; attempting to do so gives an error message. Note: 8751 The ABI of the produced assembler data itself has not changed: 8752 object files and libraries are fully compatible with older 8753 versions (except as stated below). 8754 + ABI changes: 8755 o The [24]argument passing ABI has changed for scalar dummy 8756 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 8757 which have both the VALUE and the OPTIONAL attributes. 8758 o To support finalization the virtual table associated with 8759 polymorphic variables has changed. Code containing CLASS 8760 should be recompiled, including all files which define 8761 derived types involved in the type definition used by 8762 polymorphic variables. (Note: Due to the incremented 8763 module version, trying to mix old code with new code will 8764 usually give an error message.) 8765 + GNU Fortran no longer deallocates allocatable variables or 8766 allocatable components of variables declared in the main 8767 program. Since Fortran 2008, the standard explicitly states 8768 that variables declared in the Fortran main program 8769 automatically have the SAVE attribute. 8770 + When opening files, the close-on-exec flag is set if the 8771 system supports such a feature. This is generally considered 8772 good practice these days, but if there is a need to pass file 8773 descriptors to child processes the parent process must now 8774 remember to clear the close-on-exec flag by calling fcntl(), 8775 e.g. via ISO_C_BINDING, before executing the child process. 8776 * The deprecated command-line option -fno-whole-file has been 8777 removed. (-fwhole-file is the default since GCC 4.6.) 8778 -fwhole-file/-fno-whole-file continue to be accepted but do not 8779 influence the code generation. 8780 * The compiler no longer unconditionally warns about DO loops with 8781 zero iterations. This warning is now controlled by the -Wzerotrip 8782 option, which is implied by -Wall. 8783 * The new NO_ARG_CHECK attribute of the [25]!GCC$ directive can be 8784 used to disable the type-kind-rank (TKR) argument check for a dummy 8785 argument. The feature is similar to ISO/IEC TS 29133:2012's 8786 TYPE(*), except that it additionally also disables the rank check. 8787 Variables with NO_ARG_CHECK have to be dummy arguments and may only 8788 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 8789 to another NO_ARG_CHECK dummy argument; also the other constraints 8790 of TYPE(*) apply. The dummy arguments should be declared as scalar 8791 or assumed-size variable of type type(*) (recommended) - or of type 8792 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 8793 the data without further type or shape information is passed, 8794 similar to C's void*. Note that also TS 29113's 8795 type(*),dimension(..) accepts arguments of any type and rank; 8796 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 8797 descriptor which contains the array shape and stride of the 8798 argument. 8799 * [26]Fortran 2003: 8800 + Finalization is now supported. It is currently only done for a 8801 subset of those situations in which it should occur. 8802 + Experimental support for scalar character components with 8803 deferred length (i.e. allocatable string length) in derived 8804 types has been added. (Deferred-length character variables are 8805 supported since GCC 4.6.) 8806 * [27]Fortran 2008: 8807 + When STOP or ERROR STOP are used to terminate the execution 8808 and any exception (but inexact) is signaling, a warning is 8809 printed to ERROR_UNIT, indicating which exceptions are 8810 signaling. The [28]-ffpe-summary= command-line option can be 8811 used to fine-tune for which exceptions the warning should be 8812 shown. 8813 + Rounding on input (READ) is now handled on systems where 8814 strtod honours the rounding mode. (For output, rounding is 8815 supported since GCC 4.5.) Note that for input, the compatible 8816 rounding mode is handled as nearest (i.e., rounding to an even 8817 least significant [cf. IEC 60559:1989] for a tie, while 8818 compatible rounds away from zero in that case). 8819 8820 [29]Go 8821 8822 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 8823 8824 [30]New Targets and Target Specific Improvements 8825 8826 [31]AArch64 8827 8828 * The ARMv8-A crypto and CRC instructions are now supported through 8829 intrinsics. These are enabled when the architecture supports these 8830 and are available through the -march=armv8-a+crc and 8831 -march=armv8-a+crypto options. 8832 * Initial support for ILP32 has now been added to the compiler. This 8833 is now available through the command-line option -mabi=ilp32. 8834 Support for ILP32 is considered experimental as the ABI 8835 specification is still beta. 8836 * Coverage of more of the ISA including the SIMD extensions has been 8837 added. The Advanced SIMD intrinsics have also been improved. 8838 * The new local register allocator (LRA) is now on by default for the 8839 AArch64 back end. 8840 * The REE (Redundant extension elimination) pass has now been enabled 8841 by default for the AArch64 back end. 8842 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 8843 * Initial big.LITTLE tuning support for the combination of Cortex-A57 8844 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 8845 option. 8846 * A number of structural changes have been made to both the ARM and 8847 AArch64 back ends to facilitate improved code-generation. 8848 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 8849 has been added and can be enabled by giving the 8850 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 8851 default by configuring GCC with the --enable-fix-cortex-a53-835769 8852 option. 8853 8854 [32]ARC 8855 8856 * A port for Synopsys Designware ARC has been contributed by Embecosm 8857 and Synopsys Inc. 8858 8859 [33]ARM 8860 8861 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 8862 disabled by default. This was found to generate better code in only 8863 a small number of cases. It can be turned back on with the 8864 -mneon-for-64bits option. 8865 * Further support for the ARMv8-A architecture, notably implementing 8866 the restriction around IT blocks in the Thumb32 instruction set has 8867 been added. The -mrestrict-it option can be used with 8868 -march=armv7-a or the -march=armv7ve options to make code 8869 generation fully compatible with the deprecated instructions in 8870 ARMv8-A. 8871 * Support has now been added for the ARMv7ve variant of the 8872 architecture. This can be used by the -march=armv7ve option. 8873 * The ARMv8-A crypto and CRC instructions are now supported through 8874 intrinsics and are available through the -march=armv8-a+crc and 8875 mfpu=crypto-neon-fp-armv8 options. 8876 * LRA is now on by default for the ARM target. This can be turned off 8877 using the -mno-lra option. This option is a purely transitionary 8878 command-line option and will be removed in a future release. We are 8879 interested in any bug reports regarding functional and performance 8880 regressions with LRA. 8881 * A new option -mslow-flash-data to improve performance of programs 8882 fetching data on slow flash memory has now been introduced for the 8883 ARMv7-M profile cores. 8884 * A new option -mpic-data-is-text-relative for targets that allows 8885 data segments to be relative to text segments has been added. This 8886 is on by default for all targets except VxWorks RTP. 8887 * A number of infrastructural changes have been made to both the ARM 8888 and AArch64 back ends to facilitate improved code-generation. 8889 * GCC now supports Cortex-A12 and the Cortex-R7 through the 8890 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 8891 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 8892 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 8893 * Initial big.LITTLE tuning support for the combination of Cortex-A57 8894 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 8895 option. Similar support was added for the combination of Cortex-A15 8896 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 8897 * Further performance optimizations for the Cortex-A15 and the 8898 Cortex-M4 have been added. 8899 * A number of code generation improvements for Thumb2 to reduce code 8900 size when compiling for the M-profile processors. 8901 8902 [34]AVR 8903 8904 * A new command-line option -mfract-convert-truncate has been added. 8905 It allows compiler to use truncation instead of rounding towards 8906 zero for fractional fixed-point types. 8907 8908 [35]IA-32/x86-64 8909 8910 * -mfpmath=sse is now implied by -ffast-math on all targets where 8911 SSE2 is supported. 8912 * Intel AVX-512 support was added to GCC. That includes inline 8913 assembly support, new registers and extending existing ones, new 8914 intrinsics (covered by corresponding testsuite), and basic 8915 autovectorization. AVX-512 instructions are available via the 8916 following GCC switches: AVX-512 foundation instructions: -mavx512f, 8917 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 8918 reciprocal instructions: -mavx512er, AVX-512 conflict detection 8919 instructions: -mavx512cd. 8920 * It is now possible to call x86 intrinsics from select functions in 8921 a file that are tagged with the corresponding target attribute 8922 without having to compile the entire file with the -mxxx option. 8923 This improves the usability of x86 intrinsics and is particularly 8924 useful when doing [36]Function Multiversioning. 8925 * GCC now supports the new Intel microarchitecture named Silvermont 8926 through -march=silvermont. 8927 * GCC now supports the new Intel microarchitecture named Broadwell 8928 through -march=broadwell. 8929 * Optimizing for other Intel microarchitectures have been renamed to 8930 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 8931 * -march=generic has been retuned for better support of Intel core 8932 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 8933 Pentium-M, and Pentium4 based CPUs is no longer considered 8934 important for generic. 8935 * -mtune=intel can now be used to generate code running well on the 8936 most current Intel processors, which are Haswell and Silvermont for 8937 GCC 4.9. 8938 * Support to encode 32-bit assembly instructions in 16-bit format is 8939 now available through the -m16 command-line option. 8940 * Better inlining of memcpy and memset that is aware of value ranges 8941 and produces shorter alignment prologues. 8942 * -mno-accumulate-outgoing-args is now honored when unwind 8943 information is output. Argument accumulation is also now turned off 8944 for portions of programs optimized for size. 8945 * Support for new AMD family 15h processors (Excavator core) is now 8946 available through the -march=bdver4 and -mtune=bdver4 options. 8947 8948 [37]MSP430 8949 8950 * A new command-line option -mcpu= has been added to the MSP430 back 8951 end. This option is used to specify the ISA to be used. Accepted 8952 values are msp430 (the default), msp430x and msp430xv2. The ISA is 8953 no longer deduced from the -mmcu= option as there are far too many 8954 different MCU names. The -mmcu= option is still supported, and this 8955 is still used to select linker scripts and generate a C 8956 preprocessor symbol that will be recognised by the msp430.h header 8957 file. 8958 8959 [38]NDS32 8960 8961 * A new nds32 port supports the 32-bit architecture from Andes 8962 Technology Corporation. 8963 * The port provides initial support for the V2, V3, V3m instruction 8964 set architectures. 8965 8966 [39]Nios II 8967 8968 * A port for the Altera Nios II has been contributed by Mentor 8969 Graphics. 8970 8971 [40]PowerPC / PowerPC64 / RS6000 8972 8973 * GCC now supports Power ISA 2.07, which includes support for 8974 Hardware Transactional Memory (HTM), Quadword atomics and several 8975 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 8976 integer and decimal integer operations. 8977 * Support for the POWER8 processor is now available through the 8978 -mcpu=power8 and -mtune=power8 options. 8979 * The libitm library has been modified to add a HTM fastpath that 8980 automatically uses POWER's HTM hardware instructions when it is 8981 executing on a HTM enabled processor. 8982 * Support for the new powerpc64le-linux platform has been added. It 8983 defaults to generating code that conforms to the ELFV2 ABI. 8984 8985 S/390, System z 8986 8987 * Support for the Transactional Execution Facility included with the 8988 IBM zEnterprise zEC12 processor has been added. A set of GCC style 8989 builtins as well as XLC style builtins are provided. The builtins 8990 are enabled by default when using the -march=zEC12 option but can 8991 explicitly be disabled with -mno-htm. Using the GCC builtins also 8992 libitm supports hardware transactions on S/390. 8993 * The hotpatch features allows to prepare functions for hotpatching. 8994 A certain amount of bytes is reserved before the function entry 8995 label plus a NOP is inserted at its very beginning to implement a 8996 backward jump when applying a patch. The feature can either be 8997 enabled per compilation unit via the command-line option -mhotpatch 8998 or per function using the hotpatch attribute. 8999 * The shrink wrap optimization is now supported on S/390 and enabled 9000 by default. 9001 * A major rework of the routines to determine which registers need to 9002 be saved and restored in function prologue/epilogue now allow to 9003 use floating point registers as save slots. This will happen for 9004 certain leaf function with -march=z10 or higher. 9005 * The LRA rtl pass replaces reload by default on S/390. 9006 9007 [41]RX 9008 9009 * The port now allows to specify the RX100, RX200, and RX600 9010 processors with the command-line options -mcpu=rx100, -mcpu=rx200 9011 and -mcpu=rx600. 9012 9013 [42]SH 9014 9015 * Minor improvements to code generated for integer arithmetic and 9016 code that involves the T bit. 9017 * Added support for the SH2A clips and clipu instructions. The 9018 compiler will now try to utilize them for min/max expressions such 9019 as max (-128, min (127, x)). 9020 * Added support for the cmp/str instruction through built-in 9021 functions such as __builtin_strlen. When not optimizing for size, 9022 the compiler will now expand calls to e.g. strlen as an inlined 9023 sequences which utilize the cmp/str instruction. 9024 * Improved code generated around volatile memory loads and stores. 9025 * The option -mcbranchdi has been deprecated. Specifying it will 9026 result in a warning and will not influence code generation. 9027 * The option -mcmpeqdi has been deprecated. Specifying it will result 9028 in a warning and will not influence code generation. 9029 9030 [43]GCC 4.9.1 9031 9032 This is the [44]list of problem reports (PRs) from GCC's bug tracking 9033 system that are known to be fixed in the 4.9.1 release. This list might 9034 not be complete (that is, it is possible that some PRs that have been 9035 fixed are not listed here). 9036 9037 Version 4.0 of the OpenMP specification is supported even in Fortran, 9038 not just C and C++. 9039 9040 [45]GCC 4.9.2 9041 9042 This is the [46]list of problem reports (PRs) from GCC's bug tracking 9043 system that are known to be fixed in the 4.9.2 release. This list might 9044 not be complete (that is, it is possible that some PRs that have been 9045 fixed are not listed here). 9046 9047 [47]GCC 4.9.3 9048 9049 This is the [48]list of problem reports (PRs) from GCC's bug tracking 9050 system that are known to be fixed in the 4.9.3 release. This list might 9051 not be complete (that is, it is possible that some PRs that have been 9052 fixed are not listed here). 9053 9054 [49]GCC 4.9.4 9055 9056 This is the [50]list of problem reports (PRs) from GCC's bug tracking 9057 system that are known to be fixed in the 4.9.4 release. This list might 9058 not be complete (that is, it is possible that some PRs that have been 9059 fixed are not listed here). 9060 9061 9062 For questions related to the use of GCC, please consult these web 9063 pages and the [51]GCC manuals. If that fails, the 9064 [52]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9065 web pages and the development of GCC are welcome on our developer 9066 list at [53]gcc (a] gcc.gnu.org. All of [54]our lists have public 9067 archives. 9068 9069 Copyright (C) [55]Free Software Foundation, Inc. Verbatim copying and 9070 distribution of this entire article is permitted in any medium, 9071 provided this notice is preserved. 9072 9073 These pages are [56]maintained by the GCC team. Last modified 9074 2025-06-10. 9075 9076 References 9077 9078 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 9079 2. https://gcc.gnu.org/PR60825 9080 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 9081 4. https://gcc.gnu.org/gcc-4.9/changes.html#languages 9082 5. https://www.openmp.org/specifications/ 9083 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 9084 7. https://gcc.gnu.org/gcc-4.9/changes.html#ada 9085 8. https://gcc.gnu.org/gcc-4.9/changes.html#c-family 9086 9. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 9087 10. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 9088 11. https://gcc.gnu.org/gcc-4.9/changes.html#c 9089 12. https://gcc.gnu.org/gcc-4.9/changes.html#cxx 9090 13. https://gcc.gnu.org/projects/cxx1y.html 9091 14. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 9092 15. https://gcc.gnu.org/projects/cxx1y.html 9093 16. https://gcc.gnu.org/projects/cxx1y.html 9094 17. https://gcc.gnu.org/projects/cxx1y.html 9095 18. https://gcc.gnu.org/projects/cxx1y.html 9096 19. https://gcc.gnu.org/projects/cxx1y.html 9097 20. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 9098 21. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 9099 22. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 9100 23. https://gcc.gnu.org/gcc-4.9/changes.html#fortran 9101 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 9102 25. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 9103 26. https://gcc.gnu.org/wiki/Fortran2003Status 9104 27. https://gcc.gnu.org/wiki/Fortran2008Status 9105 28. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 9106 29. https://gcc.gnu.org/gcc-4.9/changes.html#go 9107 30. https://gcc.gnu.org/gcc-4.9/changes.html#targets 9108 31. https://gcc.gnu.org/gcc-4.9/changes.html#aarch64 9109 32. https://gcc.gnu.org/gcc-4.9/changes.html#arc 9110 33. https://gcc.gnu.org/gcc-4.9/changes.html#arm 9111 34. https://gcc.gnu.org/gcc-4.9/changes.html#avr 9112 35. https://gcc.gnu.org/gcc-4.9/changes.html#x86 9113 36. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 9114 37. https://gcc.gnu.org/gcc-4.9/changes.html#msp430 9115 38. https://gcc.gnu.org/gcc-4.9/changes.html#nds32 9116 39. https://gcc.gnu.org/gcc-4.9/changes.html#nios2 9117 40. https://gcc.gnu.org/gcc-4.9/changes.html#powerpc 9118 41. https://gcc.gnu.org/gcc-4.9/changes.html#rx 9119 42. https://gcc.gnu.org/gcc-4.9/changes.html#sh 9120 43. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.1 9121 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 9122 45. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.2 9123 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 9124 47. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.3 9125 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 9126 49. https://gcc.gnu.org/gcc-4.9/changes.html#GCC4.9.4 9127 50. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 9128 51. https://gcc.gnu.org/onlinedocs/ 9129 52. mailto:gcc-help (a] gcc.gnu.org 9130 53. mailto:gcc (a] gcc.gnu.org 9131 54. https://gcc.gnu.org/lists.html 9132 55. https://www.fsf.org/ 9133 56. https://gcc.gnu.org/about.html 9134 ====================================================================== 9135 http://gcc.gnu.org/gcc-4.8/index.html 9136 9137 GCC 4.8 Release Series 9138 9139 (This release series is no longer supported.) 9140 9141 June 23, 2015 9142 9143 The [1]GNU project and the GCC developers are pleased to announce the 9144 release of GCC 4.8.5. 9145 9146 This release is a bug-fix release, containing fixes for regressions in 9147 GCC 4.8.4 relative to previous releases of GCC. 9148 9149 Release History 9150 9151 GCC 4.8.5 9152 June 23, 2015 ([2]changes, [3]documentation) 9153 9154 GCC 4.8.4 9155 December 19, 2014 ([4]changes, [5]documentation) 9156 9157 GCC 4.8.3 9158 May 22, 2014 ([6]changes, [7]documentation) 9159 9160 GCC 4.8.2 9161 October 16, 2013 ([8]changes, [9]documentation) 9162 9163 GCC 4.8.1 9164 May 31, 2013 ([10]changes, [11]documentation) 9165 9166 GCC 4.8.0 9167 March 22, 2013 ([12]changes, [13]documentation) 9168 9169 References and Acknowledgements 9170 9171 GCC used to stand for the GNU C Compiler, but since the compiler 9172 supports several other languages aside from C, it now stands for the 9173 GNU Compiler Collection. 9174 9175 A list of [14]successful builds is updated as new information becomes 9176 available. 9177 9178 The GCC developers would like to thank the numerous people that have 9179 contributed new features, improvements, bug fixes, and other changes as 9180 well as test results to GCC. This [15]amazing group of volunteers is 9181 what makes GCC successful. 9182 9183 For additional information about GCC please refer to the [16]GCC 9184 project web site or contact the [17]GCC development mailing list. 9185 9186 To obtain GCC please use [18]our mirror sites or [19]our version 9187 control system. 9188 9189 9190 For questions related to the use of GCC, please consult these web 9191 pages and the [20]GCC manuals. If that fails, the 9192 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9193 web pages and the development of GCC are welcome on our developer 9194 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 9195 archives. 9196 9197 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 9198 distribution of this entire article is permitted in any medium, 9199 provided this notice is preserved. 9200 9201 These pages are [25]maintained by the GCC team. Last modified 9202 2022-10-26. 9203 9204 References 9205 9206 1. http://www.gnu.org/ 9207 2. https://gcc.gnu.org/gcc-4.8/changes.html 9208 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 9209 4. https://gcc.gnu.org/gcc-4.8/changes.html 9210 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 9211 6. https://gcc.gnu.org/gcc-4.8/changes.html 9212 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 9213 8. https://gcc.gnu.org/gcc-4.8/changes.html 9214 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 9215 10. https://gcc.gnu.org/gcc-4.8/changes.html 9216 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 9217 12. https://gcc.gnu.org/gcc-4.8/changes.html 9218 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 9219 14. https://gcc.gnu.org/gcc-4.8/buildstat.html 9220 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9221 16. https://gcc.gnu.org/index.html 9222 17. mailto:gcc (a] gcc.gnu.org 9223 18. https://gcc.gnu.org/mirrors.html 9224 19. https://gcc.gnu.org/git.html 9225 20. https://gcc.gnu.org/onlinedocs/ 9226 21. mailto:gcc-help (a] gcc.gnu.org 9227 22. mailto:gcc (a] gcc.gnu.org 9228 23. https://gcc.gnu.org/lists.html 9229 24. https://www.fsf.org/ 9230 25. https://gcc.gnu.org/about.html 9231 ====================================================================== 9232 http://gcc.gnu.org/gcc-4.8/changes.html 9233 9234 GCC 4.8 Release Series 9235 Changes, New Features, and Fixes 9236 9237 Caveats 9238 9239 GCC now uses C++ as its implementation language. This means that to 9240 build GCC from sources, you will need a C++ compiler that understands 9241 C++ 2003. For more details on the rationale and specific changes, 9242 please refer to the [1]C++ conversion page. 9243 9244 To enable the Graphite framework for loop optimizations you now need 9245 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 9246 the [2]GCC infrastructure directory. The installation manual contains 9247 more information about requirements to build GCC. 9248 9249 GCC now uses a more aggressive analysis to derive an upper bound for 9250 the number of iterations of loops using constraints imposed by language 9251 standards. This may cause non-conforming programs to no longer work as 9252 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 9253 option, -fno-aggressive-loop-optimizations, was added to disable this 9254 aggressive analysis. In some loops that have known constant number of 9255 iterations, but undefined behavior is known to occur in the loop before 9256 reaching or during the last iteration, GCC will warn about the 9257 undefined behavior in the loop instead of deriving lower upper bound of 9258 the number of iterations for the loop. The warning can be disabled with 9259 -Wno-aggressive-loop-optimizations. 9260 9261 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 9262 for the layout of vectors that could lead to wrong code being 9263 generated. Vectors larger than 8 bytes in size are now by default 9264 aligned to an 8-byte boundary. This is an ABI change: code that makes 9265 explicit use of vector types may be incompatible with binary objects 9266 built with older versions of GCC. Auto-vectorized code is not affected 9267 by this change. 9268 9269 More information on porting to GCC 4.8 from previous versions of GCC 9270 can be found in the [3]porting guide for this release. 9271 9272 General Optimizer Improvements (and Changes) 9273 9274 * DWARF4 is now the default when generating DWARF debug information. 9275 When -g is used on a platform that uses DWARF debugging 9276 information, GCC will now default to -gdwarf-4 9277 -fno-debug-types-section. 9278 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 9279 consumers support DWARF4 by default. Before GCC 4.8 the default 9280 version used was DWARF2. To make GCC 4.8 generate an older DWARF 9281 version use -g together with -gdwarf-2 or -gdwarf-3. The default 9282 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 9283 * A new general optimization level, -Og, has been introduced. It 9284 addresses the need for fast compilation and a superior debugging 9285 experience while providing a reasonable level of run-time 9286 performance. Overall experience for development should be better 9287 than the default optimization level -O0. 9288 * A new option -ftree-partial-pre was added to control the partial 9289 redundancy elimination (PRE) optimization. This option is enabled 9290 by default at the -O3 optimization level, and it makes PRE more 9291 aggressive. 9292 * The option -fconserve-space has been removed; it was no longer 9293 useful on most targets since GCC supports putting variables into 9294 BSS without making them common. 9295 * The struct reorg and matrix reorg optimizations (command-line 9296 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 9297 removed. They did not always work correctly, nor did they work with 9298 link-time optimization (LTO), hence were only applicable to 9299 programs consisting of a single translation unit. 9300 * Several scalability bottle-necks have been removed from GCC's 9301 optimization passes. Compilation of extremely large functions, e.g. 9302 due to the use of the flatten attribute in the "Eigen" C++ linear 9303 algebra templates library, is significantly faster than previous 9304 releases of GCC. 9305 * Link-time optimization (LTO) improvements: 9306 + LTO partitioning has been rewritten for better reliability and 9307 maintanibility. Several important bugs leading to link 9308 failures have been fixed. 9309 * Interprocedural optimization improvements: 9310 + A new symbol table has been implemented. It builds on existing 9311 callgraph and varpool modules and provide a new API. Unusual 9312 symbol visibilities and aliases are handled more consistently 9313 leading to, for example, more aggressive unreachable code 9314 removal with LTO. 9315 + The inline heuristic can now bypass limits on the size of of 9316 inlined functions when the inlining is particularly 9317 profitable. This happens, for example, when loop bounds or 9318 array strides get propagated. 9319 + Values passed through aggregates (either by value or 9320 reference) are now propagated at the inter-procedural level 9321 leading to better inlining decisions (for example in the case 9322 of Fortran array descriptors) and devirtualization. 9323 * [4]AddressSanitizer , a fast memory error detector, has been added 9324 and can be enabled via -fsanitize=address. Memory access 9325 instructions will be instrumented to detect heap-, stack-, and 9326 global-buffer overflow as well as use-after-free bugs. To get nicer 9327 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 9328 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 9329 x86-64 Darwin. 9330 * [5]ThreadSanitizer has been added and can be enabled via 9331 -fsanitize=thread. Instructions will be instrumented to detect data 9332 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 9333 * A new local register allocator (LRA) has been implemented, which 9334 replaces the 26 year old reload pass and improves generated code 9335 quality. For now it is active on the IA-32 and x86-64 targets. 9336 * Support for transactional memory has been implemented on the 9337 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 9338 Alpha. 9339 9340 New Languages and Language specific improvements 9341 9342 C family 9343 9344 * Each diagnostic emitted now includes the original source line and a 9345 caret '^' indicating the column. The option 9346 -fno-diagnostics-show-caret suppresses this information. 9347 * The option -ftrack-macro-expansion=2 is now enabled by default. 9348 This allows the compiler to display the macro expansion stack in 9349 diagnostics. Combined with the caret information, an example 9350 diagnostic showing these two features is: 9351 9352 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 9353 ') 9354 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 9355 _b = (B); __a < __b ? __b : __a; }) 9356 9357 ^ 9358 t.c:7:7: note: in expansion of macro 'MYMAX' 9359 X = MYMAX(P, F); 9360 ^ 9361 9362 * A new -Wsizeof-pointer-memaccess warning has been added (also 9363 enabled by -Wall) to warn about suspicious length parameters to 9364 certain string and memory built-in functions if the argument uses 9365 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 9366 (ptr)); if ptr is not an array, but a pointer, and suggests a 9367 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 9368 * The new option -Wpedantic is an alias for -pedantic, which is now 9369 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 9370 -Wno-error=pedantic work in the same way as for any other -W 9371 option. One caveat is that -Werror=pedantic is not equivalent to 9372 -pedantic-errors, since the latter makes into errors some warnings 9373 that are not controlled by -Wpedantic, and the former only affects 9374 diagnostics that are disabled when using -Wno-pedantic. 9375 * The option -Wshadow no longer warns if a declaration shadows a 9376 function declaration, unless the former declares a function or 9377 pointer to function, because this is [6]a common and valid case in 9378 real-world code. 9379 9380 [7]C++ 9381 9382 * G++ now implements the [8]C++11 thread_local keyword; this differs 9383 from the GNU __thread keyword primarily in that it allows dynamic 9384 initialization and destruction semantics. Unfortunately, this 9385 support requires a run-time penalty for references to 9386 non-function-local thread_local variables defined in a different 9387 translation unit even if they don't need dynamic initialization, so 9388 users may want to continue to use __thread for TLS variables with 9389 static initialization semantics. 9390 If the programmer can be sure that no use of the variable in a 9391 non-defining TU needs to trigger dynamic initialization (either 9392 because the variable is statically initialized, or a use of the 9393 variable in the defining TU will be executed before any uses in 9394 another TU), they can avoid this overhead with the 9395 -fno-extern-tls-init option. 9396 OpenMP threadprivate variables now also support dynamic 9397 initialization and destruction by the same mechanism. 9398 * G++ now implements the [9]C++11 attribute syntax, e.g. 9399 9400 [[noreturn]] void f(); 9401 9402 and also the alignment specifier, e.g. 9403 9404 alignas(double) int i; 9405 9406 * G++ now implements [10]C++11 inheriting constructors, e.g. 9407 9408 struct A { A(int); }; 9409 struct B: A { using A::A; }; // defines B::B(int) 9410 B b(42); // OK 9411 9412 * As of GCC 4.8.1, G++ implements the change to decltype semantics 9413 from [11]N3276. 9414 9415 struct A f(); 9416 decltype(f()) g(); // OK, return type of f() is not required to be complete. 9417 9418 * As of GCC 4.8.1, G++ implements [12]C++11 ref-qualifiers, e.g. 9419 9420 struct A { int f() &; }; 9421 int i = A().f(); // error, f() requires an lvalue object 9422 9423 * G++ now supports a -std=c++1y option for experimentation with 9424 features proposed for the next revision of the standard, expected 9425 around 2014. Currently the only difference from -std=c++11 is 9426 support for return type deduction in normal functions, as proposed 9427 in [13]N3386. Status of C++1y features in GCC 4.8 can be found 9428 [14]here. 9429 * The G++ namespace association extension, __attribute ((strong)), 9430 has been deprecated. Inline namespaces should be used instead. 9431 * G++ now supports a -fext-numeric-literal option to control whether 9432 GNU numeric literal suffixes are accepted as extensions or 9433 processed as C++11 user-defined numeric literal suffixes. The flag 9434 is on (use suffixes for GNU literals) by default for -std=gnu++*, 9435 and -std=c++98. The flag is off (use suffixes for user-defined 9436 literals) by default for -std=c++11 and later. 9437 9438 Runtime Library (libstdc++) 9439 9440 * [15]Improved experimental support for the new ISO C++ standard, 9441 C++11, including: 9442 + forward_list meets the allocator-aware container requirements; 9443 + this_thread::sleep_for(), this_thread::sleep_until() and 9444 this_thread::yield() are defined without requiring the 9445 configure option --enable-libstdcxx-time; 9446 * Improvements to <random>: 9447 + SSE optimized normal_distribution. 9448 + Use of hardware RNG instruction for random_device on new x86 9449 processors (requires the assembler to support the 9450 instruction.) 9451 and <ext/random>: 9452 + New random number engine simd_fast_mersenne_twister_engine 9453 with an optimized SSE implementation. 9454 + New random number distributions beta_distribution, 9455 normal_mv_distribution, rice_distribution, 9456 nakagami_distribution, pareto_distribution, k_distribution, 9457 arcsine_distribution, hoyt_distribution. 9458 * Added --disable-libstdcxx-verbose configure option to disable 9459 diagnostic messages issued when a process terminates abnormally. 9460 This may be useful for embedded systems to reduce the size of 9461 executables that link statically to the library. 9462 9463 [16]Fortran 9464 9465 * Compatibility notice: 9466 + Module files: The version of module files (.mod) has been 9467 incremented. Fortran MODULEs compiled by earlier GCC versions 9468 have to be recompiled, when they are USEd by files compiled 9469 with GCC 4.8. GCC 4.8 is not able to read .mod files created 9470 by earlier versions; attempting to do so gives an error 9471 message. 9472 Note: The ABI of the produced assembler data itself has not 9473 changed; object files and libraries are fully compatible with 9474 older versions except as noted below. 9475 + ABI: Some internal names (used in the assembler/object file) 9476 have changed for symbols declared in the specification part of 9477 a module. If an affected module - or a file using it via use 9478 association - is recompiled, the module and all files which 9479 directly use such symbols have to be recompiled as well. This 9480 change only affects the following kind of module symbols: 9481 o Procedure pointers. Note: C-interoperable function 9482 pointers (type(c_funptr)) are not affected nor are 9483 procedure-pointer components. 9484 o Deferred-length character strings. 9485 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 9486 backtrace at an arbitrary place in user code; program execution 9487 continues normally afterwards. 9488 * The [18]-Wc-binding-type warning option has been added (disabled by 9489 default). It warns if the a variable might not be C interoperable; 9490 in particular, if the variable has been declared using an intrinsic 9491 type with default kind instead of using a kind parameter defined 9492 for C interoperability in the intrinsic ISO_C_Binding module. 9493 Before, this warning was always printed. The -Wc-binding-type 9494 option is enabled by -Wall. 9495 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 9496 options have been added, which diagnose when code is inserted for 9497 automatic (re)allocation of a variable during assignment. This 9498 option can be used to decide whether it is safe to use 9499 [20]-fno-realloc-lhs. Additionally, it can be used to find 9500 automatic (re)allocation in hot loops. (For arrays, replacing 9501 "var=" by "var(:)=" disables the automatic reallocation.) 9502 * The [21]-Wcompare-reals command-line option has been added. When 9503 this is set, warnings are issued when comparing REAL or COMPLEX 9504 types for equality and inequality; consider replacing a == b by 9505 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by 9506 -Wextra. 9507 * The [22]-Wtarget-lifetime command-line option has been added 9508 (enabled with -Wall), which warns if the pointer in a pointer 9509 assignment might outlive its target. 9510 * Reading floating point numbers which use "q" for the exponential 9511 (such as 4.0q0) is now supported as vendor extension for better 9512 compatibility with old data files. It is strongly recommended to 9513 use for I/O the equivalent but standard conforming "e" (such as 9514 4.0e0). 9515 (For Fortran source code, consider replacing the "q" in 9516 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 9517 suitable qp). Note that - in Fortran source code - replacing "q" by 9518 a simple "e" is not equivalent.) 9519 * The GFORTRAN_TMPDIR environment variable for specifying a 9520 non-default directory for files opened with STATUS="SCRATCH", is 9521 not used anymore. Instead gfortran checks the POSIX/GNU standard 9522 TMPDIR environment variable. If TMPDIR is not defined, gfortran 9523 falls back to other methods to determine the directory for 9524 temporary files as documented in the [23]user manual. 9525 * [24]Fortran 2003: 9526 + Support for unlimited polymorphic variables (CLASS(*)) has 9527 been added. Nonconstant character lengths are not yet 9528 supported. 9529 * [25]TS 29113: 9530 + Assumed types (TYPE(*)) are now supported. 9531 + Experimental support for assumed-rank arrays (dimension(..)) 9532 has been added. Note that currently gfortran's own array 9533 descriptor is used, which is different from the one defined in 9534 TS29113, see [26]gfortran's header file or use the [27]Chasm 9535 Language Interoperability Tools. 9536 9537 [28]Go 9538 9539 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 9540 release. 9541 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 9542 release. The library support is not quite complete. 9543 * Go has been tested on GNU/Linux and Solaris platforms for various 9544 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 9545 work on other platforms as well. 9546 9547 [29]New Targets and Target Specific Improvements 9548 9549 [30]AArch64 9550 9551 * A new port has been added to support AArch64, the new 64-bit 9552 architecture from ARM. Note that this is a separate port from the 9553 existing 32-bit ARM port. 9554 * The port provides initial support for the Cortex-A53 and the 9555 Cortex-A57 processors with the command line options 9556 -mcpu=cortex-a53 and -mcpu=cortex-a57. 9557 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 9558 has been added and can be enabled by giving the 9559 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 9560 default by configuring GCC with the --enable-fix-cortex-a53-835769 9561 option. 9562 9563 [31]ARM 9564 9565 * Initial support has been added for the AArch32 extensions defined 9566 in the ARMv8 architecture. 9567 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 9568 * A new option, -mcpu=marvell-pj4, has been added to generate code 9569 for the Marvell PJ4 processor. 9570 * The compiler can now automatically generate the VFMA, VFMS, REVSH 9571 and REV16 instructions. 9572 * A new vectorizer cost model for Advanced SIMD configurations to 9573 improve the auto-vectorization strategies used. 9574 * The scheduler now takes into account the number of live registers 9575 to reduce the amount of spilling that can occur. This should 9576 improve code performance in large functions. The limit can be 9577 removed by using the option -fno-sched-pressure. 9578 * Improvements have been made to the Marvell iWMMX code generation 9579 and support for the iWMMX2 SIMD unit has been added. The option 9580 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 9581 * A number of code generation improvements for Thumb2 to reduce code 9582 size when compiling for the M-profile processors. 9583 * The RTEMS (arm-rtems) port has been updated to use the EABI. 9584 * Code generation support for the old FPA and Maverick floating-point 9585 architectures has been removed. Ports that previously relied on 9586 these features have also been removed. This includes the targets: 9587 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 9588 + arm*-*-elf (use arm*-*-eabi) 9589 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 9590 + arm*-*-ecos-elf (no alternative) 9591 + arm*-*-freebsd (no alternative) 9592 + arm*-wince-pe* (no alternative). 9593 9594 [32]AVR 9595 9596 * Support for the command-line option -mshort-calls has been removed. 9597 It was deprecated in GCC 4.7. 9598 * The configure option --with-avrlibc supported since GCC 4.7.2 is 9599 turned on per default for all non-RTEMS configurations. This option 9600 arranges for a better integration of [33]AVR-LibC with avr-gcc. For 9601 technical details, see [34]PR54461. To turn off the option in 9602 non-AVR-LibC configurations, use --with-avrlibc=no. If the compiler 9603 is configured for RTEMS, the option is always turned off. 9604 * Support for the "Embedded C" fixed-point has been added. For 9605 details, see the [35]GCC wiki and the [36]user manual. The support 9606 is not complete. 9607 * A new print modifier %r for register operands in inline assembler 9608 is supported. It will print the raw register number without the 9609 register prefix 'r': 9610 /* Return the most significant byte of 'val', a 64-bit value. */ 9611 9612 unsigned char msb (long long val) 9613 { 9614 unsigned char c; 9615 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 9616 return c; 9617 } 9618 The inline assembler in this example will generate code like 9619 mov r24, 8+7 9620 provided c is allocated to R24 and val is allocated to R8...R15. 9621 This works because the GNU assembler accepts plain register numbers 9622 without register prefix. 9623 * Static initializers with 3-byte symbols are supported now: 9624 extern const __memx char foo; 9625 const __memx void *pfoo = &foo; 9626 This requires at least Binutils 2.23. 9627 9628 IA-32/x86-64 9629 9630 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 9631 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 9632 stack alignment, this is ABI incompatible and intended to be used 9633 in controlled environments where stack space is an important 9634 limitation. This option will lead to wrong code when functions 9635 compiled with 16 byte stack alignment (such as functions from a 9636 standard library) are called with misaligned stack. In this case, 9637 SSE instructions may lead to misaligned memory access traps. In 9638 addition, variable arguments will be handled incorrectly for 16 9639 byte aligned objects (including x87 long double and __int128), 9640 leading to wrong results. You must build all modules with 9641 -mpreferred-stack-boundary=3, including any libraries. This 9642 includes the system libraries and startup modules. 9643 * Support for the new Intel processor codename Broadwell with RDSEED, 9644 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 9645 -mrdseed command-line options. 9646 * Support for the Intel RTM and HLE intrinsics, built-in functions 9647 and code generation is available via -mrtm and -mhle. 9648 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 9649 Intrinsics and built-in functions are available via -mfxsr, -mxsave 9650 and -mxsaveopt respectively. 9651 * New -maddress-mode=[short|long] options for x32. 9652 -maddress-mode=short overrides default 64-bit addresses to 32-bit 9653 by emitting the 0x67 address-size override prefix. This is the 9654 default address mode for x32. 9655 * New built-in functions to detect run-time CPU type and ISA: 9656 + A built-in function __builtin_cpu_is has been added to detect 9657 if the run-time CPU is of a particular type. It returns a 9658 positive integer on a match and zero otherwise. It accepts one 9659 string literal argument, the CPU name. For example, 9660 __builtin_cpu_is("westmere") returns a positive integer if the 9661 run-time CPU is an Intel Core i7 Westmere processor. Please 9662 refer to the [37]user manual for the list of valid CPU names 9663 recognized. 9664 + A built-in function __builtin_cpu_supports has been added to 9665 detect if the run-time CPU supports a particular ISA feature. 9666 It returns a positive integer on a match and zero otherwise. 9667 It accepts one string literal argument, the ISA feature. For 9668 example, __builtin_cpu_supports("ssse3") returns a positive 9669 integer if the run-time CPU supports SSSE3 instructions. 9670 Please refer to the [38]user manual for the list of valid ISA 9671 names recognized. 9672 Caveat: If these built-in functions are called before any static 9673 constructors are invoked, like during IFUNC initialization, then 9674 the CPU detection initialization must be explicitly run using this 9675 newly provided built-in function, __builtin_cpu_init. The 9676 initialization needs to be done only once. For example, this is how 9677 the invocation would look like inside an IFUNC initializer: 9678 static void (*some_ifunc_resolver(void))(void) 9679 { 9680 __builtin_cpu_init(); 9681 if (__builtin_cpu_is("amdfam10h") ... 9682 if (__builtin_cpu_supports("popcnt") ... 9683 } 9684 9685 * Function Multiversioning Support with G++: 9686 It is now possible to create multiple function versions each 9687 targeting a specific processor and/or ISA. Function versions have 9688 the same signature but different target attributes. For example, 9689 here is a program with function versions: 9690 __attribute__ ((target ("default"))) 9691 int foo(void) 9692 { 9693 return 1; 9694 } 9695 9696 __attribute__ ((target ("sse4.2"))) 9697 int foo(void) 9698 { 9699 return 2; 9700 } 9701 9702 int main (void) 9703 { 9704 int (*p) = &foo; 9705 assert ((*p)() == foo()); 9706 return 0; 9707 } 9708 9709 Please refer to this [39]wiki for more information. 9710 * The x86 back end has been improved to allow option -fschedule-insns 9711 to work reliably. This option can be used to schedule instructions 9712 better and leads to improved performace in certain cases. 9713 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 9714 from the Mingw-w64 trunk. 9715 * Support for new AMD family 15h processors (Steamroller core) is now 9716 available through the -march=bdver3 and -mtune=bdver3 options. 9717 * Support for new AMD family 16h processors (Jaguar core) is now 9718 available through the -march=btver2 and -mtune=btver2 options. 9719 9720 [40]FRV 9721 9722 * This target now supports the -fstack-usage command-line option. 9723 9724 [41]MIPS 9725 9726 * GCC can now generate code specifically for the R4700, Broadcom XLP 9727 and MIPS 34kn processors. The associated -march options are 9728 -march=r4700, -march=xlp and -march=34kn respectively. 9729 * GCC now generates better DSP code for MIPS 74k cores thanks to 9730 further scheduling optimizations. 9731 * The MIPS port now supports the -fstack-check option. 9732 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 9733 * Previous versions of GCC would silently accept -fpic and -fPIC for 9734 -mno-abicalls targets like mips*-elf. This combination was not 9735 intended or supported, and did not generate position-independent 9736 code. GCC 4.8 now reports an error when this combination is used. 9737 9738 [42]PowerPC / PowerPC64 / RS6000 9739 9740 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 9741 restore or update the VRSAVE register by default. The respective 9742 operating systems manage the VRSAVE register directly. 9743 * Large TOC support has been added for AIX through the command line 9744 option -mcmodel=large. 9745 * Native Thread-Local Storage support has been added for AIX. 9746 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 9747 when targetting processors that support those hardware features on 9748 AIX 6.1 and above. 9749 9750 [43]RX 9751 9752 * This target will now issue a warning message whenever multiple fast 9753 interrupt handlers are found in the same compilation unit. This 9754 feature can be turned off by the new 9755 -mno-warn-multiple-fast-interrupts command-line option. 9756 9757 S/390, System z 9758 9759 * Support for the IBM zEnterprise zEC12 processor has been added. 9760 When using the -march=zEC12 option, the compiler will generate code 9761 making use of the following new instructions: 9762 + load and trap instructions 9763 + 2 new compare and trap instructions 9764 + rotate and insert selected bits - without CC clobber 9765 The -mtune=zEC12 option enables zEC12 specific instruction 9766 scheduling without making use of new instructions. 9767 * Register pressure sensitive instruction scheduling is enabled by 9768 default. 9769 * The ifunc function attribute is enabled by default. 9770 * memcpy and memcmp invokations on big memory chunks or with run time 9771 lengths are not generated inline anymore when tuning for z10 or 9772 higher. The purpose is to make use of the IFUNC optimized versions 9773 in Glibc. 9774 9775 [44]SH 9776 9777 * The default alignment settings have been reduced to be less 9778 aggressive. This results in more compact code for optimization 9779 levels other than -Os. 9780 * Improved support for the __atomic built-in functions: 9781 + A new option -matomic-model=model selects the model for the 9782 generated atomic sequences. The following models are 9783 supported: 9784 9785 soft-gusa 9786 Software gUSA sequences (SH3* and SH4* only). On 9787 SH4A targets this will now also partially utilize 9788 the movco.l and movli.l instructions. This is the 9789 default when the target is sh3*-*-linux* or 9790 sh4*-*-linux*. 9791 9792 hard-llcs 9793 Hardware movco.l / movli.l sequences (SH4A only). 9794 9795 soft-tcb 9796 Software thread control block sequences. 9797 9798 soft-imask 9799 Software interrupt flipping sequences (privileged 9800 mode only). This is the default when the target is 9801 sh1*-*-linux* or sh2*-*-linux*. 9802 9803 none 9804 Generates function calls to the respective __atomic 9805 built-in functions. This is the default for SH64 9806 targets or when the target is not sh*-*-linux*. 9807 9808 + The option -msoft-atomic has been deprecated. It is now an 9809 alias for -matomic-model=soft-gusa. 9810 + A new option -mtas makes the compiler generate the tas.b 9811 instruction for the __atomic_test_and_set built-in function 9812 regardless of the selected atomic model. 9813 + The __sync functions in libgcc now reflect the selected atomic 9814 model when building the toolchain. 9815 * Added support for the mov.b and mov.w instructions with 9816 displacement addressing. 9817 * Added support for the SH2A instructions movu.b and movu.w. 9818 * Various improvements to code generated for integer arithmetic. 9819 * Improvements to conditional branches and code that involves the T 9820 bit. A new option -mzdcbranch tells the compiler to favor 9821 zero-displacement branches. This is enabled by default for SH4* 9822 targets. 9823 * The pref instruction will now be emitted by the __builtin_prefetch 9824 built-in function for SH3* targets. 9825 * The fmac instruction will now be emitted by the fmaf standard 9826 function and the __builtin_fmaf built-in function. 9827 * The -mfused-madd option has been deprecated in favor of the 9828 machine-independent -ffp-contract option. Notice that the fmac 9829 instruction will now be generated by default for expressions like a 9830 * b + c. This is due to the compiler default setting 9831 -ffp-contract=fast. 9832 * Added new options -mfsrra and -mfsca to allow the compiler using 9833 the fsrra and fsca instructions on targets other than SH4A (where 9834 they are already enabled by default). 9835 * Added support for the __builtin_bswap32 built-in function. It is 9836 now expanded as a sequence of swap.b and swap.w instructions 9837 instead of a library function call. 9838 * The behavior of the -mieee option has been fixed and the negative 9839 form -mno-ieee has been added to control the IEEE conformance of 9840 floating point comparisons. By default -mieee is now enabled and 9841 the option -ffinite-math-only implicitly sets -mno-ieee. 9842 * Added support for the built-in functions __builtin_thread_pointer 9843 and __builtin_set_thread_pointer. This assumes that GBR is used to 9844 hold the thread pointer of the current thread. Memory loads and 9845 stores relative to the address returned by __builtin_thread_pointer 9846 will now also utilize GBR based displacement address modes. 9847 * The -mdiv= option for targets other than SHmedia has been fixed and 9848 documented. 9849 9850 [45]SPARC 9851 9852 * Added optimized instruction scheduling for Niagara4. 9853 9854 [46]TILE-Gx 9855 9856 * Added support for the -mcmodel=MODEL command-line option. The 9857 models supported are small and large. 9858 9859 [47]V850 9860 9861 * This target now supports the E3V5 architecture via the use of the 9862 new -mv850e3v5 command-line option. It also has experimental 9863 support for the e3v5 LOOP instruction which can be enabled via the 9864 new -mloop command-line option. 9865 9866 [48]XStormy16 9867 9868 * This target now supports the -fstack-usage command-line option. 9869 9870 [49]Operating Systems 9871 9872 [50]OpenBSD 9873 9874 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 9875 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 9876 9877 [51]Windows (Cygwin) 9878 9879 * Executables are now linked against shared libgcc by default. The 9880 previous default was to link statically, which can still be done by 9881 explicitly specifying -static or static-libgcc on the command line. 9882 However it is strongly advised against, as it will cause problems 9883 for any application that makes use of DLLs compiled by GCC. It 9884 should be alright for a monolithic stand-alone application that 9885 only links against the Windows DLLs, but offers little or no 9886 benefit. 9887 9888 [52]GCC 4.8.1 9889 9890 This is the [53]list of problem reports (PRs) from GCC's bug tracking 9891 system that are known to be fixed in the 4.8.1 release. This list might 9892 not be complete (that is, it is possible that some PRs that have been 9893 fixed are not listed here). 9894 9895 The C++11 <chrono> std::chrono::system_clock and 9896 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 9897 both are now separate (never typedefs of each other), both use 9898 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 9899 std::chrono::steady_clock is now finally monotonic, and both classes 9900 are mangled differently than in the previous GCC releases. 9901 std::chrono::system_clock::now() with std::chrono::microseconds resp. 9902 std::chrono::seconds resolution is still exported for backwards 9903 compatibility with default configured libstdc++. Note that libstdc++ 9904 configured with --enable-libstdcxx-time= used to be ABI incompatible 9905 with default configured libstdc++ for those two classes and no ABI 9906 compatibility can be offered for those configurations, so any C++11 9907 code that uses those classes and has been compiled and linked against 9908 libstdc++ configured with the non-default --enable-libstdcxx-time= 9909 configuration option needs to be recompiled. 9910 9911 [54]GCC 4.8.2 9912 9913 This is the [55]list of problem reports (PRs) from GCC's bug tracking 9914 system that are known to be fixed in the 4.8.2 release. This list might 9915 not be complete (that is, it is possible that some PRs that have been 9916 fixed are not listed here). 9917 9918 [56]GCC 4.8.3 9919 9920 This is the [57]list of problem reports (PRs) from GCC's bug tracking 9921 system that are known to be fixed in the 4.8.3 release. This list might 9922 not be complete (that is, it is possible that some PRs that have been 9923 fixed are not listed here). 9924 9925 Support for the new powerpc64le-linux platform has been added. It 9926 defaults to generating code that conforms to the ELFV2 ABI. 9927 9928 [58]GCC 4.8.4 9929 9930 This is the [59]list of problem reports (PRs) from GCC's bug tracking 9931 system that are known to be fixed in the 4.8.4 release. This list might 9932 not be complete (that is, it is possible that some PRs that have been 9933 fixed are not listed here). 9934 9935 [60]GCC 4.8.5 9936 9937 This is the [61]list of problem reports (PRs) from GCC's bug tracking 9938 system that are known to be fixed in the 4.8.5 release. This list might 9939 not be complete (that is, it is possible that some PRs that have been 9940 fixed are not listed here). 9941 9942 9943 For questions related to the use of GCC, please consult these web 9944 pages and the [62]GCC manuals. If that fails, the 9945 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9946 web pages and the development of GCC are welcome on our developer 9947 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 9948 archives. 9949 9950 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 9951 distribution of this entire article is permitted in any medium, 9952 provided this notice is preserved. 9953 9954 These pages are [67]maintained by the GCC team. Last modified 9955 2025-02-28. 9956 9957 References 9958 9959 1. https://gcc.gnu.org/wiki/cxx-conversion 9960 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 9961 3. https://gcc.gnu.org/gcc-4.8/porting_to.html 9962 4. https://github.com/google/sanitizers 9963 5. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 9964 6. https://lkml.org/lkml/2006/11/28/239 9965 7. https://gcc.gnu.org/gcc-4.8/changes.html#cxx 9966 8. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9967 9. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9968 10. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9969 11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 9970 12. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9971 13. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 9972 14. https://gcc.gnu.org/projects/cxx1y.html 9973 15. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 9974 16. https://gcc.gnu.org/gcc-4.8/changes.html#fortran 9975 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 9976 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9977 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9978 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 9979 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9980 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9981 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 9982 24. https://gcc.gnu.org/wiki/Fortran2003Status 9983 25. https://gcc.gnu.org/wiki/TS29113Status 9984 26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h 9985 27. https://chasm-interop.sourceforge.net/ 9986 28. https://gcc.gnu.org/gcc-4.8/changes.html#go 9987 29. https://gcc.gnu.org/gcc-4.8/changes.html#targets 9988 30. https://gcc.gnu.org/gcc-4.8/changes.html#aarch64 9989 31. https://gcc.gnu.org/gcc-4.8/changes.html#arm 9990 32. https://gcc.gnu.org/gcc-4.8/changes.html#avr 9991 33. http://www.nongnu.org/avr-libc/ 9992 34. https://gcc.gnu.org/PR54461 9993 35. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 9994 36. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 9995 37. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 9996 38. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 9997 39. https://gcc.gnu.org/wiki/FunctionMultiVersioning 9998 40. https://gcc.gnu.org/gcc-4.8/changes.html#frv 9999 41. https://gcc.gnu.org/gcc-4.8/changes.html#mips 10000 42. https://gcc.gnu.org/gcc-4.8/changes.html#powerpc 10001 43. https://gcc.gnu.org/gcc-4.8/changes.html#rx 10002 44. https://gcc.gnu.org/gcc-4.8/changes.html#sh 10003 45. https://gcc.gnu.org/gcc-4.8/changes.html#sparc 10004 46. https://gcc.gnu.org/gcc-4.8/changes.html#tilegx 10005 47. https://gcc.gnu.org/gcc-4.8/changes.html#v850 10006 48. https://gcc.gnu.org/gcc-4.8/changes.html#xstormy16 10007 49. https://gcc.gnu.org/gcc-4.8/changes.html#os 10008 50. https://gcc.gnu.org/gcc-4.8/changes.html#openbsd 10009 51. https://gcc.gnu.org/gcc-4.8/changes.html#windows 10010 52. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.1 10011 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 10012 54. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.2 10013 55. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 10014 56. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.3 10015 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 10016 58. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.4 10017 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 10018 60. https://gcc.gnu.org/gcc-4.8/changes.html#GCC4.8.5 10019 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 10020 62. https://gcc.gnu.org/onlinedocs/ 10021 63. mailto:gcc-help (a] gcc.gnu.org 10022 64. mailto:gcc (a] gcc.gnu.org 10023 65. https://gcc.gnu.org/lists.html 10024 66. https://www.fsf.org/ 10025 67. https://gcc.gnu.org/about.html 10026 ====================================================================== 10027 http://gcc.gnu.org/gcc-4.7/index.html 10028 10029 GCC 4.7 Release Series 10030 10031 (This release series is no longer supported.) 10032 10033 June 12, 2014 10034 10035 The [1]GNU project and the GCC developers are pleased to announce the 10036 release of GCC 4.7.4. 10037 10038 This release is a bug-fix release, containing fixes for regressions in 10039 GCC 4.7.3 relative to previous releases of GCC. 10040 10041 Release History 10042 10043 GCC 4.7.4 10044 June 12, 2014 ([2]changes, [3]documentation) 10045 10046 GCC 4.7.3 10047 April 11, 2013 ([4]changes, [5]documentation) 10048 10049 GCC 4.7.2 10050 September 20, 2012 ([6]changes, [7]documentation) 10051 10052 GCC 4.7.1 10053 June 14, 2012 ([8]changes, [9]documentation) 10054 10055 GCC 4.7.0 10056 March 22, 2012 ([10]changes, [11]documentation) 10057 10058 References and Acknowledgements 10059 10060 GCC used to stand for the GNU C Compiler, but since the compiler 10061 supports several other languages aside from C, it now stands for the 10062 GNU Compiler Collection. 10063 10064 A list of [12]successful builds is updated as new information becomes 10065 available. 10066 10067 The GCC developers would like to thank the numerous people that have 10068 contributed new features, improvements, bug fixes, and other changes as 10069 well as test results to GCC. This [13]amazing group of volunteers is 10070 what makes GCC successful. 10071 10072 For additional information about GCC please refer to the [14]GCC 10073 project web site or contact the [15]GCC development mailing list. 10074 10075 To obtain GCC please use [16]our mirror sites or [17]our version 10076 control system. 10077 10078 10079 For questions related to the use of GCC, please consult these web 10080 pages and the [18]GCC manuals. If that fails, the 10081 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10082 web pages and the development of GCC are welcome on our developer 10083 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 10084 archives. 10085 10086 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 10087 distribution of this entire article is permitted in any medium, 10088 provided this notice is preserved. 10089 10090 These pages are [23]maintained by the GCC team. Last modified 10091 2022-10-26. 10092 10093 References 10094 10095 1. http://www.gnu.org/ 10096 2. https://gcc.gnu.org/gcc-4.7/changes.html 10097 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 10098 4. https://gcc.gnu.org/gcc-4.7/changes.html 10099 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 10100 6. https://gcc.gnu.org/gcc-4.7/changes.html 10101 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 10102 8. https://gcc.gnu.org/gcc-4.7/changes.html 10103 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 10104 10. https://gcc.gnu.org/gcc-4.7/changes.html 10105 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 10106 12. https://gcc.gnu.org/gcc-4.7/buildstat.html 10107 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10108 14. https://gcc.gnu.org/index.html 10109 15. mailto:gcc (a] gcc.gnu.org 10110 16. https://gcc.gnu.org/mirrors.html 10111 17. https://gcc.gnu.org/git.html 10112 18. https://gcc.gnu.org/onlinedocs/ 10113 19. mailto:gcc-help (a] gcc.gnu.org 10114 20. mailto:gcc (a] gcc.gnu.org 10115 21. https://gcc.gnu.org/lists.html 10116 22. https://www.fsf.org/ 10117 23. https://gcc.gnu.org/about.html 10118 ====================================================================== 10119 http://gcc.gnu.org/gcc-4.7/changes.html 10120 10121 GCC 4.7 Release Series 10122 Changes, New Features, and Fixes 10123 10124 Caveats 10125 10126 * The -fconserve-space flag has been deprecated. The flag had no 10127 effect for most targets: only targets without a global .bss section 10128 and without support for switchable sections. Furthermore, the flag 10129 only had an effect for G++, where it could result in wrong 10130 semantics (please refer to the GCC manual for further details). The 10131 flag will be removed in GCC 4.8 10132 * Support for a number of older systems and recently unmaintained or 10133 untested target ports of GCC has been declared obsolete in GCC 4.7. 10134 Unless there is activity to revive them, the next release of GCC 10135 will have their sources permanently removed. 10136 All GCC ports for the following processor architectures have been 10137 declared obsolete: 10138 + picoChip (picochip-*) 10139 The following ports for individual systems on particular 10140 architectures have been obsoleted: 10141 + IRIX 6.5 (mips-sgi-irix6.5) 10142 + MIPS OpenBSD (mips*-*-openbsd*) 10143 + Solaris 8 (*-*-solaris2.8). Details can be found in the 10144 [1]announcement. 10145 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 10146 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 10147 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 10148 default, which for some sources generates code that accesses memory 10149 on unaligned addresses. This requires the kernel of those systems 10150 to enable such accesses (controlled by CP15 register c1, refer to 10151 ARM documentation). Alternatively, or for compatibility with 10152 kernels where unaligned accesses are not supported, all code has to 10153 be compiled with -mno-unaligned-access. Upstream Linux kernel 10154 releases have automatically and unconditionally supported unaligned 10155 accesses as emitted by GCC due to this option being active since 10156 version 2.6.28. 10157 * Support on ARM for the legacy floating-point accelerator (FPA) and 10158 the mixed-endian floating-point format that it used has been 10159 obsoleted. The ports that still use this format have been obsoleted 10160 as well. Many legacy ARM ports already provide an alternative that 10161 uses the VFP floating-point format. The obsolete ports will be 10162 deleted in the next release. 10163 The obsolete ports with alternatives are: 10164 + arm*-*-rtems (use arm*-*-rtemseabi) 10165 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 10166 + arm*-*-elf (use arm*-*-eabi) 10167 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 10168 Note, however, that these alternatives are not binary compatible 10169 with their legacy counterparts (although some can support running 10170 legacy applications). 10171 The obsolete ports that currently lack a modern alternative are: 10172 + arm*-*-ecos-elf 10173 + arm*-*-freebsd 10174 + arm*-wince-pe* 10175 New ports that support more recent versions of the architecture are 10176 welcome. 10177 * Support for the Maverick co-processor on ARM has been obsoleted. 10178 Code to support it will be deleted in the next release. 10179 * Support has been removed for Unix International threads on Solaris 10180 2, so the --enable-threads=solaris configure option and the 10181 -threads compiler option don't work any longer. 10182 * Support has been removed for the Solaris BSD Compatibility Package, 10183 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 10184 from Solaris 11, and was only intended as a migration aid from 10185 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 10186 recognized any longer. 10187 * The ARM port's -mwords-little-endian option has been deprecated. It 10188 will be removed in a future release. 10189 * Support has been removed for the NetWare x86 configuration 10190 obsoleted in GCC 4.6. 10191 * It is no longer possible to use the "l" constraint in MIPS16 asm 10192 statements. 10193 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 10194 library which affected the ABI in C++11 mode: a data member was 10195 added to std::list changing its size and altering the definitions 10196 of some member functions, and std::pair's move constructor was 10197 non-trivial which altered the calling convention for functions with 10198 std::pair arguments or return types. The ABI incompatibilities have 10199 been fixed for GCC version 4.7.2 but as a result C++11 code 10200 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 10201 code compiled with different GCC versions and with C++98/C++03 code 10202 compiled with any version. 10203 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 10204 rules for the layout of vectors that could lead to wrong code being 10205 generated. Vectors larger than 8 bytes in size are now by default 10206 aligned to an 8-byte boundary. This is an ABI change: code that 10207 makes explicit use of vector types may be incompatible with binary 10208 objects built with older versions of GCC. Auto-vectorized code is 10209 not affected by this change. (This change affects GCC versions 10210 4.7.2 and later.) 10211 * More information on porting to GCC 4.7 from previous versions of 10212 GCC can be found in the [2]porting guide for this release. 10213 10214 General Optimizer Improvements 10215 10216 * Support for a new parameter --param case-values-threshold=n was 10217 added to allow users to control the cutoff between doing switch 10218 statements as a series of if statements and using a jump table. 10219 * Link-time optimization (LTO) improvements: 10220 + Improved scalability and reduced memory usage. Link time 10221 optimization of Firefox now requires 3GB of RAM on a 64-bit 10222 system, while over 8GB was needed previously. Linking time has 10223 been improved, too. The serial stage of linking Firefox has 10224 been sped up by about a factor of 10. 10225 + Reduced size of object files and temporary storage used during 10226 linking. 10227 + Streaming performance (both outbound and inbound) has been 10228 improved. 10229 + ld -r is now supported with LTO. 10230 + Several bug fixes, especially in symbol table handling and 10231 merging. 10232 * Interprocedural optimization improvements: 10233 + Heuristics now take into account that after inlining code will 10234 be optimized out because of known values (or properties) of 10235 function parameters. For example: 10236 void foo(int a) 10237 { 10238 if (a > 10) 10239 ... huge code ... 10240 } 10241 void bar (void) 10242 { 10243 foo (0); 10244 } 10245 10246 The call of foo will be inlined into bar even when optimizing 10247 for code size. Constructs based on __builtin_constant_p are 10248 now understood by the inliner and code size estimates are 10249 evaluated a lot more realistically. 10250 + The representation of C++ virtual thunks and aliases (both 10251 implicit and defined via the alias attribute) has been 10252 re-engineered. Aliases no longer pose optimization barriers 10253 and calls to an alias can be inlined and otherwise optimized. 10254 + The inter-procedural constant propagation pass has been 10255 rewritten. It now performs generic function specialization. 10256 For example when compiling the following: 10257 void foo(bool flag) 10258 { 10259 if (flag) 10260 ... do something ... 10261 else 10262 ... do something else ... 10263 } 10264 void bar (void) 10265 { 10266 foo (false); 10267 foo (true); 10268 foo (false); 10269 foo (true); 10270 foo (false); 10271 foo (true); 10272 } 10273 10274 GCC will now produce two copies of foo. One with flag being 10275 true, while other with flag being false. This leads to 10276 performance improvements previously possible only by inlining 10277 all calls. Cloning causes a lot less code size growth. 10278 * A string length optimization pass has been added. It attempts to 10279 track string lengths and optimize various standard C string 10280 functions like strlen, strchr, strcpy, strcat, stpcpy and their 10281 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 10282 enabled by default at -O2 or above, unless optimizing for size, and 10283 can be disabled by the -fno-optimize-strlen option. The pass can 10284 e.g. optimize 10285 char *bar (const char *a) 10286 { 10287 size_t l = strlen (a) + 2; 10288 char *p = malloc (l); if (p == NULL) return p; 10289 strcpy (p, a); strcat (p, "/"); return p; 10290 } 10291 10292 into: 10293 char *bar (const char *a) 10294 { 10295 size_t tmp = strlen (a); 10296 char *p = malloc (tmp + 2); if (p == NULL) return p; 10297 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 10298 } 10299 10300 or for hosted compilations where stpcpy is available in the runtime 10301 and headers provide its prototype, e.g. 10302 void foo (char *a, const char *b, const char *c, const char *d) 10303 { 10304 strcpy (a, b); strcat (a, c); strcat (a, d); 10305 } 10306 10307 can be optimized into: 10308 void foo (char *a, const char *b, const char *c, const char *d) 10309 { 10310 strcpy (stpcpy (stpcpy (a, b), c), d); 10311 } 10312 10313 New Languages and Language specific improvements 10314 10315 * Version 3.1 of the OpenMP specification is now supported for the C, 10316 C++, and Fortran compilers. 10317 10318 Ada 10319 10320 * The command-line option -feliminate-unused-debug-types has been 10321 re-enabled by default, as it is for the other languages, leading to 10322 a reduction in debug info size of 12.5% and more for relevant 10323 cases, as well as to a small compilation speedup. 10324 10325 C family 10326 10327 * A new built-in, __builtin_assume_aligned, has been added, through 10328 which the compiler can be hinted about pointer alignment and can 10329 use it to improve generated code. 10330 * A new warning option -Wunused-local-typedefs was added for C, C++, 10331 Objective-C and Objective-C++. This warning diagnoses typedefs 10332 locally defined in a function, and otherwise not used. 10333 * A new experimental command-line option -ftrack-macro-expansion was 10334 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 10335 the compiler to emit diagnostic about the current macro expansion 10336 stack when a compilation error occurs in a macro expansion. 10337 * Experimental support for transactional memory has been added. It 10338 includes support in the compiler, as well as a supporting runtime 10339 library called libitm. To compile code with transactional memory 10340 constructs, use the -fgnu-tm option. 10341 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 10342 and 32-bit/64-bit x86 platforms. 10343 For more details on transactional memory see [3]the GCC WiKi. 10344 * Support for atomic operations specifying the C++11/C11 memory model 10345 has been added. These new __atomic routines replace the existing 10346 __sync built-in routines. 10347 Atomic support is also available for memory blocks. Lock-free 10348 instructions will be used if a memory block is the same size and 10349 alignment as a supported integer type. Atomic operations which do 10350 not have lock-free support are left as function calls. A set of 10351 library functions is available on the GCC atomic wiki in the 10352 "External Atomics Library" section. 10353 For more details on the memory models and features, see the 10354 [4]atomic wiki. 10355 * When a binary operation is performed on vector types and one of the 10356 operands is a uniform vector, it is possible to replace the vector 10357 with the generating element. For example: 10358 typedef int v4si __attribute__ ((vector_size (16))); 10359 v4si res, a = {1,2,3,4}; 10360 int x; 10361 10362 res = 2 + a; /* means {2,2,2,2} + a */ 10363 res = a - x; /* means a - {x,x,x,x} */ 10364 10365 [5]C 10366 10367 * There is support for some more features from the C11 revision of 10368 the ISO C standard. GCC now accepts the options -std=c11 and 10369 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 10370 + Unicode strings (previously supported only with options such 10371 as -std=gnu11, now supported with -std=c11), and the 10372 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 10373 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 10374 + Alignment support (_Alignas, _Alignof, max_align_t, 10375 <stdalign.h>). 10376 + A built-in function __builtin_complex is provided to support C 10377 library implementation of the CMPLX family of macros. 10378 10379 [6]C++ 10380 10381 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 10382 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 10383 -Wc++0x-compat, respectively. 10384 * G++ now implements [7]C++11 extended friend syntax: 10385 10386 template<class W> 10387 class Q 10388 { 10389 static const int I = 2; 10390 public: 10391 friend W; 10392 }; 10393 10394 struct B 10395 { 10396 int ar[Q<B>::I]; 10397 }; 10398 10399 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 10400 override control. 10401 10402 struct B { 10403 virtual void f() const final; 10404 virtual void f(int); 10405 }; 10406 10407 struct D : B { 10408 void f() const; // error: D::f attempts to override final B::f 10409 void f(long) override; // error: doesn't override anything 10410 void f(int) override; // ok 10411 }; 10412 10413 struct E final { }; 10414 struct F: E { }; // error: deriving from final class 10415 10416 * G++ now implements [9]C++11 non-static data member initializers. 10417 10418 struct A { 10419 int i = 42; 10420 } a; // initializes a.i to 42 10421 10422 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 10423 user-defined literals. 10424 10425 // Not actually a good approximation. :) 10426 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 10427 long double pi = 180.0_degrees; 10428 10429 * G++ now implements [11]C++11 alias-declarations. 10430 10431 template <class T> using Ptr = T*; 10432 Ptr<int> ip; // decltype(ip) is int* 10433 10434 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 10435 [12]C++11 delegating constructors. 10436 10437 struct A { 10438 A(int); 10439 A(): A(42) { } // delegate to the A(int) constructor 10440 }; 10441 10442 * G++ now fully implements C++11 atomic classes rather than just 10443 integer derived classes. 10444 10445 class POD { 10446 int a; 10447 int b; 10448 }; 10449 std::atomic<POD> my_atomic_POD; 10450 10451 * G++ now sets the predefined macro __cplusplus to the correct value, 10452 199711L for C++98/03, and 201103L for C++11. 10453 * G++ now correctly implements the two-phase lookup rules such that 10454 an unqualified name used in a template must have an appropriate 10455 declaration found either in scope at the point of definition of the 10456 template or by argument-dependent lookup at the point of 10457 instantiation. As a result, code that relies on a second 10458 unqualified lookup at the point of instantiation to find functions 10459 declared after the template or in dependent bases will be rejected. 10460 The compiler will suggest ways to fix affected code, and using the 10461 -fpermissive compiler flag will allow the code to compile with a 10462 warning. 10463 10464 template <class T> 10465 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 10466 void g(int) { } // fix by moving this declaration before the declaration of f 10467 10468 template <class T> 10469 struct A: T { 10470 // error, B::g(B) not found by argument-dependent lookup 10471 void f() { g(T()); } // fix by using this->g or A::g 10472 }; 10473 10474 struct B { void g(B); }; 10475 10476 int main() 10477 { 10478 f<int>(); 10479 A<B>().f(); 10480 } 10481 10482 * G++ now properly re-uses stack space allocated for temporary 10483 objects when their lifetime ends, which can significantly lower 10484 stack consumption for some C++ functions. As a result of this, some 10485 code with undefined behavior will now break: 10486 10487 const int &f(const int &i) { return i; } 10488 .... 10489 const int &x = f(1); 10490 const int &y = f(2); 10491 10492 Here, x refers to the temporary allocated to hold the 1 argument, 10493 which only lives until the end of the initialization; it 10494 immediately becomes a dangling reference. So the next statement 10495 re-uses the stack slot to hold the 2 argument, and users of x get 10496 that value instead. 10497 Note that this should not cause any change of behavior for 10498 temporaries of types with non-trivial destructors, as they are 10499 already destroyed at end of full-expression; the change is that now 10500 the storage is released as well. 10501 * A new command-line option -Wdelete-non-virtual-dtor has been added 10502 to warn when delete is used to destroy an instance of a class which 10503 has virtual functions and non-virtual destructor. It is unsafe to 10504 delete an instance of a derived class through a pointer to a base 10505 class if the base class does not have a virtual destructor. This 10506 warning is enabled by -Wall. 10507 * A new command-line option -Wzero-as-null-pointer-constant has been 10508 added to warn when a literal '0' is used as null pointer constant. 10509 It can be useful to facilitate the conversion to nullptr in C++11. 10510 * As per C++98, access-declarations are now deprecated by G++. 10511 Using-declarations are to be used instead. Furthermore, some 10512 efforts have been made to improve the support of class scope 10513 using-declarations. In particular, using-declarations referring to 10514 a dependent type now work as expected ([13]bug c++/14258). 10515 * The ELF symbol visibility of a template instantiation is now 10516 properly constrained by the visibility of its template arguments 10517 ([14]bug c++/35688). 10518 10519 Runtime Library (libstdc++) 10520 10521 * [15]Improved experimental support for the new ISO C++ standard, 10522 C++11, including: 10523 + using noexcept in most of the library; 10524 + implementations of pointer_traits, allocator_traits and 10525 scoped_allocator_adaptor; 10526 + uses-allocator construction for tuple; 10527 + vector meets the allocator-aware container requirements; 10528 + replacing monotonic_clock with steady_clock; 10529 + enabling the thread support library on most POSIX targets; 10530 + many small improvements to conform to the FDIS. 10531 * Added --enable-clocale=newlib configure option. 10532 * Debug Mode iterators for unordered associative containers. 10533 * Avoid polluting the global namespace and do not include <unistd.h>. 10534 10535 [16]Fortran 10536 10537 * The compile flag [17]-fstack-arrays has been added, which causes 10538 all local arrays to be put on stack memory. For some programs this 10539 will improve the performance significantly. If your program uses 10540 very large local arrays, it is possible that you will have to 10541 extend your runtime limits for stack memory. 10542 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 10543 [20]-fstack-arrays. 10544 * Front-end optimizations can now be selected by the 10545 [21]-ffrontend-optimize option and deselected by the 10546 -fno-frontend-optimize option. 10547 * When front-end optimization removes a function call, 10548 [22]-Wfunction-elimination warns about that. 10549 * When performing front-end-optimization, the 10550 [23]-faggressive-function-elimination option allows the removal of 10551 duplicate function calls even for impure functions. 10552 * The flag [24]-Wreal-q-constant has been added, which warns if 10553 floating-point literals have been specified using q (such as 10554 1.0q0); the q marker is now supported as a vendor extension to 10555 denote quad precision (REAL(16) or, if not available, REAL(10)). 10556 Consider using a kind parameter (such as in 1.0_qp) instead, which 10557 can be obtained via [25]SELECTED_REAL_KIND. 10558 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 10559 Fortran now always prints error messages to standard error. If you 10560 wish to redirect standard error, please consult the manual for your 10561 OS, shell, batch environment etc. as appropriate. 10562 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 10563 variable have been removed. When encountering a serious error, 10564 gfortran will now always abort the program. Whether a core dump is 10565 generated depends on the user environment settings; see the ulimit 10566 -c setting for POSIX shells, limit coredumpsize for C shells, and 10567 the [26]WER user-mode dumps settings on Windows. 10568 * The [27]-fbacktrace option is now enabled by default. When 10569 encountering a fatal error, gfortran will attempt to print a 10570 backtrace to standard error before aborting. It can be disabled 10571 with -fno-backtrace. Note: On POSIX targets with the addr2line 10572 utility from GNU binutils, GNU Fortran can print a backtrace with 10573 function name, file name, line number information in addition to 10574 the addresses; otherwise only the addresses are printed. 10575 * [28]Fortran 2003: 10576 + Generic interface names which have the same name as derived 10577 types are now supported, which allows to write constructor 10578 functions. Note that Fortran does not support static 10579 constructor functions; only default initialization or an 10580 explicit structure-constructor initialization are available. 10581 + [29]Polymorphic (class) arrays are now supported. 10582 * [30]Fortran 2008: 10583 + Support for the DO CONCURRENT construct has been added, which 10584 allows the user to specify that individual loop iterations 10585 have no interdependencies. 10586 + [31]Coarrays: Full single-image support except for polymorphic 10587 coarrays. Additionally, preliminary support for multiple 10588 images via an MPI-based [32]coarray communication library has 10589 been added. Note: The library version is not yet usable as 10590 remote coarray access is not yet possible. 10591 * [33]TS 29113: 10592 + New flag [34]-std=f2008ts permits programs that are expected 10593 to conform to the Fortran 2008 standard and the draft 10594 Technical Specification (TS) 29113 on Further Interoperability 10595 of Fortran with C. 10596 + The OPTIONAL attribute is now allowed for dummy arguments of 10597 BIND(C) procedures. 10598 + The RANK intrinsic has been added. 10599 + The implementation of the ASYNCHRONOUS attribute in GCC is 10600 compatible with the candidate draft of TS 29113 (since GCC 10601 4.6). 10602 10603 [35]Go 10604 10605 * GCC 4.7 implements the [36]Go 1 language standard. The library 10606 support in 4.7.0 is not quite complete, due to release timing. 10607 Release 4.7.1 includes complete support for Go 1. The Go library is 10608 from the Go 1.0.1 release. 10609 * Go has been tested on GNU/Linux and Solaris platforms. It may work 10610 on other platforms as well. 10611 10612 [37]New Targets and Target Specific Improvements 10613 10614 [38]ARM 10615 10616 * GCC now supports the Cortex-A7 processor implementing the v7-a 10617 version of the architecture using the option -mcpu=cortex-a7. 10618 * The default vector size in auto-vectorization for NEON is now 128 10619 bits. If vectorization fails thusly, the vectorizer tries again 10620 with 64-bit vectors. 10621 * A new option -mvectorize-with-neon-double was added to allow users 10622 to change the vector size to 64 bits. 10623 10624 [39]AVR 10625 10626 * The AVR port's libgcc has been improved and its multilib structure 10627 has been enhanced. As a result, all objects contributing to an 10628 application must either be compiled with GCC versions up to 4.6.x 10629 or with GCC versions 4.7.1 or later. If the compiler is used with 10630 AVR-LibC, you need a version that supports the new layout, i.e. 10631 implements [40]#35407. 10632 * The -mshort-calls command-line option has been deprecated. It will 10633 be removed in the GCC 4.8 release. See -mrelax for a replacement. 10634 * The compiled code only references startup code that clears .bss and 10635 the common section resp. initializes the .data and .rodata section 10636 provided respective sections (or subsections thereof) are not 10637 empty, see [41]PR18145. Applications that put all static storage 10638 objects into non-standard sections and / or define all static 10639 storage objects in assembler modules, must reference __do_clear_bss 10640 resp. __do_copy_data by hand or undefine the symbol(s) by means of 10641 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 10642 * GCC now supports the XMEGA architecture. This requires GNU Binutils 10643 2.22 or later. 10644 * Support for the [42]named address spaces __flash, __flash1, ..., 10645 __flash5 and __memx has been added. These address spaces locate 10646 read-only data in flash memory and allow reading from flash memory 10647 by means of ordinary C code, i.e. without the need of (inline) 10648 assembler code: 10649 10650 const __flash int values[] = { 42, 31 }; 10651 10652 int add_values (const __flash int *p, int i) 10653 { 10654 return values[i] + *p; 10655 } 10656 10657 * Support has been added for the AVR-specific configure option 10658 --with-avrlibc=yes in order to arrange for better integration of 10659 [43]AVR-Libc. This configure option is supported in avr-gcc v4.7.2 10660 and newer and will only take effect in non-RTEMS configurations. If 10661 avr-gcc is configured for RTEMS, the option will be ignored which 10662 is the same as specifying --with-avrlibc=no. See [44]PR54461 for 10663 more technical details. 10664 * Support for AVR-specific [45]built-in functions has been added. 10665 * Support has been added for the signed and unsigned 24-bit scalar 10666 integer types __int24 and __uint24. 10667 * New command-line options -maccumulate-args, -mbranch-cost=cost and 10668 -mstrict-X were added to allow better fine-tuning of code 10669 optimization. 10670 * The command-line option -fdata-sections now also takes affect on 10671 the section names of variables with the progmem attribute. 10672 * A new inline assembler print modifier %i to print a RAM address as 10673 I/O address has been added: 10674 10675 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 10676 10677 void set_portb (uint8_t value) 10678 { 10679 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 10680 } 10681 10682 The offset between an I/O address and the RAM address for that I/O 10683 location is device-specific. This offset is taken into account when 10684 printing a RAM address with the %i modifier so that the address is 10685 suitable to be used as operand in an I/O command. The address must 10686 be a constant integer known at compile time. 10687 * The inline assembler constraint "R" to represent integers in the 10688 range -6 ... 5 has been removed without replacement. 10689 * Many optimizations to: 10690 + 64-bit integer arithmetic 10691 + Widening multiplication 10692 + Integer division by a constant 10693 + Avoid constant reloading in multi-byte instructions. 10694 + Micro-optimizations for special instruction sequences. 10695 + Generic built-in functions like __builtin_ffs*, 10696 __builtin_clz*, etc. 10697 + If-else decision trees generated by switch instructions 10698 + Merging of data located in flash memory 10699 + New libgcc variants for devices with 8-bit wide stack pointer 10700 + ... 10701 * Better documentation: 10702 + Handling of EIND and indirect jumps on devices with more than 10703 128 KiB of program memory. 10704 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 10705 registers. 10706 + Function attributes OS_main and OS_task. 10707 + AVR-specific built-in macros. 10708 10709 C6X 10710 10711 * Support has been added for the Texas Instruments C6X family of 10712 processors. 10713 10714 CR16 10715 10716 * Support has been added for National Semiconductor's CR16 10717 architecture. 10718 10719 Epiphany 10720 10721 * Support has been added for Adapteva's Epiphany architecture. 10722 10723 IA-32/x86-64 10724 10725 * Support for Intel AVX2 intrinsics, built-in functions and code 10726 generation is available via -mavx2. 10727 * Support for Intel BMI2 intrinsics, built-in functions and code 10728 generation is available via -mbmi2. 10729 * Implementation and automatic generation of __builtin_clz* using the 10730 lzcnt instruction is available via -mlzcnt. 10731 * Support for Intel FMA3 intrinsics and code generation is available 10732 via -mfma. 10733 * A new -mfsgsbase command-line option is available that makes GCC 10734 generate new segment register read/write instructions through 10735 dedicated built-ins. 10736 * Support for the new Intel rdrnd instruction is available via 10737 -mrdrnd. 10738 * Two additional AVX vector conversion instructions are available via 10739 -mf16c. 10740 * Support for new Intel processor codename IvyBridge with RDRND, 10741 FSGSBASE and F16C is available through -march=core-avx-i. 10742 * Support for the new Intel processor codename Haswell with AVX2, 10743 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 10744 * Support for new AMD family 15h processors (Piledriver core) is now 10745 available through -march=bdver2 and -mtune=bdver2 options. 10746 * Support for [46]the x32 psABI is now available through the -mx32 10747 option. 10748 * Windows mingw targets are using the -mms-bitfields option by 10749 default. 10750 * Windows x86 targets are using the __thiscall calling convention for 10751 C++ class-member functions. 10752 * Support for the configure option --with-threads=posix for Windows 10753 mingw targets. 10754 10755 [47]MIPS 10756 10757 * GCC now supports thread-local storage (TLS) for MIPS16. This 10758 requires GNU binutils 2.22 or later. 10759 * GCC can now generate code specifically for the Cavium Octeon+ and 10760 Octeon2 processors. The associated command-line options are 10761 -march=octeon+ and -march=octeon2 respectively. Both options 10762 require GNU binutils 2.22 or later. 10763 * GCC can now work around certain 24k errata, under the control of 10764 the command-line option -mfix-24k. These workarounds require GNU 10765 binutils 2.20 or later. 10766 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 10767 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 10768 toolchain that generates 32-bit code by default. Use the 10769 configure-time option --enable-targets=all to select these extra 10770 multilibs. 10771 * Passing -fno-delayed-branch now also stops the assembler from 10772 automatically filling delay slots. 10773 10774 PowerPC/PowerPC64 10775 10776 * Vectors of type vector long long or vector long are passed and 10777 returned using the same method as other vectors with the VSX 10778 instruction set. Previously GCC did not adhere to the ABI for 10779 128-bit vectors with 64-bit integer base types (PR 48857). This 10780 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 10781 * A new option -mno-pointers-to-nested-functions was added to allow 10782 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 10783 that the compiler should not load up the chain register (r11) 10784 before calling a function through a pointer. If you use this 10785 option, you cannot call nested functions through a pointer, or call 10786 other languages that might use the static chain. 10787 * A new option msave-toc-indirect was added to allow AIX 10788 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 10789 save the TOC in the prologue for indirect calls or generate the 10790 save inline. This can speed up some programs that call through a 10791 function pointer a lot, but it can slow down other functions that 10792 only call through a function pointer in exceptional cases. 10793 * The PowerPC port will now enable machine-specific built-in 10794 functions when the user switches the target machine using the 10795 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 10796 sequences. In addition, the target macros are updated. However, due 10797 to the way the -save-temps switch is implemented, you won't see the 10798 effect of these additional macros being defined in preprocessor 10799 output. 10800 10801 SH 10802 10803 * A new option -msoft-atomic has been added. When it is specified, 10804 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 10805 the new __atomic routines. 10806 * Since it is neither supported by GAS nor officially documented, 10807 code generation for little endian SH2A has been disabled. 10808 Specifying -ml with -m2a* will now result in a compiler error. 10809 * The defunct -mbranch-cost option has been fixed. 10810 * Some improvements to the generated code of: 10811 + Utilization of the tst #imm,R0 instruction. 10812 + Dynamic shift instructions on SH2A. 10813 + Integer absolute value calculations. 10814 * The -mdiv= option for targets other than SHmedia has been fixed and 10815 documented. 10816 10817 SPARC 10818 10819 * The option -mflat has been reinstated. When it is specified, the 10820 compiler will generate code for a single register window model. 10821 This is essentially a new implementation and the corresponding 10822 debugger support has been added to GDB 7.4. 10823 * Support for the options -mtune=native and -mcpu=native has been 10824 added on selected native platforms (GNU/Linux and Solaris). 10825 * Support for the SPARC T3 (Niagara 3) processor has been added. 10826 * VIS: 10827 + An intrinsics header visintrin.h has been added. 10828 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 10829 compare instructions have been added. 10830 + The little-endian version of alignaddr is now supported. 10831 + When possible, VIS builtins are marked const, which should 10832 increase the compiler's ability to optimize VIS operations. 10833 + The compiler now properly tracks the %gsr register and how it 10834 behaves as an input for various VIS instructions. 10835 + Akin to fzero, the compiler can now generate fone instructions 10836 in order to set all of the bits of a floating-point register 10837 to 1. 10838 + The documentation for the VIS intrinsics in the GCC manual has 10839 been brought up to date and many inaccuracies were fixed. 10840 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 10841 non-condition-code setting edge instructions have been added. 10842 Their availability is controlled by the new -mvis2 and 10843 -mno-vis2 options. They are enabled by default on 10844 UltraSPARC-III and later CPUs. 10845 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 10846 has been added. These instructions are enabled by default on SPARC 10847 T3 (Niagara 3) and later CPUs. 10848 10849 TILE-Gx/TILEPro 10850 10851 * Support has been added for the Tilera TILE-Gx and TILEPro families 10852 of processors. 10853 10854 Other significant improvements 10855 10856 * A new option (-grecord-gcc-switches) was added that appends 10857 compiler command-line options that might affect code generation to 10858 the DW_AT_producer attribute string in the DWARF debugging 10859 information. 10860 * GCC now supports various new GNU extensions to the DWARF debugging 10861 information format, like [48]entry value and [49]call site 10862 information, [50]typed DWARF stack or [51]a more compact macro 10863 representation. Support for these extensions has been added to GDB 10864 7.4. They can be disabled through the -gstrict-dwarf command-line 10865 option. 10866 10867 [52]GCC 4.7.1 10868 10869 This is the [53]list of problem reports (PRs) from GCC's bug tracking 10870 system that are known to be fixed in the 4.7.1 release. This list might 10871 not be complete (that is, it is possible that some PRs that have been 10872 fixed are not listed here). 10873 10874 The Go front end in the 4.7.1 release fully supports the [54]Go 1 10875 language standard. 10876 10877 [55]GCC 4.7.2 10878 10879 This is the [56]list of problem reports (PRs) from GCC's bug tracking 10880 system that are known to be fixed in the 4.7.2 release. This list might 10881 not be complete (that is, it is possible that some PRs that have been 10882 fixed are not listed here). 10883 10884 [57]GCC 4.7.3 10885 10886 This is the [58]list of problem reports (PRs) from GCC's bug tracking 10887 system that are known to be fixed in the 4.7.3 release. This list might 10888 not be complete (that is, it is possible that some PRs that have been 10889 fixed are not listed here). 10890 10891 [59]GCC 4.7.4 10892 10893 This is the [60]list of problem reports (PRs) from GCC's bug tracking 10894 system that are known to be fixed in the 4.7.4 release. This list might 10895 not be complete (that is, it is possible that some PRs that have been 10896 fixed are not listed here). 10897 10898 10899 For questions related to the use of GCC, please consult these web 10900 pages and the [61]GCC manuals. If that fails, the 10901 [62]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10902 web pages and the development of GCC are welcome on our developer 10903 list at [63]gcc (a] gcc.gnu.org. All of [64]our lists have public 10904 archives. 10905 10906 Copyright (C) [65]Free Software Foundation, Inc. Verbatim copying and 10907 distribution of this entire article is permitted in any medium, 10908 provided this notice is preserved. 10909 10910 These pages are [66]maintained by the GCC team. Last modified 10911 2025-06-10. 10912 10913 References 10914 10915 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 10916 2. https://gcc.gnu.org/gcc-4.7/porting_to.html 10917 3. https://gcc.gnu.org/wiki/TransactionalMemory 10918 4. https://gcc.gnu.org/wiki/Atomic/GCCMM 10919 5. https://gcc.gnu.org/gcc-4.7/changes.html#c 10920 6. https://gcc.gnu.org/gcc-4.7/changes.html#cxx 10921 7. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10922 8. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10923 9. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10924 10. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10925 11. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10926 12. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10927 13. https://gcc.gnu.org/PR14258 10928 14. https://gcc.gnu.org/PR35688 10929 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 10930 16. https://gcc.gnu.org/gcc-4.7/changes.html#fortran 10931 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 10932 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 10933 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 10934 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 10935 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 10936 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 10937 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 10938 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 10939 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 10940 26. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 10941 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 10942 28. https://gcc.gnu.org/wiki/Fortran2003Status 10943 29. https://gcc.gnu.org/wiki/OOP 10944 30. https://gcc.gnu.org/wiki/Fortran2008Status 10945 31. https://gcc.gnu.org/wiki/Coarray 10946 32. https://gcc.gnu.org/wiki/CoarrayLib 10947 33. https://gcc.gnu.org/wiki/TS29113Status 10948 34. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 10949 35. https://gcc.gnu.org/gcc-4.7/changes.html#go 10950 36. https://go.dev/doc/go1 10951 37. https://gcc.gnu.org/gcc-4.7/changes.html#targets 10952 38. https://gcc.gnu.org/gcc-4.7/changes.html#arm 10953 39. https://gcc.gnu.org/gcc-4.7/changes.html#avr 10954 40. http://savannah.nongnu.org/bugs/?35407 10955 41. https://gcc.gnu.org/PR18145 10956 42. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 10957 43. http://www.nongnu.org/avr-libc/ 10958 44. https://gcc.gnu.org/PR54461 10959 45. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 10960 46. https://sites.google.com/site/x32abi/ 10961 47. https://gcc.gnu.org/gcc-4.7/changes.html#mips 10962 48. https://dwarfstd.org/issues/100909.1.html 10963 49. https://dwarfstd.org/issues/100909.2.html 10964 50. https://dwarfstd.org/issues/140425.1.html 10965 51. https://dwarfstd.org/issues/110722.1.html 10966 52. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.1 10967 53. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 10968 54. https://go.dev/doc/go1 10969 55. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.2 10970 56. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 10971 57. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.3 10972 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 10973 59. https://gcc.gnu.org/gcc-4.7/changes.html#GCC4.7.4 10974 60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 10975 61. https://gcc.gnu.org/onlinedocs/ 10976 62. mailto:gcc-help (a] gcc.gnu.org 10977 63. mailto:gcc (a] gcc.gnu.org 10978 64. https://gcc.gnu.org/lists.html 10979 65. https://www.fsf.org/ 10980 66. https://gcc.gnu.org/about.html 10981 ====================================================================== 10982 http://gcc.gnu.org/gcc-4.6/index.html 10983 10984 GCC 4.6 Release Series 10985 10986 (This release series is no longer supported.) 10987 10988 April 12, 2013 10989 10990 The [1]GNU project and the GCC developers are pleased to announce the 10991 release of GCC 4.6.4. 10992 10993 This release is a bug-fix release, containing fixes for regressions in 10994 GCC 4.6.3 relative to previous releases of GCC. 10995 10996 Release History 10997 10998 GCC 4.6.4 10999 April 12, 2013 ([2]changes, [3]documentation) 11000 11001 GCC 4.6.3 11002 March 1, 2012 ([4]changes, [5]documentation) 11003 11004 GCC 4.6.2 11005 October 26, 2011 ([6]changes, [7]documentation) 11006 11007 GCC 4.6.1 11008 June 27, 2011 ([8]changes, [9]documentation) 11009 11010 GCC 4.6.0 11011 March 25, 2011 ([10]changes, [11]documentation) 11012 11013 References and Acknowledgements 11014 11015 GCC used to stand for the GNU C Compiler, but since the compiler 11016 supports several other languages aside from C, it now stands for the 11017 GNU Compiler Collection. 11018 11019 A list of [12]successful builds is updated as new information becomes 11020 available. 11021 11022 The GCC developers would like to thank the numerous people that have 11023 contributed new features, improvements, bug fixes, and other changes as 11024 well as test results to GCC. This [13]amazing group of volunteers is 11025 what makes GCC successful. 11026 11027 For additional information about GCC please refer to the [14]GCC 11028 project web site or contact the [15]GCC development mailing list. 11029 11030 To obtain GCC please use [16]our mirror sites or [17]our version 11031 control system. 11032 11033 11034 For questions related to the use of GCC, please consult these web 11035 pages and the [18]GCC manuals. If that fails, the 11036 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11037 web pages and the development of GCC are welcome on our developer 11038 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 11039 archives. 11040 11041 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 11042 distribution of this entire article is permitted in any medium, 11043 provided this notice is preserved. 11044 11045 These pages are [23]maintained by the GCC team. Last modified 11046 2022-10-26. 11047 11048 References 11049 11050 1. http://www.gnu.org/ 11051 2. https://gcc.gnu.org/gcc-4.6/changes.html 11052 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 11053 4. https://gcc.gnu.org/gcc-4.6/changes.html 11054 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 11055 6. https://gcc.gnu.org/gcc-4.6/changes.html 11056 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 11057 8. https://gcc.gnu.org/gcc-4.6/changes.html 11058 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 11059 10. https://gcc.gnu.org/gcc-4.6/changes.html 11060 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 11061 12. https://gcc.gnu.org/gcc-4.6/buildstat.html 11062 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11063 14. https://gcc.gnu.org/index.html 11064 15. mailto:gcc (a] gcc.gnu.org 11065 16. https://gcc.gnu.org/mirrors.html 11066 17. https://gcc.gnu.org/git.html 11067 18. https://gcc.gnu.org/onlinedocs/ 11068 19. mailto:gcc-help (a] gcc.gnu.org 11069 20. mailto:gcc (a] gcc.gnu.org 11070 21. https://gcc.gnu.org/lists.html 11071 22. https://www.fsf.org/ 11072 23. https://gcc.gnu.org/about.html 11073 ====================================================================== 11074 http://gcc.gnu.org/gcc-4.6/changes.html 11075 11076 GCC 4.6 Release Series 11077 Changes, New Features, and Fixes 11078 11079 Caveats 11080 11081 * The options -b <machine> and -V <version> have been removed because 11082 they were unreliable. Instead, users should directly run 11083 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 11084 run a different version of gcc. 11085 * GCC now has stricter checks for invalid command-line options. In 11086 particular, when gcc was called to link object files rather than 11087 compile source code, it would previously accept and ignore all 11088 options starting with --, including linker options such as 11089 --as-needed and --export-dynamic, although such options would 11090 result in errors if any source code was compiled. Such options, if 11091 unknown to the compiler, are now rejected in all cases; if the 11092 intent was to pass them to the linker, options such as 11093 -Wl,--as-needed should be used. 11094 * Versions of the GNU C library up to and including 2.11.1 included 11095 an [1]incorrect implementation of the cproj function. GCC optimizes 11096 its builtin cproj according to the behavior specified and allowed 11097 by the ISO C99 standard. If you want to avoid discrepancies between 11098 the C library and GCC's builtin transformations when using cproj in 11099 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 11100 and actually rely on the incorrect behavior of cproj, then you can 11101 disable GCC's transformations using -fno-builtin-cproj. 11102 * The C-only intermodule optimization framework (IMA, enabled by 11103 -combine) has been removed in favor of the new generic link-time 11104 optimization framework (LTO) introduced in [2]GCC 4.5.0. 11105 * GCC now ships with the LGPL-licensed libquadmath library, which 11106 provides quad-precision mathematical functions for targets with a 11107 __float128 datatype. __float128 is available for targets on 32-bit 11108 x86, x86-64 and Itanium architectures. The libquadmath library is 11109 automatically built on such targets when building the Fortran 11110 compiler. 11111 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 11112 warnings were added for C, C++, Objective-C and Objective-C++. 11113 These warnings diagnose variables respective parameters which are 11114 only set in the code and never otherwise used. Usually such 11115 variables are useless and often even the value assigned to them is 11116 computed needlessly, sometimes expensively. The 11117 -Wunused-but-set-variable warning is enabled by default by -Wall 11118 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 11119 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 11120 rules for the layout of vectors that could lead to wrong code being 11121 generated. Vectors larger than 8 bytes in size are now by default 11122 aligned to an 8-byte boundary. This is an ABI change: code that 11123 makes explicit use of vector types may be incompatible with binary 11124 objects built with older versions of GCC. Auto-vectorized code is 11125 not affected by this change. (This change affects GCC versions 11126 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 11127 * On AVR, variables with the progmem attribute to locate data in 11128 flash memory must be qualified as const. 11129 * Support for a number of older systems and recently unmaintained or 11130 untested target ports of GCC has been declared obsolete in GCC 4.6. 11131 Unless there is activity to revive them, the next release of GCC 11132 will have their sources permanently removed. 11133 All GCC ports for the following processor architectures have been 11134 declared obsolete: 11135 + Argonaut ARC (arc-*) 11136 + National Semiconductor CRX (crx-*) 11137 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 11138 m68hc12-*-*, m6812-*-*) 11139 + Sunplus S+core (score-*) 11140 The following ports for individual systems on particular 11141 architectures have been obsoleted: 11142 + Interix (i[34567]86-*-interix3*) 11143 + NetWare x86 (i[3456x]86-*-netware*) 11144 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 11145 + MCore PE (mcore-*-pe*) 11146 + SH SymbianOS (sh*-*-symbianelf*) 11147 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 11148 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 11149 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 11150 vax-*-netbsd*, but not *-*-netbsdelf*) 11151 The i[34567]86-*-pe alias for Cygwin targets has also been 11152 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 11153 Certain configure options to control the set of libraries built 11154 with GCC on some targets have been obsoleted. On ARM targets, the 11155 options --disable-fpu, --disable-26bit, --disable-underscore, 11156 --disable-interwork, --disable-biendian and --disable-nofmult have 11157 been obsoleted. On MIPS targets, the options 11158 --disable-single-float, --disable-biendian and --disable-softfloat 11159 have been obsoleted. 11160 * Support has been removed for all the [3]configurations obsoleted in 11161 GCC 4.5. 11162 * More information on porting to GCC 4.6 from previous versions of 11163 GCC can be found in the [4]porting guide for this release. 11164 11165 General Optimizer Improvements 11166 11167 * A new general optimization level, -Ofast, has been introduced. It 11168 combines the existing optimization level -O3 with options that can 11169 affect standards compliance but result in better optimized code. 11170 For example, -Ofast enables -ffast-math. 11171 * Link-time optimization improvements: 11172 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 11173 stabilized to the point of being usable. It has become the 11174 default mode when using the LTO optimization model. Link time 11175 optimization can now split itself into multiple parallel 11176 compilations. Parallelism is controlled with -flto=n (where n 11177 specifies the number of compilations to execute in parallel). 11178 GCC can also cooperate with a GNU make job server by 11179 specifying the -flto=jobserver option and adding + to the 11180 beginning of the Makefile rule executing the linker. 11181 Classical LTO mode can be enforced by -flto-partition=none. 11182 This may result in small code quality improvements. 11183 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 11184 and other large applications can be built with LTO enabled. 11185 + The linker plugin support improvements 11186 o Linker plugin is now enabled by default when the linker 11187 is detected to have plugin support. This is the case for 11188 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 11189 the Gold linker on ELF targets. Plugin support of the 11190 Apple linker on Darwin is not compatible with GCC. The 11191 linker plugin can also be controlled by the 11192 -fuse-linker-plugin command-line option. 11193 o Resolution information from the linker plugin is used to 11194 drive whole program assumptions. Use of the linker plugin 11195 results in more aggressive optimization on binaries and 11196 on shared libraries that use the hidden visibility 11197 attribute. Consequently the use of -fwhole-program is not 11198 necessary in addition to LTO. 11199 + Hidden symbols used from non-LTO objects now have to be 11200 explicitly annotated with externally_visible when the linker 11201 plugin is not used. 11202 + C++ inline functions and virtual tables are now privatized 11203 more aggressively, leading to better inter-procedural 11204 optimization and faster dynamic linking. 11205 + Memory usage and intermediate language streaming performance 11206 have been improved. 11207 + Static constructors and destructors from individual units are 11208 inlined into a single function. This can significantly improve 11209 startup times of large C++ applications where static 11210 constructors are very common. For example, static constructors 11211 are used when including the iostream header. 11212 + Support for the Ada language has been added. 11213 * Interprocedural optimization improvements 11214 + The interprocedural framework was re-tuned for link time 11215 optimization. Several scalability issues were resolved. 11216 + Improved auto-detection of const and pure functions. Newly, 11217 noreturn functions are auto-detected. 11218 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 11219 available that informs users when adding attributes to headers 11220 might improve code generation. 11221 + A number of inlining heuristic improvements. In particular: 11222 o Partial inlining is now supported and enabled by default 11223 at -O2 and greater. The feature can be controlled via 11224 -fpartial-inlining. 11225 Partial inlining splits functions with short hot path to 11226 return. This allows more aggressive inlining of the hot 11227 path leading to better performance and often to code size 11228 reductions (because cold parts of functions are not 11229 duplicated). 11230 o Scalability for large compilation units was improved 11231 significantly. 11232 o Inlining of callbacks is now more aggressive. 11233 o Virtual methods are considered for inlining when the 11234 caller is inlined and devirtualization is then possible. 11235 o Inlining when optimizing for size (either in cold regions 11236 of a program or when compiling with -Os) was improved to 11237 better handle C++ programs with larger abstraction 11238 penalty, leading to smaller and faster code. 11239 + The IPA reference optimization pass detecting global variables 11240 used or modified by functions was strengthened and sped up. 11241 + Functions whose address was taken are now optimized out when 11242 all references to them are dead. 11243 + A new inter-procedural static profile estimation pass detects 11244 functions that are executed once or unlikely to be executed. 11245 Unlikely executed functions are optimized for size. Functions 11246 executed once are optimized for size except for the inner 11247 loops. 11248 + On most targets with named section support, functions used 11249 only at startup (static constructors and main), functions used 11250 only at exit and functions detected to be cold are placed into 11251 separate text segment subsections. This extends the 11252 -freorder-functions feature and is controlled by the same 11253 switch. The goal is to improve the startup time of large C++ 11254 programs. 11255 Proper function placement requires linker support. GNU ld 11256 2.21.51 on ELF targets was updated to place those functions 11257 together within the text section leading to better code 11258 locality and faster startup times of large C++ programs. The 11259 feature is also supported in the Apple linker. Support in the 11260 gold linker is planned. 11261 * A new switch -fstack-usage has been added. It makes the compiler 11262 output stack usage information for the program, on a per-function 11263 basis, in an auxiliary file. 11264 * A new switch -fcombine-stack-adjustments has been added. It can be 11265 used to enable or disable the compiler's stack-slot combining pass 11266 which before was enabled automatically at -O1 and above, but could 11267 not be controlled on its own. 11268 * A new switch -fstrict-volatile-bitfields has been added. Using it 11269 indicates that accesses to volatile bitfields should use a single 11270 access of the width of the field's type. This option can be useful 11271 for precisely defining and accessing memory-mapped peripheral 11272 registers from C or C++. 11273 11274 Compile time and memory usage improvements 11275 11276 * Datastructures used by the dataflow framework in GCC were 11277 reorganized for better memory usage and more cache locality. 11278 Compile time is improved especially on units with large functions 11279 (possibly resulting from a lot of inlining) not fitting into the 11280 processor cache. The compile time of the GCC C compiler binary with 11281 link-time optimization went down by over 10% (benchmarked on x86-64 11282 target). 11283 11284 New Languages and Language specific improvements 11285 11286 Ada 11287 11288 * Stack checking has been improved on selected architectures (Alpha, 11289 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 11290 overflows in all cases on these architectures. 11291 * Initial support for Ada 2012 has been added. 11292 11293 C family 11294 11295 * A new warning, enabled by -Wdouble-promotion, has been added that 11296 warns about cases where a value of type float is implicitly 11297 promoted to double. This is especially helpful for CPUs that handle 11298 the former in hardware, but emulate the latter in software. 11299 * A new function attribute leaf was introduced. This attribute allows 11300 better inter-procedural optimization across calls to functions that 11301 return to the current unit only via returning or exception 11302 handling. This is the case for most library functions that have no 11303 callbacks. 11304 * Support for a new data type __int128 for targets having wide enough 11305 machine-mode support. 11306 * The new function attribute callee_pop_aggregate allows to specify 11307 if the caller or callee is responsible for popping the aggregate 11308 return pointer value from the stack. 11309 * Support for selectively enabling and disabling warnings via #pragma 11310 GCC diagnostic has been added. For instance: 11311 #pragma GCC diagnostic error "-Wuninitialized" 11312 foo(a); /* error is given for this one */ 11313 #pragma GCC diagnostic push 11314 #pragma GCC diagnostic ignored "-Wuninitialized" 11315 foo(b); /* no diagnostic for this one */ 11316 #pragma GCC diagnostic pop 11317 foo(c); /* error is given for this one */ 11318 #pragma GCC diagnostic pop 11319 foo(d); /* depends on command-line options */ 11320 11321 * The -fmax-errors=N option is now supported. Using this option 11322 causes the compiler to exit after N errors have been issued. 11323 11324 [7]C 11325 11326 * There is now experimental support for some features from the 11327 upcoming C1X revision of the ISO C standard. This support may be 11328 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 11329 Note that this support is experimental and may change incompatibly 11330 in future releases for consistency with changes to the C1X standard 11331 draft. The following features are newly supported as described in 11332 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 11333 meeting); some other features were already supported with no 11334 compiler changes being needed, or have some support but not in full 11335 accord with N1539 (as amended). 11336 + Static assertions (_Static_assert keyword) 11337 + Typedef redefinition 11338 + New macros in <float.h> 11339 + Anonymous structures and unions 11340 * The new -fplan9-extensions option directs the compiler to support 11341 some extensions for anonymous struct fields which are implemented 11342 by the Plan 9 compiler. A pointer to a struct may be automatically 11343 converted to a pointer to an anonymous field when calling a 11344 function, in order to make the types match. An anonymous struct 11345 field whose type is a typedef name may be referred to using the 11346 typedef name. 11347 11348 [8]C++ 11349 11350 * Improved [9]experimental support for the upcoming C++0x ISO C++ 11351 standard, including support for constexpr (thanks to Gabriel Dos 11352 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 11353 noexcept, unrestricted unions, range-based for loops (thanks to 11354 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 11355 Rodrigo), implicitly deleted functions and implicit move 11356 constructors. 11357 * When an extern declaration within a function does not match a 11358 declaration in the enclosing context, G++ now properly declares the 11359 name within the namespace of the function rather than the namespace 11360 which was open just before the function definition ([10]c++/43145). 11361 * GCC now warns by default when casting integers to larger pointer 11362 types. These warnings can be disabled with the option 11363 -Wno-int-to-pointer-cast, which is now also available in C++. 11364 * G++ no longer optimizes using the assumption that a value of 11365 enumeration type will fall within the range specified by the 11366 standard, since that assumption is easily violated with a 11367 conversion from integer type ([11]c++/43680). The old behavior can 11368 be restored with -fstrict-enums. 11369 * The new -fnothrow-opt flag changes the semantics of a throw() 11370 exception specification to match the proposed semantics of the 11371 noexcept specification: just call terminate if an exception tries 11372 to propagate out of a function with such an exception 11373 specification. This dramatically reduces or eliminates the code 11374 size overhead from adding the exception specification. 11375 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 11376 a function that the compiler can tell doesn't throw if it would 11377 change the value of a noexcept expression. 11378 * The -Wshadow option now warns if a local variable or type 11379 declaration shadows another type in C++. Note that the compiler 11380 will not warn if a local variable shadows a struct/class/enum, but 11381 will warn if it shadows an explicit typedef. 11382 * When an identifier is not found in the current scope, G++ now 11383 offers suggestions about which identifier might have been intended. 11384 * G++ now issues clearer diagnostics for missing semicolons after 11385 class, struct, and union definitions. 11386 * G++ now issues clearer diagnostics for missing semicolons after 11387 class member declarations. 11388 * G++ now issues clearer diagnostics when a colon is used in a place 11389 where a double-colon was intended. 11390 * G++ no longer accepts mutable on reference members ([12]c++/33558). 11391 Use -fpermissive to allow the old, non-conforming behaviour. 11392 * A few mangling fixes have been made, to attribute const/volatile on 11393 function pointer types, decltype of a plain decl, and use of a 11394 function parameter in the declaration of another parameter. By 11395 default the compiler still uses the old mangling, but emits aliases 11396 with the new mangling on targets that support strong aliases. Users 11397 can switch over entirely to the new mangling with -fabi-version=5 11398 or -fabi-version=0. -Wabi will now warn about code that uses the 11399 old mangling. 11400 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 11401 type to be default initialized unless the type has a user-declared 11402 default constructor. In 4.6.2 G++ implements the proposed 11403 resolution of [13]DR 253, so default initialization is allowed if 11404 it initializes all subobjects. Code that fails to compile can be 11405 fixed by providing an initializer e.g. 11406 struct A { A(); }; 11407 struct B : A { int i; }; 11408 const B b = B(); 11409 Use -fpermissive to allow the old, non-conforming behaviour. 11410 11411 Runtime Library (libstdc++) 11412 11413 * [14]Improved experimental support for the upcoming ISO C++ 11414 standard, C++0x, including using constexpr and nullptr. 11415 * Performance improvements to the [15]Debug Mode, thanks to Franois 11416 Dumont. 11417 * Atomic operations used for reference-counting are annotated so that 11418 they can be understood by race detectors such as Helgrind, see 11419 [16]Data Race Hunting. 11420 * Most libstdc++ standard headers have been changed to no longer 11421 include the cstddef header as an implementation detail. Code that 11422 relied on that header being included as side-effect of including 11423 other standard headers will need to include cstddef explicitly. 11424 11425 [17]Fortran 11426 11427 * On systems supporting the libquadmath library, GNU Fortran now also 11428 supports a quad-precision, kind=16 floating-point data type 11429 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 11430 hardware, calculations might be one to two orders of magnitude 11431 slower than with the 4, 8 or 10 bytes floating-point data types. 11432 This change does not affect systems which support REAL(16) in 11433 hardware nor those which do not support libquadmath. 11434 * Much improved compile time for large array constructors. 11435 * In order to reduce execution time and memory consumption, use of 11436 temporary arrays in assignment expressions is avoided for many 11437 cases. The compiler now reverses loops in order to avoid generating 11438 a temporary array where possible. 11439 * Improved diagnostics, especially with -fwhole-file. 11440 * The -fwhole-file flag is now enabled by default. This improves code 11441 generation and diagnostics. It can be disabled using the deprecated 11442 -fno-whole-file flag. 11443 * Support the generation of Makefile dependencies via the [18]-M... 11444 flags of GCC; you may need to specify the -cpp option in addition. 11445 The dependencies take modules, Fortran's include, and CPP's 11446 #include into account. Note: Using -M for the module path is no 11447 longer supported, use -J instead. 11448 * The flag -Wconversion has been modified to only issue warnings 11449 where a conversion leads to information loss. This drastically 11450 reduces the number of warnings; -Wconversion is thus now enabled 11451 with -Wall. The flag -Wconversion-extra has been added and also 11452 warns about other conversions; -Wconversion-extra typically issues 11453 a huge number of warnings, most of which can be ignored. 11454 * A new command-line option -Wunused-dummy-argument warns about 11455 unused dummy arguments and is included in -Wall. Before, 11456 -Wunused-variable also warned about unused dummy arguments. 11457 * Fortran 2003 support has been extended: 11458 + Improved support for polymorphism between libraries and 11459 programs and for complicated inheritance patterns (cf. 11460 [19]object-oriented programming). 11461 + Experimental support of the ASSOCIATE construct. 11462 + In pointer assignments it is now possible to specify the lower 11463 bounds of the pointer and, for a rank-1 or a simply contiguous 11464 data-target, to remap the bounds. 11465 + Automatic (re)allocation: In intrinsic assignments to 11466 allocatable variables the left-hand side will be automatically 11467 allocated (if unallocated) or reallocated (if the shape or 11468 type parameter is different). To avoid the small performance 11469 penalty, you can use a(:) = ... instead of a = ... for arrays 11470 and character strings - or disable the feature using -std=f95 11471 or -fno-realloc-lhs. 11472 + Deferred type parameter: For scalar allocatable and pointer 11473 variables the character length can be deferred. 11474 + Namelist variables with allocatable and pointer attribute and 11475 nonconstant length type parameter are supported. 11476 * Fortran 2008 support has been extended: 11477 + Experimental [20]coarray support (for one image only, i.e. 11478 num_images() == 1); use the [21]-fcoarray=single flag to 11479 enable it. 11480 + The STOP and the new ERROR STOP statements now support all 11481 constant expressions. 11482 + Support for the CONTIGUOUS attribute. 11483 + Support for ALLOCATE with MOLD. 11484 + Support for the STORAGE_SIZE intrinsic inquiry function. 11485 + Support of the NORM2 and PARITY intrinsic functions. 11486 + The following bit intrinsics were added: POPCNT and POPPAR for 11487 counting the number of 1 bits and returning the parity; BGE, 11488 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 11489 for combined left and right shifts, MASKL and MASKR for simple 11490 left and right justified masks, MERGE_BITS for a bitwise merge 11491 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 11492 and the transformational bit intrinsics IALL, IANY and 11493 IPARITY. 11494 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 11495 + Support for the IMPURE attribute for procedures, which allows 11496 for ELEMENTAL procedures without the restrictions of PURE. 11497 + Null pointers (including NULL()) and not allocated variables 11498 can be used as actual argument to optional non-pointer, 11499 non-allocatable dummy arguments, denoting an absent argument. 11500 + Non-pointer variables with TARGET attribute can be used as 11501 actual argument to POINTER dummies with INTENT(IN) 11502 + Pointers including procedure pointers and those in a derived 11503 type (pointer components) can now be initialized by a target 11504 instead of only by NULL. 11505 + The EXIT statement (with construct-name) can now be used to 11506 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 11507 SELECT CASE and SELECT TYPE constructs. 11508 + Internal procedures can now be used as actual argument. 11509 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 11510 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 11511 have been added; these arrays contain the supported kind 11512 values for the respective types. 11513 + The module procedures C_SIZEOF of the intrinsic module 11514 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 11515 ISO_FORTRAN_ENV have been implemented. 11516 + Minor changes: obsolescence diagnostics for ENTRY was added 11517 for -std=f2008; a line may start with a semicolon; for 11518 internal and module procedures END can be used instead of END 11519 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 11520 a RADIX argument; intrinsic types are supported for 11521 TYPE(intrinsic-type-spec); multiple type-bound procedures can 11522 be declared in a single PROCEDURE statement; implied-shape 11523 arrays are supported for named constants (PARAMETER). The 11524 transformational, three argument versions of BESSEL_JN and 11525 BESSEL_YN were added - the elemental, two-argument version had 11526 been added in GCC 4.4; note that the transformational 11527 functions use a recurrence algorithm. 11528 11529 [22]Go 11530 11531 Support for the Go programming language has been added to GCC. It is 11532 not enabled by default when you build GCC; use the --enable-languages 11533 configure option to build it. The driver program for compiling Go code 11534 is gccgo. 11535 11536 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 11537 is in progress. It may or may not work on other platforms. 11538 11539 [23]Objective-C and Objective-C++ 11540 11541 * The -fobjc-exceptions flag is now required to enable Objective-C 11542 exception and synchronization syntax (introduced by the keywords 11543 @try, @catch, @finally and @synchronized). 11544 * A number of Objective-C 2.0 features and extensions are now 11545 supported by GCC. These features are enabled by default; you can 11546 disable them by using the new -fobjc-std=objc1 command-line option. 11547 * The Objective-C 2.0 dot-syntax is now supported. It is an 11548 alternative syntax for using getters and setters; object.count is 11549 automatically converted into [object count] or [object setCount: 11550 ...] depending on context; for example if (object.count > 0) is 11551 automatically compiled into the equivalent of if ([object count] > 11552 0) while object.count = 0; is automatically compiled into the 11553 equivalent ot [object setCount: 0];. The dot-syntax can be used 11554 with instance and class objects and with any setters or getters, no 11555 matter if they are part of a declared property or not. 11556 * Objective-C 2.0 declared properties are now supported. They are 11557 declared using the new @property keyword, and are most commonly 11558 used in conjunction with the new Objective-C 2.0 dot-syntax. The 11559 nonatomic, readonly, readwrite, assign, retain, copy, setter and 11560 getter attributes are all supported. Marking declared properties 11561 with __attribute__ ((deprecated)) is supported too. 11562 * The Objective-C 2.0 @synthesize and @dynamic keywords are 11563 supported. @synthesize causes the compiler to automatically 11564 synthesize a declared property, while @dynamic is used to disable 11565 all warnings for a declared property for which no implementation is 11566 provided at compile time. Synthesizing declared properties requires 11567 runtime support in most useful cases; to be able to use it with the 11568 GNU runtime, appropriate helper functions have been added to the 11569 GNU Objective-C runtime ABI, and are implemented by the GNU 11570 Objective-C runtime library shipped with GCC. 11571 * The Objective-C 2.0 fast enumeration syntax is supported in 11572 Objective-C. This is currently not yet available in Objective-C++. 11573 Fast enumeration requires support in the runtime, and such support 11574 has been added to the GNU Objective-C runtime library (shipped with 11575 GCC). 11576 * The Objective-C 2.0 @optional keyword is supported. It allows you 11577 to mark methods or properties in a protocol as optional as opposed 11578 to required. 11579 * The Objective-C 2.0 @package keyword is supported. It has currently 11580 the same effect as the @public keyword. 11581 * Objective-C 2.0 method attributes are supported. Currently the 11582 supported attributes are deprecated, sentinel, noreturn and format. 11583 * Objective-C 2.0 method argument attributes are supported. The most 11584 widely used attribute is unused, to mark an argument as unused in 11585 the implementation. 11586 * Objective-C 2.0 class and protocol attributes are supported. 11587 Currently the only supported attribute is deprecated. 11588 * Objective-C 2.0 class extensions are supported. A class extension 11589 has the same syntax as a category declaration with no category 11590 name, and the methods and properties declared in it are added 11591 directly to the main class. It is mostly used as an alternative to 11592 a category to add methods to a class without advertising them in 11593 the public headers, with the advantage that for class extensions 11594 the compiler checks that all the privately declared methods are 11595 actually implemented. 11596 * As a result of these enhancements, GCC can now be used to build 11597 Objective-C and Objective-C++ software that uses Foundation and 11598 other important system frameworks with the NeXT runtime on Darwin 9 11599 and Darwin 10 (OSX 10.5 and 10.6). 11600 * Many bugs in the compiler have been fixed in this release; in 11601 particular, LTO can now be used when compiling Objective-C and 11602 Objective-C++ and the parser is much more robust in dealing with 11603 invalid code. 11604 11605 Runtime Library (libobjc) 11606 11607 * The GNU Objective-C runtime library now defines the macro 11608 __GNU_LIBOBJC__ (with a value that is increased at every release 11609 where there is any change to the API) in objc/objc.h, making it 11610 easy to determine if the GNU Objective-C runtime library is being 11611 used, and if so, which version. Previous versions of the GNU 11612 Objective-C runtime library (and other Objective-C runtime 11613 libraries such as the Apple one) do not define this macro. 11614 * A new Objective-C 2.0 API, almost identical to the one implemented 11615 by the Apple Objective-C runtime, has been implemented in the GNU 11616 Objective-C runtime library. The new API hides the internals of 11617 most runtime structures but provides a more extensive set of 11618 functions to operate on them. It is much easier, for example, to 11619 create or modify classes at runtime. The new API also makes it 11620 easier to port software from Apple to GNU as almost no changes 11621 should be required. The old API is still supported for backwards 11622 compatibility; including the old objc/objc-api.h header file 11623 automatically selects the old API, while including the new 11624 objc/runtime.h header file automatically selects the new API. 11625 Support for the old API is being phased out and upgrading the 11626 software to use the new API is strongly recommended. To check for 11627 the availability of the new API, the __GNU_LIBOBJC__ macro can be 11628 used as older versions of the GNU Objective-C runtime library, 11629 which do not support the new API, do not define such a macro. 11630 * Runtime support for @synchronized has been added. 11631 * Runtime support for Objective-C 2.0 synthesized property accessors 11632 has been added. 11633 * Runtime support for Objective-C 2.0 fast enumeration has been 11634 added. 11635 11636 [24]New Targets and Target Specific Improvements 11637 11638 [25]ARM 11639 11640 * GCC now supports the Cortex-M4 processor implementing the v7-em 11641 version of the architecture using the option -mcpu=cortex-m4. 11642 * Scheduling descriptions for the Cortex-M4, the Neon and the 11643 floating point units of the Cortex-A9 and a pipeline description 11644 for the Cortex-A5 have been added. 11645 * Synchronization primitives such as __sync_fetch_and_add and friends 11646 are now inlined for supported architectures rather than calling 11647 into a kernel helper function. 11648 * SSA loop prefetching is enabled by default for the Cortex-A9 at 11649 -O3. 11650 * Several improvements were committed to improve code generation for 11651 the ARM architecture including a rewritten implementation for load 11652 and store multiples. 11653 * Several enhancements were committed to improve SIMD code generation 11654 for NEON by adding support for widening instructions, misaligned 11655 loads and stores, vector conditionals and support for 64-bit 11656 arithmetic. 11657 * Support was added for the Faraday cores fa526, fa606te, fa626te, 11658 fmp626te, fmp626 and fa726te and can be used with the respective 11659 names as parameters to the -mcpu= option. 11660 * Basic support was added for Cortex-A15 and is available through 11661 -mcpu=cortex-a15. 11662 * GCC for AAPCS configurations now more closely adheres to the AAPCS 11663 specification by enabling -fstrict-volatile-bitfields by default. 11664 11665 IA-32/x86-64 11666 11667 * The new -fsplit-stack option permits programs to use a 11668 discontiguous stack. This is useful for threaded programs, in that 11669 it is no longer necessary to specify the maximum stack size when 11670 creating a thread. This feature is currently only implemented for 11671 32-bit and 64-bit x86 GNU/Linux targets. 11672 * Support for emitting profiler counter calls before function 11673 prologues. This is enabled via a new command-line option -mfentry. 11674 * Optimization for the Intel Core 2 processors is now available 11675 through the -march=core2 and -mtune=core2 options. 11676 * Support for Intel Core i3/i5/i7 processors is now available through 11677 the -march=corei7 and -mtune=corei7 options. 11678 * Support for Intel Core i3/i5/i7 processors with AVX is now 11679 available through the -march=corei7-avx and -mtune=corei7-avx 11680 options. 11681 * Support for AMD Bobcat (family 14) processors is now available 11682 through the -march=btver1 and -mtune=btver1 options. 11683 * Support for AMD Bulldozer (family 15) processors is now available 11684 through the -march=bdver1 and -mtune=bdver1 options. 11685 * The default setting (when not optimizing for size) for 32-bit 11686 GNU/Linux and Darwin x86 targets has been changed to 11687 -fomit-frame-pointer. The default can be reverted to 11688 -fno-omit-frame-pointer by configuring GCC with the 11689 --enable-frame-pointer configure option. 11690 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 11691 __float128 on 32-bit and 64-bit x86 targets. 11692 * AVX floating-point arithmetic can now be enabled by default at 11693 configure time with the new --with-fpmath=avx option. 11694 * The SSA loop prefetching pass is enabled when using -O3 when 11695 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 11696 than K6). 11697 * Support for TBM (Trailing Bit Manipulation) built-in functions and 11698 code generation is available via -mtbm. 11699 * Support for AMD's BMI (Bit Manipulation) built-in functions and 11700 code generation is available via -mbmi. 11701 11702 [26]MicroBlaze 11703 11704 * Support has been added for the Xilinx MicroBlaze softcore processor 11705 (microblaze-elf) embedded target. This configurable processor is 11706 supported on several Xilinx Spartan and Virtex FPGAs. 11707 11708 MIPS 11709 11710 * GCC now supports the Loongson 3A processor. Its canonical -march= 11711 and -mtune= name is loongson3a. 11712 11713 [27]MN10300 / AM33 11714 11715 * The inline assembly register constraint "A" has been renamed "c". 11716 This constraint is used to select a floating-point register that 11717 can be used as the destination of a multiply-accumulate 11718 instruction. 11719 * New inline assembly register constraints "A" and "D" have been 11720 added. These constraint letters resolve to all general registers 11721 when compiling for AM33, and resolve to address registers only or 11722 data registers only when compiling for MN10300. 11723 * The MDR register is represented in the compiler. One can access the 11724 register via the "z" constraint in inline assembly. It can be 11725 marked as clobbered or used as a local register variable via the 11726 "mdr" name. The compiler uses the RETF instruction if the function 11727 does not modify the MDR register, so it is important that inline 11728 assembly properly annotate any usage of the register. 11729 11730 PowerPC/PowerPC64 11731 11732 * GCC now supports the Applied Micro Titan processor with 11733 -mcpu=titan. 11734 * The -mrecip option has been added, which indicates whether the 11735 reciprocal and reciprocal square root instructions should be used. 11736 * The -mveclibabi=mass option can be used to enable the compiler to 11737 autovectorize mathematical functions using the Mathematical 11738 Acceleration Subsystem library. 11739 * The -msingle-pic-base option has been added, which instructs the 11740 compiler to avoid loading the PIC base register in function 11741 prologues. The PIC base register must be initialized by the runtime 11742 system. 11743 * The -mblock-move-inline-limit option has been added, which enables 11744 the user to control the maximum size of inlined memcpy calls and 11745 similar. 11746 * PowerPC64 GNU/Linux support for applications requiring a large TOC 11747 section has been improved. A new command-line option, 11748 -mcmodel=MODEL, controls this feature; valid values for MODEL are 11749 small, medium, or large. 11750 * The Altivec builtin functions vec_ld and vec_st have been modified 11751 to generate the Altivec memory instructions LVX and STVX, even if 11752 the -mvsx option is used. In the initial GCC 4.5 release, these 11753 builtin functions were changed to generate VSX memory reference 11754 instructions instead of Altivec memory instructions, but there are 11755 differences between the two instructions. If the VSX instruction 11756 set is available, you can now use the new builtin functions 11757 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 11758 instructions. 11759 * The GCC compiler on AIX now defaults to a process layout with a 11760 larger data space allowing larger programs to be compiled. 11761 * The GCC long double type on AIX 6.1 and above has reverted to 11762 64-bit double precision, matching the AIX XL compiler default, 11763 because of missing C99 symbols required by the GCC runtime. 11764 * The default processor scheduling model and tuning for PowerPC64 11765 GNU/Linux and for AIX 6.1 and above now is POWER7. 11766 * Starting with GCC 4.6.1, vectors of type vector long long or vector 11767 long are passed and returned in the same method as other vectors 11768 with the VSX instruction set. Previously the GCC compiler did not 11769 adhere to the ABI for 128-bit vectors with 64-bit integer base 11770 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 11771 11772 S/390, zSeries and System z9/z10, IBM zEnterprise z196 11773 11774 * Support for the zEnterprise z196 processor has been added. When 11775 using the -march=z196 option, the compiler will generate code 11776 making use of the following instruction facilities: 11777 + Conditional load/store 11778 + Distinct-operands 11779 + Floating-point-extension 11780 + Interlocked-access 11781 + Population-count 11782 The -mtune=z196 option avoids the compare and branch instructions 11783 as well as the load address instruction with an index register as 11784 much as possible and performs instruction scheduling appropriate 11785 for the new out-of-order pipeline architecture. 11786 * When using the -m31 -mzarch options the generated code still 11787 conforms to the 32-bit ABI but uses the general purpose registers 11788 as 64-bit registers internally. This requires a Linux kernel saving 11789 the whole 64-bit registers when doing a context switch. Kernels 11790 providing that feature indicate that by the 'highgprs' string in 11791 /proc/cpuinfo. 11792 * The SSA loop prefetching pass is enabled when using -O3. 11793 11794 SPARC 11795 11796 * GCC now supports the LEON series of SPARC V8 processors. The code 11797 generated by the compiler can either be tuned to it by means of the 11798 --with-tune=leon configure option and -mtune=leon compilation 11799 option, or the compiler can be built for the sparc-leon-{elf,linux} 11800 and sparc-leon3-{elf,linux} targets directly. 11801 * GCC has stopped sign/zero-extending parameter registers in the 11802 callee for functions taking parameters with sub-word size in 32-bit 11803 mode, since this is redundant with the specification of the ABI. 11804 GCC has never done so in 64-bit mode since this is also redundant. 11805 * The command-line option -mfix-at697f has been added to enable the 11806 documented workaround for the single erratum of the Atmel AT697F 11807 processor. 11808 11809 [28]Operating Systems 11810 11811 [29]Android 11812 11813 * GCC now supports the Bionic C library and provides a convenient way 11814 of building native libraries and applications for the Android 11815 platform. Refer to the documentation of the -mandroid and -mbionic 11816 options for details on building native code. At the moment, Android 11817 support is enabled only for ARM. 11818 11819 [30]Darwin/Mac OS X 11820 11821 * General 11822 + Initial support for CFString types has been added. 11823 This allows GCC to build projects including the system Core 11824 Foundation frameworks. The GCC Objective-C family supports 11825 CFString "toll-free bridged" as per the Mac OS X system tools. 11826 CFString is also recognized in the context of format 11827 attributes and arguments (see the documentation for format 11828 attributes for limitations). At present, 8-bit character types 11829 are supported. 11830 + Object file size reduction. 11831 The Darwin zeroed memory allocators have been re-written to 11832 make more use of .zerofill sections. For non-debug code, this 11833 can reduce object file size significantly. 11834 + Objective-C family 64-bit support (NeXT ABI 2). 11835 Initial support has been added to support 64-bit Objective-C 11836 code using the Darwin/OS X native (NeXT) runtime. ABI version 11837 2 will be selected automatically when 64-bit code is built. 11838 + Objective-C family 32-bit ABI 1. 11839 For 32-bit code ABI 1 is also now also allowed. At present it 11840 must be selected manually using -fobjc-abi-version=1 where 11841 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 11842 * x86 Architecture 11843 + The -mdynamic-no-pic option has been enabled. 11844 Code supporting -mdynamic-no-pic optimization has been added 11845 and is applicable to -m32 builds. The compiler bootstrap uses 11846 the option where appropriate. 11847 + The default value for -mtune= has been changed. 11848 Since Darwin systems are primarily Xeon, Core-2 or similar the 11849 default tuning has been changed to -mtune=core2. 11850 + Enable 128-bit long double (__float128) support on Darwin. 11851 * PPC Architecture 11852 + Darwin64 ABI. 11853 Several significant bugs have been fixed, such that GCC now 11854 produces code compatible with the Darwin64 PowerPC ABI. 11855 + libffi and boehm-gc. 11856 The Darwin ports of the libffi and boehm-gc libraries have 11857 been upgraded to include a Darwin64 implementation. This means 11858 that powerpc*-*-darwin9 platforms may now, for example, build 11859 Java applications with -m64 enabled. 11860 + Plug-in support has been enabled. 11861 + The -fsection-anchors option is now available although, 11862 presently, not heavily tested. 11863 11864 [31]Solaris 2 11865 11866 New Features 11867 11868 * Support symbol versioning with the Sun linker. 11869 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 11870 * Support thread-local storage (TLS) with the Sun assembler on 11871 Solaris 2/x86. 11872 * Support TLS on Solaris 8/9 if prerequisites are met. 11873 * Support COMDAT group with the GNU assembler and recent Sun linker. 11874 * Support the Sun assembler visibility syntax. 11875 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 11876 -march=pentiumpro (Solaris 8/9). 11877 * Don't use SSE on Solaris 8/9 x86 by default. 11878 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 11879 11880 ABI Change 11881 11882 * Change the ABI for returning 8-byte vectors like __m64 in MMX 11883 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 11884 compilers. This is an incompatible change. If you use such types, 11885 you must either recompile all your code with the new compiler or 11886 use the new -mvect8-ret-in-mem option to remain compatible with 11887 previous versions of GCC and Sun Studio. 11888 11889 [32]Windows x86/x86_64 11890 11891 * Initial support for decimal floating point. 11892 * Support for the __thiscall calling-convention. 11893 * Support for hot-patchable function prologues via the 11894 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 11895 * Improvements of stack-probing and stack-allocation mechanisms. 11896 * Support of push/pop-macro pragma as preprocessor command. 11897 With #pragma push_macro("macro-name") the current definition of 11898 macro-name is saved and can be restored with #pragma 11899 pop_macro("macro-name") to its saved definition. 11900 * Enable 128-bit long double (__float128) support on MinGW and 11901 Cygwin. 11902 11903 Other significant improvements 11904 11905 Installation changes 11906 11907 * An install-strip make target is provided that installs stripped 11908 executables, and may install libraries with unneeded or debugging 11909 sections stripped. 11910 * On Power7 systems, there is a potential problem if you build the 11911 GCC compiler with a host compiler using options that enable the VSX 11912 instruction set generation. If the host compiler has been patched 11913 so that the vec_ld and vec_st builtin functions generate Altivec 11914 memory instructions instead of VSX memory instructions, then you 11915 should be able to build the compiler with VSX instruction 11916 generation. 11917 11918 Changes for GCC Developers 11919 11920 Note: these changes concern developers that develop GCC itself or 11921 software that integrates with GCC, such as plugins, and not the general 11922 GCC users. 11923 * The gengtype utility, which previously was internal to the GCC 11924 build process, has been enchanced to provide GC root information 11925 for plugins as necessary. 11926 * The old GC allocation interface of ggc_alloc and friends was 11927 replaced with a type-safe alternative. 11928 11929 [33]GCC 4.6.1 11930 11931 This is the [34]list of problem reports (PRs) from GCC's bug tracking 11932 system that are known to be fixed in the 4.6.1 release. This list might 11933 not be complete (that is, it is possible that some PRs that have been 11934 fixed are not listed here). 11935 11936 [35]GCC 4.6.2 11937 11938 This is the [36]list of problem reports (PRs) from GCC's bug tracking 11939 system that are known to be fixed in the 4.6.2 release. This list might 11940 not be complete (that is, it is possible that some PRs that have been 11941 fixed are not listed here). 11942 11943 [37]GCC 4.6.3 11944 11945 This is the [38]list of problem reports (PRs) from GCC's bug tracking 11946 system that are known to be fixed in the 4.6.3 release. This list might 11947 not be complete (that is, it is possible that some PRs that have been 11948 fixed are not listed here). 11949 11950 [39]GCC 4.6.4 11951 11952 This is the [40]list of problem reports (PRs) from GCC's bug tracking 11953 system that are known to be fixed in the 4.6.4 release. This list might 11954 not be complete (that is, it is possible that some PRs that have been 11955 fixed are not listed here). 11956 11957 11958 For questions related to the use of GCC, please consult these web 11959 pages and the [41]GCC manuals. If that fails, the 11960 [42]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11961 web pages and the development of GCC are welcome on our developer 11962 list at [43]gcc (a] gcc.gnu.org. All of [44]our lists have public 11963 archives. 11964 11965 Copyright (C) [45]Free Software Foundation, Inc. Verbatim copying and 11966 distribution of this entire article is permitted in any medium, 11967 provided this notice is preserved. 11968 11969 These pages are [46]maintained by the GCC team. Last modified 11970 2025-01-31. 11971 11972 References 11973 11974 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 11975 2. https://gcc.gnu.org/gcc-4.5/changes.html 11976 3. https://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 11977 4. https://gcc.gnu.org/gcc-4.6/porting_to.html 11978 5. https://gcc.gnu.org/projects/lto/whopr.pdf 11979 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 11980 7. https://gcc.gnu.org/gcc-4.6/changes.html#c 11981 8. https://gcc.gnu.org/gcc-4.6/changes.html#cplusplus 11982 9. https://gcc.gnu.org/gcc-4.6/cxx0x_status.html 11983 10. https://gcc.gnu.org/PR43145 11984 11. https://gcc.gnu.org/PR43680 11985 12. https://gcc.gnu.org/PR33558 11986 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 11987 14. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 11988 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 11989 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 11990 17. https://gcc.gnu.org/gcc-4.6/changes.html#fortran 11991 18. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 11992 19. https://gcc.gnu.org/wiki/OOP 11993 20. https://gcc.gnu.org/wiki/Coarray 11994 21. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 11995 22. https://gcc.gnu.org/gcc-4.6/changes.html#go 11996 23. https://gcc.gnu.org/gcc-4.6/changes.html#objective-c 11997 24. https://gcc.gnu.org/gcc-4.6/changes.html#targets 11998 25. https://gcc.gnu.org/gcc-4.6/changes.html#arm 11999 26. https://gcc.gnu.org/gcc-4.6/changes.html#microblaze 12000 27. https://gcc.gnu.org/gcc-4.6/changes.html#mn10300 12001 28. https://gcc.gnu.org/gcc-4.6/changes.html#os 12002 29. https://gcc.gnu.org/gcc-4.6/changes.html#android 12003 30. https://gcc.gnu.org/gcc-4.6/changes.html#darwin 12004 31. https://gcc.gnu.org/gcc-4.6/changes.html#solaris 12005 32. https://gcc.gnu.org/gcc-4.6/changes.html#windows 12006 33. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.1 12007 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 12008 35. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.2 12009 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 12010 37. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.3 12011 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 12012 39. https://gcc.gnu.org/gcc-4.6/changes.html#GCC4.6.4 12013 40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 12014 41. https://gcc.gnu.org/onlinedocs/ 12015 42. mailto:gcc-help (a] gcc.gnu.org 12016 43. mailto:gcc (a] gcc.gnu.org 12017 44. https://gcc.gnu.org/lists.html 12018 45. https://www.fsf.org/ 12019 46. https://gcc.gnu.org/about.html 12020 ====================================================================== 12021 http://gcc.gnu.org/gcc-4.5/index.html 12022 12023 GCC 4.5 Release Series 12024 12025 (This release series is no longer supported.) 12026 12027 Jul 2, 2012 12028 12029 The [1]GNU project and the GCC developers are pleased to announce the 12030 release of GCC 4.5.4. 12031 12032 This release is a bug-fix release, containing fixes for regressions in 12033 GCC 4.5.3 relative to previous releases of GCC. 12034 12035 Release History 12036 12037 GCC 4.5.4 12038 Jul 2, 2012 ([2]changes) 12039 12040 GCC 4.5.3 12041 Apr 28, 2011 ([3]changes) 12042 12043 GCC 4.5.2 12044 Dec 16, 2010 ([4]changes) 12045 12046 GCC 4.5.1 12047 Jul 31, 2010 ([5]changes) 12048 12049 GCC 4.5.0 12050 April 14, 2010 ([6]changes) 12051 12052 References and Acknowledgements 12053 12054 GCC used to stand for the GNU C Compiler, but since the compiler 12055 supports several other languages aside from C, it now stands for the 12056 GNU Compiler Collection. 12057 12058 A list of [7]successful builds is updated as new information becomes 12059 available. 12060 12061 The GCC developers would like to thank the numerous people that have 12062 contributed new features, improvements, bug fixes, and other changes as 12063 well as test results to GCC. This [8]amazing group of volunteers is 12064 what makes GCC successful. 12065 12066 For additional information about GCC please refer to the [9]GCC project 12067 web site or contact the [10]GCC development mailing list. 12068 12069 To obtain GCC please use [11]our mirror sites or [12]our version 12070 control system. 12071 12072 12073 For questions related to the use of GCC, please consult these web 12074 pages and the [13]GCC manuals. If that fails, the 12075 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12076 web pages and the development of GCC are welcome on our developer 12077 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 12078 archives. 12079 12080 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 12081 distribution of this entire article is permitted in any medium, 12082 provided this notice is preserved. 12083 12084 These pages are [18]maintained by the GCC team. Last modified 12085 2022-10-26. 12086 12087 References 12088 12089 1. http://www.gnu.org/ 12090 2. https://gcc.gnu.org/gcc-4.5/changes.html 12091 3. https://gcc.gnu.org/gcc-4.5/changes.html 12092 4. https://gcc.gnu.org/gcc-4.5/changes.html 12093 5. https://gcc.gnu.org/gcc-4.5/changes.html 12094 6. https://gcc.gnu.org/gcc-4.5/changes.html 12095 7. https://gcc.gnu.org/gcc-4.5/buildstat.html 12096 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12097 9. https://gcc.gnu.org/index.html 12098 10. mailto:gcc (a] gcc.gnu.org 12099 11. https://gcc.gnu.org/mirrors.html 12100 12. https://gcc.gnu.org/git.html 12101 13. https://gcc.gnu.org/onlinedocs/ 12102 14. mailto:gcc-help (a] gcc.gnu.org 12103 15. mailto:gcc (a] gcc.gnu.org 12104 16. https://gcc.gnu.org/lists.html 12105 17. https://www.fsf.org/ 12106 18. https://gcc.gnu.org/about.html 12107 ====================================================================== 12108 http://gcc.gnu.org/gcc-4.5/changes.html 12109 12110 GCC 4.5 Release Series 12111 Changes, New Features, and Fixes 12112 12113 Caveats 12114 12115 * GCC now requires the [1]MPC library in order to build. See the 12116 [2]prerequisites page for version requirements. 12117 * Support for a number of older systems and recently unmaintained or 12118 untested target ports of GCC has been declared obsolete in GCC 4.5. 12119 Unless there is activity to revive them, the next release of GCC 12120 will have their sources permanently removed. 12121 The following ports for individual systems on particular 12122 architectures have been obsoleted: 12123 + IRIX releases before 6.5 (mips-sgi-irix5*, 12124 mips-sgi-irix6.[0-4]) 12125 + Solaris 7 (*-*-solaris2.7) 12126 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 12127 alpha-dec-osf5.0*) 12128 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 12129 can be found in the [3]announcement. 12130 Support for the classic POWER architecture implemented in the 12131 original RIOS and RIOS2 processors of the old IBM RS/6000 product 12132 line has been obsoleted in the rs6000 port. This does not affect 12133 the new generation Power and PowerPC architectures. 12134 * Support has been removed for all the [4]configurations obsoleted in 12135 GCC 4.4. 12136 * Support has been removed for the protoize and unprotoize utilities, 12137 obsoleted in GCC 4.4. 12138 * Support has been removed for tuning for Itanium1 (Merced) variants. 12139 Note that code tuned for Itanium2 should also run correctly on 12140 Itanium1. 12141 * GCC now generates unwind info also for epilogues. DWARF debuginfo 12142 generated by GCC now uses more features of DWARF3 than before, and 12143 also some DWARF4 features. GDB older than 7.0 is not able to handle 12144 either of these, so to debug GCC 4.5 generated binaries or 12145 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 12146 features with the -gdwarf-3 -gstrict-dwarf options, or use 12147 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 12148 epilogue unwind info is emitted unconditionally whenever unwind 12149 info is emitted. 12150 * On x86 targets, code containing floating-point calculations may run 12151 significantly more slowly when compiled with GCC 4.5 in strict C99 12152 conformance mode than they did with earlier GCC versions. This is 12153 due to stricter standard conformance of the compiler and can be 12154 avoided by using the option -fexcess-precision=fast; also see 12155 [5]below. 12156 * The function attribute noinline no longer prevents GCC from cloning 12157 the function. A new attribute noclone has been introduced for this 12158 purpose. Cloning a function means that it is duplicated and the new 12159 copy is specialized for certain contexts (for example when a 12160 parameter is a known constant). 12161 12162 General Optimizer Improvements 12163 12164 * The -save-temps now takes an optional argument. The -save-temps and 12165 -save-temps=cwd switches write the temporary files in the current 12166 working directory based on the original source file. The 12167 -save-temps=obj switch will write files into the directory 12168 specified with the -o option, and the intermediate filenames are 12169 based on the output file. This will allow the user to get the 12170 compiler intermediate files when doing parallel builds without two 12171 builds of the same filename located in different directories from 12172 interfering with each other. 12173 * Debugging dumps are now created in the same directory as the object 12174 file rather than in the current working directory. This allows the 12175 user to get debugging dumps when doing parallel builds without two 12176 builds of the same filename interfering with each other. 12177 * GCC has been integrated with the MPC library. This allows GCC to 12178 evaluate complex arithmetic at compile time [6]more accurately. It 12179 also allows GCC to evaluate calls to complex built-in math 12180 functions having constant arguments and replace them at compile 12181 time with their mathematically equivalent results. In doing so, GCC 12182 can generate correct results regardless of the math library 12183 implementation or floating point precision of the host platform. 12184 This also allows GCC to generate identical results regardless of 12185 whether one compiles in native or cross-compile configurations to a 12186 particular target. The following built-in functions take advantage 12187 of this new capability: cacos, cacosh, casin, casinh, catan, 12188 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 12189 and ctanh. The float and long double variants of these functions 12190 (e.g. csinf and csinl) are also handled. 12191 * A new link-time optimizer has been added ([7]-flto). When this 12192 option is used, GCC generates a bytecode representation of each 12193 input file and writes it to specially-named sections in each object 12194 file. When the object files are linked together, all the function 12195 bodies are read from these named sections and instantiated as if 12196 they had been part of the same translation unit. This enables 12197 interprocedural optimizations to work across different files (and 12198 even different languages), potentially improving the performance of 12199 the generated code. To use the link-timer optimizer, -flto needs to 12200 be specified at compile time and during the final link. If the 12201 program does not require any symbols to be exported, it is possible 12202 to combine -flto and the experimental [8]-fwhopr with 12203 [9]-fwhole-program to allow the interprocedural optimizers to use 12204 more aggressive assumptions. 12205 * The automatic parallelization pass was enhanced to support 12206 parallelization of outer loops. 12207 * Automatic parallelization can be enabled as part of Graphite. In 12208 addition to -ftree-parallelize-loops=, specify 12209 -floop-parallelize-all to enable the Graphite-based optimization. 12210 * The infrastructure for optimizing based on [10]restrict qualified 12211 pointers has been rewritten and should result in code generation 12212 improvements. Optimizations based on restrict qualified pointers 12213 are now also available when using -fno-strict-aliasing. 12214 * There is a new optimization pass that attempts to change prototype 12215 of functions to avoid unused parameters, pass only relevant parts 12216 of structures and turn arguments passed by reference to arguments 12217 passed by value when possible. It is enabled by -O2 and above as 12218 well as -Os and can be manually invoked using the new command-line 12219 switch -fipa-sra. 12220 * GCC now optimize exception handling code. In particular cleanup 12221 regions that are proved to not have any effect are optimized out. 12222 12223 New Languages and Language specific improvements 12224 12225 All languages 12226 12227 * The -fshow-column option is now on by default. This means error 12228 messages now have a column associated with them. 12229 12230 Ada 12231 12232 * Compilation of programs heavily using discriminated record types 12233 with variant parts has been sped up and generates more compact 12234 code. 12235 * Stack checking now works reasonably well on most plaforms. In some 12236 specific cases, stack overflows may still fail to be detected, but 12237 a compile-time warning will be issued for these cases. 12238 12239 C family 12240 12241 * If a header named in a #include directive is not found, the 12242 compiler exits immediately. This avoids a cascade of errors arising 12243 from declarations expected to be found in that header being 12244 missing. 12245 * A new built-in function __builtin_unreachable() has been added that 12246 tells the compiler that control will never reach that point. It may 12247 be used after asm statements that terminate by transferring control 12248 elsewhere, and in other places that are known to be unreachable. 12249 * The -Wlogical-op option now warns for logical expressions such as 12250 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 12251 mistakes. This option is disabled by default. 12252 * An asm goto feature has been added to allow asm statements that 12253 jump to C labels. 12254 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 12255 * The deprecated attribute now takes an optional string argument, for 12256 example, __attribute__((deprecated("text string"))), that will be 12257 printed together with the deprecation warning. 12258 12259 C 12260 12261 * The -Wenum-compare option, which warns when comparing values of 12262 different enum types, now works for C. It formerly only worked for 12263 C++. This warning is enabled by -Wall. It may be avoided by using a 12264 type cast. 12265 * The -Wcast-qual option now warns about casts which are unsafe in 12266 that they permit const-correctness to be violated without further 12267 warnings. Specifically, it warns about cases where a qualifier is 12268 added when all the lower types are not const. For example, it warns 12269 about a cast from char ** to const char **. 12270 * The -Wc++-compat option is significantly improved. It issues new 12271 warnings for: 12272 + Using C++ reserved operator names as identifiers. 12273 + Conversions to enum types without explicit casts. 12274 + Using va_arg with an enum type. 12275 + Using different enum types in the two branches of ?:. 12276 + Using ++ or -- on a variable of enum type. 12277 + Using the same name as both a struct, union or enum tag and a 12278 typedef, unless the typedef refers to the tagged type itself. 12279 + Using a struct, union, or enum which is defined within another 12280 struct or union. 12281 + A struct field defined using a typedef if there is a field in 12282 the struct, or an enclosing struct, whose name is the typedef 12283 name. 12284 + Duplicate definitions at file scope. 12285 + Uninitialized const variables. 12286 + A global variable with an anonymous struct, union, or enum 12287 type. 12288 + Using a string constant to initialize a char array whose size 12289 is the length of the string. 12290 * The new -Wjump-misses-init option warns about cases where a goto or 12291 switch skips the initialization of a variable. This sort of branch 12292 is an error in C++ but not in C. This warning is enabled by 12293 -Wc++-compat. 12294 * GCC now ensures that a C99-conforming <stdint.h> is present on most 12295 targets, and uses information about the types in this header to 12296 implement the Fortran bindings to those types. GCC does not ensure 12297 the presence of such a header, and does not implement the Fortran 12298 bindings, on the following targets: NetBSD, VxWorks, VMS, 12299 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 12300 * GCC now implements C90- and C99-conforming rules for constant 12301 expressions. This may cause warnings or errors for some code using 12302 expressions that can be folded to a constant but are not constant 12303 expressions as defined by ISO C. 12304 * All known target-independent C90 and C90 Amendment 1 conformance 12305 bugs, and all known target-independent C99 conformance bugs not 12306 related to floating point or extended identifiers, have been fixed. 12307 * The C decimal floating point support now includes support for the 12308 FLOAT_CONST_DECIMAL64 pragma. 12309 * The named address space feature from ISO/IEC TR 18037 is now 12310 supported. This is currently only implemented for the SPU 12311 processor. 12312 12313 [11]C++ 12314 12315 * Improved [12]experimental support for the upcoming C++0x ISO C++ 12316 standard, including support for raw strings, lambda expressions and 12317 explicit type conversion operators. 12318 * When printing the name of a class template specialization, G++ will 12319 now omit any template arguments which come from default template 12320 arguments. This behavior (and the pretty-printing of function 12321 template specializations as template signature and arguments) can 12322 be disabled with the -fno-pretty-templates option. 12323 * Access control is now applied to typedef names used in a template, 12324 which may cause G++ to reject some ill-formed code that was 12325 accepted by earlier releases. The -fno-access-control option can be 12326 used as a temporary workaround until the code is corrected. 12327 * Compilation time for code that uses templates should now scale 12328 linearly with the number of instantiations rather than 12329 quadratically, as template instantiations are now looked up using 12330 hash tables. 12331 * Declarations of functions that look like builtin declarations of 12332 library functions are only considered to be redeclarations if they 12333 are declared with extern "C". This may cause problems with code 12334 that omits extern "C" on hand-written declarations of C library 12335 functions such as abort or memcpy. Such code is ill-formed, but was 12336 accepted by earlier releases. 12337 * Diagnostics that used to complain about passing non-POD types to 12338 ... or jumping past the declaration of a non-POD variable now check 12339 for triviality rather than PODness, as per C++0x. 12340 * In C++0x mode local and anonymous classes are now allowed as 12341 template arguments, and in declarations of variables and functions 12342 with linkage, so long as any such declaration that is used is also 12343 defined ([13]DR 757). 12344 * Labels may now have attributes, as has been permitted for a while 12345 in C. This is only permitted when the label definition and the 12346 attribute specifier is followed by a semicolon--i.e., the label 12347 applies to an empty statement. The only useful attribute for a 12348 label is unused. 12349 * G++ now implements [14]DR 176. Previously G++ did not support using 12350 the injected-class-name of a template base class as a type name, 12351 and lookup of the name found the declaration of the template in the 12352 enclosing scope. Now lookup of the name finds the 12353 injected-class-name, which can be used either as a type or as a 12354 template, depending on whether or not the name is followed by a 12355 template argument list. As a result of this change, some code that 12356 was previously accepted may be ill-formed because 12357 1. The injected-class-name is not accessible because it's from a 12358 private base, or 12359 2. The injected-class-name cannot be used as an argument for a 12360 template template parameter. 12361 In either of these cases, the code can be fixed by adding a 12362 nested-name-specifier to explicitly name the template. The first 12363 can be worked around with -fno-access-control; the second is only 12364 rejected with -pedantic. 12365 * A new standard mangling for SIMD vector types has been added, to 12366 avoid name clashes on systems with vectors of varying length. By 12367 default the compiler still uses the old mangling, but emits aliases 12368 with the new mangling on targets that support strong aliases. Users 12369 can switch over entirely to the new mangling with -fabi-version=4 12370 or -fabi-version=0. -Wabi will now warn about code that uses the 12371 old mangling. 12372 * The command-line option -ftemplate-depth-N is now written as 12373 -ftemplate-depth=N and the old form is deprecated. 12374 * Conversions between NULL and non-pointer types are now warned by 12375 default. The new option -Wno-conversion-null disables these 12376 warnings. Previously these warnings were only available when using 12377 -Wconversion explicitly. 12378 12379 Runtime Library (libstdc++) 12380 12381 * Improved experimental support for the upcoming ISO C++ standard, 12382 C++0x, including: 12383 + Support for <future>, <functional>, and <random>. 12384 + Existing facilities now exploit explicit operators and the 12385 newly implemented core C++0x features. 12386 + The header <cstdatomic> has been renamed to <atomic>. 12387 * An experimental [15]profile mode has been added. This is an 12388 implementation of many C++ standard library constructs with an 12389 additional analysis layer that gives performance improvement advice 12390 based on recognition of suboptimal usage patterns. For example, 12391 #include <vector> 12392 int main() 12393 { 12394 std::vector<int> v; 12395 for (int k = 0; k < 1024; ++k) 12396 v.insert(v.begin(), k); 12397 } 12398 12399 When instrumented via the profile mode, can return suggestions 12400 about the initial size and choice of the container used as follows: 12401 vector-to-list: improvement = 5: call stack = 0x804842c ... 12402 : advice = change std::vector to std::list 12403 vector-size: improvement = 3: call stack = 0x804842c ... 12404 : advice = change initial container size from 0 to 1024 12405 12406 These constructs can be substituted for the normal libstdc++ 12407 constructs on a piecemeal basis, or all existing components can be 12408 transformed via the -D_GLIBCXX_PROFILE macro. 12409 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR 12410 24733) has been added. This support is in header file 12411 <decimal/decimal>, uses namespace std::decimal, and includes 12412 classes decimal32, decimal64, and decimal128. 12413 * Sources have been audited for application of function attributes 12414 nothrow, const, pure, and noreturn. 12415 * Python pretty-printers have been added for many standard library 12416 components that simplify the internal representation and present a 12417 more intuitive view of components when used with 12418 appropriately-advanced versions of GDB. For more information, 12419 please consult the more [17]detailed description. 12420 * The default behavior for comparing typeinfo names has changed, so 12421 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 12422 * The new -static-libstdc++ option directs g++ to link the C++ 12423 library statically, even if the default would normally be to link 12424 it dynamically. 12425 12426 [18]Fortran 12427 12428 * The COMMON default padding has been changed - instead of adding the 12429 padding before a variable it is now added afterwards, which 12430 increases the compatibility with other vendors and helps to obtain 12431 the correct output in some cases. Cf. also the -falign-commons 12432 option ([19]added in 4.4). 12433 * The -finit-real= option now also supports the value snan for 12434 signaling not-a-number; to be effective, one additionally needs to 12435 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 12436 optimizations can turn a signaling NaN into a quiet one. 12437 * The new option -fcheck= has been added with the options bounds, 12438 array-temps, do, pointer, and recursive. The bounds and array-temps 12439 options are equivalent to -fbounds-check and 12440 -fcheck-array-temporaries. The do option checks for invalid 12441 modification of loop iteration variables, and the recursive option 12442 tests for recursive calls to subroutines/functions which are not 12443 marked as recursive. With pointer pointer association checks in 12444 calls are performed; however, neither undefined pointers nor 12445 pointers in expressions are handled. Using -fcheck=all enables all 12446 these run-time checks. 12447 * The run-time checking -fcheck=bounds now warns about invalid string 12448 lengths of character dummy arguments. Additionally, more 12449 compile-time checks have been added. 12450 * The new option [20]-fno-protect-parens has been added; if set, the 12451 compiler may reorder REAL and COMPLEX expressions without regard to 12452 parentheses. 12453 * GNU Fortran no longer links against libgfortranbegin. As before, 12454 MAIN__ (assembler symbol name) is the actual Fortran main program, 12455 which is invoked by the main function. However, main is now 12456 generated and put in the same object file as MAIN__. For the time 12457 being, libgfortranbegin still exists for backward compatibility. 12458 For details see the new [21]Mixed-Language Programming chapter in 12459 the manual. 12460 * The I/O library was restructured for performance and cleaner code. 12461 * Array assignments and WHERE are now run in parallel when OpenMP's 12462 WORKSHARE is used. 12463 * The experimental option -fwhole-file was added. The option allows 12464 whole-file checking of procedure arguments and allows for better 12465 optimizations. It can also be used with -fwhole-program, which is 12466 now also supported in gfortran. 12467 * More Fortran 2003 and Fortran 2008 mathematical functions can now 12468 be used as initialization expressions. 12469 * Some extended attributes such as STDCALL are now supported via the 12470 [22]GCC$ compiler directive. 12471 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 12472 intrinsic behaves now as if zero were always positive. 12473 * For legacy compatibiliy: On Cygwin and MinGW, the special files 12474 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 12475 supported. 12476 * Fortran 2003 support has been extended: 12477 + Procedure-pointer function results and procedure-pointer 12478 components (including PASS), 12479 + allocatable scalars (experimental), 12480 + DEFERRED type-bound procedures, 12481 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 12482 have been implemented. 12483 + The ALLOCATE statement supports type-specs and the SOURCE= 12484 argument. 12485 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 12486 type-bound procedure (i.e. as type-bound operators). 12487 + Rounding (ROUND=, RZ, ...) for output is now supported. 12488 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 12489 intrinsic module ISO_C_BINDING are now supported, except for 12490 the targets listed above as ones where GCC does not have 12491 <stdint.h> type information. 12492 + Extensible derived types with type-bound procedure or 12493 procedure pointer with PASS attribute now have to use CLASS in 12494 line with the Fortran 2003 standard; the workaround to use 12495 TYPE is no longer supported. 12496 + [23]Experimental, incomplete support for polymorphism, 12497 including CLASS, SELECT TYPE and dynamic dispatch of 12498 type-bound procedure calls. Some features do not work yet such 12499 as unlimited polymorphism (CLASS(*)). 12500 * Fortran 2008 support has been extended: 12501 + The OPEN statement now supports the NEWUNIT= option, which 12502 returns a unique file unit, thus preventing inadvertent use of 12503 the same unit in different parts of the program. 12504 + Support for unlimited format items has been added. 12505 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 12506 the intrinsic module ISO_FORTRAN_ENV are now supported. 12507 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 12508 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 12509 and ATANH have been added (for real and complex arguments) and 12510 ATAN(Y,X) is now an alias for ATAN2(Y,X). 12511 + The BLOCK construct has been implemented. 12512 12513 [24]New Targets and Target Specific Improvements 12514 12515 AIX 12516 12517 * Full cross-toolchain support now available with GNU Binutils 12518 12519 ARM 12520 12521 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 12522 * GCC now supports the ARM v7E-M architecture. 12523 * GCC now supports VFPv4-based FPUs and FPUs with 12524 single-precision-only VFP. 12525 * GCC has many improvements to optimization for other ARM processors, 12526 including scheduling support for the integer pipeline on Cortex-A9. 12527 * GCC now supports the IEEE 754-2008 half-precision floating-point 12528 type, and a variant ARM-specific half-precision type. This type is 12529 specified using __fp16, with the layout determined by 12530 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 12531 VFPv4 half-precision instructions will be used. 12532 * GCC now supports the variant of AAPCS that uses VFP registers for 12533 parameter passing and return values. 12534 12535 AVR 12536 12537 * The -mno-tablejump option has been removed because it has the same 12538 effect as the -fno-jump-tables option. 12539 * Added support for these new AVR devices: 12540 + ATmega8U2 12541 + ATmega16U2 12542 + ATmega32U2 12543 12544 [25]IA-32/x86-64 12545 12546 * GCC now will set the default for -march= based on the configure 12547 target. 12548 * GCC now supports handling floating-point excess precision arising 12549 from use of the x87 floating-point unit in a way that conforms to 12550 ISO C99. This is enabled with -fexcess-precision=standard and with 12551 standards conformance options such as -std=c99, and may be disabled 12552 using -fexcess-precision=fast. 12553 * Support for the Intel Atom processor is now available through the 12554 -march=atom and -mtune=atom options. 12555 * A new -mcrc32 option is now available to enable crc32 intrinsics. 12556 * A new -mmovbe option is now available to enable GCC to use the 12557 movbe instruction to implement __builtin_bswap32 and 12558 __builtin_bswap64. 12559 * SSE math now can be enabled by default at configure time with the 12560 new --with-fpmath=sse option. 12561 * There is a new intrinsic header file, <x86intrin.h>. It should be 12562 included before using any IA-32/x86-64 intrinsics. 12563 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 12564 Orochi processors are now available with the -mxop, -mfma4, and 12565 -mlwp options. 12566 * The -mabm option enables GCC to use the popcnt and lzcnt 12567 instructions on AMD processors. 12568 * The -mpopcnt option enables GCC to use the popcnt instructions on 12569 both AMD and Intel processors. 12570 12571 M68K/ColdFire 12572 12573 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 12574 and 5441x devices. 12575 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 12576 processors. 12577 12578 [26]MeP 12579 12580 Support has been added for the Toshiba Media embedded Processor (MeP, 12581 or mep-elf) embedded target. 12582 12583 MIPS 12584 12585 * GCC now supports MIPS 1004K processors. 12586 * GCC can now be configured with options --with-arch-32, 12587 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 12588 default optimization separately for 32-bit and 64-bit modes. 12589 * MIPS targets now support an alternative _mcount interface, in which 12590 register $12 points to the function's save slot for register $31. 12591 This interface is selected by the -mcount-ra-address option; see 12592 the documentation for more details. 12593 * GNU/Linux targets can now generate read-only .eh_frame sections. 12594 This optimization requires GNU binutils 2.20 or above, and is only 12595 available if GCC is configured with a suitable version of binutils. 12596 * GNU/Linux targets can now attach special relocations to indirect 12597 calls, so that the linker can turn them into direct jumps or 12598 branches. This optimization requires GNU binutils 2.20 or later, 12599 and is automatically selected if GCC is configured with an 12600 appropriate version of binutils. It can be explicitly enabled or 12601 disabled using the -mrelax-pic-calls command-line option. 12602 * GCC now generates more heavily-optimized atomic operations on 12603 Octeon processors. 12604 * MIPS targets now support the -fstack-protector option. 12605 * GCC now supports an -msynci option, which specifies that synci is 12606 enough to flush the instruction cache, without help from the 12607 operating system. GCC uses this information to optimize 12608 automatically-generated cache flush operations, such as those used 12609 for nested functions in C. There is also a --with-synci 12610 configure-time option, which makes -msynci the default. 12611 * GCC supports four new function attributes for interrupt handlers: 12612 interrupt, use_shadow_register_set, keep_interrupts_masked and 12613 use_debug_exception_return. See the documentation for more details 12614 about these attributes. 12615 12616 [27]RS/6000 (POWER/PowerPC) 12617 12618 * GCC now supports the Power ISA 2.06, which includes the VSX 12619 instructions that add vector 64-bit floating point support, new 12620 population count instructions, and conversions between floating 12621 point and unsigned types. 12622 * Support for the power7 processor is now available through the 12623 -mcpu=power7 and -mtune=power7. 12624 * GCC will now vectorize loops that contain simple math functions 12625 like copysign when generating code for altivec or VSX targets. 12626 * Support for the A2 processor is now available through the -mcpu=a2 12627 and -mtune=a2 options. 12628 * Support for the 476 processor is now available through the 12629 -mcpu={476,476fp} and -mtune={476,476fp} options. 12630 * Support for the e500mc64 processor is now available through the 12631 -mcpu=e500mc64 and -mtune=e500mc64 options. 12632 * GCC can now be configured with options --with-cpu-32, 12633 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 12634 default optimization separately for 32-bit and 64-bit modes. 12635 * Starting with GCC 4.5.4, vectors of type vector long long or vector 12636 long are passed and returned in the same method as other vectors 12637 with the VSX instruction set. Previously the GCC compiler did not 12638 adhere to the ABI for 128-bit vectors with 64-bit integer base 12639 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 12640 12641 [28]RX 12642 12643 Support has been added for the Renesas RX Processor (rx-elf) target. 12644 12645 [29]Operating Systems 12646 12647 [30]Windows (Cygwin and MinGW) 12648 12649 * GCC now installs all the major language runtime libraries as DLLs 12650 when configured with the --enable-shared option. 12651 * GCC now makes use of the new support for aligned common variables 12652 in versions of binutils >= 2.20 to fix bugs in the support for SSE 12653 data types. 12654 * Improvements to the libffi support library increase the reliability 12655 of code generated by GCJ on all Windows platforms. Libgcj is 12656 enabled by default for the first time. 12657 * Libtool improvements simplify installation by placing the generated 12658 DLLs in the correct binaries directory. 12659 * Numerous other minor bugfixes and improvements, and substantial 12660 enhancements to the Fortran language support library. 12661 12662 > 12663 12664 Other significant improvements 12665 12666 Plugins 12667 12668 * It is now possible to extend the compiler without having to modify 12669 its source code. A new option -fplugin=file.so tells GCC to load 12670 the shared object file.so and execute it as part of the compiler. 12671 The internal documentation describes the details on how plugins can 12672 interact with the compiler. 12673 12674 Installation changes 12675 12676 * The move to newer autotools changed default installation 12677 directories and switches to control them: The --with-datarootdir, 12678 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 12679 used any more. Instead, you can now use --datarootdir, --docdir, 12680 --htmldir, and --pdfdir. The default installation directories have 12681 changed as follows according to the GNU Coding Standards: 12682 12683 datarootdir read-only architecture-independent data root [PREFIX/share] 12684 localedir locale-specific message catalogs [DATAROOTDIR/locale] 12685 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 12686 htmldir html documentation [DOCDIR] 12687 dvidir dvi documentation [DOCDIR] 12688 pdfdir pdf documentation [DOCDIR] 12689 psdir ps documentation [DOCDIR] 12690 The following variables have new default values: 12691 12692 datadir read-only architecture-independent data [DATAROOTDIR] 12693 infodir info documentation [DATAROOTDIR/info] 12694 mandir man documentation [DATAROOTDIR/man] 12695 12696 [31]GCC 4.5.1 12697 12698 This is the [32]list of problem reports (PRs) from GCC's bug tracking 12699 system that are known to be fixed in the 4.5.1 release. This list might 12700 not be complete (that is, it is possible that some PRs that have been 12701 fixed are not listed here). 12702 12703 All languages 12704 12705 * GCC's new link-time optimizer ([33]-flto) now also works on a few 12706 non-ELF targets: 12707 + Cygwin (*-cygwin*) 12708 + MinGW (*-mingw*) 12709 + Darwin on x86-64 (x86_64-apple-darwin*) 12710 LTO is not enabled by default for these targets. To enable LTO, you 12711 should configure with the --enable-lto option. 12712 12713 [34]GCC 4.5.2 12714 12715 This is the [35]list of problem reports (PRs) from GCC's bug tracking 12716 system that are known to be fixed in the 4.5.2 release. This list might 12717 not be complete (that is, it is possible that some PRs that have been 12718 fixed are not listed here). 12719 12720 [36]GCC 4.5.3 12721 12722 This is the [37]list of problem reports (PRs) from GCC's bug tracking 12723 system that are known to be fixed in the 4.5.3 release. This list might 12724 not be complete (that is, it is possible that some PRs that have been 12725 fixed are not listed here). 12726 12727 On the PowerPC compiler, the Altivec builtin functions vec_ld and 12728 vec_st have been modified to generate the Altivec memory instructions 12729 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 12730 release, these builtin functions were changed to generate VSX memory 12731 reference instructions instead of Altivec memory instructions, but 12732 there are differences between the two instructions. If the VSX 12733 instruction set is available, you can now use the new builtin functions 12734 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 12735 instructions. 12736 12737 [38]GCC 4.5.4 12738 12739 This is the [39]list of problem reports (PRs) from GCC's bug tracking 12740 system that are known to be fixed in the 4.5.4 release. This list might 12741 not be complete (that is, it is possible that some PRs that have been 12742 fixed are not listed here). 12743 12744 12745 For questions related to the use of GCC, please consult these web 12746 pages and the [40]GCC manuals. If that fails, the 12747 [41]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12748 web pages and the development of GCC are welcome on our developer 12749 list at [42]gcc (a] gcc.gnu.org. All of [43]our lists have public 12750 archives. 12751 12752 Copyright (C) [44]Free Software Foundation, Inc. Verbatim copying and 12753 distribution of this entire article is permitted in any medium, 12754 provided this notice is preserved. 12755 12756 These pages are [45]maintained by the GCC team. Last modified 12757 2025-01-31. 12758 12759 References 12760 12761 1. https://www.multiprecision.org/ 12762 2. https://gcc.gnu.org/install/prerequisites.html 12763 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 12764 4. https://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 12765 5. https://gcc.gnu.org/gcc-4.5/changes.html#x86 12766 6. https://gcc.gnu.org/PR30789 12767 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 12768 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 12769 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 12770 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 12771 11. https://gcc.gnu.org/gcc-4.5/changes.html#cplusplus 12772 12. https://gcc.gnu.org/gcc-4.5/cxx0x_status.html 12773 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 12774 14. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 12775 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 12776 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 12777 17. https://sourceware.org/gdb/wiki/STLSupport 12778 18. https://gcc.gnu.org/gcc-4.5/changes.html#Fortran 12779 19. https://gcc.gnu.org/gcc-4.4/changes.html 12780 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 12781 21. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 12782 22. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 12783 23. https://gcc.gnu.org/wiki/OOP 12784 24. https://gcc.gnu.org/gcc-4.5/changes.html#targets 12785 25. https://gcc.gnu.org/gcc-4.5/changes.html#x86 12786 26. https://gcc.gnu.org/gcc-4.5/changes.html#mep 12787 27. https://gcc.gnu.org/gcc-4.5/changes.html#rs6000 12788 28. https://gcc.gnu.org/gcc-4.5/changes.html#rx 12789 29. https://gcc.gnu.org/gcc-4.5/changes.html#os 12790 30. https://gcc.gnu.org/gcc-4.5/changes.html#windows 12791 31. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.1 12792 32. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 12793 33. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 12794 34. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.2 12795 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 12796 36. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.3 12797 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 12798 38. https://gcc.gnu.org/gcc-4.5/changes.html#GCC4.5.4 12799 39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 12800 40. https://gcc.gnu.org/onlinedocs/ 12801 41. mailto:gcc-help (a] gcc.gnu.org 12802 42. mailto:gcc (a] gcc.gnu.org 12803 43. https://gcc.gnu.org/lists.html 12804 44. https://www.fsf.org/ 12805 45. https://gcc.gnu.org/about.html 12806 ====================================================================== 12807 http://gcc.gnu.org/gcc-4.4/index.html 12808 12809 GCC 4.4 Release Series 12810 12811 This release series is no longer maintained. 12812 12813 March 13, 2012 12814 12815 The [1]GNU project and the GCC developers are pleased to announce the 12816 release of GCC 4.4.7. 12817 12818 This release is a bug-fix release, containing fixes for regressions in 12819 GCC 4.4.6 relative to previous releases of GCC. 12820 12821 Release History 12822 12823 GCC 4.4.7 12824 March 13, 2012 ([2]changes) 12825 12826 GCC 4.4.6 12827 April 16, 2011 ([3]changes) 12828 12829 GCC 4.4.5 12830 October 1, 2010 ([4]changes) 12831 12832 GCC 4.4.4 12833 April 29, 2010 ([5]changes) 12834 12835 GCC 4.4.3 12836 January 21, 2010 ([6]changes) 12837 12838 GCC 4.4.2 12839 October 15, 2009 ([7]changes) 12840 12841 GCC 4.4.1 12842 July 22, 2009 ([8]changes) 12843 12844 GCC 4.4.0 12845 April 21, 2009 ([9]changes) 12846 12847 References and Acknowledgements 12848 12849 GCC used to stand for the GNU C Compiler, but since the compiler 12850 supports several other languages aside from C, it now stands for the 12851 GNU Compiler Collection. 12852 12853 A list of [10]successful builds is updated as new information becomes 12854 available. 12855 12856 The GCC developers would like to thank the numerous people that have 12857 contributed new features, improvements, bug fixes, and other changes as 12858 well as test results to GCC. This [11]amazing group of volunteers is 12859 what makes GCC successful. 12860 12861 For additional information about GCC please refer to the [12]GCC 12862 project web site or contact the [13]GCC development mailing list. 12863 12864 To obtain GCC please use [14]our mirror sites or [15]our version 12865 control system. 12866 12867 12868 For questions related to the use of GCC, please consult these web 12869 pages and the [16]GCC manuals. If that fails, the 12870 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12871 web pages and the development of GCC are welcome on our developer 12872 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 12873 archives. 12874 12875 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 12876 distribution of this entire article is permitted in any medium, 12877 provided this notice is preserved. 12878 12879 These pages are [21]maintained by the GCC team. Last modified 12880 2022-10-26. 12881 12882 References 12883 12884 1. http://www.gnu.org/ 12885 2. https://gcc.gnu.org/gcc-4.4/changes.html 12886 3. https://gcc.gnu.org/gcc-4.4/changes.html 12887 4. https://gcc.gnu.org/gcc-4.4/changes.html 12888 5. https://gcc.gnu.org/gcc-4.4/changes.html 12889 6. https://gcc.gnu.org/gcc-4.4/changes.html 12890 7. https://gcc.gnu.org/gcc-4.4/changes.html 12891 8. https://gcc.gnu.org/gcc-4.4/changes.html 12892 9. https://gcc.gnu.org/gcc-4.4/changes.html 12893 10. https://gcc.gnu.org/gcc-4.4/buildstat.html 12894 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12895 12. https://gcc.gnu.org/index.html 12896 13. mailto:gcc (a] gcc.gnu.org 12897 14. https://gcc.gnu.org/mirrors.html 12898 15. https://gcc.gnu.org/git.html 12899 16. https://gcc.gnu.org/onlinedocs/ 12900 17. mailto:gcc-help (a] gcc.gnu.org 12901 18. mailto:gcc (a] gcc.gnu.org 12902 19. https://gcc.gnu.org/lists.html 12903 20. https://www.fsf.org/ 12904 21. https://gcc.gnu.org/about.html 12905 ====================================================================== 12906 http://gcc.gnu.org/gcc-4.4/changes.html 12907 12908 GCC 4.4 Release Series 12909 Changes, New Features, and Fixes 12910 12911 The latest release in the 4.4 release series is [1]GCC 4.4.7. 12912 12913 Caveats 12914 12915 * __builtin_stdarg_start has been completely removed from GCC. 12916 Support for <varargs.h> had been deprecated since GCC 4.0. Use 12917 __builtin_va_start as a replacement. 12918 * Some of the errors issued by the C++ front end that could be 12919 downgraded to warnings in previous releases by using -fpermissive 12920 are now warnings by default. They can be converted into errors by 12921 using -pedantic-errors. 12922 * Use of the cpp assertion extension will now emit a warning when 12923 -Wdeprecated or -pedantic is used. This extension has been 12924 deprecated for many years, but never warned about. 12925 * Packed bit-fields of type char were not properly bit-packed on many 12926 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 12927 causes an ABI change. For example there is no longer a 4-bit 12928 padding between field a and b in this structure: 12929 struct foo 12930 { 12931 char a:4; 12932 char b:8; 12933 } __attribute__ ((packed)); 12934 There is a new warning to help identify fields that are affected: 12935 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 12936 The warning can be disabled with -Wno-packed-bitfield-compat. 12937 * On ARM EABI targets, the C++ mangling of the va_list type has been 12938 changed to conform to the current revision of the EABI. This does 12939 not affect the libstdc++ library included with GCC. 12940 * The SCOUNT and POS bits of the MIPS DSP control register are now 12941 treated as global. Previous versions of GCC treated these fields as 12942 call-clobbered instead. 12943 * The MIPS port no longer recognizes the h asm constraint. It was 12944 necessary to remove this constraint in order to avoid generating 12945 unpredictable code sequences. 12946 One of the main uses of the h constraint was to extract the high 12947 part of a multiplication on 64-bit targets. For example: 12948 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 12949 You can now achieve the same effect using 128-bit types: 12950 typedef unsigned int uint128_t __attribute__((mode(TI))); 12951 result = ((uint128_t) x * y) >> 64; 12952 The second sequence is better in many ways. For example, if x and y 12953 are constants, the compiler can perform the multiplication at 12954 compile time. If x and y are not constants, the compiler can 12955 schedule the runtime multiplication better than it can schedule an 12956 asm statement. 12957 * Support for a number of older systems and recently unmaintained or 12958 untested target ports of GCC has been declared obsolete in GCC 4.4. 12959 Unless there is activity to revive them, the next release of GCC 12960 will have their sources permanently removed. 12961 The following ports for individual systems on particular 12962 architectures have been obsoleted: 12963 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 12964 m68k-*-aout*) 12965 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 12966 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 12967 sh-*-*). This does not affect other more specific targets 12968 using the COFF object format on those architectures, or the 12969 more specific H8300 and SH targets (h8300-*-rtems*, 12970 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 12971 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 12972 + 2BSD on PDP-11 (pdp11-*-bsd) 12973 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 12974 powerpc-ibm-aix4.[12]*) 12975 + Tuning support for Itanium1 (Merced) variants. Note that code 12976 tuned for Itanium2 should also run correctly on Itanium1. 12977 * The protoize and unprotoize utilities have been obsoleted and will 12978 be removed in GCC 4.5. These utilities have not been installed by 12979 default since GCC 3.0. 12980 * Support has been removed for all the [2]configurations obsoleted in 12981 GCC 4.3. 12982 * Unknown -Wno-* options are now silently ignored by GCC if no other 12983 diagnostics are issued. If other diagnostics are issued, then GCC 12984 warns about the unknown options. 12985 * More information on porting to GCC 4.4 from previous versions of 12986 GCC can be found in the [3]porting guide for this release. 12987 12988 General Optimizer Improvements 12989 12990 * A new command-line switch -findirect-inlining has been added. When 12991 turned on it allows the inliner to also inline indirect calls that 12992 are discovered to have known targets at compile time thanks to 12993 previous inlining. 12994 * A new command-line switch -ftree-switch-conversion has been added. 12995 This new pass turns simple initializations of scalar variables in 12996 switch statements into initializations from a static array, given 12997 that all the values are known at compile time and the ratio between 12998 the new array size and the original switch branches does not exceed 12999 the parameter --param switch-conversion-max-branch-ratio (default 13000 is eight). 13001 * A new command-line switch -ftree-builtin-call-dce has been added. 13002 This optimization eliminates unnecessary calls to certain builtin 13003 functions when the return value is not used, in cases where the 13004 calls can not be eliminated entirely because the function may set 13005 errno. This optimization is on by default at -O2 and above. 13006 * A new command-line switch -fconserve-stack directs the compiler to 13007 minimize stack usage even if it makes the generated code slower. 13008 This affects inlining decisions. 13009 * When the assembler supports it, the compiler will now emit unwind 13010 information using assembler .cfi directives. This makes it possible 13011 to use such directives in inline assembler code. The new option 13012 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 13013 directives. 13014 * The [4]Graphite branch has been merged. This merge has brought in a 13015 new framework for loop optimizations based on a polyhedral 13016 intermediate representation. These optimizations apply to all the 13017 languages supported by GCC. The following new code transformations 13018 are available in GCC 4.4: 13019 + -floop-interchange performs loop interchange transformations 13020 on loops. Interchanging two nested loops switches the inner 13021 and outer loops. For example, given a loop like: 13022 DO J = 1, M 13023 DO I = 1, N 13024 A(J, I) = A(J, I) * C 13025 ENDDO 13026 ENDDO 13027 13028 loop interchange will transform the loop as if the user had 13029 written: 13030 DO I = 1, N 13031 DO J = 1, M 13032 A(J, I) = A(J, I) * C 13033 ENDDO 13034 ENDDO 13035 13036 which can be beneficial when N is larger than the caches, 13037 because in Fortran, the elements of an array are stored in 13038 memory contiguously by column, and the original loop iterates 13039 over rows, potentially creating at each access a cache miss. 13040 + -floop-strip-mine performs loop strip mining transformations 13041 on loops. Strip mining splits a loop into two nested loops. 13042 The outer loop has strides equal to the strip size and the 13043 inner loop has strides of the original loop within a strip. 13044 For example, given a loop like: 13045 DO I = 1, N 13046 A(I) = A(I) + C 13047 ENDDO 13048 13049 loop strip mining will transform the loop as if the user had 13050 written: 13051 DO II = 1, N, 4 13052 DO I = II, min (II + 3, N) 13053 A(I) = A(I) + C 13054 ENDDO 13055 ENDDO 13056 13057 + -floop-block performs loop blocking transformations on loops. 13058 Blocking strip mines each loop in the loop nest such that the 13059 memory accesses of the element loops fit inside caches. For 13060 example, given a loop like: 13061 DO I = 1, N 13062 DO J = 1, M 13063 A(J, I) = B(I) + C(J) 13064 ENDDO 13065 ENDDO 13066 13067 loop blocking will transform the loop as if the user had 13068 written: 13069 DO II = 1, N, 64 13070 DO JJ = 1, M, 64 13071 DO I = II, min (II + 63, N) 13072 DO J = JJ, min (JJ + 63, M) 13073 A(J, I) = B(I) + C(J) 13074 ENDDO 13075 ENDDO 13076 ENDDO 13077 ENDDO 13078 13079 which can be beneficial when M is larger than the caches, 13080 because the innermost loop will iterate over a smaller amount 13081 of data that can be kept in the caches. 13082 * A new register allocator has replaced the old one. It is called 13083 integrated register allocator (IRA) because coalescing, register 13084 live range splitting, and hard register preferencing are done 13085 on-the-fly during coloring. It also has better integration with the 13086 reload pass. IRA is a regional register allocator which uses modern 13087 Chaitin-Briggs coloring instead of Chow's priority coloring used in 13088 the old register allocator. More info about IRA internals and 13089 options can be found in the GCC manuals. 13090 * A new instruction scheduler and software pipeliner, based on the 13091 selective scheduling approach, has been added. The new pass 13092 performs instruction unification, register renaming, substitution 13093 through register copies, and speculation during scheduling. The 13094 software pipeliner is able to pipeline non-countable loops. The new 13095 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 13096 it is available for the Intel Itanium platform working by default 13097 as the second scheduling pass (after register allocation) at the 13098 -O3 optimization level. 13099 * When using -fprofile-generate with a multi-threaded program, the 13100 profile counts may be slightly wrong due to race conditions. The 13101 new -fprofile-correction option directs the compiler to apply 13102 heuristics to smooth out the inconsistencies. By default the 13103 compiler will give an error message when it finds an inconsistent 13104 profile. 13105 * The new -fprofile-dir=PATH option permits setting the directory 13106 where profile data files are stored when using -fprofile-generate 13107 and friends, and the directory used when reading profile data files 13108 using -fprofile-use and friends. 13109 13110 New warning options 13111 13112 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 13113 warning if any stack frame is larger than NUMBER bytes. This may be 13114 used to help ensure that code fits within a limited amount of stack 13115 space. 13116 * The command-line option -Wlarger-than-N is now written as 13117 -Wlarger-than=N and the old form is deprecated. 13118 * The new -Wno-mudflap option disables warnings about constructs 13119 which can not be instrumented when using -fmudflap. 13120 13121 New Languages and Language specific improvements 13122 13123 * Version 3.0 of the OpenMP specification is now supported for the C, 13124 C++, and Fortran compilers. 13125 * New character data types, per [5]TR 19769: New character types in 13126 C, are now supported for the C compiler in -std=gnu99 mode, as 13127 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 13128 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 13129 13130 C family 13131 13132 * A new optimize attribute was added to allow programmers to change 13133 the optimization level and particular optimization options for an 13134 individual function. You can also change the optimization options 13135 via the GCC optimize pragma for functions defined after the pragma. 13136 The GCC push_options pragma and the GCC pop_options pragma allow 13137 you temporarily save and restore the options used. The GCC 13138 reset_options pragma restores the options to what was specified on 13139 the command line. 13140 * Uninitialized warnings do not require enabling optimization 13141 anymore, that is, -Wuninitialized can be used together with -O0. 13142 Nonetheless, the warnings given by -Wuninitialized will probably be 13143 more accurate if optimization is enabled. 13144 * -Wparentheses now warns about expressions such as (!x | y) and (!x 13145 & y). Using explicit parentheses, such as in ((!x) | y), silences 13146 this warning. 13147 * -Wsequence-point now warns within if, while,do while and for 13148 conditions, and within for begin/end expressions. 13149 * A new option -dU is available to dump definitions of preprocessor 13150 macros that are tested or expanded. 13151 13152 C++ 13153 13154 * [6]Improved experimental support for the upcoming ISO C++ standard, 13155 C++0x. Including support for auto, inline namespaces, generalized 13156 initializer lists, defaulted and deleted functions, new character 13157 types, and scoped enums. 13158 * Those errors that may be downgraded to warnings to build legacy 13159 code now mention -fpermissive when -fdiagnostics-show-option is 13160 enabled. 13161 * -Wconversion now warns if the result of a static_cast to enumeral 13162 type is unspecified because the value is outside the range of the 13163 enumeral type. 13164 * -Wuninitialized now warns if a non-static reference or non-static 13165 const member appears in a class without constructors. 13166 * G++ now properly implements value-initialization, so objects with 13167 an initializer of () and an implicitly defined default constructor 13168 will be zero-initialized before the default constructor is called. 13169 13170 Runtime Library (libstdc++) 13171 13172 * Improved experimental support for the upcoming ISO C++ standard, 13173 C++0x, including: 13174 + Support for <chrono>, <condition_variable>, <cstdatomic>, 13175 <forward_list>, <initializer_list>, <mutex>, <ratio>, 13176 <system_error>, and <thread>. 13177 + unique_ptr, <algorithm> additions, exception propagation, and 13178 support for the new character types in <string> and <limits>. 13179 + Existing facilities now exploit initializer lists, defaulted 13180 and deleted functions, and the newly implemented core C++0x 13181 features. 13182 + Some standard containers are more efficient together with 13183 stateful allocators, i.e., no allocator is constructed on the 13184 fly at element construction time. 13185 * Experimental support for non-standard pointer types in containers. 13186 * The long standing libstdc++/30928 has been fixed for targets 13187 running glibc 2.10 or later. 13188 * As usual, many small and larger bug fixes, in particular quite a 13189 few corner cases in <locale>. 13190 13191 Fortran 13192 13193 * GNU Fortran now employs libcpp directly instead of using cc1 as an 13194 external preprocessor. The [7]-cpp option was added to allow manual 13195 invocation of the preprocessor without relying on filename 13196 extensions. 13197 * The [8]-Warray-temporaries option warns about array temporaries 13198 generated by the compiler, as an aid to optimization. 13199 * The [9]-fcheck-array-temporaries option has been added, printing a 13200 notification at run time, when an array temporary had to be created 13201 for an function argument. Contrary to -Warray-temporaries the 13202 warning is only printed if the array is noncontiguous. 13203 * Improved generation of DWARF debugging symbols 13204 * If using an intrinsic not part of the selected standard (via -std= 13205 and -fall-intrinsics) gfortran will now treat it as if this 13206 procedure were declared EXTERNAL and try to link to a user-supplied 13207 procedure. -Wintrinsics-std will warn whenever this happens. The 13208 now-useless option -Wnonstd-intrinsic was removed. 13209 * The flag -falign-commons has been added to control the alignment of 13210 variables in COMMON blocks, which is enabled by default in line 13211 with previous GCC version. Using -fno-align-commons one can force 13212 commons to be contiguous in memory as required by the Fortran 13213 standard, however, this slows down the memory access. The option 13214 -Walign-commons, which is enabled by default, warns when padding 13215 bytes were added for alignment. The proper solution is to sort the 13216 common objects by decreasing storage size, which avoids the 13217 alignment problems. 13218 * Fortran 2003 support has been extended: 13219 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 13220 now supported (except internal reads from/writes to wide 13221 strings). [10]-fbackslash now supports also \unnnn and 13222 \Unnnnnnnn to enter Unicode characters. 13223 + Asynchronous I/O (implemented as synchronous I/O) and the 13224 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 13225 are now supported in I/O statements. 13226 + Support for Fortran 2003 structure constructors and for array 13227 constructor with typespec has been added. 13228 + Procedure Pointers (but not yet as component in derived types 13229 and as function results) are now supported. 13230 + Abstract types, type extension, and type-bound procedures 13231 (both PROCEDURE and GENERIC but not as operators). Note: As 13232 CLASS/polymorphyic types are not implemented, type-bound 13233 procedures with PASS accept as non-standard extension TYPE 13234 arguments. 13235 * Fortran 2008 support has been added: 13236 + The -std=f2008 option and support for the file extensions 13237 .f2008 and .F2008 has been added. 13238 + The g0 format descriptor is now supported. 13239 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 13240 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 13241 are now available (some of them existed as GNU extension 13242 before). Note: The hyperbolic functions are not yet supporting 13243 complex arguments and the three- argument version of BESSEL_*N 13244 is not available. 13245 + The bit intrinsics LEADZ and TRAILZ have been added. 13246 13247 Java (GCJ) 13248 13249 Ada 13250 13251 * The Ada runtime now supports multilibs on many platforms including 13252 x86_64, SPARC and PowerPC. Their build is enabled by default. 13253 13254 [11]New Targets and Target Specific Improvements 13255 13256 ARM 13257 13258 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 13259 Cortex-R4F processors and has many other improvements to 13260 optimization for ARM processors. 13261 * GCC now supports the VFPv3 variant with 16 double-precision 13262 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 13263 renamed to -mfpu=vfpv3. 13264 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 13265 erratum on Cortex-M3 processors. 13266 * GCC now supports the __sync_* atomic operations for ARM EABI 13267 GNU/Linux. 13268 * The section anchors optimization is now enabled by default when 13269 optimizing for ARM. 13270 * GCC now uses a new EABI-compatible profiling interface for EABI 13271 targets. This requires a function __gnu_mcount_nc, which is 13272 provided by GNU libc versions 2.8 and later. 13273 13274 AVR 13275 13276 * The -mno-tablejump option has been deprecated because it has the 13277 same effect as the -fno-jump-tables option. 13278 * Added support for these new AVR devices: 13279 + ATA6289 13280 + ATtiny13A 13281 + ATtiny87 13282 + ATtiny167 13283 + ATtiny327 13284 + ATmega8C1 13285 + ATmega16C1 13286 + ATmega32C1 13287 + ATmega8M1 13288 + ATmega16M1 13289 + ATmega32M1 13290 + ATmega32U4 13291 + ATmega16HVB 13292 + ATmega4HVD 13293 + ATmega8HVD 13294 + ATmega64C1 13295 + ATmega64M1 13296 + ATmega16U4 13297 + ATmega32U6 13298 + ATmega128RFA1 13299 + AT90PWM81 13300 + AT90SCR100 13301 + M3000F 13302 + M3000S 13303 + M3001B 13304 13305 IA-32/x86-64 13306 13307 * Support for Intel AES built-in functions and code generation is 13308 available via -maes. 13309 * Support for Intel PCLMUL built-in function and code generation is 13310 available via -mpclmul. 13311 * Support for Intel AVX built-in functions and code generation is 13312 available via -mavx. 13313 * Automatically align the stack for local variables with alignment 13314 requirement. 13315 * GCC can now utilize the SVML library for vectorizing calls to a set 13316 of C99 functions if -mveclibabi=svml is specified and you link to 13317 an SVML ABI compatible library. 13318 * On x86-64, the ABI has been changed in the following cases to 13319 conform to the x86-64 ABI: 13320 + Passing/returning structures with flexible array member: 13321 struct foo 13322 { 13323 int i; 13324 int flex[]; 13325 }; 13326 + Passing/returning structures with complex float member: 13327 struct foo 13328 { 13329 int i; 13330 __complex__ float f; 13331 }; 13332 + Passing/returning unions with long double member: 13333 union foo 13334 { 13335 int x; 13336 long double ld; 13337 }; 13338 Code built with previous versions of GCC that uses any of these is 13339 not compatible with code built with GCC 4.4.0 or later. 13340 * A new target attribute was added to allow programmers to change the 13341 target options like -msse2 or -march=k8 for an individual function. 13342 You can also change the target options via the GCC target pragma 13343 for functions defined after the pragma. 13344 * GCC can now be configured with options --with-arch-32, 13345 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 13346 --with-tune-64 to control the default optimization separately for 13347 32-bit and 64-bit modes. 13348 13349 IA-32/IA64 13350 13351 * Support for __float128 (TFmode) IEEE quad type and corresponding 13352 TCmode IEEE complex quad type is available via the soft-fp library 13353 on IA-32/IA64 targets. This includes basic arithmetic operations 13354 (addition, subtraction, negation, multiplication and division) on 13355 __float128 real and TCmode complex values, the full set of IEEE 13356 comparisons between __float128 values, conversions to and from 13357 float, double and long double floating point types, as well as 13358 conversions to and from signed or unsigned integer, signed or 13359 unsigned long integer and signed or unsigned quad (TImode, IA64 13360 only) integer types. Additionally, all operations generate the full 13361 set of IEEE exceptions and support the full set of IEEE rounding 13362 modes. 13363 13364 M68K/ColdFire 13365 13366 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 13367 processors. (Scheduling support for ColdFire V2 processors was 13368 added in GCC 4.3.) 13369 * GCC now supports the -mxgot option to support programs requiring 13370 many GOT entries on ColdFire. 13371 * The m68k-*-linux-gnu target now builds multilibs by default. 13372 13373 MIPS 13374 13375 * MIPS Technologies have extended the original MIPS SVR4 ABI to 13376 include support for procedure linkage tables (PLTs) and copy 13377 relocations. These extensions allow GNU/Linux executables to use a 13378 significantly more efficient code model than the one defined by the 13379 original ABI. 13380 GCC support for this code model is available via a new command-line 13381 option, -mplt. There is also a new configure-time option, 13382 --with-mips-plt, to make -mplt the default. 13383 The new code model requires support from the assembler, the linker, 13384 and the runtime C library. This support is available in binutils 13385 2.19 and GLIBC 2.9. 13386 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 13387 and 32-bit GNU/Linux shared libraries. This feature requires GNU 13388 binutils 2.19 or above. 13389 * Support for RMI's XLR processor is now available through the 13390 -march=xlr and -mtune=xlr options. 13391 * 64-bit targets can now perform 128-bit multiplications inline, 13392 instead of relying on a libgcc function. 13393 * Native GNU/Linux toolchains now support -march=native and 13394 -mtune=native, which select the host processor. 13395 * GCC now supports the R10K, R12K, R14K and R16K processors. The 13396 canonical -march= and -mtune= names for these processors are 13397 r10000, r12000, r14000 and r16000 respectively. 13398 * GCC can now work around the side effects of speculative execution 13399 on R10K processors. Please see the documentation of the 13400 -mr10k-cache-barrier option for details. 13401 * Support for the MIPS64 Release 2 instruction set has been added. 13402 The option -march=mips64r2 enables generation of these 13403 instructions. 13404 * GCC now supports Cavium Networks' Octeon processor. This support is 13405 available through the -march=octeon and -mtune=octeon options. 13406 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 13407 canonical -march= and -mtune= names for these processors are 13408 loongson2e and loongson2f. 13409 13410 [12]picochip 13411 13412 Picochip is a 16-bit processor. A typical picoChip contains over 250 13413 small cores, each with small amounts of memory. There are three 13414 processor variants (STAN, MEM and CTRL) with different instruction sets 13415 and memory configurations and they can be chosen using the -mae option. 13416 13417 This port is intended to be a "C" only port. 13418 13419 Power Architecture and PowerPC 13420 13421 * GCC now supports the e300c2, e300c3 and e500mc processors. 13422 * GCC now supports Xilinx processors with a single-precision FPU. 13423 * Decimal floating point is now supported for e500 processors. 13424 13425 S/390, zSeries and System z9/z10 13426 13427 * Support for the IBM System z10 EC/BC processor has been added. When 13428 using the -march=z10 option, the compiler will generate code making 13429 use of instructions provided by the General-Instruction-Extension 13430 Facility and the Execute-Extension Facility. 13431 13432 VxWorks 13433 13434 * GCC now supports the thread-local storage mechanism used on 13435 VxWorks. 13436 13437 Xtensa 13438 13439 * GCC now supports thread-local storage (TLS) for Xtensa processor 13440 configurations that include the Thread Pointer option. TLS also 13441 requires support from the assembler and linker; this support is 13442 provided in the GNU binutils beginning with version 2.19. 13443 13444 Documentation improvements 13445 13446 Other significant improvements 13447 13448 [13]GCC 4.4.1 13449 13450 This is the [14]list of problem reports (PRs) from GCC's bug tracking 13451 system that are known to be fixed in the 4.4.1 release. This list might 13452 not be complete (that is, it is possible that some PRs that have been 13453 fixed are not listed here). 13454 13455 [15]GCC 4.4.2 13456 13457 This is the [16]list of problem reports (PRs) from GCC's bug tracking 13458 system that are known to be fixed in the 4.4.2 release. This list might 13459 not be complete (that is, it is possible that some PRs that have been 13460 fixed are not listed here). 13461 13462 [17]GCC 4.4.3 13463 13464 This is the [18]list of problem reports (PRs) from GCC's bug tracking 13465 system that are known to be fixed in the 4.4.3 release. This list might 13466 not be complete (that is, it is possible that some PRs that have been 13467 fixed are not listed here). 13468 13469 [19]GCC 4.4.4 13470 13471 This is the [20]list of problem reports (PRs) from GCC's bug tracking 13472 system that are known to be fixed in the 4.4.4 release. This list might 13473 not be complete (that is, it is possible that some PRs that have been 13474 fixed are not listed here). 13475 13476 [21]GCC 4.4.5 13477 13478 This is the [22]list of problem reports (PRs) from GCC's bug tracking 13479 system that are known to be fixed in the 4.4.5 release. This list might 13480 not be complete (that is, it is possible that some PRs that have been 13481 fixed are not listed here). 13482 13483 [23]GCC 4.4.6 13484 13485 This is the [24]list of problem reports (PRs) from GCC's bug tracking 13486 system that are known to be fixed in the 4.4.6 release. This list might 13487 not be complete (that is, it is possible that some PRs that have been 13488 fixed are not listed here). 13489 13490 [25]GCC 4.4.7 13491 13492 This is the [26]list of problem reports (PRs) from GCC's bug tracking 13493 system that are known to be fixed in the 4.4.7 release. This list might 13494 not be complete (that is, it is possible that some PRs that have been 13495 fixed are not listed here). 13496 13497 13498 For questions related to the use of GCC, please consult these web 13499 pages and the [27]GCC manuals. If that fails, the 13500 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13501 web pages and the development of GCC are welcome on our developer 13502 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 13503 archives. 13504 13505 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 13506 distribution of this entire article is permitted in any medium, 13507 provided this notice is preserved. 13508 13509 These pages are [32]maintained by the GCC team. Last modified 13510 2025-01-31. 13511 13512 References 13513 13514 1. https://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 13515 2. https://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 13516 3. https://gcc.gnu.org/gcc-4.4/porting_to.html 13517 4. https://gcc.gnu.org/wiki/Graphite 13518 5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 13519 6. https://gcc.gnu.org/gcc-4.4/cxx0x_status.html 13520 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 13521 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 13522 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 13523 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 13524 11. https://gcc.gnu.org/gcc-4.4/changes.html#targets 13525 12. https://gcc.gnu.org/gcc-4.4/changes.html#picochip 13526 13. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.1 13527 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 13528 15. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.2 13529 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 13530 17. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.3 13531 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 13532 19. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.4 13533 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 13534 21. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.5 13535 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 13536 23. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.6 13537 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 13538 25. https://gcc.gnu.org/gcc-4.4/changes.html#GCC4.4.7 13539 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 13540 27. https://gcc.gnu.org/onlinedocs/ 13541 28. mailto:gcc-help (a] gcc.gnu.org 13542 29. mailto:gcc (a] gcc.gnu.org 13543 30. https://gcc.gnu.org/lists.html 13544 31. https://www.fsf.org/ 13545 32. https://gcc.gnu.org/about.html 13546 ====================================================================== 13547 http://gcc.gnu.org/gcc-4.3/index.html 13548 13549 GCC 4.3 Release Series 13550 13551 (This release series is no longer supported.) 13552 13553 Jun 27, 2011 13554 13555 The [1]GNU project and the GCC developers are pleased to announce the 13556 release of GCC 4.3.6. 13557 13558 This release is a bug-fix release, containing fixes for regressions in 13559 GCC 4.3.5 relative to previous releases of GCC. 13560 13561 Release History 13562 13563 GCC 4.3.6 13564 Jun 27, 2011 ([2]changes) 13565 13566 GCC 4.3.5 13567 May 22, 2010 ([3]changes) 13568 13569 GCC 4.3.4 13570 August 4, 2009 ([4]changes) 13571 13572 GCC 4.3.3 13573 January 24, 2009 ([5]changes) 13574 13575 GCC 4.3.2 13576 August 27, 2008 ([6]changes) 13577 13578 GCC 4.3.1 13579 June 6, 2008 ([7]changes) 13580 13581 GCC 4.3.0 13582 March 5, 2008 ([8]changes) 13583 13584 References and Acknowledgements 13585 13586 GCC used to stand for the GNU C Compiler, but since the compiler 13587 supports several other languages aside from C, it now stands for the 13588 GNU Compiler Collection. 13589 13590 A list of [9]successful builds is updated as new information becomes 13591 available. 13592 13593 The GCC developers would like to thank the numerous people that have 13594 contributed new features, improvements, bug fixes, and other changes as 13595 well as test results to GCC. This [10]amazing group of volunteers is 13596 what makes GCC successful. 13597 13598 For additional information about GCC please refer to the [11]GCC 13599 project web site or contact the [12]GCC development mailing list. 13600 13601 To obtain GCC please use [13]our mirror sites or [14]our version 13602 control system. 13603 13604 13605 For questions related to the use of GCC, please consult these web 13606 pages and the [15]GCC manuals. If that fails, the 13607 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13608 web pages and the development of GCC are welcome on our developer 13609 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 13610 archives. 13611 13612 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 13613 distribution of this entire article is permitted in any medium, 13614 provided this notice is preserved. 13615 13616 These pages are [20]maintained by the GCC team. Last modified 13617 2022-10-26. 13618 13619 References 13620 13621 1. http://www.gnu.org/ 13622 2. https://gcc.gnu.org/gcc-4.3/changes.html 13623 3. https://gcc.gnu.org/gcc-4.3/changes.html 13624 4. https://gcc.gnu.org/gcc-4.3/changes.html 13625 5. https://gcc.gnu.org/gcc-4.3/changes.html 13626 6. https://gcc.gnu.org/gcc-4.3/changes.html 13627 7. https://gcc.gnu.org/gcc-4.3/changes.html 13628 8. https://gcc.gnu.org/gcc-4.3/changes.html 13629 9. https://gcc.gnu.org/gcc-4.3/buildstat.html 13630 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13631 11. https://gcc.gnu.org/index.html 13632 12. mailto:gcc (a] gcc.gnu.org 13633 13. https://gcc.gnu.org/mirrors.html 13634 14. https://gcc.gnu.org/git.html 13635 15. https://gcc.gnu.org/onlinedocs/ 13636 16. mailto:gcc-help (a] gcc.gnu.org 13637 17. mailto:gcc (a] gcc.gnu.org 13638 18. https://gcc.gnu.org/lists.html 13639 19. https://www.fsf.org/ 13640 20. https://gcc.gnu.org/about.html 13641 ====================================================================== 13642 http://gcc.gnu.org/gcc-4.3/changes.html 13643 13644 GCC 4.3 Release Series 13645 Changes, New Features, and Fixes 13646 13647 The latest release in the 4.3 release series is [1]GCC 4.3.5. 13648 13649 Caveats 13650 13651 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 13652 various front-end languages it supports. See the [4]prerequisites 13653 page for version requirements. 13654 * ColdFire targets now treat long double as having the same format as 13655 double. In earlier versions of GCC, they used the 68881 long double 13656 format instead. 13657 * The m68k-uclinux target now uses the same calling conventions as 13658 m68k-linux-gnu. You can select the original calling conventions by 13659 configuring for m68k-uclinuxoldabi instead. Note that 13660 m68k-uclinuxoldabi also retains the original 80-bit long double on 13661 ColdFire targets. 13662 * The -fforce-mem option has been removed because it has had no 13663 effect in the last few GCC releases. 13664 * The i386 -msvr3-shlib option has been removed since it is no longer 13665 used. 13666 * Fastcall for i386 has been changed not to pass aggregate arguments 13667 in registers, following Microsoft compilers. 13668 * Support for the AOF assembler has been removed from the ARM back 13669 end; this affects only the targets arm-semi-aof and armel-semi-aof, 13670 which are no longer recognized. We removed these targets without a 13671 deprecation period because we discovered that they have been 13672 unusable since GCC 4.0.0. 13673 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 13674 has been removed. This support had been deprecated since GCC 4.0.0. 13675 * Support for a number of older systems and recently unmaintained or 13676 untested target ports of GCC has been declared obsolete in GCC 4.3. 13677 Unless there is activity to revive them, the next release of GCC 13678 will have their sources permanently removed. 13679 All GCC ports for the following processor architectures have been 13680 declared obsolete: 13681 + Morpho MT (mt-*) 13682 The following aliases for processor architectures have been 13683 declared obsolete. Users should use the indicated generic target 13684 names instead, with compile-time options such as -mcpu or 13685 configure-time options such as --with-cpu to control the 13686 configuration more precisely. 13687 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 13688 instead). 13689 + parisc*-*-* (use hppa*-*-* instead). 13690 + m680[012]0-*-* (use m68k-*-* instead). 13691 All GCC ports for the following operating systems have been 13692 declared obsolete: 13693 + BeOS (*-*-beos*) 13694 + kaOS (*-*-kaos*) 13695 + GNU/Linux using the a.out object format (*-*-linux*aout*) 13696 + GNU/Linux using version 1 of the GNU C Library 13697 (*-*-linux*libc1*) 13698 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 13699 *-*-solaris2.[0-6].*) 13700 + Miscellaneous System V (*-*-sysv*) 13701 + WindISS (*-*-windiss*) 13702 Also, those for some individual systems on particular architectures 13703 have been obsoleted: 13704 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 13705 + CRIS with a.out object format (cris-*-aout) 13706 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 13707 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 13708 + PRO on PA-RISC (hppa1.1-*-pro*) 13709 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 13710 i[34567]86-sequent-sysv4*) 13711 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 13712 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 13713 was previously [5]removed in 2001, leaving only the support 13714 for UWIN as a target now being deprecated) 13715 + ChorusOS on PowerPC (powerpc-*-chorusos*) 13716 + All VAX configurations apart from NetBSD and OpenBSD 13717 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 13718 * The [6]-Wconversion option has been modified. Its purpose now is to 13719 warn for implicit conversions that may alter a value. This new 13720 behavior is available for both C and C++. Warnings about 13721 conversions between signed and unsigned integers can be disabled by 13722 using -Wno-sign-conversion. In C++, they are disabled by default 13723 unless -Wsign-conversion is explicitly requested. The old behavior 13724 of -Wconversion, that is, warn for prototypes causing a type 13725 conversion that is different from what would happen to the same 13726 argument in the absence of a prototype, has been moved to a new 13727 option -Wtraditional-conversion, which is only available for C. 13728 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 13729 been removed because they were deprecated for more than 3 GCC major 13730 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 13731 -mtune=pentiumpro as a replacement. 13732 * The -funsafe-math-optimizations option now automatically turns on 13733 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 13734 reassociation and thus may introduce or remove traps. 13735 * The -ftree-vectorize option is now on by default under -O3. In 13736 order to generate code for a SIMD extension, it has to be enabled 13737 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 13738 i?86 and x86_64. 13739 * More information on porting to GCC 4.3 from previous versions of 13740 GCC can be found in the [7]porting guide for this release. 13741 13742 General Optimizer Improvements 13743 13744 * The GCC middle-end has been integrated with the MPFR library. This 13745 allows GCC to evaluate and replace at compile-time calls to 13746 built-in math functions having constant arguments with their 13747 mathematically equivalent results. In making use of MPFR, GCC can 13748 generate correct results regardless of the math library 13749 implementation or floating point precision of the host platform. 13750 This also allows GCC to generate identical results regardless of 13751 whether one compiles in native or cross-compile configurations to a 13752 particular target. The following built-in functions take advantage 13753 of this new capability: acos, acosh, asin, asinh, atan2, atan, 13754 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 13755 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 13756 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 13757 tan, tanh, tgamma, y0, y1 and yn. The float and long double 13758 variants of these functions (e.g. sinf and sinl) are also handled. 13759 The sqrt and cabs functions with constant arguments were already 13760 optimized in prior GCC releases. Now they also use MPFR. 13761 * A new forward propagation pass on RTL was added. The new pass 13762 replaces several slower transformations, resulting in compile-time 13763 improvements as well as better code generation in some cases. 13764 * A new command-line switch -frecord-gcc-switches has been added to 13765 GCC, although it is only enabled for some targets. The switch 13766 causes the command line that was used to invoke the compiler to be 13767 recorded into the object file that is being created. The exact 13768 format of this recording is target and binary file format 13769 dependent, but it usually takes the form of a note section 13770 containing ASCII text. The switch is related to the -fverbose-asm 13771 switch, but that one only records the information in the assembler 13772 output file as comments, so the information never reaches the 13773 object file. 13774 * The inliner heuristic is now aware of stack frame consumption. New 13775 command-line parameters --param large-stack-frame and --param 13776 large-stack-frame-growth can be used to limit stack frame size 13777 growth caused by inlining. 13778 * During feedback directed optimizations, the expected block size the 13779 memcpy, memset and bzero functions operate on is discovered and for 13780 cases of commonly used small sizes, specialized inline code is 13781 generated. 13782 * __builtin_expect no longer requires its argument to be a compile 13783 time constant. 13784 * Interprocedural optimization was reorganized to work on functions 13785 in SSA form. This enables more precise and cheaper dataflow 13786 analysis and makes writing interprocedural optimizations easier. 13787 The following improvements have been implemented on top of this 13788 framework: 13789 + Pre-inline optimization: Selected local optimization passes 13790 are run before the inliner (and other interprocedural passes) 13791 are executed. This significantly improves the accuracy of code 13792 growth estimates used by the inliner and reduces the overall 13793 memory footprint for large compilation units. 13794 + Early inlining (a simple bottom-up inliner pass inlining only 13795 functions whose body is smaller than the expected call 13796 overhead) is now executed with the early optimization passes, 13797 thus inlining already optimized function bodies into an 13798 unoptimized function that is subsequently optimized by early 13799 optimizers. This enables the compiler to quickly eliminate 13800 abstraction penalty in C++ programs. 13801 + Interprocedural constant propagation now operate on SSA form 13802 increasing accuracy of the analysis. 13803 * A new internal representation for GIMPLE statements has been 13804 contributed, resulting in compile-time memory savings. 13805 * The vectorizer was enhanced to support vectorization of outer 13806 loops, intra-iteration parallelism (loop-aware SLP), vectorization 13807 of strided accesses and loops with multiple data-types. Run-time 13808 dependency testing using loop versioning was added. The cost model, 13809 turned on by -fvect-cost-model, was developed. 13810 13811 New Languages and Language specific improvements 13812 13813 * We have added new command-line options 13814 -finstrument-functions-exclude-function-list and 13815 -finstrument-functions-exclude-file-list. They provide more control 13816 over which functions are annotated by the -finstrument-functions 13817 option. 13818 13819 C family 13820 13821 * Implicit conversions between generic vector types are now only 13822 permitted when the two vectors in question have the same number of 13823 elements and compatible element types. (Note that the restriction 13824 involves compatible element types, not implicitly-convertible 13825 element types: thus, a vector type with element type int may not be 13826 implicitly converted to a vector type with element type unsigned 13827 int.) This restriction, which is in line with specifications for 13828 SIMD architectures such as AltiVec, may be relaxed using the flag 13829 -flax-vector-conversions. This flag is intended only as a 13830 compatibility measure and should not be used for new code. 13831 * -Warray-bounds has been added and is now enabled by default for 13832 -Wall . It produces warnings for array subscripts that can be 13833 determined at compile time to be always out of bounds. 13834 -Wno-array-bounds will disable the warning. 13835 * The constructor and destructor function attributes now accept 13836 optional priority arguments which control the order in which the 13837 constructor and destructor functions are run. 13838 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 13839 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 13840 -Wignored-qualifiers have been added for finer control of the 13841 diverse warnings enabled by -Wextra. 13842 * A new function attribute alloc_size has been added to mark up 13843 malloc style functions. For constant sized allocations this can be 13844 used to find out the size of the returned pointer using the 13845 __builtin_object_size() function for buffer overflow checking and 13846 similar. This supplements the already built-in malloc and calloc 13847 constant size handling. 13848 * Integer constants written in binary are now supported as a GCC 13849 extension. They consist of a prefix 0b or 0B, followed by a 13850 sequence of 0 and 1 digits. 13851 * A new predefined macro __COUNTER__ has been added. It expands to 13852 sequential integral values starting from 0. In conjunction with the 13853 ## operator, this provides a convenient means to generate unique 13854 identifiers. 13855 * A new command-line option -fdirectives-only has been added. It 13856 enables a special preprocessing mode which improves the performance 13857 of applications like distcc and ccache. 13858 * Fixed-point data types and operators have been added. They are 13859 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 13860 Currently, only MIPS targets are supported. 13861 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 13862 N1241, is now supported as a GCC extension to C for targets 13863 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 13864 and x86_64-*-linux-gnu. The feature introduces new data types 13865 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 13866 DD, and DL. 13867 13868 C++ 13869 13870 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 13871 * -Wc++0x-compat has been added and is now enabled by default for 13872 -Wall. It produces warnings for constructs whose meaning differs 13873 between ISO C++ 1998 and C++0x. 13874 * The -Wparentheses option now works for C++ as it does for C. It 13875 warns if parentheses are omitted when operators with confusing 13876 precedence are nested. It also warns about ambiguous else 13877 statements. Since -Wparentheses is enabled by -Wall, this may cause 13878 additional warnings with existing C++ code which uses -Wall. These 13879 new warnings may be disabled by using -Wall -Wno-parentheses. 13880 * The -Wmissing-declarations now works for C++ as it does for C. 13881 * The -fvisibility-ms-compat flag was added, to make it easier to 13882 port larger projects using shared libraries from Microsoft's Visual 13883 Studio to ELF and Mach-O systems. 13884 * C++ attribute handling has been overhauled for template arguments 13885 (ie dependent types). In particular, __attribute__((aligned(T))); 13886 works for C++ types. 13887 13888 Runtime Library (libstdc++) 13889 13890 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 13891 * Support for TR1 mathematical special functions and regular 13892 expressions. 13893 * Default what implementations give more elaborate exception strings 13894 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 13895 * Header dependencies have been streamlined, reducing unnecessary 13896 includes and pre-processed bloat. 13897 * Variadic template implementations of items in <tuple> and 13898 <functional>. 13899 * An experimental [11]parallel mode has been added. This is a 13900 parallel implementation of many C++ Standard library algorithms, 13901 like std::accumulate, std::for_each, std::transform, or std::sort, 13902 to give but four examples. These algorithms can be substituted for 13903 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 13904 or all existing algorithms can be transformed via the 13905 -D_GLIBCXX_PARALLEL macro. 13906 * Debug mode versions of classes in <unordered_set> and 13907 <unordered_map>. 13908 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 13909 now <backward/hash_set> and <backward/hash_map>. This code: 13910 #include <ext/hash_set> 13911 __gnu_cxx::hash_set<int> s; 13912 13913 Can be transformed (in order of preference) to: 13914 #include <tr1/unordered_set> 13915 std::tr1::unordered_set<int> s; 13916 13917 or 13918 #include <backward/hash_set> 13919 __gnu_cxx::hash_set<int> s; 13920 13921 Similar transformations apply to __gnu_cxx::hash_map, 13922 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 13923 __gnu_cxx::hash_multiset. 13924 13925 Fortran 13926 13927 * Due to the fact that the GMP and MPFR libraries are required for 13928 all languages, Fortran is no longer special in this regard and is 13929 available by default. 13930 * The [12]-fexternal-blas option has been added, which generates 13931 calls to BLAS routines for intrinsic matrix operations such as 13932 matmul rather than using the built-in algorithms. 13933 * Support to give a backtrace (compiler flag -fbacktrace or 13934 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 13935 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 13936 run-time error occured. 13937 * GNU Fortran now defines __GFORTRAN__ when it runs the C 13938 preprocessor (CPP). 13939 * The [13]-finit-local-zero, -finit-real, -finit-integer, 13940 -finit-character, and -finit-logical options have been added, which 13941 can be used to initialize local variables. 13942 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 13943 which calculate the Gamma function and its logarithm. Use EXTERNAL 13944 gamma if you want to use your own gamma function. 13945 * GNU Fortran now regards the backslash character as literal (as 13946 required by the Fortran 2003 standard); using [16]-fbackslash GNU 13947 Fortran interprets backslashes as C-style escape characters. 13948 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 13949 literal constants has been changed. Before they were always 13950 interpreted as integer; now they are bit-wise transferred as 13951 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 13952 2003 standard, and for real and complex variables in DATA 13953 statements or when directly assigned to real and complex variables. 13954 Everywhere else and especially in expressions they are still 13955 regarded as integer constants. 13956 * Fortran 2003 support has been extended: 13957 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 13958 + Pointer intent 13959 + Intrinsic module ISO_ENV_FORTRAN 13960 + Interoperability with C (ISO C Bindings) 13961 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 13962 attribute) 13963 + Fortran 2003 BOZ 13964 13965 Java (GCJ) 13966 13967 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 13968 This enables the use of all 1.5 language features, and fixes most 13969 existing front end bugs. 13970 * libgcj now supports all 1.5 language features which require runtime 13971 support: foreach, enum, annotations, generics, and auto-boxing. 13972 * We've made many changes to the tools shipped with gcj. 13973 + The old jv-scan tool has been removed. This tool never really 13974 worked properly. There is no replacement. 13975 + gcjh has been rewritten. Some of its more obscure options no 13976 longer work, but are still recognized in an attempt at 13977 compatibility. gjavah is a new program with similar 13978 functionality but different command-line options. 13979 + grmic and grmiregistry have been rewritten. grmid has been 13980 added. 13981 + gjar replaces the old fastjar. 13982 + gjarsigner (used for signing jars), gkeytool (used for key 13983 management), gorbd (for CORBA), gserialver (computes 13984 serialization UIDs), and gtnameserv (also for CORBA) are now 13985 installed. 13986 * The ability to dump the contents of the java run time heap to a 13987 file for off-line analysis has been added. The heap dumps may be 13988 analyzed with the new gc-analyze tool. They may be generated on 13989 out-of-memory conditions or on demand and are controlled by the new 13990 run time class gnu.gcj.util.GCInfo. 13991 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 13992 provide correct, updated, timezone information. This means that 13993 packagers no longer have to update libgcj when a time zone change 13994 is published. 13995 13996 [18]New Targets and Target Specific Improvements 13997 13998 IA-32/x86-64 13999 14000 * Tuning for Intel Core 2 processors is available via -mtune=core2 14001 and -march=core2. 14002 * Tuning for AMD Geode processors is available via -mtune=geode and 14003 -march=geode. 14004 * Code generation of block move (memcpy) and block set (memset) was 14005 rewritten. GCC can now pick the best algorithm (loop, unrolled 14006 loop, instruction with rep prefix or a library call) based on the 14007 size of the block being copied and the CPU being optimized for. A 14008 new option -minline-stringops-dynamically has been added. With this 14009 option string operations of unknown size are expanded such that 14010 small blocks are copied by in-line code, while for large blocks a 14011 library call is used. This results in faster code than 14012 -minline-all-stringops when the library implementation is capable 14013 of using cache hierarchy hints. The heuristic choosing the 14014 particular algorithm can be overwritten via -mstringop-strategy. 14015 Newly also memset of values different from 0 is inlined. 14016 * GCC no longer places the cld instruction before string operations. 14017 Both i386 and x86-64 ABI documents mandate the direction flag to be 14018 clear at the entry of a function. It is now invalid to set the flag 14019 in asm statement without reseting it afterward. 14020 * Support for SSSE3 built-in functions and code generation are 14021 available via -mssse3. 14022 * Support for SSE4.1 built-in functions and code generation are 14023 available via -msse4.1. 14024 * Support for SSE4.2 built-in functions and code generation are 14025 available via -msse4.2. 14026 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 14027 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 14028 allow explicit control of x87 floating point precision. 14029 * Support for __float128 (TFmode) IEEE quad type and corresponding 14030 TCmode IEEE complex quad type is available via the soft-fp library 14031 on x86_64 targets. This includes basic arithmetic operations 14032 (addition, subtraction, negation, multiplication and division) on 14033 __float128 real and TCmode complex values, the full set of IEEE 14034 comparisons between __float128 values, conversions to and from 14035 float, double and long double floating point types, as well as 14036 conversions to and from signed or unsigned integer, signed or 14037 unsigned long integer and signed or unsigned quad (TImode) integer 14038 types. Additionally, all operations generate the full set of IEEE 14039 exceptions and support the full set of IEEE rounding modes. 14040 * GCC can now utilize the ACML library for vectorizing calls to a set 14041 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 14042 link to an ACML ABI compatible library. 14043 14044 ARM 14045 14046 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 14047 has been added. 14048 14049 [19]CRIS 14050 14051 New features 14052 14053 * Compiler and Library support for the CRIS v32 architecture, as 14054 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 14055 added. 14056 14057 Configuration changes 14058 14059 * The cris-*-elf target now includes support for CRIS v32, including 14060 libraries, through the -march=v32 option. 14061 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 14062 * A new crisv32-*-linux* target defaults to generate code for CRIS 14063 v32. 14064 * The cris-*-aout target has been obsoleted. 14065 14066 Improved support for built-in functions 14067 14068 * GCC can now use the lz and swapwbr instructions to implement the 14069 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 14070 * __builtin_bswap32 is now implemented using the swapwb instruction, 14071 when available. 14072 14073 [20]m68k and ColdFire 14074 14075 New features 14076 14077 * Support for several new ColdFire processors has been added. You can 14078 generate code for them using the new -mcpu option. 14079 * All targets now support ColdFire processors. 14080 * m68k-uclinux targets have improved support for C++ constructors and 14081 destructors, and for shared libraries. 14082 * It is now possible to set breakpoints on the first or last line of 14083 a function, even if there are no statements on that line. 14084 14085 Optimizations 14086 14087 * Support for sibling calls has been added. 14088 * More use is now made of the ColdFire mov3q instruction. 14089 * __builtin_clz is now implemented using the ff1 ColdFire 14090 instruction, when available. 14091 * GCC now honors the -m68010 option. 68010 code now uses clr rather 14092 than move to zero volatile memory. 14093 * 68020 targets and above can now use symbol(index.size*scale) 14094 addresses for indexed array accesses. Earlier compilers would 14095 always load the symbol into a base register first. 14096 14097 Configuration changes 14098 14099 * All m68k and ColdFire targets now allow the default processor to be 14100 set at configure time using --with-cpu. 14101 * A --with-arch configuration option has been added. This option 14102 allows you to restrict a target to ColdFire or non-ColdFire 14103 processors. 14104 14105 Preprocessor macros 14106 14107 * An __mcfv*__ macro is now defined for all ColdFire targets. 14108 (Earlier versions of GCC only defined __mcfv4e__.) 14109 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 14110 * All targets now define __mc68010 and __mc68010__ when generating 14111 68010 code. 14112 14113 Command-line changes 14114 14115 * New command-line options -march, -mcpu, -mtune and -mhard-float 14116 have been added. These options apply to both m68k and ColdFire 14117 targets. 14118 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 14119 versions of -mshort, etc. 14120 * -fforce-addr has been removed. It is now ignored by the compiler. 14121 14122 Other improvements 14123 14124 * ColdFire targets now try to maintain a 4-byte-aligned stack where 14125 possible. 14126 * m68k-uclinux targets now try to avoid situations that lead to the 14127 load-time error: BINFMT_FLAT: reloc outside program. 14128 14129 MIPS 14130 14131 Changes to existing configurations 14132 14133 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 14134 and n64. Every GNU/Linux configuration now builds these libraries 14135 by default. 14136 * GNU/Linux configurations now generate -mno-shared code unless 14137 overridden by -fpic, -fPIC, -fpie or -fPIE. 14138 * mipsisa32*-linux-gnu configurations now generate hard-float code by 14139 default, just like other mipsisa32* and mips*-linux-gnu 14140 configurations. You can build a soft-float version of any 14141 mips*-linux-gnu configuration by passing --with-float=soft to 14142 configure. 14143 * mips-wrs-vxworks now supports run-time processes (RTPs). 14144 14145 Changes to existing command-line options 14146 14147 * The -march and -mtune options no longer accept 24k as a processor 14148 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 14149 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 14150 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 14151 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 14152 24kx, 24kex and 34kx. 14153 14154 New configurations 14155 14156 GCC now supports the following configurations: 14157 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 14158 default. Earlier releases also recognized this configuration, but 14159 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 14160 you can customize any mips*-linux-gnu* configuration to a 14161 particular ISA or processor by passing an appropriate --with-arch 14162 option to configure. 14163 * mipsisa*-sde-elf*, which provides compatibility with MIPS 14164 Technologies' SDE toolchains. The configuration uses the SDE 14165 libraries by default, but you can use it like other newlib-based 14166 ELF configurations by passing --with-newlib to configure. It is the 14167 only configuration besides mips64vr*-elf* to build MIPS16 as well 14168 as non-MIPS16 libraries. 14169 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 14170 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 14171 and 64-bit forms of the EABI. 14172 14173 New processors and application-specific extensions 14174 14175 * Support for the SmartMIPS ASE is available through the new 14176 -msmartmips option. 14177 * Support for revision 2 of the DSP ASE is available through the new 14178 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 14179 indicates the revision of the ASE in use. 14180 * Support for the 4KS and 74K families of processors is available 14181 through the -march and -mtune options. 14182 14183 Improved support for built-in functions 14184 14185 * GCC can now use load-linked, store-conditional and sync 14186 instructions to implement atomic built-in functions such as 14187 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 14188 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 14189 * GCC can now use the clz and dclz instructions to implement the 14190 __builtin_ctz and __builtin_ffs families of functions. 14191 * There is a new __builtin___clear_cache function for flushing the 14192 instruction cache. GCC expands this function inline on MIPS32 14193 revision 2 targets, otherwise it calls the function specified by 14194 -mcache-flush-func. 14195 14196 MIPS16 improvements 14197 14198 * GCC can now compile objects that contain a mixture of MIPS16 and 14199 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 14200 for specifying which mode a function should use. 14201 * A new option called -minterlink-mips16 makes non-MIPS16 code 14202 link-compatible with MIPS16 code. 14203 * After many bug fixes, the long-standing MIPS16 -mhard-float support 14204 should now work fairly reliably. 14205 * GCC can now use the MIPS16e save and restore instructions. 14206 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 14207 with -G0 -fsection-anchors is often smaller than code compiled with 14208 -G8. However, please note that you must usually compile all objects 14209 in your application with the same -G option; see the documentation 14210 of -G for details. 14211 * A new option called-mcode-readable specifies which instructions are 14212 allowed to load from the code segment. -mcode-readable=yes is the 14213 default and says that any instruction may load from the code 14214 segment. The other alternatives are -mcode-readable=pcrel, which 14215 says that only PC-relative MIPS16 instructions may load from the 14216 code segment, and -mcode-readable=no, which says that no 14217 instruction may do so. Please see the documentation for more 14218 details, including example uses. 14219 14220 Small-data improvements 14221 14222 There are three new options for controlling small data: 14223 * -mno-extern-sdata, which disables small-data accesses for 14224 externally-defined variables. Code compiled with -Gn 14225 -mno-extern-sdata will be link-compatible with any -G setting 14226 between -G0 and -Gn inclusive. 14227 * -mno-local-sdata, which disables the use of small-data sections for 14228 data that is not externally visible. This option can be a useful 14229 way of reducing small-data usage in less performance-critical parts 14230 of an application. 14231 * -mno-gpopt, which disables the use of the $gp register while still 14232 honoring the -G limit when placing externally-visible data. This 14233 option implies -mno-extern-sdata and -mno-local-sdata and it can be 14234 useful in situations where $gp does not necessarily hold the 14235 expected value. 14236 14237 Miscellaneous improvements 14238 14239 * There is a new option called -mbranch-cost for tweaking the 14240 perceived cost of branches. 14241 * If GCC is configured to use a version of GAS that supports the 14242 .gnu_attribute directive, it will use that directive to record 14243 certain properties of the output code. .gnu_attribute is new to GAS 14244 2.18. 14245 * There are two new function attributes, near and far, for overriding 14246 the command-line setting of -mlong-calls on a function-by-function 14247 basis. 14248 * -mfp64, which previously required a 64-bit target, now works with 14249 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 14250 mipsisa*-sde-elf* configurations provide suitable library support. 14251 * GCC now recognizes the -mdmx and -mmt options and passes them down 14252 to the assembler. It does nothing else with the options at present. 14253 14254 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 14255 (BEA) 14256 14257 * Support has been added for this new architecture. 14258 14259 RS6000 (POWER/PowerPC) 14260 14261 * Support for the PowerPC 750CL paired-single instructions has been 14262 added with a new powerpc-*-linux*paired* target configuration. It 14263 is enabled by an associated -mpaired option and can be accessed 14264 using new built-in functions. 14265 * Support for auto-detecting architecture and system configuration to 14266 auto-select processor optimization tuning. 14267 * Support for VMX on AIX 5.3 has been added. 14268 * Support for AIX Version 6.1 has been added. 14269 14270 S/390, zSeries and System z9 14271 14272 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 14273 added. When using the -march=z9-ec option, the compiler will 14274 generate code making use of instructions provided by the decimal 14275 floating point facility and the floating point conversion facility 14276 (pfpo). Besides the instructions used to implement decimal floating 14277 point operations these facilities also contain instructions to move 14278 between general purpose and floating point registers and to modify 14279 and copy the sign-bit of floating point values. 14280 * When the -march=z9-ec option is used the new 14281 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 14282 decimal floating point hardware instructions will be used or not. 14283 If none of them is given the hardware support is enabled by 14284 default. 14285 * The -mstack-guard option can now be omitted when using stack 14286 checking via -mstack-size in order to let GCC choose a sensible 14287 stack guard value according to the frame size of each function. 14288 * Various changes to improve performance of generated code have been 14289 implemented, including: 14290 + The condition code set by an add logical with carry 14291 instruction is now available for overflow checks like: a + b + 14292 carry < b. 14293 + The test data class instruction is now used to implement 14294 sign-bit and infinity checks of binary and decimal floating 14295 point numbers. 14296 14297 SPARC 14298 14299 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 14300 added. 14301 14302 Xtensa 14303 14304 * Stack unwinding for exception handling now uses by default a 14305 specialized version of DWARF unwinding. This is not 14306 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 14307 Xtensa with previous versions of GCC. 14308 * For Xtensa processors that include the Conditional Store option, 14309 the built-in functions for atomic memory access are now implemented 14310 using S32C1I instructions. 14311 * If the Xtensa NSA option is available, GCC will use it to implement 14312 the __builtin_ctz and __builtin_clz functions. 14313 14314 Documentation improvements 14315 14316 * Existing libstdc++ documentation has been edited and restructured 14317 into a single DocBook XML manual. The results can be viewed online 14318 [21]here. 14319 14320 Other significant improvements 14321 14322 * The compiler's --help command-line option has been extended so that 14323 it now takes an optional set of arguments. These arguments restrict 14324 the information displayed to specific classes of command-line 14325 options, and possibly only a subset of those options. It is also 14326 now possible to replace the descriptive text associated with each 14327 displayed option with an indication of its current value, or for 14328 binary options, whether it has been enabled or disabled. 14329 Here are some examples. The following will display all the options 14330 controlling warning messages: 14331 --help=warnings 14332 14333 Whereas this will display all the undocumented, target specific 14334 options: 14335 --help=target,undocumented 14336 14337 This sequence of commands will display the binary optimizations 14338 that are enabled by -O3: 14339 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 14340 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 14341 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 14342 14343 * The configure options --with-pkgversion and --with-bugurl have been 14344 added. These allow distributors of GCC to include a 14345 distributor-specific string in manuals and --version output and to 14346 specify the URL for reporting bugs in their versions of GCC. 14347 14348 [22]GCC 4.3.1 14349 14350 This is the [23]list of problem reports (PRs) from GCC's bug tracking 14351 system that are known to be fixed in the 4.3.1 release. This list might 14352 not be complete (that is, it is possible that some PRs that have been 14353 fixed are not listed here). 14354 14355 Target Specific Changes 14356 14357 IA-32/x86-64 14358 14359 ABI changes 14360 14361 * Starting with GCC 4.3.1, decimal floating point variables are 14362 aligned to their natural boundaries when they are passed on the 14363 stack for i386. 14364 14365 Command-line changes 14366 14367 * Starting with GCC 4.3.1, the -mcld option has been added to 14368 automatically generate a cld instruction in the prologue of 14369 functions that use string instructions. This option is used for 14370 backward compatibility on some operating systems and can be enabled 14371 by default for 32-bit x86 targets by configuring GCC with the 14372 --enable-cld configure option. 14373 14374 [24]GCC 4.3.2 14375 14376 This is the [25]list of problem reports (PRs) from GCC's bug tracking 14377 system that are known to be fixed in the 4.3.2 release. This list might 14378 not be complete (that is, it is possible that some PRs that have been 14379 fixed are not listed here). 14380 14381 [26]GCC 4.3.3 14382 14383 This is the [27]list of problem reports (PRs) from GCC's bug tracking 14384 system that are known to be fixed in the 4.3.3 release. This list might 14385 not be complete (that is, it is possible that some PRs that have been 14386 fixed are not listed here). 14387 14388 [28]GCC 4.3.4 14389 14390 This is the [29]list of problem reports (PRs) from GCC's bug tracking 14391 system that are known to be fixed in the 4.3.4 release. This list might 14392 not be complete (that is, it is possible that some PRs that have been 14393 fixed are not listed here). 14394 14395 [30]GCC 4.3.5 14396 14397 This is the [31]list of problem reports (PRs) from GCC's bug tracking 14398 system that are known to be fixed in the 4.3.5 release. This list might 14399 not be complete (that is, it is possible that some PRs that have been 14400 fixed are not listed here). 14401 14402 [32]GCC 4.3.6 14403 14404 This is the [33]list of problem reports (PRs) from GCC's bug tracking 14405 system that are known to be fixed in the 4.3.6 release. This list might 14406 not be complete (that is, it is possible that some PRs that have been 14407 fixed are not listed here). 14408 14409 14410 For questions related to the use of GCC, please consult these web 14411 pages and the [34]GCC manuals. If that fails, the 14412 [35]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14413 web pages and the development of GCC are welcome on our developer 14414 list at [36]gcc (a] gcc.gnu.org. All of [37]our lists have public 14415 archives. 14416 14417 Copyright (C) [38]Free Software Foundation, Inc. Verbatim copying and 14418 distribution of this entire article is permitted in any medium, 14419 provided this notice is preserved. 14420 14421 These pages are [39]maintained by the GCC team. Last modified 14422 2025-01-31. 14423 14424 References 14425 14426 1. https://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 14427 2. https://gmplib.org/ 14428 3. https://www.mpfr.org/ 14429 4. https://gcc.gnu.org/install/prerequisites.html 14430 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 14431 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 14432 7. https://gcc.gnu.org/gcc-4.3/porting_to.html 14433 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 14434 9. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 14435 10. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 14436 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 14437 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 14438 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 14439 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 14440 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 14441 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 14442 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 14443 18. https://gcc.gnu.org/gcc-4.3/changes.html#targets 14444 19. https://gcc.gnu.org/gcc-4.3/changes.html#cris 14445 20. https://gcc.gnu.org/gcc-4.3/changes.html#m68k 14446 21. https://gcc.gnu.org/onlinedocs/libstdc++/ 14447 22. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.1 14448 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 14449 24. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.2 14450 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 14451 26. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.3 14452 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 14453 28. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.4 14454 29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 14455 30. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.5 14456 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 14457 32. https://gcc.gnu.org/gcc-4.3/changes.html#GCC4.3.6 14458 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 14459 34. https://gcc.gnu.org/onlinedocs/ 14460 35. mailto:gcc-help (a] gcc.gnu.org 14461 36. mailto:gcc (a] gcc.gnu.org 14462 37. https://gcc.gnu.org/lists.html 14463 38. https://www.fsf.org/ 14464 39. https://gcc.gnu.org/about.html 14465 ====================================================================== 14466 http://gcc.gnu.org/gcc-4.2/index.html 14467 14468 GCC 4.2 Release Series 14469 14470 (This release series is no longer supported.) 14471 14472 May 19, 2008 14473 14474 The [1]GNU project and the GCC developers are pleased to announce the 14475 release of GCC 4.2.4. 14476 14477 This release is a bug-fix release, containing fixes for regressions in 14478 GCC 4.2.3 relative to previous releases of GCC. 14479 14480 Release History 14481 14482 GCC 4.2.4 14483 May 19, 2008 ([2]changes) 14484 14485 GCC 4.2.3 14486 February 1, 2008 ([3]changes) 14487 14488 GCC 4.2.2 14489 October 7, 2007 ([4]changes) 14490 14491 GCC 4.2.1 14492 July 18, 2007 ([5]changes) 14493 14494 GCC 4.2.0 14495 May 13, 2007 ([6]changes) 14496 14497 References and Acknowledgements 14498 14499 GCC used to stand for the GNU C Compiler, but since the compiler 14500 supports several other languages aside from C, it now stands for the 14501 GNU Compiler Collection. 14502 14503 A list of [7]successful builds is updated as new information becomes 14504 available. 14505 14506 The GCC developers would like to thank the numerous people that have 14507 contributed new features, improvements, bug fixes, and other changes as 14508 well as test results to GCC. This [8]amazing group of volunteers is 14509 what makes GCC successful. 14510 14511 For additional information about GCC please refer to the [9]GCC project 14512 web site or contact the [10]GCC development mailing list. 14513 14514 To obtain GCC please use [11]our mirror sites or [12]our version 14515 control system. 14516 14517 14518 For questions related to the use of GCC, please consult these web 14519 pages and the [13]GCC manuals. If that fails, the 14520 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14521 web pages and the development of GCC are welcome on our developer 14522 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 14523 archives. 14524 14525 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14526 distribution of this entire article is permitted in any medium, 14527 provided this notice is preserved. 14528 14529 These pages are [18]maintained by the GCC team. Last modified 14530 2022-10-26. 14531 14532 References 14533 14534 1. http://www.gnu.org/ 14535 2. https://gcc.gnu.org/gcc-4.2/changes.html 14536 3. https://gcc.gnu.org/gcc-4.2/changes.html 14537 4. https://gcc.gnu.org/gcc-4.2/changes.html 14538 5. https://gcc.gnu.org/gcc-4.2/changes.html 14539 6. https://gcc.gnu.org/gcc-4.2/changes.html 14540 7. https://gcc.gnu.org/gcc-4.2/buildstat.html 14541 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14542 9. https://gcc.gnu.org/index.html 14543 10. mailto:gcc (a] gcc.gnu.org 14544 11. https://gcc.gnu.org/mirrors.html 14545 12. https://gcc.gnu.org/git.html 14546 13. https://gcc.gnu.org/onlinedocs/ 14547 14. mailto:gcc-help (a] gcc.gnu.org 14548 15. mailto:gcc (a] gcc.gnu.org 14549 16. https://gcc.gnu.org/lists.html 14550 17. https://www.fsf.org/ 14551 18. https://gcc.gnu.org/about.html 14552 ====================================================================== 14553 http://gcc.gnu.org/gcc-4.2/changes.html 14554 14555 GCC 4.2 Release Series 14556 Changes, New Features, and Fixes 14557 14558 Caveats 14559 14560 * GCC no longer accepts the -fshared-data option. This option has had 14561 no effect in any GCC 4 release; the targets to which the option 14562 used to apply had been removed before GCC 4.0. 14563 14564 General Optimizer Improvements 14565 14566 * New command-line options specify the possible relationships among 14567 parameters and between parameters and global data. For example, 14568 -fargument-noalias-anything specifies that arguments do not alias 14569 any other storage. 14570 Each language will automatically use whatever option is required by 14571 the language standard. You should not need to use these options 14572 yourself. 14573 14574 New Languages and Language specific improvements 14575 14576 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 14577 * New command-line options -fstrict-overflow and -Wstrict-overflow 14578 have been added. -fstrict-overflow tells the compiler that it may 14579 assume that the program follows the strict signed overflow 14580 semantics permitted for the language: for C and C++ this means that 14581 the compiler may assume that signed overflow does not occur. For 14582 example, a loop like 14583 for (i = 1; i > 0; i *= 2) 14584 14585 is presumably intended to continue looping until i overflows. With 14586 -fstrict-overflow, the compiler may assume that signed overflow 14587 will not occur, and transform this into an infinite loop. 14588 -fstrict-overflow is turned on by default at -O2, and may be 14589 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 14590 be used to warn about cases where the compiler assumes that signed 14591 overflow will not occur. It takes five different levels: 14592 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 14593 -Wstrict-overflow=1 is enabled by -Wall. 14594 * The new command-line option -fno-toplevel-reorder directs GCC to 14595 emit top-level functions, variables, and asm statements in the same 14596 order that they appear in the input file. This is intended to 14597 support existing code which relies on a particular ordering (for 14598 example, code which uses top-level asm statements to switch 14599 sections). For new code, it is generally better to use function and 14600 variable attributes. The -fno-toplevel-reorder option may be used 14601 for most cases which currently use -fno-unit-at-a-time. The 14602 -fno-unit-at-a-time option will be removed in some future version 14603 of GCC. If you know of a case which requires -fno-unit-at-a-time 14604 which is not fixed by -fno-toplevel-reorder, please open a bug 14605 report. 14606 14607 C family 14608 14609 * The pragma redefine_extname will now macro expand its tokens for 14610 compatibility with SunPRO. 14611 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 14612 GCC to handle inline functions as specified in the C99 standard. In 14613 preparation for this, GCC 4.2 will warn about any use of non-static 14614 inline functions in gnu99 or c99 mode. This new warning may be 14615 disabled with the new gnu_inline function attribute or the new 14616 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 14617 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 14618 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 14619 in the current compilation. 14620 * A new command-line option -Waddress has been added to warn about 14621 suspicious uses of memory addresses as, for example, using the 14622 address of a function in a conditional expression, and comparisons 14623 against the memory address of a string literal. This warning is 14624 enabled by -Wall. 14625 14626 C++ 14627 14628 * C++ visibility handling has been overhauled. 14629 Restricted visiblity is propagated from classes to members, from 14630 functions to local statics, and from templates and template 14631 arguments to instantiations, unless the latter has explicitly 14632 declared visibility. 14633 The visibility attribute for a class must come between the 14634 class-key and the name, not after the closing brace. 14635 Attributes are now allowed for enums and elaborated-type-specifiers 14636 that only declare a type. 14637 Members of the anonymous namespace are now local to a particular 14638 translation unit, along with any other declarations which use them, 14639 though they are still treated as having external linkage for 14640 language semantics. 14641 * The (undocumented) extension which permitted templates with default 14642 arguments to be bound to template template parameters with fewer 14643 parameters has been removed. For example: 14644 template <template <typename> class C> 14645 void f(C<double>) {} 14646 14647 template <typename T, typename U = int> 14648 struct S {}; 14649 14650 template void f(S<double>); 14651 14652 is no longer accepted by G++. The reason this code is not accepted 14653 is that S is a template with two parameters; therefore, it cannot 14654 be bound to C which has only one parameter. 14655 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 14656 releases, have been removed. 14657 * The command-line option -fconst-strings, deprecated in previous GCC 14658 releases, has been removed. 14659 * The configure variable enable-__cxa_atexit is now enabled by 14660 default for more targets. Enabling this variable is necessary in 14661 order for static destructors to be executed in the correct order, 14662 but it depends upon the presence of a non-standard C library in the 14663 target library in order to work. The variable is now enabled for 14664 more targets which are known to have suitable C libraries. 14665 * -Wextra will produce warnings for if statements with a semicolon as 14666 the only body, to catch code like: 14667 if (a); 14668 return 1; 14669 return 0; 14670 14671 To suppress the warning in valid cases, use { } instead. 14672 * The C++ front end now also produces strict aliasing warnings when 14673 -fstrict-aliasing -Wstrict-aliasing is in effect. 14674 14675 Runtime Library (libstdc++) 14676 14677 * Added support for TR1 <random>, <complex>, and C compatibility 14678 headers. In addition, a lock-free version of shared_ptr was 14679 contributed as part of Phillip Jordan's Google Summer of Code 14680 project on lock-free containers. 14681 * In association with the Summer of Code work on lock-free 14682 containers, the interface for atomic builtins was adjusted, 14683 creating simpler alternatives for non-threaded code paths. Also, 14684 usage was consolidated and all elements were moved from namespace 14685 std to namespace__gnu_cxx. Affected interfaces are the functions 14686 __exchange_and_add, __atomic_add, and the objects __mutex, 14687 __recursive_mutex, and __scoped_lock. 14688 * Support for versioning weak symbol names via namespace association 14689 was added. However, as this changes the names of exported symbols, 14690 this is turned off by default in the current ABI. Intrepid users 14691 can enable this feature by using 14692 --enable-symvers=gnu-versioned-namespace during configuration. 14693 * Revised, simplified, and expanded policy-based associative 14694 containers, including data types for tree and trie forms 14695 (basic_tree, tree, trie), lists (list_update), and both 14696 collision-chaining and probing hash-based containers 14697 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 14698 the [3]documentation. 14699 * The implementation of the debug mode was modified, whereby the 14700 debug namespaces were nested inside of namespace std and namespace 14701 __gnu_cxx in order to resolve some long standing corner cases 14702 involving name lookup. Debug functionality from the policy-based 14703 data structures was consolidated and enabled with the single macro, 14704 _GLIBCXX_DEBUG. See PR 26142 for more information. 14705 * Added extensions for type traits: __conditional_type, 14706 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 14707 * Added a typelist implementation for compile-time meta-programming. 14708 Elements for typelist construction and operation can be found 14709 within namespace __gnu_cxx::typelist. 14710 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 14711 exception-safety. 14712 * Enabled library-wide visibility control, allowing -fvisibility to 14713 be used. 14714 * Consolidated all nested namespaces and the conversion of 14715 __gnu_internal implementation-private details to anonymous 14716 namespaces whenever possible. 14717 * Implemented LWG resolutions DR 431 and DR 538. 14718 14719 Fortran 14720 14721 * Support for allocatable components has been added (TR 15581 and 14722 Fortran 2003). 14723 * Support for the Fortran 2003 streaming IO extension has been added. 14724 * The GNU Fortran compiler now uses 4-byte record markers by default 14725 for unformatted files to be compatible with g77 and most other 14726 compilers. The implementation allows for records greater than 2 GB 14727 and is compatible with several other compilers. Older versions of 14728 gfortran used 8-byte record markers by default (on most systems). 14729 In order to change the length of the record markers, e.g. to read 14730 unformatted files created by older gfortran versions, the 14731 [4]-frecord-marker=8 option can be used. 14732 14733 Java (GCJ) 14734 14735 * A new command-line option -static-libgcj has been added for targets 14736 that use a linker compatible with GNU Binutils. As its name 14737 implies, this causes libgcj to be linked statically. In some cases 14738 this causes the resulting executable to start faster and use less 14739 memory than if the shared version of libgcj were used. However 14740 caution should be used as it can also cause essential parts of the 14741 library to be omitted. Some of these issues are discussed in: 14742 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 14743 * fastjar is no longer bundled with GCC. To build libgcj, you will 14744 need either InfoZIP (both zip and unzip) or an external jar 14745 program. In the former case, the GCC build will install a jar shell 14746 script that is based on InfoZIP and provides the same functionality 14747 as fastjar. 14748 14749 [6]New Targets and Target Specific Improvements 14750 14751 IA-32/x86-64 14752 14753 * -mtune=generic can now be used to generate code running well on 14754 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 14755 Pentium-M, Intel Pentium 4 and Intel Core 2. 14756 * -mtune=native and -march=native will produce code optimized for the 14757 host architecture as detected using the cpuid instruction. 14758 * Added a new command-line option -fstackrealign and and 14759 __attribute__ ((force_align_arg_pointer)) to realign the stack at 14760 runtime. This allows functions compiled with a vector-aligned stack 14761 to be invoked from legacy objects that keep only word-alignment. 14762 14763 SPARC 14764 14765 * The default CPU setting has been changed from V7 to V9 in 32-bit 14766 mode on Solaris 7 and above. This is already the case in 64-bit 14767 mode. It can be overridden by specifying --with-cpu at configure 14768 time. 14769 * Back-end support of built-in functions for atomic memory access has 14770 been implemented. 14771 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 14772 added. 14773 14774 M32C 14775 14776 * Various bug fixes have made some functions (notably, functions 14777 returning structures) incompatible with previous releases. 14778 Recompiling all libraries is recommended. Note that code quality 14779 has considerably improved since 4.1, making a recompile even more 14780 beneficial. 14781 14782 MIPS 14783 14784 * Added support for the Broadcom SB-1A core. 14785 14786 IA-64 14787 14788 * Added support for IA-64 data and control speculation. By default 14789 speculation is enabled only during second scheduler pass. A number 14790 of machine flags was introduced to control the usage of speculation 14791 for both scheduler passes. 14792 14793 HPPA 14794 14795 * Added Java language support (libffi and libjava) for 32-bit HP-UX 14796 11 target. 14797 14798 [7]Obsolete Systems 14799 14800 Documentation improvements 14801 14802 PDF Documentation 14803 14804 * A make pdf target has been added to the top-level makefile, 14805 enabling automated production of PDF documentation files. 14806 (Front-ends external to GCC should modify their Make-lang.in file 14807 to add a lang.pdf: target.) 14808 14809 Other significant improvements 14810 14811 Build system improvements 14812 14813 * All the components of the compiler are now bootstrapped by default. 14814 This improves the resilience to bugs in the system compiler or 14815 binary compatibility problems, as well as providing better testing 14816 of GCC 4.2 itself. In addition, if you build the compiler from a 14817 combined tree, the assembler, linker, etc. will also be 14818 bootstrapped (i.e. built with themselves). 14819 You can disable this behavior, and go back to the pre-GCC 4.2 set 14820 up, by configuring GCC with --disable-bootstrap. 14821 * The rules that configure follows to find target tools resemble more 14822 closely the locations that the built compiler will search. In 14823 addition, you can use the new configure option --with-target-tools 14824 to specify where to find the target tools used during the build, 14825 without affecting what the built compiler will use. 14826 This can be especially useful when building packages of GCC. For 14827 example, you may want to build GCC with GNU as or ld, even if the 14828 resulting compiler to work with the native assembler and linker. To 14829 do so, you can use --with-target-tools to point to the native 14830 tools. 14831 14832 Incompatible changes to the build system 14833 14834 * Front-ends external to GCC should modify their Make-lang.in file to 14835 replace double-colon rules (e.g. dvi::) with normal rules (like 14836 lang.dvi:). Front-end makefile hooks do not use double-colon rules 14837 anymore. 14838 * Up to GCC 4.1, a popular way to specify the target tools used 14839 during the build was to create directories named gas, binutils, 14840 etc. in the build tree, and create links to the tools from there. 14841 This does not work any more when the compiler is bootstrapped. The 14842 new configure option --with-target-tools provides a better way to 14843 achieve the same effect, and works for all native and cross 14844 settings. 14845 14846 14847 For questions related to the use of GCC, please consult these web 14848 pages and the [8]GCC manuals. If that fails, the 14849 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14850 web pages and the development of GCC are welcome on our developer 14851 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 14852 archives. 14853 14854 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 14855 distribution of this entire article is permitted in any medium, 14856 provided this notice is preserved. 14857 14858 These pages are [13]maintained by the GCC team. Last modified 14859 2025-01-31. 14860 14861 References 14862 14863 1. https://gcc.gnu.org/projects/gomp/ 14864 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 14865 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 14866 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 14867 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 14868 6. https://gcc.gnu.org/gcc-4.2/changes.html#targets 14869 7. https://gcc.gnu.org/gcc-4.2/changes.html#obsolete_systems 14870 8. https://gcc.gnu.org/onlinedocs/ 14871 9. mailto:gcc-help (a] gcc.gnu.org 14872 10. mailto:gcc (a] gcc.gnu.org 14873 11. https://gcc.gnu.org/lists.html 14874 12. https://www.fsf.org/ 14875 13. https://gcc.gnu.org/about.html 14876 ====================================================================== 14877 http://gcc.gnu.org/gcc-4.1/index.html 14878 14879 GCC 4.1 Release Series 14880 14881 (This release series is no longer supported.) 14882 14883 February 13, 2007 14884 14885 The [1]GNU project and the GCC developers are pleased to announce the 14886 release of GCC 4.1.2. 14887 14888 This release is a bug-fix release, containing fixes for regressions in 14889 GCC 4.1.1 relative to previous releases of GCC. 14890 14891 Release History 14892 14893 GCC 4.1.2 14894 February 13, 2007 ([2]changes) 14895 14896 GCC 4.1.1 14897 May 24, 2006 ([3]changes) 14898 14899 GCC 4.1.0 14900 February 28, 2006 ([4]changes) 14901 14902 References and Acknowledgements 14903 14904 GCC used to stand for the GNU C Compiler, but since the compiler 14905 supports several other languages aside from C, it now stands for the 14906 GNU Compiler Collection. 14907 14908 A list of [5]successful builds is updated as new information becomes 14909 available. 14910 14911 The GCC developers would like to thank the numerous people that have 14912 contributed new features, improvements, bug fixes, and other changes as 14913 well as test results to GCC. This [6]amazing group of volunteers is 14914 what makes GCC successful. 14915 14916 For additional information about GCC please refer to the [7]GCC project 14917 web site or contact the [8]GCC development mailing list. 14918 14919 To obtain GCC please use [9]our mirror sites or [10]our version control 14920 system. 14921 14922 14923 For questions related to the use of GCC, please consult these web 14924 pages and the [11]GCC manuals. If that fails, the 14925 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14926 web pages and the development of GCC are welcome on our developer 14927 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 14928 archives. 14929 14930 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 14931 distribution of this entire article is permitted in any medium, 14932 provided this notice is preserved. 14933 14934 These pages are [16]maintained by the GCC team. Last modified 14935 2022-10-26. 14936 14937 References 14938 14939 1. http://www.gnu.org/ 14940 2. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 14941 3. https://gcc.gnu.org/gcc-4.1/changes.html 14942 4. https://gcc.gnu.org/gcc-4.1/changes.html 14943 5. https://gcc.gnu.org/gcc-4.1/buildstat.html 14944 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14945 7. https://gcc.gnu.org/index.html 14946 8. mailto:gcc (a] gcc.gnu.org 14947 9. https://gcc.gnu.org/mirrors.html 14948 10. https://gcc.gnu.org/git.html 14949 11. https://gcc.gnu.org/onlinedocs/ 14950 12. mailto:gcc-help (a] gcc.gnu.org 14951 13. mailto:gcc (a] gcc.gnu.org 14952 14. https://gcc.gnu.org/lists.html 14953 15. https://www.fsf.org/ 14954 16. https://gcc.gnu.org/about.html 14955 ====================================================================== 14956 http://gcc.gnu.org/gcc-4.1/changes.html 14957 14958 GCC 4.1 Release Series 14959 Changes, New Features, and Fixes 14960 14961 The latest release in the 4.1 release series is [1]GCC 4.1.2. 14962 14963 Caveats 14964 14965 General Optimizer Improvements 14966 14967 * GCC now has infrastructure for inter-procedural optimizations and 14968 the following inter-procedural optimizations are implemented: 14969 + Profile guided inlining. When doing profile feedback guided 14970 optimization, GCC can now use the profile to make better 14971 informed decisions on whether inlining of a function is 14972 profitable or not. This means that GCC will no longer inline 14973 functions at call sites that are not executed very often, and 14974 that functions at hot call sites are more likely to be 14975 inlined. 14976 A new parameter min-inline-recursive-probability is also now 14977 available to throttle recursive inlining of functions with 14978 small average recursive depths. 14979 + Discovery of pure and const functions, a form of side-effects 14980 analysis. While older GCC releases could also discover such 14981 special functions, the new IPA-based pass runs earlier so that 14982 the results are available to more optimizers. The pass is also 14983 simply more powerful than the old one. 14984 + Analysis of references to static variables and type escape 14985 analysis, also forms of side-effects analysis. The results of 14986 these passes allow the compiler to be less conservative about 14987 call-clobbered variables and references. This results in more 14988 redundant loads being eliminated and in making static 14989 variables candidates for register promotion. 14990 + Improvement of RTL-based alias analysis. The results of type 14991 escape analysis are fed to the RTL type-based alias analyzer, 14992 allowing it to disambiguate more memory references. 14993 + Interprocedural constant propagation and function versioning. 14994 This pass looks for functions that are always called with the 14995 same constant value for one or more of the function arguments, 14996 and propagates those constants into those functions. 14997 + GCC will now eliminate static variables whose usage was 14998 optimized out. 14999 + -fwhole-program --combine can now be used to make all 15000 functions in program static allowing whole program 15001 optimization. As an exception, the main function and all 15002 functions marked with the new externally_visible attribute are 15003 kept global so that programs can link with runtime libraries. 15004 * GCC can now do a form of partial dead code elimination (PDCE) that 15005 allows code motion of expressions to the paths where the result of 15006 the expression is actually needed. This is not always a win, so the 15007 pass has been limited to only consider profitable cases. Here is an 15008 example: 15009 int foo (int *, int *); 15010 int 15011 bar (int d) 15012 { 15013 int a, b, c; 15014 b = d + 1; 15015 c = d + 2; 15016 a = b + c; 15017 if (d) 15018 { 15019 foo (&b, &c); 15020 a = b + c; 15021 } 15022 printf ("%d\n", a); 15023 } 15024 15025 The a = b + c can be sunk to right before the printf. Normal code 15026 sinking will not do this, it will sink the first one above into the 15027 else-branch of the conditional jump, which still gives you two 15028 copies of the code. 15029 * GCC now has a value range propagation pass. This allows the 15030 compiler to eliminate bounds checks and branches. The results of 15031 the pass can also be used to accurately compute branch 15032 probabilities. 15033 * The pass to convert PHI nodes to straight-line code (a form of 15034 if-conversion for GIMPLE) has been improved significantly. The two 15035 most significant improvements are an improved algorithm to 15036 determine the order in which the PHI nodes are considered, and an 15037 improvement that allow the pass to consider if-conversions of basic 15038 blocks with more than two predecessors. 15039 * Alias analysis improvements. GCC can now differentiate between 15040 different fields of structures in Tree-SSA's virtual operands form. 15041 This lets stores/loads from non-overlapping structure fields not 15042 conflict. A new algorithm to compute points-to sets was contributed 15043 that can allows GCC to see now that p->a and p->b, where p is a 15044 pointer to a structure, can never point to the same field. 15045 * Various enhancements to auto-vectorization: 15046 + Incrementally preserve SSA form when vectorizing. 15047 + Incrementally preserve loop-closed form when vectorizing. 15048 + Improvements to peeling for alignment: generate better code 15049 when the misalignment of an access is known at compile time, 15050 or when different accesses are known to have the same 15051 misalignment, even if the misalignment amount itself is 15052 unknown. 15053 + Consider dependence distance in the vectorizer. 15054 + Externalize generic parts of data reference analysis to make 15055 this analysis available to other passes. 15056 + Vectorization of conditional code. 15057 + Reduction support. 15058 * GCC can now partition functions in sections of hot and cold code. 15059 This can significantly improve performance due to better 15060 instruction cache locality. This feature works best together with 15061 profile feedback driven optimization. 15062 * A new pass to avoid saving of unneeded arguments to the stack in 15063 vararg functions if the compiler can prove that they will not be 15064 needed. 15065 * Transition of basic block profiling to tree level implementation 15066 has been completed. The new implementation should be considerably 15067 more reliable (hopefully avoiding profile mismatch errors when 15068 using -fprofile-use or -fbranch-probabilities) and can be used to 15069 drive higher level optimizations, such as inlining. 15070 The -ftree-based-profiling command-line option was removed and 15071 -fprofile-use now implies disabling old RTL level loop optimizer 15072 (-fno-loop-optimize). Speculative prefetching optimization 15073 (originally enabled by -fspeculative-prefetching) was removed. 15074 15075 New Languages and Language specific improvements 15076 15077 C and Objective-C 15078 15079 * The old Bison-based C and Objective-C parser has been replaced by a 15080 new, faster hand-written recursive-descent parser. 15081 15082 Ada 15083 15084 * The build infrastructure for the Ada runtime library and tools has 15085 been changed to be better integrated with the rest of the build 15086 infrastructure of GCC. This should make doing cross builds of Ada a 15087 bit easier. 15088 15089 [2]C++ 15090 15091 * ARM-style name-injection of friend declarations is no longer the 15092 default. For example: 15093 struct S { 15094 friend void f(); 15095 }; 15096 15097 void g() { f(); } 15098 will not be accepted; instead a declaration of f will need to be 15099 present outside of the scope of S. The new -ffriend-injection 15100 option will enable the old behavior. 15101 * The (undocumented) extension which permitted templates with default 15102 arguments to be bound to template template parameters with fewer 15103 parameters has been deprecated, and will be removed in the next 15104 major release of G++. For example: 15105 template <template <typename> class C> 15106 void f(C<double>) {} 15107 15108 template <typename T, typename U = int> 15109 struct S {}; 15110 15111 template void f(S<double>); 15112 15113 makes use of the deprecated extension. The reason this code is not 15114 valid ISO C++ is that S is a template with two parameters; 15115 therefore, it cannot be bound to C which has only one parameter. 15116 15117 Runtime Library (libstdc++) 15118 15119 * Optimization work: 15120 + A new implementation of std::search_n is provided, better 15121 performing in case of random access iterators. 15122 + Added further efficient specializations of istream functions, 15123 i.e., character array and string extractors. 15124 + Other smaller improvements throughout. 15125 * Policy-based associative containers, designed for high-performance, 15126 flexibility and semantic safety are delivered in ext/pb_assoc. 15127 * A versatile string class, __gnu_cxx::__versa_string, providing 15128 facilities conforming to the standard requirements for 15129 basic_string, is delivered in <ext/vstring.h>. In particular: 15130 + Two base classes are provided: the default one avoids 15131 reference counting and is optimized for short strings; the 15132 alternate one, still uses it while improving in a few low 15133 level areas (e.g., alignment). See vstring_fwd.h for some 15134 useful typedefs. 15135 + Various algorithms have been rewritten (e.g., replace), the 15136 code streamlined and simple optimizations added. 15137 + Option 3 of DR 431 is implemented for both available bases, 15138 thus improving the support for stateful allocators. 15139 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 15140 libstdc++/23953) and LWG resolutions put into effect for the first 15141 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 15142 Issue 6.19). The implementation status of TR1 is now tracked in the 15143 docs in tr1.html. 15144 15145 Objective-C++ 15146 15147 * A new language front end for Objective-C++ has been added. This 15148 language allows users to mix the object oriented features of 15149 Objective-C with those of C++. 15150 15151 Java (GCJ) 15152 15153 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 15154 features (plus some 0.20 bug-fixes) 15155 + Networking 15156 o The java.net.HttpURLConnection implementation no longer 15157 buffers the entire response body in memory. This means 15158 that response bodies larger than available memory can now 15159 be handled. 15160 + (N)IO 15161 o NIO FileChannel.map implementation, fast bulk put 15162 implementation for DirectByteBuffer (speeds up this 15163 method 10x). 15164 o FileChannel.lock() and FileChannel.force() implemented. 15165 + XML 15166 o gnu.xml fix for nodes created outside a namespace 15167 context. 15168 o Add support for output indenting and 15169 cdata-section-elements output instruction in 15170 xml.transform. 15171 o xml.xpath corrections for cases where elements/attributes 15172 might have been created in non-namespace-aware mode. 15173 Corrections to handling of XSL variables and minor 15174 conformance updates. 15175 + AWT 15176 o GNU JAWT implementation, the AWT Native Interface, which 15177 allows direct access to native screen resources from 15178 within a Canvas's paint method. GNU Classpath Examples 15179 comes with a Demo, see libjava/classpath/examples/README. 15180 o awt.datatransfer updated to 1.5 with support for 15181 FlavorEvents. The gtk+ awt peers now allow copy/paste of 15182 text, images, URIs/files and serialized objects with 15183 other applications and tracking clipboard change events 15184 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 15185 objects are supported). A GNU Classpath Examples 15186 datatransfer Demo was added to show the new 15187 functionality. 15188 o Split gtk+ awt peers event handling in two threads and 15189 improve gdk lock handling (solves several awt lock ups). 15190 o Speed up awt Image loading. 15191 o Better gtk+ scrollbar peer implementation when using gtk+ 15192 >= 2.6. 15193 o Handle image loading errors correctly for gdkpixbuf and 15194 MediaTracker. 15195 o Better handle GDK lock. Properly prefix gtkpeer native 15196 functions (cp_gtk). 15197 o GdkGraphics2D has been updated to use Cairo 0.5.x or 15198 higher. 15199 o BufferedImage and GtkImage rewrites. All image drawing 15200 operations should now work correctly (flipping requires 15201 gtk+ >= 2.6) 15202 o When gtk+ 2.6 or higher is installed the default log 15203 handler will produce stack traces whenever a WARNING, 15204 CRITICAL or ERROR message is produced. 15205 + Free Swing 15206 o The RepaintManager has been reworked for more efficient 15207 painting, especially for large GUIs. 15208 o The layout manager OverlayLayout has been implemented, 15209 the BoxLayout has been rewritten to make use of the 15210 SizeRequirements utility class and caching for more 15211 efficient layout. 15212 o Improved accessibility support. 15213 o Significant progress has been made in the implementation 15214 of the javax.swing.plaf.metal package, with most UI 15215 delegates in a working state now. Please test this with 15216 your own applications and provide feedback that will help 15217 us to improve this package. 15218 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 15219 extended to highlight various features in our Free Swing 15220 implementation. And it includes a look and feel switcher 15221 for Metal (default), Ocean and GNU themes. 15222 o The javax.swing.plaf.multi package is now implemented. 15223 o Editing and several key actions for JTree and JTable were 15224 implemented. 15225 o Lots of icons and look and feel improvements for Free 15226 Swing basic and metal themes were added. Try running the 15227 GNU Classpath Swing Demo in examples 15228 (gnu.classpath.examples.swing.Demo) with: 15229 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 15230 l or 15231 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 15232 l 15233 o Start of styled text capabilites for java.swing.text. 15234 o DefaultMutableTreeNode pre-order, post-order, depth-first 15235 and breadth-first traversal enumerations implemented. 15236 o JInternalFrame colors and titlebar draw properly. 15237 o JTree is working up to par (icons, selection and keyboard 15238 traversal). 15239 o JMenus were made more compatible in visual and 15240 programmatic behavior. 15241 o JTable changeSelection and multiple selections 15242 implemented. 15243 o JButton and JToggleButton change states work properly 15244 now. 15245 o JFileChooser fixes. 15246 o revalidate() and repaint() fixes which make Free Swing 15247 much more responsive. 15248 o MetalIconFactory implemented. 15249 o Free Swing Top-Level Compatibility. JFrame, JDialog, 15250 JApplet, JInternalFrame, and JWindow are now 1.5 15251 compatible in the sense that you can call add() and 15252 setLayout() directly on them, which will have the same 15253 effect as calling getContentPane().add() and 15254 getContentPane().setLayout(). 15255 o The JTree interface has been completed. JTrees now 15256 recognizes mouse clicks and selections work. 15257 o BoxLayout works properly now. 15258 o Fixed GrayFilter to actually work. 15259 o Metal SplitPane implemented. 15260 o Lots of Free Swing text and editor stuff work now. 15261 + Free RMI and Corba 15262 o Andrew Watson, Vice President and Technical Director of 15263 the Object Management Group, has officially assigned us 15264 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 15265 mark remote classpath-specific system exceptions. 15266 Obtaining the VMCID means that GNU Classpath now is a 15267 recogniseable type of node in a highly interoperable 15268 CORBA world. 15269 o GNU Classpath now includes the first working draft to 15270 support the RMI over IIOP protocol. The current 15271 implementation is capable of remote invocations, 15272 transferring various Serializables and Externalizables 15273 via RMI-IIOP protocol. It can flatten graphs and, at 15274 least for the simple cases, is interoperable with 1.5 15275 JDKs. 15276 o org.omg.PortableInterceptor and related functionality in 15277 other packages is now implemented: 15278 # The sever and client interceptors work as required 15279 since 1.4. 15280 # The IOR interceptor works as needed for 1.5. 15281 o The org.omg.DynamicAny package is completed and passes 15282 the prepared tests. 15283 o The Portable Object Adapter should now support the output 15284 of the recent IDL to java compilers. These compilers now 15285 generate servants and not CORBA objects as before, making 15286 the output depend on the existing POA implementation. 15287 Completing POA means that such code can already be tried 15288 to run on Classpath. Our POA is tested for the following 15289 usager scenarios: 15290 # POA converts servant to the CORBA object. 15291 # Servant provides to the CORBA object. 15292 # POA activates new CORBA object with the given Object 15293 Id (byte array) that is later accessible for the 15294 servant. 15295 # During the first call, the ServantActivator provides 15296 servant for this and all subsequent calls on the 15297 current object. 15298 # During each call, the ServantLocator provides 15299 servant for this call only. 15300 # ServantLocator or ServantActivator forwards call to 15301 another server. 15302 # POA has a single servant, responsible for all 15303 objects. 15304 # POA has a default servant, but some objects are 15305 explicitly connected to they specific servants. 15306 The POA is verified using tests from the former 15307 cost.omg.org. 15308 o The CORBA implementation is now a working prototype that 15309 should support features up to 1.3 inclusive. We invite 15310 groups writing CORBA dependent applications to try 15311 Classpath implementation, reporting any possible bugs. 15312 The CORBA prototype is interoperable with Sun's 15313 implementation v 1.4, transferring object references, 15314 primitive types, narrow and wide strings, arrays, 15315 structures, trees, abstract interfaces and value types 15316 (feature of CORBA 2.3) between these two platforms. 15317 Remote exceptions are transferred and handled correctly. 15318 The stringified object references (IORs) from various 15319 sources are parsed as required. The transient (for 15320 current session) and permanent (till jre restart) 15321 redirections work. Both Little and Big Endian encoded 15322 messages are accepted. The implementation is verified 15323 using tests from the former cost.omg.org. The current 15324 release includes working examples (see the examples 15325 directory), demonstrating the client-server 15326 communication, using either CORBA Request or IDL-based 15327 stub (usually generated by a IDL to java compiler). These 15328 examples also show how to use the Classpath CORBA naming 15329 service. The IDL to java compiler is not yet written, but 15330 as our library must be compatible, it naturally accepts 15331 the output of other idlj implementations. 15332 + Misc 15333 o Updated TimeZone data against Olson tzdata2005l. 15334 o Make zip and jar packages UTF-8 clean. 15335 o "native" code builds and compiles (warning free) on 15336 Darwin and Solaris. 15337 o java.util.logging.FileHandler now rotates files. 15338 o Start of a generic JDWP framework in gnu/classpath/jdwp. 15339 This is unfinished, but feedback (at classpath (a] gnu.org) 15340 from runtime hackers is greatly appreciated. Although 15341 most of the work is currently being done around gcj/gij 15342 we want this framework to be as VM neutral as possible. 15343 Early design is described in: 15344 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 15345 o QT4 AWT peers, enable by giving configure 15346 --enable-qt-peer. Included, but not ready for production 15347 yet. They are explicitly disabled and not supported. But 15348 if you want to help with the development of these new 15349 features we are interested in feedback. You will have to 15350 explicitly enable them to try them out (and they will 15351 most likely contain bugs). 15352 o Documentation fixes all over the place. See 15353 [4]https://developer.classpath.org/doc/ 15354 15355 [5]New Targets and Target Specific Improvements 15356 15357 IA-32/x86-64 15358 15359 * The x86-64 medium model (that allows building applications whose 15360 data segment exceeds 4GB) was redesigned to match latest ABI draft. 15361 New implementation split large datastructures into separate segment 15362 improving performance of accesses to small datastructures and also 15363 allows linking of small model libraries into medium model programs 15364 as long as the libraries are not accessing the large datastructures 15365 directly. Medium model is also supported in position independent 15366 code now. 15367 The ABI change results in partial incompatibility among medium 15368 model objects. Linking medium model libraries (or objects) compiled 15369 with new compiler into medium model program compiled with older 15370 will likely result in exceeding ranges of relocations. 15371 Binutils 2.16.91 or newer are required for compiling medium model 15372 now. 15373 15374 RS6000 (POWER/PowerPC) 15375 15376 * The AltiVec vector primitives in <altivec.h> are now implemented in 15377 a way that puts a smaller burden on the preprocessor, instead 15378 processing the "overloading" in the front ends. This should benefit 15379 compilation speed on AltiVec vector code. 15380 * AltiVec initializers now are generated more efficiently. 15381 * The popcountb instruction available on POWER5 now is generated. 15382 * The floating point round to integer instructions available on 15383 POWER5+ now is generated. 15384 * Floating point divides can be synthesized using the floating point 15385 reciprocal estimate instructions. 15386 * Double precision floating point constants are initialized as single 15387 precision values if they can be represented exactly. 15388 15389 S/390, zSeries and System z9 15390 15391 * Support for the IBM System z9 109 processor has been added. When 15392 using the -march=z9-109 option, the compiler will generate code 15393 making use of instructions provided by the extended immediate 15394 facility. 15395 * Support for 128-bit IEEE floating point has been added. When using 15396 the -mlong-double-128 option, the compiler will map the long double 15397 data type to 128-bit IEEE floating point. Using this option 15398 constitutes an ABI change, and requires glibc support. 15399 * Various changes to improve performance of generated code have been 15400 implemented, including: 15401 + In functions that do not require a literal pool, register %r13 15402 (which is traditionally reserved as literal pool pointer), can 15403 now be freely used for other purposes by the compiler. 15404 + More precise tracking of register use allows the compiler to 15405 generate more efficient function prolog and epilog code in 15406 certain cases. 15407 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 15408 instructions are now used to implement C string functions. 15409 + The MOVE CHARACTER instruction with single byte overlap is now 15410 used to implement the memset function with non-zero fill byte. 15411 + The LOAD ZERO instructions are now used where appropriate. 15412 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 15413 and INSERT IMMEDIATE instructions are now used more frequently 15414 to optimize bitfield operations. 15415 + The BRANCH ON COUNT instruction is now used more frequently. 15416 In particular, the fact that a loop contains a subroutine call 15417 no longer prevents the compiler from using this instruction. 15418 + The compiler is now aware that all shift and rotate 15419 instructions implicitly truncate the shift count to six bits. 15420 * Back-end support for the following generic features has been 15421 implemented: 15422 + The full set of [6]built-in functions for atomic memory 15423 access. 15424 + The -fstack-protector feature. 15425 + The optimization pass avoiding unnecessary stores of incoming 15426 argument registers in functions with variable argument list. 15427 15428 SPARC 15429 15430 * The default code model in 64-bit mode has been changed from 15431 Medium/Anywhere to Medium/Middle on Solaris. 15432 * TLS support is disabled by default on Solaris prior to release 10. 15433 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 15434 and later) by specifying --enable-tls at configure time. 15435 15436 MorphoSys 15437 15438 * Support has been added for this new architecture. 15439 15440 [7]Obsolete Systems 15441 15442 Documentation improvements 15443 15444 Other significant improvements 15445 15446 * GCC can now emit code for protecting applications from 15447 stack-smashing attacks. The protection is realized by buffer 15448 overflow detection and reordering of stack variables to avoid 15449 pointer corruption. 15450 * Some built-in functions have been fortified to protect them against 15451 various buffer overflow (and format string) vulnerabilities. 15452 Compared to the mudflap bounds checking feature, the safe builtins 15453 have far smaller overhead. This means that programs built using 15454 safe builtins should not experience any measurable slowdown. 15455 15456 [8]GCC 4.1.2 15457 15458 This is the [9]list of problem reports (PRs) from GCC's bug tracking 15459 system that are known to be fixed in the 4.1.2 release. This list might 15460 not be complete (that is, it is possible that some PRs that have been 15461 fixed are not listed here). 15462 15463 When generating code for a shared library, GCC now recognizes that 15464 global functions may be replaced when the program runs. Therefore, it 15465 is now more conservative in deducing information from the bodies of 15466 functions. For example, in this example: 15467 void f() {} 15468 void g() { 15469 try { f(); } 15470 catch (...) { 15471 cout << "Exception"; 15472 } 15473 } 15474 15475 G++ would previously have optimized away the catch clause, since it 15476 would have concluded that f cannot throw exceptions. Because users may 15477 replace f with another function in the main body of the program, this 15478 optimization is unsafe, and is no longer performed. If you wish G++ to 15479 continue to optimize as before, you must add a throw() clause to the 15480 declaration of f to make clear that it does not throw exceptions. 15481 15482 15483 For questions related to the use of GCC, please consult these web 15484 pages and the [10]GCC manuals. If that fails, the 15485 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15486 web pages and the development of GCC are welcome on our developer 15487 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 15488 archives. 15489 15490 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 15491 distribution of this entire article is permitted in any medium, 15492 provided this notice is preserved. 15493 15494 These pages are [15]maintained by the GCC team. Last modified 15495 2025-01-31. 15496 15497 References 15498 15499 1. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 15500 2. https://gcc.gnu.org/gcc-4.1/changes.html#cplusplus 15501 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 15502 4. https://developer.classpath.org/doc/ 15503 5. https://gcc.gnu.org/gcc-4.1/changes.html#targets 15504 6. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 15505 7. https://gcc.gnu.org/gcc-4.1/changes.html#obsolete_systems 15506 8. https://gcc.gnu.org/gcc-4.1/changes.html#GCC4.1.2 15507 9. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 15508 10. https://gcc.gnu.org/onlinedocs/ 15509 11. mailto:gcc-help (a] gcc.gnu.org 15510 12. mailto:gcc (a] gcc.gnu.org 15511 13. https://gcc.gnu.org/lists.html 15512 14. https://www.fsf.org/ 15513 15. https://gcc.gnu.org/about.html 15514 ====================================================================== 15515 http://gcc.gnu.org/gcc-4.0/index.html 15516 15517 GCC 4.0 Release Series 15518 15519 (This release series is no longer supported.) 15520 15521 January 31, 2007 15522 15523 The [1]GNU project and the GCC developers are pleased to announce the 15524 release of GCC 4.0.4. 15525 15526 This release is a bug-fix release, containing fixes for regressions in 15527 GCC 4.0.3 relative to previous releases of GCC. 15528 15529 Release History 15530 15531 GCC 4.0.4 15532 January 31, 2007 ([2]changes) 15533 15534 GCC 4.0.3 15535 March 10, 2006 ([3]changes) 15536 15537 GCC 4.0.2 15538 September 28, 2005 ([4]changes) 15539 15540 GCC 4.0.1 15541 July 7, 2005 ([5]changes) 15542 15543 GCC 4.0.0 15544 April 20, 2005 ([6]changes) 15545 15546 References and Acknowledgements 15547 15548 GCC used to stand for the GNU C Compiler, but since the compiler 15549 supports several other languages aside from C, it now stands for the 15550 GNU Compiler Collection. 15551 15552 A list of [7]successful builds is updated as new information becomes 15553 available. 15554 15555 The GCC developers would like to thank the numerous people that have 15556 contributed new features, improvements, bug fixes, and other changes as 15557 well as test results to GCC. This [8]amazing group of volunteers is 15558 what makes GCC successful. 15559 15560 For additional information about GCC please refer to the [9]GCC project 15561 web site or contact the [10]GCC development mailing list. 15562 15563 To obtain GCC please use [11]our mirror sites, or [12]our version 15564 control system. 15565 15566 15567 For questions related to the use of GCC, please consult these web 15568 pages and the [13]GCC manuals. If that fails, the 15569 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15570 web pages and the development of GCC are welcome on our developer 15571 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 15572 archives. 15573 15574 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 15575 distribution of this entire article is permitted in any medium, 15576 provided this notice is preserved. 15577 15578 These pages are [18]maintained by the GCC team. Last modified 15579 2022-10-26. 15580 15581 References 15582 15583 1. http://www.gnu.org/ 15584 2. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 15585 3. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 15586 4. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 15587 5. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 15588 6. https://gcc.gnu.org/gcc-4.0/changes.html 15589 7. https://gcc.gnu.org/gcc-4.0/buildstat.html 15590 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15591 9. https://gcc.gnu.org/index.html 15592 10. mailto:gcc (a] gcc.gnu.org 15593 11. https://gcc.gnu.org/mirrors.html 15594 12. https://gcc.gnu.org/git.html 15595 13. https://gcc.gnu.org/onlinedocs/ 15596 14. mailto:gcc-help (a] gcc.gnu.org 15597 15. mailto:gcc (a] gcc.gnu.org 15598 16. https://gcc.gnu.org/lists.html 15599 17. https://www.fsf.org/ 15600 18. https://gcc.gnu.org/about.html 15601 ====================================================================== 15602 http://gcc.gnu.org/gcc-4.0/changes.html 15603 15604 GCC 4.0 Release Series 15605 Changes, New Features, and Fixes 15606 15607 The latest release in the 4.0 release series is [1]GCC 4.0.4. 15608 15609 Caveats 15610 15611 * GCC now generates location lists by default when compiling with 15612 debug info and optimization. 15613 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 15614 or later is needed to debug binaries containing location 15615 lists. 15616 + When you are trying to view a value of a variable in a part of 15617 a function where it has no location (for example when the 15618 variable is no longer used and thus its location was used for 15619 something else) GDB will say that it is not available. 15620 You can disable generating location lists by -fno-var-tracking. 15621 * GCC no longer accepts the -fwritable-strings option. Use named 15622 character arrays when you need a writable string. 15623 * The options -freduce-all-givs and -fmove-all-movables have been 15624 discontinued. They were used to circumvent a shortcoming in the 15625 heuristics of the old loop optimization code with respect to common 15626 Fortran constructs. The new (tree) loop optimizer works differently 15627 and doesn't need those work-arounds. 15628 * The graph-coloring register allocator, formerly enabled by the 15629 option -fnew-ra, has been discontinued. 15630 * -I- has been deprecated. -iquote is meant to replace the need for 15631 this option. 15632 * The MIPS -membedded-pic and -mrnames options have been removed. 15633 * All MIPS targets now require the GNU assembler. In particular, IRIX 15634 configurations can no longer use the MIPSpro assemblers, although 15635 they do still support the MIPSpro linkers. 15636 * The SPARC option -mflat has been removed. 15637 * English-language diagnostic messages will now use Unicode quotation 15638 marks in UTF-8 locales. (Non-English messages already used the 15639 quotes appropriate for the language in previous releases.) If your 15640 terminal does not support UTF-8 but you are using a UTF-8 locale 15641 (such locales are the default on many GNU/Linux systems) then you 15642 should set LC_CTYPE=C in the environment to disable that locale. 15643 Programs that parse diagnostics and expect plain ASCII 15644 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 15645 explanation of Unicode quotation marks for more information. 15646 * The specs file is no longer installed on most platforms. Most users 15647 will be totally unaffected. However, if you are accustomed to 15648 editing the specs file yourself, you will now have to use the 15649 -dumpspecs option to generate the specs file, and then edit the 15650 resulting file. 15651 15652 General Optimizer Improvements 15653 15654 * The [3]tree ssa branch has been merged. This merge has brought in a 15655 completely new optimization framework based on a higher level 15656 intermediate representation than the existing RTL representation. 15657 Numerous new code transformations based on the new framework are 15658 available in GCC 4.0, including: 15659 + Scalar replacement of aggregates 15660 + Constant propagation 15661 + Value range propagation 15662 + Partial redundancy elimination 15663 + Load and store motion 15664 + Strength reduction 15665 + Dead store elimination 15666 + Dead and unreachable code elimination 15667 + [4]Autovectorization 15668 + Loop interchange 15669 + Tail recursion by accumulation 15670 Many of these passes outperform their counterparts from previous 15671 GCC releases. 15672 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 15673 scheduling optimization intended for loops that perform heavy 15674 computations. 15675 15676 New Languages and Language specific improvements 15677 15678 C family 15679 15680 * The sentinel attribute has been added to GCC. This function 15681 attribute allows GCC to warn when variadic functions such as execl 15682 are not NULL terminated. See the GCC manual for a complete 15683 description of its behavior. 15684 * Given __attribute__((alias("target"))) it is now an error if target 15685 is not a symbol, defined in the same translation unit. This also 15686 applies to aliases created by #pragma weak alias=target. This is 15687 because it's meaningless to define an alias to an undefined symbol. 15688 On Solaris, the native assembler would have caught this error, but 15689 GNU as does not. 15690 15691 C and Objective-C 15692 15693 * The -Wstrict-aliasing=2 option has been added. This warning catches 15694 all unsafe cases, but it may also give a warning for some cases 15695 that are safe. 15696 * The cast-as-lvalue, conditional-expression-as-lvalue and 15697 compound-expression-as-lvalue extensions, which were deprecated in 15698 3.3.4 and 3.4, have been removed. 15699 * The -fwritable-strings option, which was deprecated in 3.4, has 15700 been removed. 15701 * #pragma pack() semantics have been brought closer to those used by 15702 other compilers. This also applies to C++. 15703 * Taking the address of a variable with register storage is invalid 15704 in C. GCC now issues an error instead of a warning. 15705 * Arrays of incomplete element type are invalid in C. GCC now issues 15706 an error for such arrays. Declarations such as extern struct s x[]; 15707 (where struct s has not been defined) can be moved after the 15708 definition of struct s. Function parameters declared as arrays of 15709 incomplete type can instead be declared as pointers. 15710 15711 C++ 15712 15713 * When compiling without optimizations (-O0), the C++ front end is 15714 much faster than in any previous versions of GCC. Independent 15715 testers have measured speed-ups up to 25% in real-world production 15716 code, compared to the 3.4 family (which was already the fastest 15717 version to date). Upgrading from older versions might show even 15718 bigger improvements. 15719 * ELF visibility attributes can now be applied to a class type, so 15720 that it affects every member function of a class at once, without 15721 having to specify each individually: 15722 class __attribute__ ((visibility("hidden"))) Foo 15723 { 15724 int foo1(); 15725 void foo2(); 15726 }; 15727 The syntax is deliberately similar to the __declspec() system used 15728 by Microsoft Windows based compilers, allowing cross-platform 15729 projects to easily reuse their existing macro system for denoting 15730 exports and imports. By explicitly marking internal classes never 15731 used outside a binary as hidden, one can completely avoid PLT 15732 indirection overheads during their usage by the compiler. You can 15733 find out more about the advantages of this at 15734 [6]https://www.akkadia.org/drepper/dsohowto.pdf 15735 * The -fvisibility-inlines-hidden option has been added which marks 15736 all inlineable functions as having hidden ELF visibility, thus 15737 removing their symbol and typeinfo from the exported symbol table 15738 of the output ELF binary. Using this option can reduce the exported 15739 symbol count of template-heavy code by up to 40% with no code 15740 change at all, thus notably improving link and load times for the 15741 binary as well as a reduction in size of up to 10%. Also, check the 15742 new [7]-fvisibility option. 15743 * The compiler now uses the library interface specified by the [8]C++ 15744 ABI for thread-safe initialization of function-scope static 15745 variables. Most users should leave this alone, but embedded 15746 programmers may want to disable this by specifying 15747 -fno-threadsafe-statics for a small savings in code size. 15748 * Taking the address of an explicit register variable is no longer 15749 supported. Note that C++ allows taking the address of variables 15750 with register storage so this will continue to compile with a 15751 warning. For example, assuming that r0 is a machine register: 15752 register int foo asm ("r0"); 15753 register int bar; 15754 &foo; // error, no longer accepted 15755 &bar; // OK, with a warning 15756 * G++ has an undocumented extension to virtual function covariancy 15757 rules that allowed the overrider to return a type that was 15758 implicitly convertable to the overridden function's return type. 15759 For instance a function returning void * could be overridden by a 15760 function returning T *. This is now deprecated and will be removed 15761 in a future release. 15762 * The G++ minimum and maximum operators (<? and >?) and their 15763 compound forms (<?=) and >?=) have been deprecated and will be 15764 removed in a future version. Code using these operators should be 15765 modified to use std::min and std::max instead. 15766 * Declaration of nested classes of class templates as friends are 15767 supported: 15768 template <typename T> struct A { 15769 class B {}; 15770 }; 15771 class C { 15772 template <typename T> friend class A<T>::B; 15773 }; 15774 This complements the feature member functions of class templates as 15775 friends introduced in GCC 3.4.0. 15776 * When declaring a friend class using an unqualified name, classes 15777 outside the innermost non-class scope are not searched: 15778 class A; 15779 namespace N { 15780 class B { 15781 friend class A; // Refer to N::A which has not been declared yet 15782 // because name outside namespace N are not searched 15783 friend class ::A; // Refer to ::A 15784 }; 15785 } 15786 Hiding the friend name until declaration is still not implemented. 15787 * Friends of classes defined outside their namespace are correctly 15788 handled: 15789 namespace N { 15790 class A; 15791 } 15792 class N::A { 15793 friend class B; // Refer to N::B in GCC 4.0.0 15794 // but ::B in earlier versions of GCC 15795 }; 15796 15797 Runtime Library (libstdc++) 15798 15799 * Optimization work: 15800 + Added efficient specializations of istream functions for char 15801 and wchar_t. 15802 + Further performance tuning of strings, in particular wrt 15803 single-char append and getline. 15804 + iter_swap - and therefore most of the mutating algorithms - 15805 now makes an unqualified call to swap when the value_type of 15806 the two iterators is the same. 15807 * A large subset of the features in Technical Report 1 (TR1 for 15808 short) is experimentally delivered (i.e., no guarantees about the 15809 implementation are provided. In particular it is not promised that 15810 the library will remain link-compatible when code using TR1 is 15811 used): 15812 + General utilities such as reference_wrapper and shared_ptr. 15813 + Function objects, i.e., result_of, mem_fn, bind, function. 15814 + Support for metaprogramming. 15815 + New containers such as tuple, array, unordered_set, 15816 unordered_map, unordered_multiset, unordered_multimap. 15817 * As usual, many bugs have been fixed and LWG resolutions implemented 15818 for the first time (e.g., DR 409). 15819 15820 Java 15821 15822 * In order to prevent naming conflicts with other implementations of 15823 these tools, some GCJ binaries have been renamed: 15824 + rmic is now grmic, 15825 + rmiregistry is now grmiregistry, and 15826 + jar is now fastjar. 15827 In particular, these names were problematic for the jpackage.org 15828 packaging conventions which install symlinks in /usr/bin that point 15829 to the preferred versions of these tools. 15830 * The -findirect-dispatch argument to the compiler now works and 15831 generates code following a new "binary compatibility" ABI. Code 15832 compiled this way follows the binary compatibility rules of the 15833 Java Language Specification. 15834 * libgcj now has support for using GCJ as a JIT, using the 15835 gnu.gcj.jit family of system properties. 15836 * libgcj can now find a shared library corresponding to the bytecode 15837 representation of a class. See the documentation for the new 15838 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 15839 property. 15840 * There have been many improvements to the class library. Here are 15841 some highlights: 15842 + Much more of AWT and Swing exist. 15843 + Many new packages and classes were added, including 15844 java.util.regex, java.net.URI, javax.crypto, 15845 javax.crypto.interfaces, javax.crypto.spec, javax.net, 15846 javax.net.ssl, javax.security.auth, 15847 javax.security.auth.callback, javax.security.auth.login, 15848 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 15849 javax.imageio, javax.imageio.event, javax.imageio.spi, 15850 javax.print, javax.print.attribute, 15851 javax.print.attribute.standard, javax.print.event, and 15852 javax.xml 15853 + Updated SAX and DOM, and imported GNU JAXP 15854 15855 Fortran 15856 15857 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 15858 front end. The new front end supports Fortran 90 and Fortran 95. It 15859 may not yet be as stable as the old Fortran front end. 15860 15861 Ada 15862 15863 * Ada (with tasking and Zero Cost Exceptions) is now available on 15864 many more targets, including but not limited to: alpha-linux, 15865 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 15866 s390x-linux, sparc-linux. 15867 * Some of the new Ada 2005 features are now implemented like 15868 Wide_Wide_Character and Ada.Containers. 15869 * Many bugs have been fixed, tools and documentation improved. 15870 * To compile Ada from the sources, install an older working Ada 15871 compiler and then use --enable-languages=ada at configuration time, 15872 since the Ada front end is not currently activated by default. See 15873 the [10]Installing GCC for details. 15874 15875 [11]New Targets and Target Specific Improvements 15876 15877 H8/300 15878 15879 * The frame layout has changed. In the new layout, the prologue of a 15880 function first saves registers and then allocate space for locals, 15881 resulting in an 1% improvement on code size. 15882 15883 IA-32/x86-64 (AMD64) 15884 15885 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 15886 log1p, log2, logb and tan mathematical builtins (and their float 15887 and long double variants) are now implemented as inline x87 15888 intrinsics when using -ffast-math. 15889 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 15890 (and their float and long double variants) are now implemented as 15891 inline x87 intrinsics when using -ffast-math. 15892 * The x87's fsincos instruction is now used automatically with 15893 -ffast-math when calculating both the sin and cos of the same 15894 argument. 15895 * Instruction selection for multiplication and division by constants 15896 has been improved. 15897 15898 IA-64 15899 15900 * Floating point division, integer division and sqrt are now inlined, 15901 resulting in significant performance improvements on some codes. 15902 15903 MIPS 15904 15905 * Division by zero checks now use conditional traps if the target 15906 processor supports them. This decreases code size by one word per 15907 division operation. The old behavior (branch and break) can be 15908 obtained either at configure time by passing --with-divide=breaks 15909 to configure or at runtime by passing -mdivide-breaks to GCC. 15910 * Support for MIPS64 paired-single instructions has been added. It is 15911 enabled by -mpaired-single and can be accessed using both the 15912 target-independent vector extensions and new MIPS-specific built-in 15913 functions. 15914 * Support for the MIPS-3D ASE has been added. It is enabled by 15915 -mips3d and provides new MIPS-3D-specific built-in functions. 15916 * The -mexplicit-relocs option now supports static n64 code (as is 15917 used, for example, in 64-bit linux kernels). -mexplicit-relocs 15918 should now be feature-complete and is enabled by default when GCC 15919 is configured to use a compatible assembler. 15920 * Support for the NEC VR4130 series has been added. This support 15921 includes the use of VR-specific instructions and a new VR4130 15922 scheduler. Full VR4130 support can be selected with -march=vr4130 15923 while code for any ISA can be tuned for the VR4130 using 15924 -mtune=vr4130. There is also a new -mvr4130-align option that 15925 produces better schedules at the cost of increased code size. 15926 * Support for the Broadcom SB-1 has been extended. There is now an 15927 SB-1 scheduler as well as support for the SB-1-specific 15928 paired-single instructions. Full SB-1 support can be selected with 15929 -march=sb1 while code for any ISA can be optimized for the SB-1 15930 using -mtune=sb1. 15931 * The compiler can now work around errata in R4000, R4400, VR4120 and 15932 VR4130 processors. These workarounds are enabled by -mfix-r4000, 15933 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 15934 and VR4130 workarounds need binutils 2.16 or above. 15935 * IRIX shared libraries are now installed into the standard library 15936 directories: o32 libraries go into lib/, n32 libraries go into 15937 lib32/ and n64 libraries go into lib64/. 15938 * The compiler supports a new -msym32 option. It can be used to 15939 optimize n64 code in which all symbols are known to have 32-bit 15940 values. 15941 15942 S/390 and zSeries 15943 15944 * New command-line options help to generate code intended to run in 15945 an environment where stack space is restricted, e.g. Linux kernel 15946 code: 15947 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 15948 warnings for single functions that require large or dynamic 15949 stack frames. 15950 + -mstack-size and -mstack-guard generate code that checks for 15951 stack overflow at run time. 15952 + -mpacked-stack generates code that reduces the stack frame 15953 size of many functions by reusing unneeded parts of the stack 15954 bias area. 15955 * The -msoft-float option now ensures that generated code never 15956 accesses floating point registers. 15957 * The s390x-ibm-tpf target now fully supports C++, including 15958 exceptions and threads. 15959 * Various changes to improve performance of the generated code have 15960 been implemented, including: 15961 + GCC now uses sibling calls where possible. 15962 + Condition code handling has been optimized, allowing GCC to 15963 omit redundant comparisons in certain cases. 15964 + The cost function guiding many optimizations has been refined 15965 to more accurately represent the z900 and z990 processors. 15966 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 15967 instructions are now used to avoid conditional branches in 15968 certain cases. 15969 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 15970 optimize address arithmetic required to access large stack 15971 frames. 15972 + GCC now makes more efficient use of memory-to-memory type 15973 instructions (MVC, CLC, ...). 15974 + More precise tracking of special register use allows better 15975 instruction scheduling, in particular of the function prologue 15976 and epilogue sequences. 15977 + The Java front end now generates inline code to implement 15978 integer division, instead of calling library routines. 15979 15980 SPARC 15981 15982 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 15983 -mf934 have been removed. They have been replaced with -mcpu=xxx. 15984 * The internal model used to estimate the relative cost of each 15985 instruction has been updated. It is expected to give better results 15986 on recent UltraSPARC processors. 15987 * Code generation for function prologues and epilogues has been 15988 improved, resulting in better scheduling and allowing multiple exit 15989 points in functions. 15990 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 15991 It is enabled by -mvis and provides new built-in functions for VIS 15992 instructions on UltraSPARC processors. 15993 * The option -mapp-regs has been turned on by default on Solaris too. 15994 15995 NetWare 15996 15997 * Novell NetWare (on ix86, no other hardware platform was ever really 15998 supported by this OS) has been re-enabled and the ABI supported by 15999 GCC has been brought into sync with that of MetroWerks CodeWarrior 16000 (the ABI previously supported was that of some Unix systems, which 16001 NetWare never tried to support). 16002 16003 [12]Obsolete Systems 16004 16005 Support for a number of older systems has been declared obsolete in GCC 16006 4.0. Unless there is activity to revive them, the next release of GCC 16007 will have their sources permanently removed. 16008 16009 All GCC ports for the following processor architectures have been 16010 declared obsolete: 16011 * Intel i860 16012 * Ubicom IP2022 16013 * National Semiconductor NS32K (ns32k) 16014 * Texas Instruments TMS320C[34]x 16015 16016 Also, those for some individual systems have been obsoleted: 16017 * SPARC family 16018 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 16019 sparc86x-*-elf) 16020 + OpenBSD 32-bit (sparc-*-openbsd*) 16021 16022 Documentation improvements 16023 16024 Other significant improvements 16025 16026 * Location lists are now generated by default when compiling with 16027 debug info and optimization. Location lists provide more accurate 16028 debug info about locations of variables and they allow debugging 16029 code compiled with -fomit-frame-pointer. 16030 * The -fvisibility option has been added which allows the default ELF 16031 visibility of all symbols to be set per compilation and the new 16032 #pragma GCC visibility preprocessor command allows the setting of 16033 default ELF visibility for a region of code. Using 16034 -fvisibility=hidden especially in combination with the new 16035 -fvisibility-inlines-hidden can yield substantial improvements in 16036 output binary quality including avoiding PLT indirection overheads, 16037 reduction of the exported symbol count by up to 60% (with resultant 16038 improvements to link and load times), better scope for the 16039 optimizer to improve code and up to a 20% reduction in binary size. 16040 Using these options correctly yields a binary with a similar symbol 16041 count to a Windows DLL. 16042 Perhaps more importantly, this new feature finally allows (with 16043 careful planning) complete avoidance of symbol clashes when 16044 manually loading shared objects with RTLD_GLOBAL, thus finally 16045 solving problems many projects such as python were forced to use 16046 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 16047 can find more information about using these options at 16048 [13]https://gcc.gnu.org/wiki/Visibility. 16049 __________________________________________________________________ 16050 16051 [14]GCC 4.0.1 16052 16053 This is the [15]list of problem reports (PRs) from GCC's bug tracking 16054 system that are known to be fixed in the 4.0.1 release. This list might 16055 not be complete (that is, it is possible that some PRs that have been 16056 fixed are not listed here). 16057 16058 [16]GCC 4.0.2 16059 16060 This is the [17]list of problem reports (PRs) from GCC's bug tracking 16061 system that are known to be fixed in the 4.0.2 release. This list might 16062 not be complete (that is, it is possible that some PRs that have been 16063 fixed are not listed here). 16064 16065 Unfortunately, due to a release engineering failure, this release has a 16066 regression on Solaris that will affect some C++ programs. We suggest 16067 that Solaris users apply a [18]patch that corrects the problem. Users 16068 who do not wish to apply the patch should explicitly link C++ programs 16069 with the -pthreads option, even if they do not use threads. This 16070 problem has been corrected in the current 4.0 branch sources and will 16071 not be present in GCC 4.0.3. 16072 16073 [19]GCC 4.0.3 16074 16075 Starting with this release, the function getcontext is recognized by 16076 the compiler as having the same semantics as the setjmp function. In 16077 particular, the compiler will ensure that all registers are dead before 16078 calling such a function and will emit a warning about the variables 16079 that may be clobbered after the second return from the function. 16080 16081 [20]GCC 4.0.4 16082 16083 This is the [21]list of problem reports (PRs) from GCC's bug tracking 16084 system that are known to be fixed in the 4.0.4 release. This list might 16085 not be complete (that is, it is possible that some PRs that have been 16086 fixed are not listed here). 16087 16088 The 4.0.4 release is provided for those that require a high degree of 16089 binary compatibility with previous 4.0.x releases. For most users, the 16090 GCC team recommends that version 4.1.1 or later be used instead." 16091 16092 16093 For questions related to the use of GCC, please consult these web 16094 pages and the [22]GCC manuals. If that fails, the 16095 [23]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16096 web pages and the development of GCC are welcome on our developer 16097 list at [24]gcc (a] gcc.gnu.org. All of [25]our lists have public 16098 archives. 16099 16100 Copyright (C) [26]Free Software Foundation, Inc. Verbatim copying and 16101 distribution of this entire article is permitted in any medium, 16102 provided this notice is preserved. 16103 16104 These pages are [27]maintained by the GCC team. Last modified 16105 2025-01-31. 16106 16107 References 16108 16109 1. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 16110 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 16111 3. https://gcc.gnu.org/projects/tree-ssa/ 16112 4. https://gcc.gnu.org/projects/tree-ssa/vectorization.html 16113 5. https://gcc.gnu.org/news/sms.html 16114 6. https://www.akkadia.org/drepper/dsohowto.pdf 16115 7. https://gcc.gnu.org/gcc-4.0/changes.html#visibility 16116 8. https://itanium-cxx-abi.github.io/cxx-abi/ 16117 9. https://gcc.gnu.org/fortran/ 16118 10. https://gcc.gnu.org/install/ 16119 11. https://gcc.gnu.org/gcc-4.0/changes.html#targets 16120 12. https://gcc.gnu.org/gcc-4.0/changes.html#obsolete_systems 16121 13. https://gcc.gnu.org/wiki/Visibility 16122 14. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.1 16123 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 16124 16. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.2 16125 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 16126 18. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 16127 19. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.3 16128 20. https://gcc.gnu.org/gcc-4.0/changes.html#GCC4.0.4 16129 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 16130 22. https://gcc.gnu.org/onlinedocs/ 16131 23. mailto:gcc-help (a] gcc.gnu.org 16132 24. mailto:gcc (a] gcc.gnu.org 16133 25. https://gcc.gnu.org/lists.html 16134 26. https://www.fsf.org/ 16135 27. https://gcc.gnu.org/about.html 16136 ====================================================================== 16137 http://gcc.gnu.org/gcc-3.4/index.html 16138 16139 GCC 3.4 Release Series 16140 16141 (This release series is no longer supported.) 16142 16143 May 26, 2006 16144 16145 The [1]GNU project and the GCC developers are pleased to announce the 16146 release of GCC 3.4.6. 16147 16148 This release is a bug-fix release, containing fixes for regressions in 16149 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 16150 3.4.x series. 16151 16152 The GCC 3.4 release series includes numerous [2]new features, 16153 improvements, bug fixes, and other changes, thanks to an [3]amazing 16154 group of volunteers. 16155 16156 Release History 16157 16158 GCC 3.4.6 16159 March 6, 2006 ([4]changes) 16160 16161 GCC 3.4.5 16162 November 30, 2005 ([5]changes) 16163 16164 GCC 3.4.4 16165 May 18, 2005 ([6]changes) 16166 16167 GCC 3.4.3 16168 November 4, 2004 ([7]changes) 16169 16170 GCC 3.4.2 16171 September 6, 2004 ([8]changes) 16172 16173 GCC 3.4.1 16174 July 1, 2004 ([9]changes) 16175 16176 GCC 3.4.0 16177 April 18, 2004 ([10]changes) 16178 16179 References and Acknowledgements 16180 16181 GCC used to stand for the GNU C Compiler, but since the compiler 16182 supports several other languages aside from C, it now stands for the 16183 GNU Compiler Collection. 16184 16185 A list of [11]successful builds is updated as new information becomes 16186 available. 16187 16188 The GCC developers would like to thank the numerous people that have 16189 contributed new features, improvements, bug fixes, and other changes as 16190 well as test results to GCC. This [12]amazing group of volunteers is 16191 what makes GCC successful. 16192 16193 For additional information about GCC please refer to the [13]GCC 16194 project web site or contact the [14]GCC development mailing list. 16195 16196 To obtain GCC please use [15]our mirror sites, or [16]our version 16197 control system. 16198 16199 16200 For questions related to the use of GCC, please consult these web 16201 pages and the [17]GCC manuals. If that fails, the 16202 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16203 web pages and the development of GCC are welcome on our developer 16204 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 16205 archives. 16206 16207 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 16208 distribution of this entire article is permitted in any medium, 16209 provided this notice is preserved. 16210 16211 These pages are [22]maintained by the GCC team. Last modified 16212 2022-10-26. 16213 16214 References 16215 16216 1. http://www.gnu.org/ 16217 2. https://gcc.gnu.org/gcc-3.4/changes.html 16218 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16219 4. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 16220 5. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 16221 6. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 16222 7. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 16223 8. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 16224 9. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 16225 10. https://gcc.gnu.org/gcc-3.4/changes.html 16226 11. https://gcc.gnu.org/gcc-3.4/buildstat.html 16227 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16228 13. https://gcc.gnu.org/index.html 16229 14. mailto:gcc (a] gcc.gnu.org 16230 15. https://gcc.gnu.org/mirrors.html 16231 16. https://gcc.gnu.org/git.html 16232 17. https://gcc.gnu.org/onlinedocs/ 16233 18. mailto:gcc-help (a] gcc.gnu.org 16234 19. mailto:gcc (a] gcc.gnu.org 16235 20. https://gcc.gnu.org/lists.html 16236 21. https://www.fsf.org/ 16237 22. https://gcc.gnu.org/about.html 16238 ====================================================================== 16239 http://gcc.gnu.org/gcc-3.4/changes.html 16240 16241 GCC 3.4 Release Series 16242 Changes, New Features, and Fixes 16243 16244 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 16245 is now closed. 16246 16247 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 16248 a bug, please make sure it's really GCC, and not your code, that is 16249 broken. 16250 16251 Caveats 16252 16253 * GNU Make is now required to build GCC. 16254 * With -nostdinc the preprocessor used to ignore both standard 16255 include paths and include paths contained in environment variables. 16256 It was neither documented nor intended that environment variable 16257 paths be ignored, so this has been corrected. 16258 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 16259 -fvolatile-static. It is unlikely that they worked correctly in any 16260 3.x release. 16261 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 16262 * Support for all the systems [3]obsoleted in GCC 3.3 has been 16263 removed from GCC 3.4. See below for a [4]list of systems which are 16264 obsoleted in this release. 16265 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 16266 compilers will not work. 16267 * The implementation of the [5]MIPS ABIs has changed. As a result, 16268 the code generated for certain MIPS targets will not be binary 16269 compatible with earlier releases. 16270 * In previous releases, the MIPS port had a fake "hilo" register with 16271 the user-visible name accum. This register has been removed. 16272 * The implementation of the [6]SPARC ABIs has changed. As a result, 16273 the code generated will not be binary compatible with earlier 16274 releases in certain cases. 16275 * The configure option --enable-threads=pthreads has been removed; 16276 use --enable-threads=posix instead, which should have the same 16277 effect. 16278 * Code size estimates used by inlining heuristics for C, Objective-C, 16279 C++ and Java have been redesigned significantly. As a result the 16280 parameters of -finline-insns, --param max-inline-insns-single and 16281 --param max-inline-insns-auto need to be reconsidered. 16282 * --param max-inline-slope and --param min-inline-insns have been 16283 removed; they are not needed for the new bottom-up inlining 16284 heuristics. 16285 * The new unit-at-a-time compilation scheme has several compatibility 16286 issues: 16287 + The order in which functions, variables, and top-level asm 16288 statements are emitted may have changed. Code relying on some 16289 particular ordering needs to be updated. The majority of such 16290 top-level asm statements can be replaced by section 16291 attributes. 16292 + Unreferenced static variables and functions are removed. This 16293 may result in undefined references when an asm statement 16294 refers to the variable/function directly. In that case either 16295 the variable/function shall be listed in asm statement operand 16296 or in the case of top-level asm statements the attribute used 16297 shall be used to force function/variable to be always output 16298 and considered as a possibly used by unknown code. 16299 For variables the attribute is accepted only by GCC 3.4 and 16300 newer, while for earlier versions it is sufficient to use 16301 unused to silence warnings about the variables not being 16302 referenced. To keep code portable across different GCC 16303 versions, you can use appropriate preprocessor conditionals. 16304 + Static functions now can use non-standard passing conventions 16305 that may break asm statements calling functions directly. 16306 Again the attribute used shall be used to prevent this 16307 behavior. 16308 As a temporary workaround, -fno-unit-at-a-time can be used, but 16309 this scheme may not be supported by future releases of GCC. 16310 * GCC 3.4 automatically places zero-initialized variables in the .bss 16311 section on some operating systems. Versions of GNU Emacs up to (and 16312 including) 21.3 will not work correctly when using this 16313 optimization; you can use -fno-zero-initialized-in-bss to disable 16314 it. 16315 * If GCC 3.4 is configured with --enable-threads=posix (the default 16316 on most targets that support pthreads) then _REENTRANT will be 16317 defined unconditionally by some libstdc++ headers. C++ code which 16318 relies on that macro to detect whether multi-threaded code is being 16319 compiled might change in meaning, possibly resulting in linker 16320 errors for single-threaded programs. Affected users of [7]Boost 16321 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 16322 See Bugzilla for [8]more information. 16323 16324 General Optimizer Improvements 16325 16326 * Usability of the profile feedback and coverage testing has been 16327 improved. 16328 + Performance of profiled programs has been improved by faster 16329 profile merging code. 16330 + Better use of the profile feedback for optimization (loop 16331 unrolling and loop peeling). 16332 + File locking support allowing fork() calls and parallel runs 16333 of profiled programs. 16334 + Coverage file format has been redesigned. 16335 + gcov coverage tool has been improved. 16336 + make profiledbootstrap available to build a faster compiler. 16337 Experiments made on i386 hardware showed an 11% speedup on -O0 16338 and a 7.5% speedup on -O2 compilation of a [9]large C++ 16339 testcase. 16340 + New value profiling pass enabled via -fprofile-values 16341 + New value profile transformations pass enabled via -fvpt aims 16342 to optimize some code sequences by exploiting knowledge about 16343 value ranges or other properties of the operands. At the 16344 moment a conversion of expensive divisions into cheaper 16345 operations has been implemented. 16346 + New -fprofile-generate and -fprofile-use command-line options 16347 to simplify the use of profile feedback. 16348 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 16349 Java which is enabled via -funit-at-a-time (and implied by -O2). In 16350 this scheme a whole file is parsed first and optimized later. The 16351 following basic inter-procedural optimizations are implemented: 16352 + Removal of unreachable functions and variables 16353 + Discovery of local functions (functions with static linkage 16354 whose address is never taken) 16355 + On i386, these local functions use register parameter passing 16356 conventions. 16357 + Reordering of functions in topological order of the call graph 16358 to enable better propagation of optimizing hints (such as the 16359 stack alignments needed by functions) in the back end. 16360 + Call graph based out-of-order inlining heuristics which allows 16361 to limit overall compilation unit growth (--param 16362 inline-unit-growth). 16363 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 16364 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 16365 CPU). 16366 * More realistic code size estimates used by inlining for C, 16367 Objective-C, C++ and Java. The growth of large functions can now be 16368 limited via --param large-function-insns and --param 16369 large-function-growth. 16370 * A new cfg-level loop optimizer pass replaces the old loop unrolling 16371 pass and adds two other loop transformations -- loop peeling and 16372 loop unswitching -- and also uses the profile feedback to limit 16373 code growth. (The three optimizations are enabled by 16374 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 16375 respectively). 16376 The old loop unroller still can be enabled by -fold-unroll-loops 16377 and may produce better code in some cases, especially when the 16378 webizer optimization pass is not run. 16379 * A new web construction pass enabled via -fweb (and implied by -O3) 16380 improves the quality of register allocation, CSE, first scheduling 16381 pass and some other optimization passes by avoiding re-use of 16382 pseudo registers with non-overlapping live ranges. The pass almost 16383 always improves code quality but does make debugging difficult and 16384 thus is not enabled by default by -O2 16385 The pass is especially effective as cleanup after code duplication 16386 passes, such as the loop unroller or the tracer. 16387 * Experimental implementations of superblock or trace scheduling in 16388 the second scheduling pass can be enabled via 16389 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 16390 16391 New Languages and Language specific improvements 16392 16393 Ada 16394 16395 * The Ada front end has been updated to include numerous bug fixes 16396 and enhancements. These include: 16397 + Improved project file support 16398 + Additional set of warnings about potential wrong code 16399 + Improved error messages 16400 + Improved code generation 16401 + Improved cross reference information 16402 + Improved inlining 16403 + Better run-time check elimination 16404 + Better error recovery 16405 + More efficient implementation of unbounded strings 16406 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 16407 ... 16408 + New GNAT.xxxx packages (e.g. GNAT.Strings, 16409 GNAT.Exception_Action) 16410 + New pragmas 16411 + New -gnatS switch replacing gnatpsta 16412 + Implementation of new Ada features (in particular limited 16413 with, limited aggregates) 16414 16415 C/Objective-C/C++ 16416 16417 * Precompiled headers are now supported. Precompiled headers can 16418 dramatically speed up compilation of some projects. There are some 16419 known defects in the current precompiled header implementation that 16420 will result in compiler crashes in relatively rare situations. 16421 Therefore, precompiled headers should be considered a "technology 16422 preview" in this release. Read the manual for details about how to 16423 use precompiled headers. 16424 * File handling in the preprocessor has been rewritten. GCC no longer 16425 gets confused by symlinks and hardlinks, and now has a correct 16426 implementation of #import and #pragma once. These two directives 16427 have therefore been un-deprecated. 16428 * The undocumented extension that allowed C programs to have a label 16429 at the end of a compound statement, which has been deprecated since 16430 GCC 3.0, has been removed. 16431 * The cast-as-lvalue extension has been removed for C++ and 16432 deprecated for C and Objective-C. In particular, code like this: 16433 int i; 16434 (char) i = 5; 16435 16436 or this: 16437 char *p; 16438 ((int *) p)++; 16439 16440 is no longer accepted for C++ and will not be accepted for C and 16441 Objective-C in a future version. 16442 * The conditional-expression-as-lvalue extension has been deprecated 16443 for C and Objective-C. In particular, code like this: 16444 int a, b, c; 16445 (a ? b : c) = 2; 16446 16447 will not be accepted for C and Objective-C in a future version. 16448 * The compound-expression-as-lvalue extension has been deprecated for 16449 C and Objective-C. In particular, code like this: 16450 int a, b; 16451 (a, b) = 2; 16452 16453 will not be accepted for C and Objective-C in a future version. A 16454 possible non-intrusive workaround is the following: 16455 (*(a, &b)) = 2; 16456 16457 * Several [10]built-in functions such as __builtin_popcount for 16458 counting bits, finding the highest and lowest bit in a word, and 16459 parity have been added. 16460 * The -fwritable-strings option has been deprecated and will be 16461 removed. 16462 * Many C math library functions are now recognized as built-ins and 16463 optimized. 16464 * The C, C++, and Objective-C compilers can now handle source files 16465 written in any character encoding supported by the host C library. 16466 The default input character set is taken from the current locale, 16467 and may be overridden with the -finput-charset command line option. 16468 In the future we will add support for inline encoding markers. 16469 16470 [11]C++ 16471 16472 * G++ is now much closer to full conformance to the ISO/ANSI C++ 16473 standard. This means, among other things, that a lot of invalid 16474 constructs which used to be accepted in previous versions will now 16475 be rejected. It is very likely that existing C++ code will need to 16476 be fixed. This document lists some of the most common issues. 16477 * A hand-written recursive-descent C++ parser has replaced the 16478 YACC-derived C++ parser from previous GCC releases. The new parser 16479 contains much improved infrastructure needed for better parsing of 16480 C++ source codes, handling of extensions, and clean separation 16481 (where possible) between proper semantics analysis and parsing. The 16482 new parser fixes many bugs that were found in the old parser. 16483 * You must now use the typename and template keywords to disambiguate 16484 dependent names, as required by the C++ standard. 16485 struct K { 16486 typedef int mytype_t; 16487 }; 16488 16489 template <class T1> struct A { 16490 template <class T2> struct B { 16491 void callme(void); 16492 }; 16493 16494 template <int N> void bar(void) 16495 { 16496 // Use 'typename' to tell the parser that T1::mytype_t names 16497 // a type. This is needed because the name is dependent (in 16498 // this case, on template parameter T1). 16499 typename T1::mytype_t x; 16500 x = 0; 16501 } 16502 }; 16503 16504 template <class T> void template_func(void) 16505 { 16506 // Use 'template' to prefix member templates within 16507 // dependent types (a has type A<T>, which depends on 16508 // the template parameter T). 16509 A<T> a; 16510 a.template bar<0>(); 16511 16512 // Use 'template' to tell the parser that B is a nested 16513 // template class (dependent on template parameter T), and 16514 // 'typename' because the whole A<T>::B<int> is 16515 // the name of a type (again, dependent). 16516 typename A<T>::template B<int> b; 16517 b.callme(); 16518 } 16519 16520 void non_template_func(void) 16521 { 16522 // Outside of any template class or function, no names can be 16523 // dependent, so the use of the keyword 'typename' and 'template' 16524 // is not needed (and actually forbidden). 16525 A<K> a; 16526 a.bar<0>(); 16527 A<K>::B<float> b; 16528 b.callme(); 16529 } 16530 * In a template definition, unqualified names will no longer find 16531 members of a dependent base (as specified by [temp.dep]/3 in the 16532 C++ standard). For example, 16533 template <typename T> struct B { 16534 int m; 16535 int n; 16536 int f (); 16537 int g (); 16538 }; 16539 int n; 16540 int g (); 16541 template <typename T> struct C : B<T> { 16542 void h () 16543 { 16544 m = 0; // error 16545 f (); // error 16546 n = 0; // ::n is modified 16547 g (); // ::g is called 16548 } 16549 }; 16550 You must make the names dependent, e.g. by prefixing them with 16551 this->. Here is the corrected definition of C<T>::h, 16552 template <typename T> void C<T>::h () 16553 { 16554 this->m = 0; 16555 this->f (); 16556 this->n = 0 16557 this->g (); 16558 } 16559 As an alternative solution (unfortunately not backwards compatible 16560 with GCC 3.3), you may use using declarations instead of this->: 16561 template <typename T> struct C : B<T> { 16562 using B<T>::m; 16563 using B<T>::f; 16564 using B<T>::n; 16565 using B<T>::g; 16566 void h () 16567 { 16568 m = 0; 16569 f (); 16570 n = 0; 16571 g (); 16572 } 16573 }; 16574 * In templates, all non-dependent names are now looked up and bound 16575 at definition time (while parsing the code), instead of later when 16576 the template is instantiated. For instance: 16577 void foo(int); 16578 16579 template <int> struct A { 16580 static void bar(void){ 16581 foo('a'); 16582 } 16583 }; 16584 16585 void foo(char); 16586 16587 int main() 16588 { 16589 A<0>::bar(); // Calls foo(int), used to call foo(char). 16590 } 16591 16592 * In an explicit instantiation of a class template, you must use 16593 class or struct before the template-id: 16594 template <int N> 16595 class A {}; 16596 16597 template A<0>; // error, not accepted anymore 16598 template class A<0>; // OK 16599 * The "named return value" and "implicit typename" extensions have 16600 been removed. 16601 * Default arguments in function types have been deprecated and will 16602 be removed. 16603 * ARM-style name-injection of friend declarations has been deprecated 16604 and will be removed. For example: struct S { friend void f(); }; 16605 void g() { f(); } will not be accepted by future versions of G++; 16606 instead a declaration of "f" will need to be present outside of the 16607 scope of "S". 16608 * Covariant returns are implemented for all but varadic functions 16609 that require an adjustment. 16610 * When -pedantic is used, G++ now issues errors about spurious 16611 semicolons. For example, 16612 namespace N {}; // Invalid semicolon. 16613 void f() {}; // Invalid semicolon. 16614 * G++ no longer accepts attributes for a declarator after the 16615 initializer associated with that declarator. For example, 16616 X x(1) __attribute__((...)); 16617 is no longer accepted. Instead, use: 16618 X x __attribute__((...)) (1); 16619 * Inside the scope of a template class, the name of the class itself 16620 can be treated as either a class or a template. So GCC used to 16621 accept the class name as argument of type template, and template 16622 template parameter. However this is not C++ standard compliant. Now 16623 the name is not treated as a valid template template argument 16624 unless you qualify the name by its scope. For example, the code 16625 below no longer compiles. 16626 template <template <class> class TT> class X {}; 16627 template <class T> class Y { 16628 X<Y> x; // Invalid, Y is always a type template parameter. 16629 }; 16630 The valid code for the above example is 16631 X< ::Y> x; // Valid. 16632 (Notice the space between < and : to prevent GCC to interpret this 16633 as a digraph for [.) 16634 * Friend declarations that refer to template specializations are 16635 rejected if the template has not already been declared. For 16636 example, 16637 template <typename T> 16638 class C { 16639 friend void f<> (C&); 16640 }; 16641 is rejected. You must first declare f as a template, 16642 template <typename T> 16643 void f(T); 16644 * In case of friend declarations, every name used in the friend 16645 declaration must be accessible at the point of that declaration. 16646 Previous versions of G++ used to be less strict about this and 16647 allowed friend declarations for private class members, for example. 16648 See the ISO C++ Standard Committee's [12]defect report #209 for 16649 details. 16650 * Declaration of member functions of class templates as friends are 16651 supported. For example, 16652 template <typename T> struct A { 16653 void f(); 16654 }; 16655 class C { 16656 template <typename T> friend void A<T>::f(); 16657 }; 16658 * You must use template <> to introduce template specializations, as 16659 required by the standard. For example, 16660 template <typename T> 16661 struct S; 16662 16663 struct S<int> { }; 16664 is rejected. You must write, 16665 template <> struct S<int> {}; 16666 * G++ used to accept code like this, 16667 struct S { 16668 int h(); 16669 void f(int i = g()); 16670 int g(int i = h()); 16671 }; 16672 This behavior is not mandated by the standard. Now G++ issues an 16673 error about this code. To avoid the error, you must move the 16674 declaration of g before the declaration of f. The default arguments 16675 for g must be visible at the point where it is called. 16676 * The C++ ABI Section 3.3.3 specifications for the array construction 16677 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 16678 NULL when the allocator argument returns NULL. These changes are 16679 incorporated into the libstdc++ runtime library. 16680 * Using a name introduced by a typedef in a friend declaration or in 16681 an explicit instantiation is now rejected, as specified by the ISO 16682 C++ standard. 16683 class A; 16684 typedef A B; 16685 class C { 16686 friend class B; // error, no typedef name here 16687 friend B; // error, friend always needs class/struct/enum 16688 friend class A; // OK 16689 }; 16690 16691 template <int> class Q {}; 16692 typedef Q<0> R; 16693 template class R; // error, no typedef name here 16694 template class Q<0>; // OK 16695 * When allocating an array with a new expression, GCC used to allow 16696 parentheses around the type name. This is actually ill-formed and 16697 it is now rejected: 16698 int* a = new (int)[10]; // error, not accepted anymore 16699 int* a = new int[10]; // OK 16700 * When binding an rvalue of class type to a reference, the copy 16701 constructor of the class must be accessible. For instance, consider 16702 the following code: 16703 class A 16704 { 16705 public: 16706 A(); 16707 16708 private: 16709 A(const A&); // private copy ctor 16710 }; 16711 16712 A makeA(void); 16713 void foo(const A&); 16714 16715 void bar(void) 16716 { 16717 foo(A()); // error, copy ctor is not accessible 16718 foo(makeA()); // error, copy ctor is not accessible 16719 16720 A a1; 16721 foo(a1); // OK, a1 is a lvalue 16722 } 16723 This might be surprising at first sight, especially since most 16724 popular compilers do not correctly implement this rule ([13]further 16725 details). 16726 * When forming a pointer to member or a pointer to member function, 16727 access checks for class visibility (public, protected, private) are 16728 now performed using the qualifying scope of the name itself. This 16729 is better explained with an example: 16730 class A 16731 { 16732 public: 16733 void pub_func(); 16734 protected: 16735 void prot_func(); 16736 private: 16737 void priv_func(); 16738 }; 16739 16740 class B : public A 16741 { 16742 public: 16743 void foo() 16744 { 16745 &A::pub_func; // OK, pub_func is accessible through A 16746 &A::prot_func; // error, cannot access prot_func through A 16747 &A::priv_func; // error, cannot access priv_func through A 16748 16749 &B::pub_func; // OK, pub_func is accessible through B 16750 &B::prot_func; // OK, can access prot_func through B (within B) 16751 &B::priv_func; // error, cannot access priv_func through B 16752 } 16753 }; 16754 16755 Runtime Library (libstdc++) 16756 16757 * Optimization work: 16758 + Streamlined streambuf, filebuf, separate synched with C 16759 Standard I/O streambuf. 16760 + All formatted I/O now uses cached locale information. 16761 + STL optimizations (memory/speed for list, red-black trees as 16762 used by sets and maps). 16763 + More use of GCC builtins. 16764 + String optimizations (avoid contention on 16765 increment/decrement-and-test of the reference count in the 16766 empty-string object, constructor from input_iterators 16767 speedup). 16768 * Static linkage size reductions. 16769 * Large File Support (files larger than 2 GB on 32-bit systems). 16770 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 16771 * Generic character traits. 16772 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 16773 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 16774 * The allocator class is now standard-conformant, and two additional 16775 extension allocators have been added, mt_alloc and 16776 bitmap_allocator. 16777 * PCH support: -include bits/stdc++.h (2x compile speedup). 16778 * Rewrote __cxa_demangle with support for C++ style allocators. 16779 * New debug modes for STL containers and iterators. 16780 * Testsuite rewrite: five times as many tests, plus increasingly 16781 sophisticated tests, including I/O, MT, multi-locale, wide and 16782 narrow characters. 16783 * Use current versions of GNU "autotools" for build/configuration. 16784 16785 Objective-C 16786 16787 * The Objective-C front end has been updated to include the numerous 16788 bug fixes and enhancements previously available only in Apple's 16789 version of GCC. These include: 16790 + Structured exception (@try... @catch... @finally, @throw) and 16791 synchronization (@synchronized) support. These are accessible 16792 via the -fobjc-exceptions switch; as of this writing, they may 16793 only be used in conjunction with -fnext-runtime on Mac OS X 16794 10.3 and later. See [14]Options Controlling Objective-C 16795 Dialect for more information. 16796 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 16797 may now be encoded as 'B'. In addition, the back-end/codegen 16798 dependencies have been removed. 16799 + An overhaul of message dispatch construction, ensuring that 16800 the various receiver types (and casts thereof) are handled 16801 properly, and that correct diagnostics are issued. 16802 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 16803 (-freplace-objc-classes) debugging modes, currently available 16804 on Mac OS X 10.3 and later. See [15]Options Controlling 16805 Objective-C Dialect for more information. 16806 + Access to optimized runtime entry points (-fno-nil-receivers ) 16807 on the assumption that message receivers are never nil. This 16808 is currently available on Mac OS X 10.3 and later. See 16809 [16]Options Controlling Objective-C Dialect for more 16810 information. 16811 16812 Java 16813 16814 * Compiling a .jar file will now cause non-.class entries to be 16815 automatically compiled as resources. 16816 * libgcj has been ported to Darwin. 16817 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 16818 to gcj. 16819 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 16820 code from shared libraries. 16821 * libgcj has been much more completely merged with [17]GNU Classpath. 16822 * Class loading is now much more correct; in particular the caller's 16823 class loader is now used when that is required. 16824 * [18]Eclipse 2.x will run out of the box using gij. 16825 * Parts of java.nio have been implemented. Direct and indirect 16826 buffers work, as do fundamental file and socket operations. 16827 * java.awt has been improved, though it is still not ready for 16828 general use. 16829 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 16830 method. 16831 * The MinGW port has matured. Enhancements include socket timeout 16832 support, thread interruption, improved Runtime.exec() handling and 16833 support for accented characters in filenames. 16834 16835 Fortran 16836 16837 * Fortran improvements are listed in the [19]Fortran documentation. 16838 16839 New Targets and Target Specific Improvements 16840 16841 Alpha 16842 16843 * Several [20]built-in functions have been added such as 16844 __builtin_alpha_zap to allow utilizing the more obscure 16845 instructions of the CPU. 16846 * Parameter passing of complex arguments has changed to match the 16847 ABI. This change is incompatible with previous GCC versions, but 16848 does fix compatibility with the Tru64 compiler and several corner 16849 cases where GCC was incompatible with itself. 16850 16851 ARM 16852 16853 * Nicolas Pitre has contributed his hand-coded floating-point support 16854 code for ARM. It is both significantly smaller and faster than the 16855 existing C-based implementation, even when building applications 16856 for Thumb. The arm-elf configuration has been converted to use the 16857 new code. 16858 * Support for the Intel's iWMMXt architecture, a second generation 16859 XScale processor, has been added. Enabled at run time with the 16860 -mcpu=iwmmxt command line switch. 16861 * A new ARM target has been added: arm-wince-pe. This is similar to 16862 the arm-pe target, but it defaults to using the APCS32 ABI. 16863 * The existing ARM pipeline description has been converted to the use 16864 the [21]DFA processor pipeline model. There is not much change in 16865 code performance, but the description is now [22]easier to 16866 understand. 16867 * Support for the Cirrus EP9312 Maverick floating point co-processor 16868 added. Enabled at run time with the -mcpu=ep9312 command line 16869 switch. Note however that the multilibs to support this chip are 16870 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 16871 enable their production you will have to uncomment the entries in 16872 that file. 16873 16874 H8/300 16875 16876 * Support for long long has been added. 16877 * Support for saveall attribute has been added. 16878 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 16879 for H8/300H and H8S, which is much faster than the previous 16880 implementation. 16881 * A lot of small performance improvements. 16882 16883 IA-32/AMD64 (x86-64) 16884 16885 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 16886 -march=k8 and -mcpu=k8. 16887 * Scalar SSE code generation carefully avoids reformatting penalties, 16888 hidden dependencies and minimizes the number of uops generated on 16889 both Intel and AMD CPUs. 16890 * Vector MMX and SSE operands are now passed in registers to improve 16891 performance and match the argument passing convention used by the 16892 Intel C++ Compiler. As a result it is not possible to call 16893 functions accepting vector arguments compiled by older GCC version. 16894 * Conditional jump elimination is now more aggressive on modern CPUs. 16895 * The Athlon ports has been converted to use the DFA processor 16896 pipeline description. 16897 * Optimization of indirect tail calls is now possible in a similar 16898 fashion as direct sibcall optimization. 16899 * Further small performance improvements. 16900 * -m128bit-long-double is now less buggy. 16901 * __float128 support in 64-bit compilation. 16902 * Support for data structures exceeding 2GB in 64-bit mode. 16903 * -mcpu has been renamed to -mtune. 16904 16905 IA-64 16906 16907 * Tuning code for the Itanium 2 processor has been added. The 16908 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 16909 enabled by default now. To generate code tuned for Itanium 1 the 16910 option -mtune=itanium1 should be used. 16911 * [23]DFA processor pipeline descriptions for the IA-64 processors 16912 have been added. This resulted in about 3% improvement on the 16913 SPECInt2000 benchmark for Itanium 2. 16914 * Instruction bundling for the IA-64 processors has been rewritten 16915 using the DFA pipeline hazard recognizer. It resulted in about 60% 16916 compiler speedup on the SPECInt2000 C programs. 16917 16918 M32R 16919 16920 * Support for the M32R/2 processor has been added by Renesas. 16921 * Support for an M32R GNU/Linux target and PIC code generation has 16922 been added by Renesas. 16923 16924 M68000 16925 16926 * Bernardo Innocenti (Develer S.r.l.) has contributed the 16927 m68k-uclinux target, based on former work done by Paul Dale 16928 (SnapGear Inc.). Code generation for the ColdFire processors family 16929 has been enhanced and extended to support the MCF 53xx and MCF 54xx 16930 cores, integrating former work done by Peter Barada (Motorola). 16931 16932 MIPS 16933 16934 Processor-specific changes 16935 16936 * Support for the RM7000 and RM9000 processors has been added. It can 16937 be selected using the -march compiler option and should work with 16938 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 16939 * Support for revision 2 of the MIPS32 ISA has been added. It can be 16940 selected with the command-line option -march=mips32r2. 16941 * There is a new option, -mfix-sb1, to work around certain SB-1 16942 errata. 16943 16944 Configuration 16945 16946 * It is possible to customize GCC using the following configure-time 16947 options: 16948 + --with-arch, which specifies the default value of the -march 16949 option. 16950 + --with-tune, which specifies the default value of the -mtune 16951 option. 16952 + --with-abi, which specifies the default ABI. 16953 + --with-float=soft, which tells GCC to use software floating 16954 point by default. 16955 + --with-float=hard, which tells GCC to use hardware floating 16956 point by default. 16957 * A 64-bit GNU/Linux port has been added. The associated 16958 configurations are mips64-linux-gnu and mips64el-linux-gnu. 16959 * The 32-bit GNU/Linux port now supports Java. 16960 * The IRIX 6 configuration now supports the o32 ABI and will build 16961 o32 multilibs by default. This support is compatible with both 16962 binutils and the SGI tools, but note that several features, 16963 including debugging information and DWARF2 exception handling, are 16964 only available when using the GNU assembler. Use of the GNU 16965 assembler and linker (version 2.15 or above) is strongly 16966 recommended. 16967 * The IRIX 6 configuration now supports 128-bit long doubles. 16968 * There are two new RTEMS-specific configurations, mips-rtems and 16969 mipsel-rtems. 16970 * There are two new *-elf configurations, mipsisa32r2-elf and 16971 mipsisa32r2el-elf. 16972 16973 General 16974 16975 * Several [24]ABI bugs have been fixed. Unfortunately, these changes 16976 will break binary compatibility with earlier releases. 16977 * GCC can now use explicit relocation operators when generating 16978 -mabicalls code. This behavior is controlled by -mexplicit-relocs 16979 and can have several performance benefits. For example: 16980 + It allows for more optimization of GOT accesses, including 16981 better scheduling and redundancy elimination. 16982 + It allows sibling calls to be implemented as jumps. 16983 + n32 and n64 leaf functions can use a call-clobbered global 16984 pointer instead of $28. 16985 + The code to set up $gp can be removed from functions that 16986 don't need it. 16987 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 16988 option is equivalent to the assembler's -xgot option and should be 16989 used instead of -Wa,-xgot. 16990 * Frame pointer elimination is now supported when generating 64-bit 16991 MIPS16 code. 16992 * Inline block moves have been optimized to take more account of 16993 alignment information. 16994 * Many internal changes have been made to the MIPS port, mostly aimed 16995 at reducing the reliance on assembler macros. 16996 16997 PowerPC 16998 16999 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 17000 [25]ABI incompatibilities regarding the way parameters are passed 17001 during functions calls. These changes may result in incompatibility 17002 between code compiled with GCC 3.3 and GCC 3.4. 17003 17004 PowerPC Darwin 17005 17006 * Support for shared/dylib gcc libraries has been added. It is 17007 enabled by default on powerpc-apple-darwin7.0.0 and up. 17008 * Libgcj is enabled by default. On systems older than 17009 powerpc-apple-darwin7.0.0 you need to install dlcompat. 17010 * 128-bit IBM extended precision format support added for long 17011 double. 17012 17013 PowerPC64 GNU/Linux 17014 17015 * By default, PowerPC64 GNU/Linux now uses natural alignment of 17016 structure elements. The old four byte alignment for double, with 17017 special rules for a struct starting with a double, can be chosen 17018 with -malign-power. This change may result in incompatibility 17019 between code compiled with GCC 3.3 and GCC 3.4. 17020 * -mabi=altivec is now the default rather than -mabi=no-altivec. 17021 * 128-bit IBM extended precision format support added for long 17022 double. 17023 17024 S/390 and zSeries 17025 17026 * New command-line options allow to specify the intended execution 17027 environment for generated code: 17028 + -mesa/-mzarch allows to specify whether to generate code 17029 running in ESA/390 mode or in z/Architecture mode (this is 17030 applicable to 31-bit code only). 17031 + -march allows to specify a minimum processor architecture 17032 level (g5, g6, z900, or z990). 17033 + -mtune allows to specify which processor to tune for. 17034 * It is possible to customize GCC using the following configure-time 17035 options: 17036 + --with-mode, which specifies whether to default to assuming 17037 ESA/390 or z/Architecture mode. 17038 + --with-arch, which specifies the default value of the -march 17039 option. 17040 + --with-tune, which specifies the default value of the -mtune 17041 option. 17042 * Support for the z990 processor has been added, and can be selected 17043 using -march=z990 or -mtune=z990. This includes instruction 17044 scheduling tuned for the superscalar instruction pipeline of the 17045 z990 processor as well as support for all new instructions provided 17046 by the long-displacement facility. 17047 * Support to generate 31-bit code optimized for zSeries processors 17048 (running in ESA/390 or in z/Architecture mode) has been added. This 17049 can be selected using -march=z900 and -mzarch respectively. 17050 * Instruction scheduling for the z900 and z990 processors now uses 17051 the DFA pipeline hazard recognizer. 17052 * GCC no longer generates code to maintain a stack backchain, 17053 previously used to generate stack backtraces for debugging 17054 purposes. As replacement that does not incur runtime overhead, 17055 DWARF-2 call frame information is provided by GCC; this is 17056 supported by GDB 6.1. The old behavior can be restored using the 17057 -mbackchain option. 17058 * The stack frame size of functions may now exceed 2 GB in 64-bit 17059 code. 17060 * A port for the 64-bit IBM TPF operating system has been added; the 17061 configuration is s390x-ibm-tpf. This configuration is supported as 17062 cross-compilation target only. 17063 * Various changes to improve the generated code have been 17064 implemented, including: 17065 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 17066 instructions to significantly speed up many floating-point 17067 applications. 17068 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 17069 WITH BORROW instructions to speed up long long arithmetic. 17070 + GCC now uses the SEARCH STRING instruction to implement 17071 strlen(). 17072 + In many cases, function call overhead for 31-bit code has been 17073 reduced by placing the literal pool after the function code 17074 instead of after the function prolog. 17075 + Register 14 is no longer reserved in 64-bit code. 17076 + Handling of global register variables has been improved. 17077 17078 SPARC 17079 17080 * The option -mflat is deprecated. 17081 * Support for large (> 2GB) frames has been added to the 64-bit port. 17082 * Several [26]ABI bugs have been fixed. Unfortunately, these changes 17083 will break binary compatibility with earlier releases. 17084 * The default debugging format has been switched from STABS to 17085 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 17086 the default debugging format for 64-bit code on Solaris. 17087 17088 SuperH 17089 17090 * Support for the SH2E processor has been added. Enabled at run time 17091 with the -m2e command line switch, or at configure time by 17092 specifying sh2e as the machine part of the target triple. 17093 17094 V850 17095 17096 * Support for the Mitsubishi V850E1 processor has been added. This is 17097 a variant of the V850E processor with some additional debugging 17098 instructions. 17099 17100 Xtensa 17101 17102 * Several ABI bugs have been fixed. Unfortunately, these changes 17103 break binary compatibility with earlier releases. 17104 + For big-endian processors, the padding of aggregate return 17105 values larger than a word has changed. If the size of an 17106 aggregate return value is not a multiple of 32 bits, previous 17107 versions of GCC inserted padding in the most-significant bytes 17108 of the first return value register. Aggregates larger than a 17109 word are now padded in the least-significant bytes of the last 17110 return value register used. Aggregates smaller than a word are 17111 still padded in the most-significant bytes. The return value 17112 padding has not changed for little-endian processors. 17113 + Function arguments with 16-byte alignment are now properly 17114 aligned. 17115 + The implementation of the va_list type has changed. A va_list 17116 value created by va_start from a previous release cannot be 17117 used with va_arg from this release, or vice versa. 17118 * More processor configuration options for Xtensa processors are 17119 supported: 17120 + the ABS instruction is now optional; 17121 + the ADDX* and SUBX* instructions are now optional; 17122 + an experimental CONST16 instruction can be used to synthesize 17123 constants instead of loading them from constant pools. 17124 These and other Xtensa processor configuration options can no 17125 longer be enabled or disabled by command-line options; the 17126 processor configuration must be specified by the xtensa-config.h 17127 header file when building GCC. Additionally, the 17128 -mno-serialize-volatile option is no longer supported. 17129 17130 [27]Obsolete Systems 17131 17132 Support for a number of older systems has been declared obsolete in GCC 17133 3.4. Unless there is activity to revive them, the next release of GCC 17134 will have their sources permanently removed. 17135 17136 All configurations of the following processor architectures have been 17137 declared obsolete: 17138 * Mitsubishi D30V, d30v-* 17139 * AT&T DSP1600 and DSP1610, dsp16xx-* 17140 * Intel 80960, i960 17141 17142 Also, some individual systems have been obsoleted: 17143 * ARM Family 17144 + Support for generating code for operation in APCS/26 mode 17145 (-mapcs-26). 17146 * IBM ESA/390 17147 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 17148 maintained and supported.) 17149 * Intel 386 family 17150 + MOSS, i?86-moss-msdos and i?86-*-moss* 17151 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 17152 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 17153 i?86-*-freebsd2* 17154 + GNU/Linux with a.out object format, i?86-linux*aout* 17155 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 17156 + Interix versions before Interix 3, i?86-*-interix 17157 + Mach microkernel, i?86-mach* 17158 + SCO UnixWare with UDK, i?86-*-udk* 17159 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 17160 + VSTa microkernel, i386-*-vsta 17161 * Motorola M68000 family 17162 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 17163 + NetBSD with a.out object format (before NetBSD 1.4), 17164 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 17165 + Generic System V r.4, m68k-*-sysv4* 17166 * VAX 17167 + Generic VAX, vax-*-* (This is generic VAX only; we have not 17168 obsoleted any VAX triples for specific operating systems.) 17169 17170 Documentation improvements 17171 17172 Other significant improvements 17173 17174 * The build system has undergone several significant cleanups. 17175 Subdirectories will only be configured if they are being built, and 17176 all subdirectory configures are run from the make command. The top 17177 level has been autoconfiscated. 17178 * Building GCC no longer writes to its source directory. This should 17179 help those wishing to share a read-only source directory over NFS 17180 or build from a CD. The exceptions to this feature are if you 17181 configure with either --enable-maintainer-mode or 17182 --enable-generated-files-in-srcdir. 17183 * The -W warning option has been renamed to -Wextra, which is more 17184 easily understood. The older spelling will be retained for 17185 backwards compatibility. 17186 * Substantial improvements in compile time have been made, 17187 particularly for non-optimizing compilations. 17188 __________________________________________________________________ 17189 17190 [28]GCC 3.4.0 17191 17192 Bug Fixes 17193 17194 A vast number of bugs have been fixed in 3.4.0, too many to publish a 17195 complete list here. [29]Follow this link to query the Bugzilla database 17196 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 17197 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 17198 regressions. 17199 __________________________________________________________________ 17200 17201 [30]GCC 3.4.1 17202 17203 Bug Fixes 17204 17205 This section lists the problem reports (PRs) from GCC's bug tracking 17206 system that are known to be fixed in the 3.4.1 release. This list might 17207 not be complete (that is, it is possible that some PRs that have been 17208 fixed are not listed here). 17209 17210 Bootstrap failures 17211 17212 * [31]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 17213 emitted - PIC related 17214 * [32]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 17215 * [33]14760 A bug in configure.in prevents using both 17216 --program-suffix and --program-prefix 17217 * [34]14671 [hppa64] bootstrap fails: ICE in 17218 save_call_clobbered_regs, in caller_save.c 17219 * [35]15093 [alpha][Java] make bootstrap fails to configure libffi on 17220 Alpha 17221 * [36]15178 Solaris 9/x86 fails linking after stage 3 17222 17223 Multi-platform internal compiler errors (ICEs) 17224 17225 * [37]12753 (preprocessor) Memory corruption in preprocessor on bad 17226 input 17227 * [38]13985 ICE in gcc.c-torture/compile/930621-1.c 17228 * [39]14810 (c++) tree check failures with invalid code involving 17229 templates 17230 * [40]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 17231 cp/parser.c 17232 * [41]15044 (c++) ICE on syntax error, template header 17233 * [42]15057 (c++) Compiling of conditional value throw constructs 17234 cause a segmentation violation 17235 * [43]15064 (c++) typeid of template parameter gives ICE 17236 * [44]15142 (c++) ICE when passing a string where a char* is expected 17237 in a throw statement 17238 * [45]15159 ICE in rtl_verify_flow_info_1 17239 * [46]15165 (c++) ICE in instantiate_template 17240 * [47]15193 Unary minus using pointer to V4SF vector causes 17241 -fforce-mem to exhaust all memory 17242 * [48]15209 (c++) Runs out of memory with packed structs 17243 * [49]15227 (c++) Trouble with invalid function definition 17244 * [50]15285 (c++) instantiate_type ICE when forming pointer to 17245 template function 17246 * [51]15299 (c++) ICE in resolve_overloaded_unification 17247 * [52]15329 (c++) ICE on constructor of member template 17248 * [53]15550 ICE in extract_insn, in recog.c 17249 * [54]15554 (c++) ICE in tsubst_copy, in cp/pt.c 17250 * [55]15640 (c++) ICE on invalid code in arg_assoc, in 17251 cp/name-lookup.c 17252 * [56]15666 [unit-at-a-time] Gcc abort on valid code 17253 * [57]15696 (c++) ICE with bad pointer-to-member code 17254 * [58]15701 (c++) ICE with friends and template template parameter 17255 * [59]15761 ICE in do_SUBST, in combine.c 17256 * [60]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 17257 17258 Ada 17259 17260 * [61]14538 All RTEMS targets broken for gnat 17261 17262 C front end 17263 17264 * [62]12391 missing warning about assigning to an incomplete type 17265 * [63]14649 atan(1.0) should not be a constant expression 17266 * [64]15004 [unit-at-a-time] no warning for unused paramater in 17267 static function 17268 * [65]15749 --pedantic-errors behaves differently from --pedantic 17269 with C-compiler on GNU/Linux 17270 17271 C++ compiler and library 17272 17273 * [66]10646 non-const reference is incorrectly matched in a "const T" 17274 partial specialization 17275 * [67]12077 wcin.rdbuf()->in_avail() return value too high 17276 * [68]13598 enc_filebuf doesn't work 17277 * [69]14211 const_cast returns lvalue but should be rvalue 17278 * [70]14220 num_put::do_put() undesired float/double behavior 17279 * [71]14245 problem with user-defined allocators in std::basic_string 17280 * [72]14340 libstdc++ Debug mode: failure to convert iterator to 17281 const_iterator 17282 * [73]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 17283 FILE* 17284 * [74]14668 no warning anymore for reevaluation of declaration 17285 * [75]14775 LFS (large file support) tests missing 17286 * [76]14821 Duplicate namespace alias declaration should not conflict 17287 * [77]14930 Friend declaration ignored 17288 * [78]14932 cannot use offsetof to get offsets of array elements in 17289 g++ 3.4.0 17290 * [79]14950 [non unit-at-a-time] always_inline does not mix with 17291 templates and -O0 17292 * [80]14962 g++ ignores #pragma redefine_extname 17293 * [81]14975 Segfault on low-level write error during imbue 17294 * [82]15002 Linewise stream input is unusably slow (std::string slow) 17295 * [83]15025 compiler accepts redeclaration of template as 17296 non-template 17297 * [84]15046 [arm] Math functions misdetected by cross configuration 17298 * [85]15069 a bit test on a variable of enum type is miscompiled 17299 * [86]15074 g++ -lsupc++ still links against libstdc++ 17300 * [87]15083 spurious "statement has no effect" warning 17301 * [88]15096 parse error with templates and pointer to const member 17302 * [89]15287 combination of operator[] and operator .* fails in 17303 templates 17304 * [90]15317 __attribute__ unused in first parameter of constructor 17305 gives error 17306 * [91]15337 sizeof on incomplete type diagnostic 17307 * [92]15361 bitset<>::_Find_next fails 17308 * [93]15412 _GLIBCXX_ symbols symbols defined and used in different 17309 namespaces 17310 * [94]15427 valid code results in incomplete type error 17311 * [95]15471 Incorrect member pointer offsets in anonymous 17312 structs/unions 17313 * [96]15503 nested template problem 17314 * [97]15507 compiler hangs while laying out union 17315 * [98]15542 operator & and template definitions 17316 * [99]15565 SLES9: leading + sign for unsigned int with showpos 17317 * [100]15625 friend defined inside a template fails to find static 17318 function 17319 * [101]15629 Function templates, overloads, and friend name injection 17320 * [102]15742 'noreturn' attribute ignored in method of template 17321 functions. 17322 * [103]15775 Allocator::pointer consistently ignored 17323 * [104]15821 Duplicate namespace alias within namespace rejected 17324 * [105]15862 'enum yn' fails (confict with undeclared builtin) 17325 * [106]15875 rejects pointer to member in template 17326 * [107]15877 valid code using templates and anonymous enums is 17327 rejected 17328 * [108]15947 Puzzling error message for wrong destructor declaration 17329 in template class 17330 * [109]16020 cannot copy __gnu_debug::bitset 17331 * [110]16154 input iterator concept too restrictive 17332 * [111]16174 deducing top-level consts 17333 17334 Java 17335 17336 * [112]14315 Java compiler is not parallel make safe 17337 17338 Fortran 17339 17340 * [113]15151 [g77] incorrect logical i/o in 64-bit mode 17341 17342 Objective-C 17343 17344 * [114]7993 private variables cannot be shadowed in subclasses 17345 17346 Optimization bugs 17347 17348 * [115]15228 useless copies of floating point operands 17349 * [116]15345 [non-unit-at-a-time] unreferenced nested inline 17350 functions not optimized away 17351 * [117]15945 Incorrect floating point optimization 17352 * [118]15526 ftrapv aborts on 0 * (-1) 17353 * [119]14690 Miscompiled POOMA tests 17354 * [120]15112 GCC generates code to write to unchanging memory 17355 17356 Preprocessor 17357 17358 * [121]15067 Minor glitch in the source of cpp 17359 17360 Main driver program bugs 17361 17362 * [122]1963 collect2 interprets -oldstyle_liblookup as -o 17363 ldstyle_liblookup 17364 17365 x86-specific (Intel/AMD) 17366 17367 * [123]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 17368 section} 17369 17370 HPPA-specific 17371 17372 * [124]14782 GCC produces an unaligned data access at -O2 17373 * [125]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 17374 * [126]15202 ICE in reload_cse_simplify_operands, in postreload.c 17375 17376 IA64-specific 17377 17378 * [127]14610 __float80 constants incorrectly emitted 17379 * [128]14813 init_array sections are initialized in the wrong order 17380 * [129]14857 GCC segfault on duplicated asm statement 17381 * [130]15598 Gcc 3.4 ICE on valid code 17382 * [131]15653 Gcc 3.4 ICE on valid code 17383 17384 MIPS-specific 17385 17386 * [132]15189 wrong filling of delay slot with -march=mips1 -G0 17387 -mno-split-addresses -mno-explicit-relocs 17388 * [133]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 17389 2.14.91 17390 * [134]16144 Bogus reference to __divdf3 when -O1 17391 * [135]16176 Miscompilation of unaligned data in MIPS backend 17392 17393 PowerPC-specific 17394 17395 * [136]11591 ICE in gcc.dg/altivec-5.c 17396 * [137]12028 powerpc-eabispe produces bad sCOND operation 17397 * [138]14478 rs6000 geu/ltu patterns generate incorrect code 17398 * [139]14567 long double and va_arg complex args 17399 * [140]14715 Altivec stack layout may overlap gpr save with stack 17400 temps 17401 * [141]14902 (libstdc++) Stream checking functions fail when -pthread 17402 option is used. 17403 * [142]14924 Compiler ICE on valid code 17404 * [143]14960 -maltivec affects vector return with -mabi=no-altivec 17405 * [144]15106 vector varargs failure passing from altivec to 17406 non-altivec code for -m32 17407 * [145]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 17408 half-word operation 17409 * [146]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 17410 and stvx 17411 * [147]15662 Segmentation fault when an exception is thrown - even if 17412 try and catch are specified 17413 17414 s390-specific 17415 17416 * [148]15054 Bad code due to overlapping stack temporaries 17417 17418 SPARC-specific 17419 17420 * [149]15783 ICE with union assignment in 64-bit mode 17421 * [150]15626 GCC 3.4 emits "ld: warning: relocation error: 17422 R_SPARC_UA32" 17423 17424 x86-64-specific 17425 17426 * [151]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 17427 * [152]14723 Backported -march=nocona from mainline 17428 * [153]15290 __float128 failed to pass to function properly 17429 17430 Cygwin/Mingw32-specific 17431 17432 * [154]15250 Option -mms-bitfields support on GCC 3.4 is not 17433 conformant to MS layout 17434 * [155]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 17435 worker on windows32 targets 17436 17437 Bugs specific to embedded processors 17438 17439 * [156]8309 [m68k] -m5200 produces erroneous SImode set of short 17440 varaible on stack 17441 * [157]13250 [SH] Gcc code for rotation clobbers the register, but 17442 gcc continues to use the register as if it was not clobbered 17443 * [158]13803 [coldfire] movqi operand constraints too restrictivefor 17444 TARGET_COLDFIRE 17445 * [159]14093 [SH] ICE for code when using -mhitachi option in SH 17446 * [160]14457 [m6811hc] ICE with simple c++ source 17447 * [161]14542 [m6811hc] ICE on simple source 17448 * [162]15100 [SH] cc1plus got hang-up on 17449 libstdc++-v3/testsuite/abi_check.cc 17450 * [163]15296 [CRIS] Delayed branch scheduling causing invalid code on 17451 cris-* 17452 * [164]15396 [SH] ICE with -O2 -fPIC 17453 * [165]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 17454 ColdFire 17455 17456 Testsuite problems (compiler not affected) 17457 17458 * [166]11610 libstdc++ testcases 27_io/* don't work properly remotely 17459 * [167]15488 (libstdc++) possibly insufficient file permissions for 17460 executing test suite 17461 * [168]15489 (libstdc++) testsuite_files determined incorrectly 17462 17463 Documentation bugs 17464 17465 * [169]13928 (libstdc++) no whatis info in some man pages generated 17466 by doxygen 17467 * [170]14150 Ada documentation out of date 17468 * [171]14949 (c++) Need to document method visibility changes 17469 * [172]15123 libstdc++-doc: Allocators.3 manpage is empty 17470 __________________________________________________________________ 17471 17472 [173]GCC 3.4.2 17473 17474 Bug Fixes 17475 17476 This section lists the problem reports (PRs) from GCC's bug tracking 17477 system that are known to be fixed in the 3.4.2 release. This list might 17478 not be complete (that is, it is possible that some PRs that have been 17479 fixed are not listed here). 17480 17481 Bootstrap failures and issues 17482 17483 * [174]16469 [mips-sgi-irix5.3] bootstrap fails in 17484 libstdc++-v3/testsuite 17485 * [175]16344 [hppa-linux-gnu] libstdc++'s PCH built by 17486 profiledbootstrap does not work with the built compiler 17487 * [176]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 17488 17489 Multi-platform internal compiler errors (ICEs) 17490 17491 * [177]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 17492 cp_parser_class_specifier, in cp/parser.c 17493 * [178]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 17494 * [179]15461 (c++) ICE due to NRV and inlining 17495 * [180]15890 (c++) ICE in c_expand_expr, in c-common.c 17496 * [181]16180 ICE: segmentation fault in RTL optimization 17497 * [182]16224 (c++) ICE in write_unscoped_name (template/namespace) 17498 * [183]16408 ICE: in delete_insn, in cfgrtl.c 17499 * [184]16529 (c++) ICE for: namespace-alias shall not be declared as 17500 the name of any other entity 17501 * [185]16698 (c++) ICE with exceptions and declaration of __cxa_throw 17502 * [186]16706 (c++) ICE in finish_member_declaration, in 17503 cp/semantics.c 17504 * [187]16810 (c++) Legal C++ program with cast gives ICE in 17505 build_ptrmemfunc 17506 * [188]16851 (c++) ICE when throwing a comma expression 17507 * [189]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 17508 * [190]16904 (c++) ICE in finish_class_member_access_expr, in 17509 cp/typeck.c 17510 * [191]16905 (c++) ICE (segfault) with exceptions 17511 * [192]16964 (c++) ICE in cp_parser_class_specifier due to 17512 redefinition 17513 * [193]17068 (c++) ICE: tree check: expected class 'd', have 'x' 17514 (identifier_node) in dependent_template_p, in cp/pt.c 17515 17516 Preprocessor bugs 17517 17518 * [194]16366 Preprocessor option -remap causes memory corruption 17519 17520 Optimization 17521 17522 * [195]15345 unreferenced nested inline functions not optimized away 17523 * [196]16590 Incorrect execution when compiling with -O2 17524 * [197]16693 Bitwise AND is lost when used within a cast to an enum 17525 of the same precision 17526 * [198]17078 Jump into if(0) substatement fails 17527 17528 Problems in generated debug information 17529 17530 * [199]13956 incorrect stabs for nested local variables 17531 17532 C front end bugs 17533 17534 * [200]16684 GCC should not warn about redundant redeclarations of 17535 built-ins 17536 17537 C++ compiler and library 17538 17539 * [201]12658 Thread safety problems in locale::global() and 17540 locale::locale() 17541 * [202]13092 g++ accepts invalid pointer-to-member conversion 17542 * [203]15320 Excessive memory consumption 17543 * [204]16246 Incorrect template argument deduction 17544 * [205]16273 Memory exhausted when using nested classes and virtual 17545 functions 17546 * [206]16401 ostringstream in gcc 3.4.x very slow for big data 17547 * [207]16411 undefined reference to 17548 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 17549 >::file() 17550 * [208]16489 G++ incorrectly rejects use of a null constant integral 17551 expression as a null constant pointer 17552 * [209]16618 offsetof fails with constant member 17553 * [210]16637 syntax error reported for valid input code 17554 * [211]16717 __attribute__((constructor)) broken in C++ 17555 * [212]16813 compiler error in DEBUG version of range insertion 17556 std::map::insert 17557 * [213]16853 pointer-to-member initialization from incompatible one 17558 accepted 17559 * [214]16889 ambiguity is not detected 17560 * [215]16959 Segmentation fault in ios_base::sync_with_stdio 17561 17562 Java compiler and library 17563 17564 * [216]7587 direct threaded interpreter not thread-safe 17565 * [217]16473 ServerSocket accept() leaks file descriptors 17566 * [218]16478 Hash synchronization deadlock with finalizers 17567 17568 Alpha-specific 17569 17570 * [219]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 17571 * [220]16974 could not split insn (ice in final_scan_insn, in 17572 final.c) 17573 17574 x86-specific 17575 17576 * [221]16298 ICE in output_operand 17577 * [222]17113 ICE with SSE2 intrinsics 17578 17579 x86-64 specific 17580 17581 * [223]14697 libstdc++ couldn't find 32bit libgcc_s 17582 17583 MIPS-specific 17584 17585 * [224]15869 [mips64] No NOP after LW (with -mips1 -O0) 17586 * [225]16325 [mips64] value profiling clobbers gp on mips 17587 * [226]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 17588 char[]s 17589 * [227]16380 [mips64] Use of uninitialised register after dbra 17590 conversion 17591 * [228]16407 [mips64] Unaligned access to local variables 17592 * [229]16643 [mips64] verify_local_live_at_start ICE after 17593 crossjumping & cfgcleanup 17594 17595 ARM-specific 17596 17597 * [230]15927 THUMB -O2: strength-reduced iteration variable ends up 17598 off by 1 17599 * [231]15948 THUMB: ICE with non-commutative cbranch 17600 * [232]17019 THUMB: bad switch statement in md code for 17601 addsi3_cbranch_scratch 17602 17603 IA64-specific 17604 17605 * [233]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 17606 (-mtune=merced) 17607 * [234]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 17608 (-mtune=itanium) 17609 * [235]16278 Gcc failed to build Linux kernel with -mtune=merced 17610 * [236]16414 ICE on valid code: typo in comparison of asm_noperands 17611 result 17612 * [237]16445 ICE on valid code: don't count ignored insns 17613 * [238]16490 ICE (segfault) while compiling with -fprofile-use 17614 * [239]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 17615 17616 PowerPC-specific 17617 17618 * [240]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 17619 * [241]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 17620 issue) 17621 17622 SPARC-specific 17623 17624 * [242]16199 ICE while compiling apache 2.0.49 17625 * [243]16416 -m64 doesn't imply -mcpu=v9 anymore 17626 * [244]16430 ICE when returning non-C aggregates larger than 16 bytes 17627 17628 Bugs specific to embedded processors 17629 17630 * [245]16379 [m32r] can't output large model function call of memcpy 17631 * [246]17093 [m32r] ICE with -msdata=use -O0 17632 * [247]17119 [m32r] ICE at switch case 0x8000 17633 17634 DJGPP-specific 17635 17636 * [248]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 17637 17638 Alpha Tru64-specific 17639 17640 * [249]16210 libstdc++ gratuitously omits "long long" I/O 17641 17642 Testsuite, documentation issues (compiler is not affected): 17643 17644 * [250]15488 (libstdc++) possibly insufficient file permissions for 17645 executing test suite 17646 * [251]16250 ada/doctools runs makeinfo even in release tarball 17647 __________________________________________________________________ 17648 17649 [252]GCC 3.4.3 17650 17651 This is the [253]list of problem reports (PRs) from GCC's bug tracking 17652 system that are known to be fixed in the 3.4.3 release. This list might 17653 not be complete (that is, it is possible that some PRs that have been 17654 fixed are not listed here). 17655 17656 Bootstrap failures 17657 17658 * [254]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 17659 * [255]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 17660 when undeclared 17661 17662 Internal compiler errors (ICEs) affecting multiple platforms 17663 17664 * [256]13948 (java) GCJ segmentation fault while compiling GL4Java 17665 .class files 17666 * [257]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 17667 * [258]16301 (c++) ICE when "strong" attribute is attached to a using 17668 directive 17669 * [259]16566 ICE with flexible arrays 17670 * [260]17023 ICE with nested functions in parameter declaration 17671 * [261]17027 ICE with noreturn function in loop at -O2 17672 * [262]17524 ICE in grokdeclarator, in cp/decl.c 17673 * [263]17826 (c++) ICE in cp_tree_equal 17674 17675 C and optimization bugs 17676 17677 * [264]15526 -ftrapv aborts on 0 * (-1) 17678 * [265]16999 #ident stopped working 17679 * [266]17503 quadratic behaviour in invalid_mode_change_p 17680 * [267]17581 Long long arithmetic fails inside a switch/case 17681 statement when compiled with -O2 17682 * [268]18129 -fwritable-strings doesn't work 17683 17684 C++ compiler and library bugs 17685 17686 * [269]10975 incorrect initial ostringstream::tellp() 17687 * [270]11722 Unbuffered filebuf::sgetn is slow 17688 * [271]14534 Unrecognizing static function as a template parameter 17689 when its return value is also templated 17690 * [272]15172 Copy constructor optimization in aggregate 17691 initialization 17692 * [273]15786 Bad error message for frequently occuring error. 17693 * [274]16162 Rejects valid member-template-definition 17694 * [275]16612 empty basic_strings can't live in shared memory 17695 * [276]16715 std::basic_iostream is instantiated when used, even 17696 though instantiations are already contained in libstdc++ 17697 * [277]16848 code in /ext/demangle.h appears broken 17698 * [278]17132 GCC fails to eliminate function template specialization 17699 when argument deduction fails 17700 * [279]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 17701 in ropeimpl.h 17702 * [280]17327 use of `enumeral_type' in template type unification 17703 * [281]17393 "unused variable '._0'" warning with -Wall 17704 * [282]17501 Confusion with member templates 17705 * [283]17537 g++ not passing -lstdc++ to linker when all command line 17706 arguments are libraries 17707 * [284]17585 usage of unqualified name of static member from within 17708 class not allowed 17709 * [285]17821 Poor diagnostic for using "." instead of "->" 17710 * [286]17829 wrong error: call of overloaded function is ambiguous 17711 * [287]17851 Misleading diagnostic for invalid function declarations 17712 with undeclared types 17713 * [288]17976 Destructor is called twice 17714 * [289]18020 rejects valid definition of enum value in template 17715 * [290]18093 bogus conflict in namespace aliasing 17716 * [291]18140 C++ parser bug when using >> in templates 17717 17718 Fortran 17719 17720 * [292]17541 data statements with double precision constants fail 17721 17722 x86-specific 17723 17724 * [293]17853 -O2 ICE for MMX testcase 17725 17726 SPARC-specific 17727 17728 * [294]17245 ICE compiling gsl-1.5 statistics/lag1.c 17729 17730 Darwin-specific 17731 17732 * [295]17167 FATAL:Symbol L_foo$stub already defined. 17733 17734 AIX-specific 17735 17736 * [296]17277 could not catch an exception when specified -maix64 17737 17738 Solaris-specific 17739 17740 * [297]17505 <cmath> calls acosf(), ceilf(), and other functions 17741 missing from system libraries 17742 17743 HP/UX specific: 17744 17745 * [298]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 17746 17747 ARM-specific 17748 17749 * [299]17384 ICE with mode attribute on structures 17750 17751 MIPS-specific 17752 17753 * [300]17770 No NOP after LWL with -mips1 17754 17755 Other embedded target specific 17756 17757 * [301]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 17758 * [302]14064 [avr-elf] -fdata-sections triggers ICE 17759 * [303]14678 [m68hc11-elf] gcc ICE 17760 * [304]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 17761 * [305]15790 [i686-coff] Alignment error building gcc with i686-coff 17762 target 17763 * [306]15886 [SH] Miscompilation with -O2 -fPIC 17764 * [307]16884 [avr-elf] [fweb related] bug while initializing 17765 variables 17766 17767 Bugs relating to debugger support 17768 17769 * [308]13841 missing debug info for _Complex function arguments 17770 * [309]15860 [big-endian targets] No DW_AT_location debug info is 17771 emitted for formal arguments to a function that uses "register" 17772 qualifiers 17773 17774 Testsuite issues (compiler not affected) 17775 17776 * [310]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 17777 * [311]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 17778 * [312]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 17779 testsuite 17780 17781 Documentation 17782 17783 * [313]15498 typo in gcc manual: non-existing locale example en_UK, 17784 should be en_GB 17785 * [314]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 17786 document broken shell 17787 * [315]16406 USE_LD_AS_NEEDED undocumented 17788 __________________________________________________________________ 17789 17790 [316]GCC 3.4.4 17791 17792 This is the [317]list of problem reports (PRs) from GCC's bug tracking 17793 system that are known to be fixed in the 3.4.4 release. This list might 17794 not be complete (that is, it is possible that some PRs that have been 17795 fixed are not listed here). 17796 __________________________________________________________________ 17797 17798 [318]GCC 3.4.5 17799 17800 This is the [319]list of problem reports (PRs) from GCC's bug tracking 17801 system that are known to be fixed in the 3.4.5 release. This list might 17802 not be complete (that is, it is possible that some PRs that have been 17803 fixed are not listed here). 17804 17805 Bootstrap issues 17806 17807 * [320]24688 sco_math fixincl breaks math.h 17808 17809 C compiler bugs 17810 17811 * [321]17188 struct Foo { } redefinition 17812 * [322]20187 wrong code for ((unsigned char)(unsigned long 17813 long)((a?a:1)&(a*b)))?0:1) 17814 * [323]21873 infinite warning loop on bad array initializer 17815 * [324]21899 enum definition accepts values to be overriden 17816 * [325]22061 ICE in find_function_data, in function.c 17817 * [326]22308 Failure to diagnose violation of constraint 6.516p2 17818 * [327]22458 ICE on missing brace 17819 * [328]22589 ICE casting to long long 17820 * [329]24101 Segfault with preprocessed source 17821 17822 C++ compiler and library bugs 17823 17824 * [330]10611 operations on vector mode not recognized in C++ 17825 * [331]13377 unexpected behavior of namespace usage directive 17826 * [332]16002 Strange error message with new parser 17827 * [333]17413 local classes as template argument 17828 * [334]17609 spurious error message after using keyword 17829 * [335]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 17830 * [336]18124 ICE with invalid template template parameter 17831 * [337]18155 typedef in template declaration not rejected 17832 * [338]18177 ICE with const_cast for undeclared variable 17833 * [339]18368 C++ error message regression 17834 * [340]16378 ICE when returning a copy of a packed member 17835 * [341]18466 int ::i; accepted 17836 * [342]18512 ICE on invalid usage of template base class 17837 * [343]18454 ICE when returning undefined type 17838 * [344]18738 typename not allowed with non-dependent qualified name 17839 * [345]18803 rejects access to operator() in template 17840 * [346]19004 ICE in uses_template_parms, in cp/pt.c 17841 * [347]19208 Spurious error about variably modified type 17842 * [348]18253 bad error message / ICE for invalid template parameter 17843 * [349]19608 ICE after friend function definition in local class 17844 * [350]19884 ICE on explicit instantiation of a non-template 17845 constructor 17846 * [351]20153 ICE when C++ template function contains anonymous union 17847 * [352]20563 Infinite loop in diagnostic (and ice after error 17848 message) 17849 * [353]20789 ICE with incomplete type in template 17850 * [354]21336 Internal compiler error when using custom new operators 17851 * [355]21768 ICE in error message due to violation of coding 17852 conventions 17853 * [356]21853 constness of pointer to data member ignored 17854 * [357]21903 Default argument of template function causes a 17855 compile-time error 17856 * [358]21983 multiple diagnostics 17857 * [359]21987 New testsuite failure 17858 g++.dg/warn/conversion-function-1.C 17859 * [360]22153 ICE on invalid template specialization 17860 * [361]22172 Internal compiler error, seg fault. 17861 * [362]21286 filebuf::xsgetn vs pipes 17862 * [363]22233 ICE with wrong number of template parameters 17863 * [364]22508 ICE after invalid operator new 17864 * [365]22545 ICE with pointer to class member & user defined 17865 conversion operator 17866 * [366]23528 Wrong default allocator in ext/hash_map 17867 * [367]23550 char_traits requirements/1.cc test bad math 17868 * [368]23586 Bad diagnostic for invalid namespace-name 17869 * [369]23624 ICE in invert_truthvalue, in fold-const.c 17870 * [370]23639 Bad error message: not a member of '<declaration error>' 17871 * [371]23797 ICE on typename outside template 17872 * [372]23965 Bogus error message: no matching function for call to 17873 'foo(<type error>)' 17874 * [373]24052 &#`label_decl' not supported by dump_expr#<expression 17875 error> 17876 * [374]24580 virtual base class cause exception not to be caught 17877 17878 Problems in generated debug information 17879 17880 * [375]24267 Bad DWARF for altivec vectors 17881 17882 Optimizations issues 17883 17884 * [376]17810 ICE in verify_local_live_at_start 17885 * [377]17860 Wrong generated code for loop with varying bound 17886 * [378]21709 ICE on compile-time complex NaN 17887 * [379]21964 broken tail call at -O2 or more 17888 * [380]22167 Strange optimization bug when using -Os 17889 * [381]22619 Compilation failure for real_const_1.f and 17890 real_const_2.f90 17891 * [382]23241 Invalid code generated for comparison of uchar to 255 17892 * [383]23478 Miscompilation due to reloading of a var that is also 17893 used in EH pad 17894 * [384]24470 segmentation fault in cc1plus when compiling with -O 17895 * [385]24950 ICE in operand_subword_force 17896 17897 Precompiled headers problems 17898 17899 * [386]14400 Cannot compile qt-x11-free-3.3.0 17900 * [387]14940 PCH largefile test fails on various platforms 17901 17902 Preprocessor bugs 17903 17904 * [388]20239 ICE on empty preprocessed input 17905 * [389]15220 "gcc -E -MM -MG" reports missing system headers in 17906 source directory 17907 17908 Testsuite issues 17909 17910 * [390]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 17911 i686-pc-linux-gnu 17912 17913 Alpha specific 17914 17915 * [391]21888 bootstrap failure with linker relaxation enabled 17916 17917 ARM specific 17918 17919 * [392]15342 [arm-linux]: ICE in verify_local_live_at_start 17920 * [393]23985 Memory aliasing information incorrect in inlined memcpy 17921 17922 ColdFile specific 17923 17924 * [394]16719 Illegal move of byte into address register causes 17925 compiler to ICE 17926 17927 HPPA specific 17928 17929 * [395]21723 ICE while building libgfortran 17930 * [396]21841 -mhp-ld/-mgnu-ld documentation 17931 17932 IA-64 specific 17933 17934 * [397]23644 IA-64 hardware models and configuration options 17935 documentation error 17936 * [398]24718 Shared libgcc not used for linking by default 17937 17938 M68000 specific 17939 17940 * [399]18421 ICE in reload_cse_simplify_operands, in postreload.c 17941 17942 MIPS specific 17943 17944 * [400]20621 ICE in change_address_1, in emit-rtl.c 17945 17946 PowerPC and PowerPC64 specific 17947 17948 * [401]18583 error on valid code: const 17949 __attribute__((altivec(vector__))) doesn't work in arrays 17950 * [402]20191 ICE in reload_cse_simplify_operands 17951 * [403]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 17952 * [404]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 17953 * [405]23404 gij trashes args of functions with more than 8 fp args 17954 * [406]23539 C & C++ compiler generating misaligned references 17955 regardless of compiler flags 17956 * [407]24102 floatdisf2_internal2 broken 17957 * [408]24465 -mminimal-toc miscompilation of __thread vars 17958 17959 Solaris specific 17960 17961 * [409]19933 Problem with define of HUGE_VAL in math_c99 17962 * [410]21889 Native Solaris assembler cannot grok DTP-relative debug 17963 symbols 17964 17965 SPARC specific 17966 17967 * [411]19300 PCH failures on sparc-linux 17968 * [412]20301 Assembler labels have a leading "-" 17969 * [413]20673 C PCH testsuite assembly comparison failure 17970 17971 x86 and x86_64 specific 17972 17973 * [414]18582 ICE with arrays of type V2DF 17974 * [415]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 17975 -fsched2-use-traces 17976 * [416]21716 ICE in reg-stack.c's swap_rtx_condition 17977 * [417]24315 amd64 fails -fpeephole2 17978 __________________________________________________________________ 17979 17980 [418]GCC 3.4.6 17981 17982 This is the [419]list of problem reports (PRs) from GCC's bug tracking 17983 system that are known to be fixed in the 3.4.6 release. This list might 17984 not be complete (that is, it is possible that some PRs that have been 17985 fixed are not listed here). 17986 17987 17988 For questions related to the use of GCC, please consult these web 17989 pages and the [420]GCC manuals. If that fails, the 17990 [421]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17991 web pages and the development of GCC are welcome on our developer 17992 list at [422]gcc (a] gcc.gnu.org. All of [423]our lists have public 17993 archives. 17994 17995 Copyright (C) [424]Free Software Foundation, Inc. Verbatim copying and 17996 distribution of this entire article is permitted in any medium, 17997 provided this notice is preserved. 17998 17999 These pages are [425]maintained by the GCC team. Last modified 18000 2025-01-31. 18001 18002 References 18003 18004 1. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 18005 2. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 18006 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 18007 4. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 18008 5. https://gcc.gnu.org/gcc-3.4/mips-abi.html 18009 6. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 18010 7. https://www.boost.org/ 18011 8. https://gcc.gnu.org/PR11953 18012 9. https://gcc.gnu.org/PR8361 18013 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 18014 11. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 18015 12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 18016 13. https://gcc.gnu.org/bugs/#cxx_rvalbind 18017 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 18018 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 18019 16. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 18020 17. https://www.gnu.org/software/classpath/ 18021 18. https://www.eclipse.org/ 18022 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 18023 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 18024 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 18025 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 18026 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 18027 24. https://gcc.gnu.org/gcc-3.4/mips-abi.html 18028 25. https://gcc.gnu.org/gcc-3.4/powerpc-abi.html 18029 26. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 18030 27. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 18031 28. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.0 18032 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 18033 30. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.1 18034 31. https://gcc.gnu.org/PR10129 18035 32. https://gcc.gnu.org/PR14576 18036 33. https://gcc.gnu.org/PR14760 18037 34. https://gcc.gnu.org/PR14671 18038 35. https://gcc.gnu.org/PR15093 18039 36. https://gcc.gnu.org/PR15178 18040 37. https://gcc.gnu.org/PR12753 18041 38. https://gcc.gnu.org/PR13985 18042 39. https://gcc.gnu.org/PR14810 18043 40. https://gcc.gnu.org/PR14883 18044 41. https://gcc.gnu.org/PR15044 18045 42. https://gcc.gnu.org/PR15057 18046 43. https://gcc.gnu.org/PR15064 18047 44. https://gcc.gnu.org/PR15142 18048 45. https://gcc.gnu.org/PR15159 18049 46. https://gcc.gnu.org/PR15165 18050 47. https://gcc.gnu.org/PR15193 18051 48. https://gcc.gnu.org/PR15209 18052 49. https://gcc.gnu.org/PR15227 18053 50. https://gcc.gnu.org/PR15285 18054 51. https://gcc.gnu.org/PR15299 18055 52. https://gcc.gnu.org/PR15329 18056 53. https://gcc.gnu.org/PR15550 18057 54. https://gcc.gnu.org/PR15554 18058 55. https://gcc.gnu.org/PR15640 18059 56. https://gcc.gnu.org/PR15666 18060 57. https://gcc.gnu.org/PR15696 18061 58. https://gcc.gnu.org/PR15701 18062 59. https://gcc.gnu.org/PR15761 18063 60. https://gcc.gnu.org/PR15829 18064 61. https://gcc.gnu.org/PR14538 18065 62. https://gcc.gnu.org/PR12391 18066 63. https://gcc.gnu.org/PR14649 18067 64. https://gcc.gnu.org/PR15004 18068 65. https://gcc.gnu.org/PR15749 18069 66. https://gcc.gnu.org/PR10646 18070 67. https://gcc.gnu.org/PR12077 18071 68. https://gcc.gnu.org/PR13598 18072 69. https://gcc.gnu.org/PR14211 18073 70. https://gcc.gnu.org/PR14220 18074 71. https://gcc.gnu.org/PR14245 18075 72. https://gcc.gnu.org/PR14340 18076 73. https://gcc.gnu.org/PR14600 18077 74. https://gcc.gnu.org/PR14668 18078 75. https://gcc.gnu.org/PR14775 18079 76. https://gcc.gnu.org/PR14821 18080 77. https://gcc.gnu.org/PR14930 18081 78. https://gcc.gnu.org/PR14932 18082 79. https://gcc.gnu.org/PR14950 18083 80. https://gcc.gnu.org/PR14962 18084 81. https://gcc.gnu.org/PR14975 18085 82. https://gcc.gnu.org/PR15002 18086 83. https://gcc.gnu.org/PR15025 18087 84. https://gcc.gnu.org/PR15046 18088 85. https://gcc.gnu.org/PR15069 18089 86. https://gcc.gnu.org/PR15074 18090 87. https://gcc.gnu.org/PR15083 18091 88. https://gcc.gnu.org/PR15096 18092 89. https://gcc.gnu.org/PR15287 18093 90. https://gcc.gnu.org/PR15317 18094 91. https://gcc.gnu.org/PR15337 18095 92. https://gcc.gnu.org/PR15361 18096 93. https://gcc.gnu.org/PR15412 18097 94. https://gcc.gnu.org/PR15427 18098 95. https://gcc.gnu.org/PR15471 18099 96. https://gcc.gnu.org/PR15503 18100 97. https://gcc.gnu.org/PR15507 18101 98. https://gcc.gnu.org/PR15542 18102 99. https://gcc.gnu.org/PR15565 18103 100. https://gcc.gnu.org/PR15625 18104 101. https://gcc.gnu.org/PR15629 18105 102. https://gcc.gnu.org/PR15742 18106 103. https://gcc.gnu.org/PR15775 18107 104. https://gcc.gnu.org/PR15821 18108 105. https://gcc.gnu.org/PR15862 18109 106. https://gcc.gnu.org/PR15875 18110 107. https://gcc.gnu.org/PR15877 18111 108. https://gcc.gnu.org/PR15947 18112 109. https://gcc.gnu.org/PR16020 18113 110. https://gcc.gnu.org/PR16154 18114 111. https://gcc.gnu.org/PR16174 18115 112. https://gcc.gnu.org/PR14315 18116 113. https://gcc.gnu.org/PR15151 18117 114. https://gcc.gnu.org/PR7993 18118 115. https://gcc.gnu.org/PR15228 18119 116. https://gcc.gnu.org/PR15345 18120 117. https://gcc.gnu.org/PR15945 18121 118. https://gcc.gnu.org/PR15526 18122 119. https://gcc.gnu.org/PR14690 18123 120. https://gcc.gnu.org/PR15112 18124 121. https://gcc.gnu.org/PR15067 18125 122. https://gcc.gnu.org/PR1963 18126 123. https://gcc.gnu.org/PR15717 18127 124. https://gcc.gnu.org/PR14782 18128 125. https://gcc.gnu.org/PR14828 18129 126. https://gcc.gnu.org/PR15202 18130 127. https://gcc.gnu.org/PR14610 18131 128. https://gcc.gnu.org/PR14813 18132 129. https://gcc.gnu.org/PR14857 18133 130. https://gcc.gnu.org/PR15598 18134 131. https://gcc.gnu.org/PR15653 18135 132. https://gcc.gnu.org/PR15189 18136 133. https://gcc.gnu.org/PR15331 18137 134. https://gcc.gnu.org/PR16144 18138 135. https://gcc.gnu.org/PR16176 18139 136. https://gcc.gnu.org/PR11591 18140 137. https://gcc.gnu.org/PR12028 18141 138. https://gcc.gnu.org/PR14478 18142 139. https://gcc.gnu.org/PR14567 18143 140. https://gcc.gnu.org/PR14715 18144 141. https://gcc.gnu.org/PR14902 18145 142. https://gcc.gnu.org/PR14924 18146 143. https://gcc.gnu.org/PR14960 18147 144. https://gcc.gnu.org/PR15106 18148 145. https://gcc.gnu.org/PR16026 18149 146. https://gcc.gnu.org/PR15191 18150 147. https://gcc.gnu.org/PR15662 18151 148. https://gcc.gnu.org/PR15054 18152 149. https://gcc.gnu.org/PR15783 18153 150. https://gcc.gnu.org/PR15626 18154 151. https://gcc.gnu.org/PR14326 18155 152. https://gcc.gnu.org/PR14723 18156 153. https://gcc.gnu.org/PR15290 18157 154. https://gcc.gnu.org/PR15250 18158 155. https://gcc.gnu.org/PR15551 18159 156. https://gcc.gnu.org/PR8309 18160 157. https://gcc.gnu.org/PR13250 18161 158. https://gcc.gnu.org/PR13803 18162 159. https://gcc.gnu.org/PR14093 18163 160. https://gcc.gnu.org/PR14457 18164 161. https://gcc.gnu.org/PR14542 18165 162. https://gcc.gnu.org/PR15100 18166 163. https://gcc.gnu.org/PR15296 18167 164. https://gcc.gnu.org/PR15396 18168 165. https://gcc.gnu.org/PR15782 18169 166. https://gcc.gnu.org/PR11610 18170 167. https://gcc.gnu.org/PR15488 18171 168. https://gcc.gnu.org/PR15489 18172 169. https://gcc.gnu.org/PR13928 18173 170. https://gcc.gnu.org/PR14150 18174 171. https://gcc.gnu.org/PR14949 18175 172. https://gcc.gnu.org/PR15123 18176 173. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.2 18177 174. https://gcc.gnu.org/PR16469 18178 175. https://gcc.gnu.org/PR16344 18179 176. https://gcc.gnu.org/PR16842 18180 177. https://gcc.gnu.org/PR12608 18181 178. https://gcc.gnu.org/PR14492 18182 179. https://gcc.gnu.org/PR15461 18183 180. https://gcc.gnu.org/PR15890 18184 181. https://gcc.gnu.org/PR16180 18185 182. https://gcc.gnu.org/PR16224 18186 183. https://gcc.gnu.org/PR16408 18187 184. https://gcc.gnu.org/PR16529 18188 185. https://gcc.gnu.org/PR16698 18189 186. https://gcc.gnu.org/PR16706 18190 187. https://gcc.gnu.org/PR16810 18191 188. https://gcc.gnu.org/PR16851 18192 189. https://gcc.gnu.org/PR16870 18193 190. https://gcc.gnu.org/PR16904 18194 191. https://gcc.gnu.org/PR16905 18195 192. https://gcc.gnu.org/PR16964 18196 193. https://gcc.gnu.org/PR17068 18197 194. https://gcc.gnu.org/PR16366 18198 195. https://gcc.gnu.org/PR15345 18199 196. https://gcc.gnu.org/PR16590 18200 197. https://gcc.gnu.org/PR16693 18201 198. https://gcc.gnu.org/PR17078 18202 199. https://gcc.gnu.org/PR13956 18203 200. https://gcc.gnu.org/PR16684 18204 201. https://gcc.gnu.org/PR12658 18205 202. https://gcc.gnu.org/PR13092 18206 203. https://gcc.gnu.org/PR15320 18207 204. https://gcc.gnu.org/PR16246 18208 205. https://gcc.gnu.org/PR16273 18209 206. https://gcc.gnu.org/PR16401 18210 207. https://gcc.gnu.org/PR16411 18211 208. https://gcc.gnu.org/PR16489 18212 209. https://gcc.gnu.org/PR16618 18213 210. https://gcc.gnu.org/PR16637 18214 211. https://gcc.gnu.org/PR16717 18215 212. https://gcc.gnu.org/PR16813 18216 213. https://gcc.gnu.org/PR16853 18217 214. https://gcc.gnu.org/PR16889 18218 215. https://gcc.gnu.org/PR16959 18219 216. https://gcc.gnu.org/PR7587 18220 217. https://gcc.gnu.org/PR16473 18221 218. https://gcc.gnu.org/PR16478 18222 219. https://gcc.gnu.org/PR10695 18223 220. https://gcc.gnu.org/PR16974 18224 221. https://gcc.gnu.org/PR16298 18225 222. https://gcc.gnu.org/PR17113 18226 223. https://gcc.gnu.org/PR14697 18227 224. https://gcc.gnu.org/PR15869 18228 225. https://gcc.gnu.org/PR16325 18229 226. https://gcc.gnu.org/PR16357 18230 227. https://gcc.gnu.org/PR16380 18231 228. https://gcc.gnu.org/PR16407 18232 229. https://gcc.gnu.org/PR16643 18233 230. https://gcc.gnu.org/PR15927 18234 231. https://gcc.gnu.org/PR15948 18235 232. https://gcc.gnu.org/PR17019 18236 233. https://gcc.gnu.org/PR16130 18237 234. https://gcc.gnu.org/PR16142 18238 235. https://gcc.gnu.org/PR16278 18239 236. https://gcc.gnu.org/PR16414 18240 237. https://gcc.gnu.org/PR16445 18241 238. https://gcc.gnu.org/PR16490 18242 239. https://gcc.gnu.org/PR16683 18243 240. https://gcc.gnu.org/PR16195 18244 241. https://gcc.gnu.org/PR16239 18245 242. https://gcc.gnu.org/PR16199 18246 243. https://gcc.gnu.org/PR16416 18247 244. https://gcc.gnu.org/PR16430 18248 245. https://gcc.gnu.org/PR16379 18249 246. https://gcc.gnu.org/PR17093 18250 247. https://gcc.gnu.org/PR17119 18251 248. https://gcc.gnu.org/PR15928 18252 249. https://gcc.gnu.org/PR16210 18253 250. https://gcc.gnu.org/PR15488 18254 251. https://gcc.gnu.org/PR16250 18255 252. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.3 18256 253. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 18257 254. https://gcc.gnu.org/PR17369 18258 255. https://gcc.gnu.org/PR17850 18259 256. https://gcc.gnu.org/PR13948 18260 257. https://gcc.gnu.org/PR14492 18261 258. https://gcc.gnu.org/PR16301 18262 259. https://gcc.gnu.org/PR16566 18263 260. https://gcc.gnu.org/PR17023 18264 261. https://gcc.gnu.org/PR17027 18265 262. https://gcc.gnu.org/PR17524 18266 263. https://gcc.gnu.org/PR17826 18267 264. https://gcc.gnu.org/PR15526 18268 265. https://gcc.gnu.org/PR16999 18269 266. https://gcc.gnu.org/PR17503 18270 267. https://gcc.gnu.org/PR17581 18271 268. https://gcc.gnu.org/PR18129 18272 269. https://gcc.gnu.org/PR10975 18273 270. https://gcc.gnu.org/PR11722 18274 271. https://gcc.gnu.org/PR14534 18275 272. https://gcc.gnu.org/PR15172 18276 273. https://gcc.gnu.org/PR15786 18277 274. https://gcc.gnu.org/PR16162 18278 275. https://gcc.gnu.org/PR16612 18279 276. https://gcc.gnu.org/PR16715 18280 277. https://gcc.gnu.org/PR16848 18281 278. https://gcc.gnu.org/PR17132 18282 279. https://gcc.gnu.org/PR17259 18283 280. https://gcc.gnu.org/PR17327 18284 281. https://gcc.gnu.org/PR17393 18285 282. https://gcc.gnu.org/PR17501 18286 283. https://gcc.gnu.org/PR17537 18287 284. https://gcc.gnu.org/PR17585 18288 285. https://gcc.gnu.org/PR17821 18289 286. https://gcc.gnu.org/PR17829 18290 287. https://gcc.gnu.org/PR17851 18291 288. https://gcc.gnu.org/PR17976 18292 289. https://gcc.gnu.org/PR18020 18293 290. https://gcc.gnu.org/PR18093 18294 291. https://gcc.gnu.org/PR18140 18295 292. https://gcc.gnu.org/PR17541 18296 293. https://gcc.gnu.org/PR17853 18297 294. https://gcc.gnu.org/PR17245 18298 295. https://gcc.gnu.org/PR17167 18299 296. https://gcc.gnu.org/PR17277 18300 297. https://gcc.gnu.org/PR17505 18301 298. https://gcc.gnu.org/PR17684 18302 299. https://gcc.gnu.org/PR17384 18303 300. https://gcc.gnu.org/PR17770 18304 301. https://gcc.gnu.org/PR11476 18305 302. https://gcc.gnu.org/PR14064 18306 303. https://gcc.gnu.org/PR14678 18307 304. https://gcc.gnu.org/PR15583 18308 305. https://gcc.gnu.org/PR15790 18309 306. https://gcc.gnu.org/PR15886 18310 307. https://gcc.gnu.org/PR16884 18311 308. https://gcc.gnu.org/PR13841 18312 309. https://gcc.gnu.org/PR15860 18313 310. https://gcc.gnu.org/PR17465 18314 311. https://gcc.gnu.org/PR17469 18315 312. https://gcc.gnu.org/PR18138 18316 313. https://gcc.gnu.org/PR15498 18317 314. https://gcc.gnu.org/PR15747 18318 315. https://gcc.gnu.org/PR16406 18319 316. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.4 18320 317. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 18321 318. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.5 18322 319. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 18323 320. https://gcc.gnu.org/PR24688 18324 321. https://gcc.gnu.org/PR17188 18325 322. https://gcc.gnu.org/PR20187 18326 323. https://gcc.gnu.org/PR21873 18327 324. https://gcc.gnu.org/PR21899 18328 325. https://gcc.gnu.org/PR22061 18329 326. https://gcc.gnu.org/PR22208 18330 327. https://gcc.gnu.org/PR22458 18331 328. https://gcc.gnu.org/PR22589 18332 329. https://gcc.gnu.org/PR24101 18333 330. https://gcc.gnu.org/PR10611 18334 331. https://gcc.gnu.org/PR13377 18335 332. https://gcc.gnu.org/PR16002 18336 333. https://gcc.gnu.org/PR17413 18337 334. https://gcc.gnu.org/PR17609 18338 335. https://gcc.gnu.org/PR17618 18339 336. https://gcc.gnu.org/PR18124 18340 337. https://gcc.gnu.org/PR18155 18341 338. https://gcc.gnu.org/PR18177 18342 339. https://gcc.gnu.org/PR18368 18343 340. https://gcc.gnu.org/PR18378 18344 341. https://gcc.gnu.org/PR18466 18345 342. https://gcc.gnu.org/PR18512 18346 343. https://gcc.gnu.org/PR18545 18347 344. https://gcc.gnu.org/PR18738 18348 345. https://gcc.gnu.org/PR18803 18349 346. https://gcc.gnu.org/PR19004 18350 347. https://gcc.gnu.org/PR19208 18351 348. https://gcc.gnu.org/PR19253 18352 349. https://gcc.gnu.org/PR19608 18353 350. https://gcc.gnu.org/PR19884 18354 351. https://gcc.gnu.org/PR20153 18355 352. https://gcc.gnu.org/PR20563 18356 353. https://gcc.gnu.org/PR20789 18357 354. https://gcc.gnu.org/PR21336 18358 355. https://gcc.gnu.org/PR21768 18359 356. https://gcc.gnu.org/PR21853 18360 357. https://gcc.gnu.org/PR21903 18361 358. https://gcc.gnu.org/PR21983 18362 359. https://gcc.gnu.org/PR21987 18363 360. https://gcc.gnu.org/PR22153 18364 361. https://gcc.gnu.org/PR22172 18365 362. https://gcc.gnu.org/PR21286 18366 363. https://gcc.gnu.org/PR22233 18367 364. https://gcc.gnu.org/PR22508 18368 365. https://gcc.gnu.org/PR22545 18369 366. https://gcc.gnu.org/PR23528 18370 367. https://gcc.gnu.org/PR23550 18371 368. https://gcc.gnu.org/PR23586 18372 369. https://gcc.gnu.org/PR23624 18373 370. https://gcc.gnu.org/PR23639 18374 371. https://gcc.gnu.org/PR23797 18375 372. https://gcc.gnu.org/PR23965 18376 373. https://gcc.gnu.org/PR24052 18377 374. https://gcc.gnu.org/PR24580 18378 375. https://gcc.gnu.org/PR24267 18379 376. https://gcc.gnu.org/PR17810 18380 377. https://gcc.gnu.org/PR17860 18381 378. https://gcc.gnu.org/PR21709 18382 379. https://gcc.gnu.org/PR21964 18383 380. https://gcc.gnu.org/PR22167 18384 381. https://gcc.gnu.org/PR22619 18385 382. https://gcc.gnu.org/PR23241 18386 383. https://gcc.gnu.org/PR23478 18387 384. https://gcc.gnu.org/PR24470 18388 385. https://gcc.gnu.org/PR24950 18389 386. https://gcc.gnu.org/PR14400 18390 387. https://gcc.gnu.org/PR14940 18391 388. https://gcc.gnu.org/PR20239 18392 389. https://gcc.gnu.org/PR15220 18393 390. https://gcc.gnu.org/PR19275 18394 391. https://gcc.gnu.org/PR21888 18395 392. https://gcc.gnu.org/PR15342 18396 393. https://gcc.gnu.org/PR23985 18397 394. https://gcc.gnu.org/PR16719 18398 395. https://gcc.gnu.org/PR21723 18399 396. https://gcc.gnu.org/PR21841 18400 397. https://gcc.gnu.org/PR23644 18401 398. https://gcc.gnu.org/PR24718 18402 399. https://gcc.gnu.org/PR18421 18403 400. https://gcc.gnu.org/PR20621 18404 401. https://gcc.gnu.org/PR18583 18405 402. https://gcc.gnu.org/PR20191 18406 403. https://gcc.gnu.org/PR22083 18407 404. https://gcc.gnu.org/PR23070 18408 405. https://gcc.gnu.org/PR23404 18409 406. https://gcc.gnu.org/PR23539 18410 407. https://gcc.gnu.org/PR24102 18411 408. https://gcc.gnu.org/PR24465 18412 409. https://gcc.gnu.org/PR19933 18413 410. https://gcc.gnu.org/PR21889 18414 411. https://gcc.gnu.org/PR19300 18415 412. https://gcc.gnu.org/PR20301 18416 413. https://gcc.gnu.org/PR20673 18417 414. https://gcc.gnu.org/PR18582 18418 415. https://gcc.gnu.org/PR19340 18419 416. https://gcc.gnu.org/PR21716 18420 417. https://gcc.gnu.org/PR24315 18421 418. https://gcc.gnu.org/gcc-3.4/changes.html#GCC3.4.6 18422 419. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 18423 420. https://gcc.gnu.org/onlinedocs/ 18424 421. mailto:gcc-help (a] gcc.gnu.org 18425 422. mailto:gcc (a] gcc.gnu.org 18426 423. https://gcc.gnu.org/lists.html 18427 424. https://www.fsf.org/ 18428 425. https://gcc.gnu.org/about.html 18429 ====================================================================== 18430 http://gcc.gnu.org/gcc-3.3/index.html 18431 18432 GCC 3.3 Release Series 18433 18434 (This release series is no longer supported.) 18435 18436 May 03, 2005 18437 18438 The [1]GNU project and the GCC developers are pleased to announce the 18439 release of GCC 3.3.6. 18440 18441 This release is a bug-fix release, containing fixes for regressions in 18442 GCC 3.3.5 relative to previous releases of GCC. 18443 18444 This release is the last of the series 3.3.x. 18445 18446 The GCC 3.3 release series includes numerous [2]new features, 18447 improvements, bug fixes, and other changes, thanks to an [3]amazing 18448 group of volunteers. 18449 18450 Release History 18451 18452 GCC 3.3.6 18453 May 3, 2005 ([4]changes) 18454 18455 GCC 3.3.5 18456 September 30, 2004 ([5]changes) 18457 18458 GCC 3.3.4 18459 May 31, 2004 ([6]changes) 18460 18461 GCC 3.3.3 18462 February 14, 2004 ([7]changes) 18463 18464 GCC 3.3.2 18465 October 16, 2003 ([8]changes) 18466 18467 GCC 3.3.1 18468 August 8, 2003 ([9]changes) 18469 18470 GCC 3.3 18471 May 14, 2003 ([10]changes) 18472 18473 References and Acknowledgements 18474 18475 GCC used to stand for the GNU C Compiler, but since the compiler 18476 supports several other languages aside from C, it now stands for the 18477 GNU Compiler Collection. 18478 18479 A list of [11]successful builds is updated as new information becomes 18480 available. 18481 18482 The GCC developers would like to thank the numerous people that have 18483 contributed new features, improvements, bug fixes, and other changes as 18484 well as test results to GCC. This [12]amazing group of volunteers is 18485 what makes GCC successful. 18486 18487 For additional information about GCC please refer to the [13]GCC 18488 project web site or contact the [14]GCC development mailing list. 18489 18490 To obtain GCC please use [15]our mirror sites, or our CVS server. 18491 18492 18493 For questions related to the use of GCC, please consult these web 18494 pages and the [16]GCC manuals. If that fails, the 18495 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18496 web pages and the development of GCC are welcome on our developer 18497 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 18498 archives. 18499 18500 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 18501 distribution of this entire article is permitted in any medium, 18502 provided this notice is preserved. 18503 18504 These pages are [21]maintained by the GCC team. Last modified 18505 2022-10-26. 18506 18507 References 18508 18509 1. http://www.gnu.org/ 18510 2. https://gcc.gnu.org/gcc-3.3/changes.html 18511 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18512 4. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 18513 5. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 18514 6. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 18515 7. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 18516 8. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 18517 9. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 18518 10. https://gcc.gnu.org/gcc-3.3/changes.html 18519 11. https://gcc.gnu.org/gcc-3.3/buildstat.html 18520 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18521 13. https://gcc.gnu.org/index.html 18522 14. mailto:gcc (a] gcc.gnu.org 18523 15. https://gcc.gnu.org/mirrors.html 18524 16. https://gcc.gnu.org/onlinedocs/ 18525 17. mailto:gcc-help (a] gcc.gnu.org 18526 18. mailto:gcc (a] gcc.gnu.org 18527 19. https://gcc.gnu.org/lists.html 18528 20. https://www.fsf.org/ 18529 21. https://gcc.gnu.org/about.html 18530 ====================================================================== 18531 http://gcc.gnu.org/gcc-3.3/changes.html 18532 18533 GCC 3.3 Release Series 18534 Changes, New Features, and Fixes 18535 18536 The latest release in the 3.3 release series is [1]GCC 3.3.6. 18537 18538 Caveats 18539 18540 * The preprocessor no longer accepts multi-line string literals. They 18541 were deprecated in 3.0, 3.1, and 3.2. 18542 * The preprocessor no longer supports the -A- switch when appearing 18543 alone. -A- followed by an assertion is still supported. 18544 * Support for all the systems [2]obsoleted in GCC 3.1 has been 18545 removed from GCC 3.3. See below for a [3]list of systems which are 18546 obsoleted in this release. 18547 * Checking for null format arguments has been decoupled from the rest 18548 of the format checking mechanism. Programs which use the format 18549 attribute may regain this functionality by using the new [4]nonnull 18550 function attribute. Note that all functions for which GCC has a 18551 built-in format attribute, an appropriate built-in nonnull 18552 attribute is also applied. 18553 * The DWARF (version 1) debugging format has been deprecated and will 18554 be removed in a future version of GCC. Version 2 of the DWARF 18555 debugging format will continue to be supported for the foreseeable 18556 future. 18557 * The C and Objective-C compilers no longer accept the "Naming Types" 18558 extension (typedef foo = bar); it was already unavailable in C++. 18559 Code which uses it will need to be changed to use the "typeof" 18560 extension instead: typedef typeof(bar) foo. (We have removed this 18561 extension without a period of deprecation because it has caused the 18562 compiler to crash since version 3.0 and no one noticed until very 18563 recently. Thus we conclude it is not in widespread use.) 18564 * The -traditional C compiler option has been removed. It was 18565 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 18566 available.) The <varargs.h> header, used for writing variadic 18567 functions in traditional C, still exists but will produce an error 18568 message if used. 18569 * GCC 3.3.1 automatically places zero-initialized variables in the 18570 .bss section on some operating systems. Versions of GNU Emacs up to 18571 (and including) 21.3 will not work correctly when using this 18572 optimization; you can use -fno-zero-initialized-in-bss to disable 18573 it. 18574 18575 General Optimizer Improvements 18576 18577 * A new scheme for accurately describing processor pipelines, the 18578 [5]DFA scheduler, has been added. 18579 * Pavel Nejedly, Charles University Prague, has contributed new file 18580 format used by the edge coverage profiler (-fprofile-arcs). 18581 The new format is robust and diagnoses common mistakes where 18582 profiles from different versions (or compilations) of the program 18583 are combined resulting in nonsensical profiles and slow code to 18584 produced with profile feedback. Additionally this format allows 18585 extra data to be gathered. Currently, overall statistics are 18586 produced helping optimizers to identify hot spots of a program 18587 globally replacing the old intra-procedural scheme and resulting in 18588 better code. Note that the gcov tool from older GCC versions will 18589 not be able to parse the profiles generated by GCC 3.3 and vice 18590 versa. 18591 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 18592 pass enabled using -ftracer. This pass simplifies the control flow 18593 of functions allowing other optimizations to do better job. 18594 He also contributed the function reordering pass 18595 (-freorder-functions) to optimize function placement using profile 18596 feedback. 18597 18598 New Languages and Language specific improvements 18599 18600 C/ObjC/C++ 18601 18602 * The preprocessor now accepts directives within macro arguments. It 18603 processes them just as if they had not been within macro arguments. 18604 * The separate ISO and traditional preprocessors have been completely 18605 removed. The front end handles either type of preprocessed output 18606 if necessary. 18607 * In C99 mode preprocessor arithmetic is done in the precision of the 18608 target's intmax_t, as required by that standard. 18609 * The preprocessor can now copy comments inside macros to the output 18610 file when the macro is expanded. This feature, enabled using the 18611 -CC option, is intended for use by applications which place 18612 metadata or directives inside comments, such as lint. 18613 * The method of constructing the list of directories to be searched 18614 for header files has been revised. If a directory named by a -I 18615 option is a standard system include directory, the option is 18616 ignored to ensure that the default search order for system 18617 directories and the special treatment of system header files are 18618 not defeated. 18619 * A few more [6]ISO C99 features now work correctly. 18620 * A new function attribute, nonnull, has been added which allows 18621 pointer arguments to functions to be specified as requiring a 18622 non-null value. The compiler currently uses this information to 18623 issue a warning when it detects a null value passed in such an 18624 argument slot. 18625 * A new type attribute, may_alias, has been added. Accesses to 18626 objects with types with this attribute are not subjected to 18627 type-based alias analysis, but are instead assumed to be able to 18628 alias any other type of objects, just like the char type. 18629 18630 C++ 18631 18632 * Type based alias analysis has been implemented for C++ aggregate 18633 types. 18634 18635 Objective-C 18636 18637 * Generate an error if Objective-C objects are passed by value in 18638 function and method calls. 18639 * When -Wselector is used, check the whole list of selectors at the 18640 end of compilation, and emit a warning if a @selector() is not 18641 known. 18642 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 18643 * No longer need to include objc/objc-class.h to compile self calls 18644 in class methods (NeXT runtime only). 18645 * New -Wundeclared-selector option. 18646 * Removed selector bloating which was causing object files to be 10% 18647 bigger on average (GNU runtime only). 18648 * Using at run time @protocol() objects has been fixed in certain 18649 situations (GNU runtime only). 18650 * Type checking has been fixed and improved in many situations 18651 involving protocols. 18652 18653 Java 18654 18655 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 18656 1.4) API. 18657 * The JDK 1.4 assert facility has been implemented. 18658 * The bytecode interpreter is now direct threaded and thus faster. 18659 18660 Fortran 18661 18662 * Fortran improvements are listed in [7]the Fortran documentation. 18663 18664 Ada 18665 18666 * Ada tasking now works with glibc 2.3.x threading libraries. 18667 18668 New Targets and Target Specific Improvements 18669 18670 * The following changes have been made to the HP-PA port: 18671 + The port now defaults to scheduling for the PA8000 series of 18672 processors. 18673 + Scheduling support for the PA7300 processor has been added. 18674 + The 32-bit port now supports weak symbols under HP-UX 11. 18675 + The handling of initializers and finalizers has been improved 18676 under HP-UX 11. The 64-bit port no longer uses collect2. 18677 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 18678 + ABI fixes to correct the passing of small structures by value. 18679 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 18680 use the DFA processor pipeline description. 18681 * The following NetBSD configurations for the SuperH processor family 18682 have been added: 18683 + SH3, big-endian, sh-*-netbsdelf* 18684 + SH3, little-endian, shle-*-netbsdelf* 18685 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 18686 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 18687 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 18688 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 18689 * The following changes have been made to the IA-32/x86-64 port: 18690 + SSE2 and 3dNOW! intrinsics are now supported. 18691 + Support for thread local storage has been added to the IA-32 18692 and x86-64 ports. 18693 + The x86-64 port has been significantly improved. 18694 * The following changes have been made to the MIPS port: 18695 + All configurations now accept the -mabi switch. Note that you 18696 will need appropriate multilibs for this option to work 18697 properly. 18698 + ELF configurations will always pass an ABI flag to the 18699 assembler, except when the MIPS EABI is selected. 18700 + -mabi=64 no longer selects MIPS IV code. 18701 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 18702 been removed from this release. 18703 + -march now changes the core ISA level. In previous releases, 18704 it would change the use of processor-specific extensions, but 18705 would leave the core ISA unchanged. For example, mips64-elf 18706 -march=r8000 will now generate MIPS IV code. 18707 + Under most configurations, -mipsN now acts as a synonym for 18708 -march. 18709 + There are some new preprocessor macros to describe the -march 18710 and -mtune settings. See the documentation of those options 18711 for details. 18712 + Support for the NEC VR-Series processors has been added. This 18713 includes the 54xx, 5500, and 41xx series. 18714 + Support for the Sandcraft sr71k processor has been added. 18715 * The following changes have been made to the S/390 port: 18716 + Support to build the Java runtime libraries has been added. 18717 Java is now enabled by default on s390-*-linux* and 18718 s390x-*-linux* targets. 18719 + Multilib support for the s390x-*-linux* target has been added; 18720 this allows to build 31-bit binaries using the -m31 option. 18721 + Support for thread local storage has been added. 18722 + Inline assembler code may now use the 'Q' constraint to 18723 specify memory operands without index register. 18724 + Various platform-specific performance improvements have been 18725 implemented; in particular, the compiler now uses the BRANCH 18726 ON COUNT family of instructions and makes more frequent use of 18727 the TEST UNDER MASK family of instructions. 18728 * The following changes have been made to the PowerPC port: 18729 + Support for IBM Power4 processor added. 18730 + Support for Motorola e500 SPE added. 18731 + Support for AIX 5.2 added. 18732 + Function and Data sections now supported on AIX. 18733 + Sibcall optimizations added. 18734 * The support for H8 Tiny is added to the H8/300 port with -mn. 18735 18736 [8]Obsolete Systems 18737 18738 Support for a number of older systems has been declared obsolete in GCC 18739 3.3. Unless there is activity to revive them, the next release of GCC 18740 will have their sources permanently removed. 18741 18742 All configurations of the following processor architectures have been 18743 declared obsolete: 18744 * Matsushita MN10200, mn10200-*-* 18745 * Motorola 88000, m88k-*-* 18746 * IBM ROMP, romp-*-* 18747 18748 Also, some individual systems have been obsoleted: 18749 * Alpha 18750 + Interix, alpha*-*-interix* 18751 + Linux libc1, alpha*-*-linux*libc1* 18752 + Linux ECOFF, alpha*-*-linux*ecoff* 18753 * ARM 18754 + Generic a.out, arm*-*-aout* 18755 + Conix, arm*-*-conix* 18756 + "Old ABI," arm*-*-oabi 18757 + StrongARM/COFF, strongarm-*-coff* 18758 * HPPA (PA-RISC) 18759 + Generic OSF, hppa1.0-*-osf* 18760 + Generic BSD, hppa1.0-*-bsd* 18761 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 18762 + HiUX, hppa*-*-hiux* 18763 + Mach Lites, hppa*-*-lites* 18764 * Intel 386 family 18765 + Windows NT 3.x, i?86-*-win32 18766 * MC68000 family 18767 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 18768 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 18769 m68k-sun-mach* 18770 + AT&T systems, m68000-att-sysv* 18771 + Atari systems, m68k-atari-sysv* 18772 + Motorola systems, m68k-motorola-sysv* 18773 + NCR systems, m68k-ncr-sysv* 18774 + Plexus systems, m68k-plexus-sysv* 18775 + Commodore systems, m68k-cbm-sysv* 18776 + Citicorp TTI, m68k-tti-* 18777 + Unos, m68k-crds-unos* 18778 + Concurrent RTU, m68k-ccur-rtu* 18779 + Linux a.out, m68k-*-linux*aout* 18780 + Linux libc1, m68k-*-linux*libc1* 18781 + pSOS, m68k-*-psos* 18782 * MIPS 18783 + Generic ECOFF, mips*-*-ecoff* 18784 + SINIX, mips-sni-sysv4 18785 + Orion RTEMS, mips64orion-*-rtems* 18786 * National Semiconductor 32000 18787 + OpenBSD, ns32k-*-openbsd* 18788 * POWER (aka RS/6000) and PowerPC 18789 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 18790 + Bull BOSX, rs6000-bull-bosx 18791 + Generic Mach, rs6000-*-mach* 18792 + Generic SysV, powerpc*-*-sysv* 18793 + Linux libc1, powerpc*-*-linux*libc1* 18794 * Sun SPARC 18795 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 18796 sparclite-*-aout*, and sparc86x-*-aout* 18797 + NetBSD a.out, sparc-*-netbsd*aout* 18798 + Generic BSD, sparc-*-bsd* 18799 + ChorusOS, sparc-*-chorusos* 18800 + Linux a.out, sparc-*-linux*aout* 18801 + Linux libc1, sparc-*-linux*libc1* 18802 + LynxOS, sparc-*-lynxos* 18803 + Solaris on HAL hardware, sparc-hal-solaris2* 18804 + SunOS versions 3 and 4, sparc-*-sunos[34]* 18805 * NEC V850 18806 + RTEMS, v850-*-rtems* 18807 * VAX 18808 + VMS, vax-*-vms* 18809 18810 Documentation improvements 18811 18812 Other significant improvements 18813 18814 * Almost all front-end dependencies in the compiler have been 18815 separated out into a set of language hooks. This should make adding 18816 a new front end clearer and easier. 18817 * One effect of removing the separate preprocessor is a small 18818 increase in the robustness of the compiler in general, and the 18819 maintainability of target descriptions. Previously target-specific 18820 built-in macros and others, such as __FAST_MATH__, had to be 18821 handled with so-called specs that were hard to maintain. Often they 18822 would fail to behave properly when conflicting options were 18823 supplied on the command line, and define macros in the user's 18824 namespace even when strict ISO compliance was requested. 18825 Integrating the preprocessor has cleanly solved these issues. 18826 * The Makefile suite now supports redirection of make install by 18827 means of the variable DESTDIR. 18828 __________________________________________________________________ 18829 18830 GCC 3.3 18831 18832 Detailed release notes for the GCC 3.3 release follow. 18833 18834 Bug Fixes 18835 18836 bootstrap failures 18837 18838 * [9]10140 cross compiler build failures: missing __mempcpy (DUP: 18839 [10]10198,[11]10338) 18840 18841 Internal compiler errors (multi-platform) 18842 18843 * [12]3581 large string causes segmentation fault in cc1 18844 * [13]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 18845 * [14]5533 (c++) ICE when processing std::accumulate(begin, end, 18846 init, invalid_op) 18847 * [15]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 18848 * [16]6412 (c++) ICE in retrieve_specialization 18849 * [17]6620 (c++) partial template specialization causes an ICE 18850 (segmentation fault) 18851 * [18]6663 (c++) ICE with attribute aligned 18852 * [19]7068 ICE with incomplete types 18853 * [20]7083 (c++) ICE using -gstabs with dodgy class derivation 18854 * [21]7647 (c++) ICE when data member has the name of the enclosing 18855 class 18856 * [22]7675 ICE in fixup_var_refs_1 18857 * [23]7718 'complex' template instantiation causes ICE 18858 * [24]8116 (c++) ICE in member template function 18859 * [25]8358 (ada) Ada compiler accesses freed memory, crashes 18860 * [26]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 18861 fault 18862 * [27]8564 (c++) ICE in find_function_data, in function.c 18863 * [28]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 18864 * [29]8766 (c++) ICE after failed initialization of static template 18865 variable 18866 * [30]8803 ICE in instantiate_virtual_regs_1, in function.c 18867 * [31]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 18868 * [32]8906 (c++) ICE (Segmentation fault) when parsing nested-class 18869 definition 18870 * [33]9216 (c++) ICE on missing template parameter 18871 * [34]9261 (c++) ICE in arg_assoc, in cp/decl2.c 18872 * [35]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 18873 loop 18874 * [36]9429 (c++) ICE in template instantiation with a pointered new 18875 operator 18876 * [37]9516 Internal error when using a big array 18877 * [38]9600 (c++) ICE with typedefs in template class 18878 * [39]9629 (c++) virtual inheritance segfault 18879 * [40]9672 (c++) ICE: Error reporting routines re-entered 18880 * [41]9749 (c++) ICE in write_expression on invalid function 18881 prototype 18882 * [42]9794 (fortran) ICE: floating point exception during constant 18883 folding 18884 * [43]9829 (c++) Missing colon in nested namespace usage causes ICE 18885 * [44]9916 (c++) ICE with noreturn function in ?: statement 18886 * [45]9936 ICE with local function and variable-length 2d array 18887 * [46]10262 (c++) cc1plus crashes with large generated code 18888 * [47]10278 (c++) ICE in parser for invalid code 18889 * [48]10446 (c++) ICE on definition of nonexistent member function of 18890 nested class in a class template 18891 * [49]10451 (c++) ICE in grokdeclarator on spurious mutable 18892 declaration 18893 * [50]10506 (c++) ICE in build_new at cp/init.c with 18894 -fkeep-inline-functions and multiple inheritance 18895 * [51]10549 (c++) ICE in store_bit_field on bitfields that exceed the 18896 precision of the declared type 18897 18898 Optimization bugs 18899 18900 * [52]2001 Inordinately long compile times in reload CSE regs 18901 * [53]2391 Exponential compilation time explosion in combine 18902 * [54]2960 Duplicate loop conditions even with -Os 18903 * [55]4046 redundant conditional branch 18904 * [56]6405 Loop-unrolling related performance regressions 18905 * [57]6798 very long compile time with large case-statement 18906 * [58]6871 const objects shouldn't be moved to .bss 18907 * [59]6909 problem w/ -Os on modified loop-2c.c test case 18908 * [60]7189 gcc -O2 -Wall does not print ``control reaches end of 18909 non-void function'' warning 18910 * [61]7642 optimization problem with signbit() 18911 * [62]8634 incorrect code for inlining of memcpy under -O2 18912 * [63]8750 Cygwin prolog generation erroneously emitting __alloca as 18913 regular function call 18914 18915 C front end 18916 18917 * [64]2161 long if-else cascade overflows parser stack 18918 * [65]4319 short accepted on typedef'd char 18919 * [66]8602 incorrect line numbers in warning messages when using 18920 inline functions 18921 * [67]9177 -fdump-translation-unit: C front end deletes function_decl 18922 AST nodes and breaks debugging dumps 18923 * [68]9853 miscompilation of non-constant structure initializer 18924 18925 c++ compiler and library 18926 18927 * [69]45 legal template specialization code is rejected (DUP: 18928 [70]3784) 18929 * [71]764 lookup failure: friend operator and dereferencing a pointer 18930 and templates (DUP: [72]5116) 18931 * [73]2862 gcc accepts invalid explicit instantiation syntax (DUP: 18932 2863) 18933 * [74]3663 G++ doesn't check access control during template 18934 instantiation 18935 * [75]3797 gcc fails to emit explicit specialization of a template 18936 member 18937 * [76]3948 Two destructors are called when no copy destructor is 18938 defined (ABI change) 18939 * [77]4137 Conversion operator within template is not accepted 18940 * [78]4361 bogus ambiguity taking the address of a member template 18941 * [79]4802 g++ accepts illegal template code (access to private 18942 member; DUP: [80]5837) 18943 * [81]4803 inline function is used but never defined, and g++ does 18944 not object 18945 * [82]5094 Partial specialization cannot be friend? 18946 * [83]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 18947 * [84]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 18948 time 18949 * [85]7015 certain __asm__ constructs rejected 18950 * [86]7086 compile time regression (quadratic behavior in 18951 fixup_var_refs) 18952 * [87]7099 G++ doesn't set the noreturn attribute on std::exit and 18953 std::abort 18954 * [88]7247 copy constructor missing when inlining enabled (invalid 18955 optimization?) 18956 * [89]7441 string array initialization compilation time regression 18957 from seconds to minutes 18958 * [90]7768 __PRETTY_FUNCTION__ for template destructor is wrong 18959 * [91]7804 bad printing of floating point constant in warning message 18960 * [92]8099 Friend classes and template specializations 18961 * [93]8117 member function pointers and multiple inheritance 18962 * [94]8205 using declaration and multiple inheritance 18963 * [95]8645 unnecessary non-zero checks in stl_tree.h 18964 * [96]8724 explicit destructor call for incomplete class allowed 18965 * [97]8805 compile time regression with many member variables 18966 * [98]8691 -O3 and -fno-implicit-templates are incompatible 18967 * [99]8700 unhelpful error message for binding temp to reference 18968 * [100]8724 explicit destructor call for incomplete class allowed 18969 * [101]8949 numeric_limits<>::denorm_min() and is_iec559 problems 18970 * [102]9016 Failure to consistently constant fold "constant" C++ 18971 objects 18972 * [103]9053 g++ confused about ambiguity of overloaded function 18973 templates 18974 * [104]9152 undefined virtual thunks 18975 * [105]9182 basic_filebuf<> does not report errors in codecvt<>::out 18976 * [106]9297 data corruption due to codegen bug (when copying.) 18977 * [107]9318 i/ostream::operator>>/<<(streambuf*) broken 18978 * [108]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 18979 * [109]9400 bogus -Wshadow warning: shadowed declaration of this in 18980 local classes 18981 * [110]9424 i/ostream::operator>>/<<(streambuf*) drops characters 18982 * [111]9425 filebuf::pbackfail broken (DUP: [112]9439) 18983 * [113]9474 GCC freezes in compiling a weird code mixing <iostream> 18984 and <iostream.h> 18985 * [114]9548 Incorrect results from setf(ios::fixed) and precision(-1) 18986 [115][DR 231] 18987 * [116]9555 ostream inserters fail to set badbit on exception 18988 * [117]9561 ostream inserters rethrow exception of wrong type 18989 * [118]9563 ostream::sentry returns true after a failed preparation 18990 * [119]9582 one-definition rule violation in std::allocator 18991 * [120]9622 __PRETTY_FUNCTION__ incorrect in template destructors 18992 * [121]9683 bug in initialization chains for static const variables 18993 from template classes 18994 * [122]9791 -Woverloaded-virtual reports hiding of destructor 18995 * [123]9817 collate::compare doesn't handle nul characters 18996 * [124]9825 filebuf::sputbackc breaks sbumpc 18997 * [125]9826 operator>>(basic_istream, basic_string) fails to compile 18998 with custom traits 18999 * [126]9924 Multiple using statements for builtin functions not 19000 allowed 19001 * [127]9946 destructor is not called for temporary object 19002 * [128]9964 filebuf::close() sometimes fails to close file 19003 * [129]9988 filebuf::overflow writes EOF to file 19004 * [130]10033 optimization breaks polymorphic references w/ typeid 19005 operator 19006 * [131]10097 filebuf::underflow drops characters 19007 * [132]10132 filebuf destructor can throw exceptions 19008 * [133]10180 gcc fails to warn about non-inlined function 19009 * [134]10199 method parametrized by template does not work everywhere 19010 * [135]10300 use of array-new (nothrow) in segfaults on NULL return 19011 * [136]10427 Stack corruption with variable-length automatic arrays 19012 and virtual destructors 19013 * [137]10503 Compilation never stops in fixed_type_or_null 19014 19015 Objective-C 19016 19017 * [138]5956 selectors aren't matched properly when added to the 19018 selector table 19019 19020 Fortran compiler and library 19021 19022 * [139]1832 list directed i/o overflow hangs, -fbounds-check doesn't 19023 detect 19024 * [140]3924 g77 generates code that is rejected by GAS if COFF debug 19025 info requested 19026 * [141]5634 doc: explain that configure --prefix=~/... does not work 19027 * [142]6367 multiple repeat counts confuse namelist read into array 19028 * [143]6491 Logical operations error on logicals when using 19029 -fugly-logint 19030 * [144]6742 Generation of C++ Prototype for FORTRAN and extern "C" 19031 * [145]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 19032 on irix6.5 19033 * [146]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 19034 assume a direct access file 19035 * [147]7278 g77 "bug"; the executable misbehaves (with -O2 19036 -fno-automatic) 19037 * [148]7384 DATE_AND_TIME milliseconds field inactive on Windows 19038 * [149]7388 Incorrect output with 0-based array of characters 19039 * [150]8587 Double complex zero ** double precision number -> NaN 19040 instead of zero 19041 * [151]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 19042 unknown register name line-length-none 19043 * [152]10197 Direct access files not unformatted by default 19044 19045 Java compiler and library 19046 19047 * [153]6005 gcj fails to build rhug on alpha 19048 * [154]6389 System.getProperty("") should always throw an 19049 IllegalArgumentException 19050 * [155]6576 java.util.ResourceBundle.getResource ignores locale 19051 * [156]6652 new java.io.File("").getCanonicalFile() throws exception 19052 * [157]7060 getMethod() doesn't search super interface 19053 * [158]7073 bytecode interpreter gives wrong answer for interface 19054 getSuperclass() 19055 * [159]7180 possible bug in 19056 javax.naming.spi.NamingManager.getPlusPath() 19057 * [160]7416 java.security startup refs "GNU libgcj.security" 19058 * [161]7570 Runtime.exec with null envp: child doesn't inherit parent 19059 env (DUP: [162]7578) 19060 * [163]7611 Internal error while compiling libjava with -O 19061 * [164]7709 NullPointerException in _Jv_ResolvePoolEntry 19062 * [165]7766 ZipInputStream.available returns 0 immediately after 19063 construction 19064 * [166]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 19065 * [167]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 19066 * [168]8142 '$' in class names vs. dlopen 'dynamic string tokens' 19067 * [169]8234 ZipInputStream chokes when InputStream.read() returns 19068 small chunks 19069 * [170]8415 reflection bug: exception info for Method 19070 * [171]8481 java.Random.nextInt(int) may return negative 19071 * [172]8593 Error reading GZIPped files with BufferedReader 19072 * [173]8759 java.beans.Introspector has no flushCaches() or 19073 flushFromCaches() methods 19074 * [174]8997 spin() calls Thread.sleep 19075 * [175]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 19076 instead of the root content of C: 19077 * [176]9254 java::lang::Object::wait(), threads-win32.cc returns 19078 wrong return codes 19079 * [177]9271 Severe bias in java.security.SecureRandom 19080 19081 Ada compiler and library 19082 19083 * [178]6767 make gnatlib-shared fails on -laddr2line 19084 * [179]9911 gnatmake fails to link when GCC configured with 19085 --with-sjlj-exceptions=yes 19086 * [180]10020 Can't bootstrap gcc on AIX with Ada enabled 19087 * [181]10546 Ada tasking not working on Red Hat 9 19088 19089 preprocessor 19090 19091 * [182]7029 preprocessor should ignore #warning with -M 19092 19093 ARM-specific 19094 19095 * [183]2903 [arm] Optimization bug with long long arithmetic 19096 * [184]7873 arm-linux-gcc fails when assigning address to a bit field 19097 19098 FreeBSD-specific 19099 19100 * [185]7680 float functions undefined in math.h/cmath with #define 19101 _XOPEN_SOURCE 19102 19103 HP-UX or HP-PA-specific 19104 19105 * [186]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 19106 * [187]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 19107 fputc_unlocked 19108 * [188]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 19109 19110 m68hc11-specific 19111 19112 * [189]6744 Bad assembler code generated: reference to pseudo 19113 register z 19114 * [190]7361 Internal compiler error in reload_cse_simplify_operands, 19115 in reload1.c 19116 19117 MIPS-specific 19118 19119 * [191]9496 [mips-linux] bug in optimizer? 19120 19121 PowerPC-specific 19122 19123 * [192]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 19124 space 19125 * [193]8480 reload ICEs for LAPACK code on powerpc64-linux 19126 * [194]8784 [AIX] Internal compiler error in simplify_gen_subreg 19127 * [195]10315 [powerpc] ICE: in extract_insn, in recog.c 19128 19129 SPARC-specific 19130 19131 * [196]10267 (documentation) Wrong build instructions for 19132 *-*-solaris2* 19133 19134 x86-specific (Intel/AMD) 19135 19136 * [197]7916 ICE in instantiate_virtual_register_1 19137 * [198]7926 (c++) i486 instructions in header files make c++ programs 19138 crash on i386 19139 * [199]8555 ICE in gen_split_1231 19140 * [200]8994 ICE with -O -march=pentium4 19141 * [201]9426 ICE with -fssa -funroll-loops -fprofile-arcs 19142 * [202]9806 ICE in inline assembly with -fPIC flag 19143 * [203]10077 gcc -msse2 generates movd to move dwords between xmm 19144 regs 19145 * [204]10233 64-bit comparison only comparing bottom 32-bits 19146 * [205]10286 type-punning doesn't work with __m64 and -O 19147 * [206]10308 [x86] ICE with -O -fgcse or -O2 19148 __________________________________________________________________ 19149 19150 [207]GCC 3.3.1 19151 19152 Bug Fixes 19153 19154 This section lists the problem reports (PRs) from GCC's bug tracking 19155 system that are known to be fixed in the 3.3.1 release. This list might 19156 not be complete (that is, it is possible that some PRs that have been 19157 fixed are not listed here). 19158 19159 Bootstrap failures 19160 19161 * [208]11272 [Solaris] make bootstrap fails while building libstdc++ 19162 19163 Internal compiler errors (multi-platform) 19164 19165 * [209]5754 ICE on invalid nested template class 19166 * [210]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 19167 and --enable-checking 19168 * [211]6949 (c++) ICE in tsubst_decl, in cp/pt.c 19169 * [212]7053 (c++) ICE when declaring a function already defined as a 19170 friend method of a template class 19171 * [213]8164 (c++) ICE when using different const expressions as 19172 template parameter 19173 * [214]8384 (c++) ICE in is_base_type, in dwarf2out.c 19174 * [215]9559 (c++) ICE with invalid initialization of a static const 19175 * [216]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 19176 when redeclaring a static member variable 19177 * [217]9864 (fortran) ICE in add_abstract_origin_attribute, in 19178 dwarfout.c with -g -O -finline-functions 19179 * [218]10432 (c++) ICE in poplevel, in cp/decl.c 19180 * [219]10475 ICE in subreg_highpart_offset for code with long long 19181 * [220]10635 (c++) ICE when dereferencing an incomplete type casted 19182 from a void pointer 19183 * [221]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 19184 instantiating static member variables 19185 * [222]10700 ICE in copy_to_mode_reg on 64-bit targets 19186 * [223]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 19187 * [224]10796 (c++) ICE when defining an enum with two values: -1 and 19188 MAX_INT_64BIT 19189 * [225]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 19190 sched.c 19191 * [226]10939 (c++) ICE with template code 19192 * [227]10956 (c++) ICE when specializing a template member function 19193 of a template class, in tsubst, in cp/pt.c 19194 * [228]11041 (c++) ICE: const myclass &x = *x; (when operator*() 19195 defined) 19196 * [229]11059 (c++) ICE with empty union 19197 * [230]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 19198 -O2 -fnon-call-exceptions 19199 * [231]11105 (c++) ICE in mangle_conv_op_name_for_type 19200 * [232]11149 (c++) ICE on error when instantiation with call function 19201 of a base type 19202 * [233]11228 (c++) ICE on new-expression using array operator new and 19203 default-initialization 19204 * [234]11282 (c++) Infinite memory usage after syntax error 19205 * [235]11301 (fortran) ICE with -fno-globals 19206 * [236]11308 (c++) ICE when using an enum type name as if it were a 19207 class or namespace 19208 * [237]11473 (c++) ICE with -gstabs when empty struct inherits from 19209 an empty struct 19210 * [238]11503 (c++) ICE when instantiating template with ADDR_EXPR 19211 * [239]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 19212 template member functions 19213 19214 Optimization bugs 19215 19216 * [240]11198 -O2 -frename-registers generates wrong code (aliasing 19217 problem) 19218 * [241]11304 Wrong code production with -fomit-frame-pointer 19219 * [242]11381 volatile memory access optimized away 19220 * [243]11536 [strength-reduce] -O2 optimization produces wrong code 19221 * [244]11557 constant folding bug generates wrong code 19222 19223 C front end 19224 19225 * [245]5897 No warning for statement after return 19226 * [246]11279 DWARF-2 output mishandles large enums 19227 19228 Preprocessor bugs 19229 19230 * [247]11022 no warning for non-compatible macro redefinition 19231 19232 C++ compiler and library 19233 19234 * [248]2330 static_cast<>() to a private base is allowed 19235 * [249]5388 Incorrect message "operands to ?: have different types" 19236 * [250]5390 Libiberty fails to demangle multi-digit template 19237 parameters 19238 * [251]7877 Incorrect parameter passing to specializations of member 19239 function templates 19240 * [252]9393 Anonymous namespaces and compiling the same file twice 19241 * [253]10032 -pedantic converts some errors to warnings 19242 * [254]10468 const typeof(x) is non-const, but only in templates 19243 * [255]10527 confused error message with "new int()" parameter 19244 initializer 19245 * [256]10679 parameter MIN_INLINE_INSNS is not honored 19246 * [257]10682 gcc chokes on a typedef for an enum inside a class 19247 template 19248 * [258]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 19249 0. 19250 * [259]10845 template member function (with nested template as 19251 parameter) cannot be called anymore if another unrelated template 19252 member function is defined 19253 * [260]10849 Cannot define an out-of-class specialization of a 19254 private nested template class 19255 * [261]10888 Suppress -Winline warnings for system headers 19256 * [262]10929 -Winline warns about functions for which no definition 19257 is visible 19258 * [263]10931 valid conversion static_cast<const unsigned 19259 int&>(lvalue-of-type-int) is rejected 19260 * [264]10940 Bad code with explicit specialization 19261 * [265]10968 If member function implicitly instantiated, explicit 19262 instantiation of class fails to instantiate it 19263 * [266]10990 Cannot convert with dynamic_cast<> to a private base 19264 class from within a member function 19265 * [267]11039 Bad interaction between implicit typename deprecation 19266 and friendship 19267 * [268]11062 (libstdc++) avoid __attribute__ ((unused)); say 19268 "__unused__" instead 19269 * [269]11095 C++ iostream manipulator causes segfault when called 19270 with negative argument 19271 * [270]11098 g++ doesn't emit complete debugging information for 19272 local variables in destructors 19273 * [271]11137 GNU/Linux shared library constructors not called unless 19274 there's one global object 19275 * [272]11154 spurious ambiguity report for template class 19276 specialization 19277 * [273]11329 Compiler cannot find user defined implicit typecast 19278 * [274]11332 Spurious error with casts in ?: expression 19279 * [275]11431 static_cast behavior with subclasses when default 19280 constructor available 19281 * [276]11528 money_get facet does not accept "$.00" as valid 19282 * [277]11546 Type lookup problems in out-of-line definition of a 19283 class doubly nested from a template class 19284 * [278]11567 C++ code containing templated member function with same 19285 name as pure virtual member function results in linking failure 19286 * [279]11645 Failure to deal with using and private inheritance 19287 19288 Java compiler and library 19289 19290 * [280]5179 Qualified static field access doesn't initialize its 19291 class 19292 * [281]8204 gcj -O2 to native reorders certain instructions 19293 improperly 19294 * [282]10838 java.io.ObjectInputStream syntax error 19295 * [283]10886 The RMI registry that comes with GCJ does not work 19296 correctly 19297 * [284]11349 JNDI URL context factories not located correctly 19298 19299 x86-specific (Intel/AMD) 19300 19301 * [285]4823 ICE on inline assembly code 19302 * [286]8878 miscompilation with -O and SSE 19303 * [287]9815 (c++ library) atomicity.h - fails to compile with -O3 19304 -masm=intel 19305 * [288]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 19306 in reload1.c 19307 * [289]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 19308 * [290]10673 ICE for x86-64 on freebsd libc vfprintf.c source 19309 * [291]11044 [x86] out of range loop instructions for FP code on K6 19310 * [292]11089 ICE: instantiate_virtual_regs_lossage while using SSE 19311 built-ins 19312 * [293]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 19313 is used 19314 19315 SPARC- or Solaris- specific 19316 19317 * [294]9362 solaris 'as' dies when fed .s and "-gstabs" 19318 * [295]10142 [SPARC64] gcc produces wrong code when passing 19319 structures by value 19320 * [296]10663 New configure check aborts with Sun tools. 19321 * [297]10835 combinatorial explosion in scheduler on HyperSPARC 19322 * [298]10876 ICE in calculate_giv_inc when building KDE 19323 * [299]10955 wrong code at -O3 for structure argument in context of 19324 structure return 19325 * [300]11018 -mcpu=ultrasparc busts tar-1.13.25 19326 * [301]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 19327 Linux kernel 19328 19329 ia64 specific 19330 19331 * [302]10907 gcc violates the ia64 ABI (GP must be preserved) 19332 * [303]11320 scheduler bug (in machine depended reorganization pass) 19333 * [304]11599 bug with conditional and __builtin_prefetch 19334 19335 PowerPC specific 19336 19337 * [305]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 19338 during loop) 19339 * [306]10871 error in rs6000_stack_info save_size computation 19340 * [307]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 19341 cures it 19342 19343 m68k-specific 19344 19345 * [308]7594 [m68k] ICE on legal code associated with simplify-rtx 19346 * [309]10557 [m68k] ICE in subreg_offset_representable_p 19347 * [310]11054 [m68k] ICE in reg_overlap_mentioned_p 19348 19349 ARM-specific 19350 19351 * [311]10834 [arm] GCC 3.3 still generates incorrect instructions for 19352 functions with __attribute__ ((interrupt ("IRQ"))) 19353 * [312]10842 [arm] Clobbered link register is copied to pc under 19354 certain circumstances 19355 * [313]11052 [arm] noce_process_if_block() can lose REG_INC notes 19356 * [314]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 19357 (3.4) 19358 19359 MIPS-specific 19360 19361 * [315]11084 ICE in propagate_one_insn, in flow.c 19362 19363 SH-specific 19364 19365 * [316]10331 can't compile c++ part of gcc cross compiler for sh-elf 19366 * [317]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 19367 * [318]11096 i686-linux to sh-linux cross compiler fails to compile 19368 C++ files 19369 19370 GNU/Linux (or Hurd?) specific 19371 19372 * [319]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 19373 19374 UnixWare specific 19375 19376 * [320]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 19377 7.1.1 19378 19379 Cygwin (or mingw) specific 19380 19381 * [321]5287 ICE with dllimport attribute 19382 * [322]10148 [MingW/CygWin] Compiler dumps core 19383 19384 DJGPP specific 19385 19386 * [323]8787 GCC fails to emit .intel_syntax when invoked with 19387 -masm=intel on DJGPP 19388 19389 Darwin (and MacOS X) specific 19390 19391 * [324]10900 trampolines crash 19392 19393 Documentation 19394 19395 * [325]1607 (c++) Format attributes on methods undocumented 19396 * [326]4252 Invalid option `-fdump-translation-unit' 19397 * [327]4490 Clarify restrictions on -m96bit-long-double, 19398 -m128bit-long-double 19399 * [328]10355 document an issue with regparm attribute on some systems 19400 (e.g. Solaris) 19401 * [329]10726 (fortran) Documentation for function "IDate Intrinsic 19402 (Unix)" is wrong 19403 * [330]10805 document bug in old version of Sun assembler 19404 * [331]10815 warn against GNU binutils on AIX 19405 * [332]10877 document need for newer binutils on i?86-*-linux-gnu 19406 * [333]11280 Manual incorrect with respect to -freorder-blocks 19407 * [334]11466 Document -mlittle-endian and its restrictions for the 19408 sparc64 port 19409 19410 Testsuite bugs (compiler itself is not affected) 19411 19412 * [335]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 19413 report failure 19414 * [336]10810 gcc-3.3 fails make check: buffer overrun in 19415 test_demangle.c 19416 __________________________________________________________________ 19417 19418 [337]GCC 3.3.2 19419 19420 Bug Fixes 19421 19422 This section lists the problem reports (PRs) from GCC's bug tracker 19423 that are known to be fixed in the 3.3.2 release. This list might not be 19424 complete (that is, it is possible that some PRs that have been fixed 19425 are not listed here). 19426 19427 Bootstrap failures and problems 19428 19429 * [338]8336 [SCO5] bootstrap config still tries to use COFF options 19430 * [339]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 19431 --enable-threads=posix 19432 * [340]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 19433 * [341]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 19434 7.1.1) 19435 * [342]11687 xstormy16-elf build fails in libf2c 19436 * [343]12263 [SGI IRIX] bootstrap fails during compile of 19437 libf2c/libI77/backspace.c 19438 * [344]12490 buffer overflow in scan-decls.c (during Solaris 9 19439 fix-header processing) 19440 19441 Internal compiler errors (multi-platform) 19442 19443 * [345]7277 Casting integers to vector types causes ICE 19444 * [346]7939 (c++) ICE on invalid function template specialization 19445 * [347]11063 (c++) ICE on parsing initialization list of const array 19446 member 19447 * [348]11207 ICE with negative index in array element designator 19448 * [349]11522 (fortran) g77 dwarf-2 ICE in 19449 add_abstract_origin_attribute 19450 * [350]11595 (c++) ICE on duplicate label definition 19451 * [351]11646 (c++) ICE in commit_one_edge_insertion with 19452 -fnon-call-exceptions -fgcse -O 19453 * [352]11665 ICE in struct initializer when taking address 19454 * [353]11852 (c++) ICE with bad struct initializer. 19455 * [354]11878 (c++) ICE in cp_expr_size 19456 * [355]11883 ICE with any -O on mercury-generated C code 19457 * [356]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 19458 cp/typeck2.c when applying typeid operator to template template 19459 parameter 19460 * [357]12146 ICE in lookup_template_function, in cp/pt.c 19461 * [358]12215 ICE in make_label_edge with -fnon-call-exceptions 19462 -fno-gcse -O2 19463 * [359]12369 (c++) ICE with templates and friends 19464 * [360]12446 ICE in emit_move_insn on complicated array reference 19465 * [361]12510 ICE in final_scan_insn 19466 * [362]12544 ICE with large parameters used in nested functions 19467 19468 C and optimization bugs 19469 19470 * [363]9862 spurious warnings with -W -finline-functions 19471 * [364]10962 lookup_field is a linear search on a linked list (can be 19472 slow if large struct) 19473 * [365]11370 -Wunreachable-code gives false complaints 19474 * [366]11637 invalid assembly with -fnon-call-exceptions 19475 * [367]11885 Problem with bitfields in packed structs 19476 * [368]12082 Inappropriate unreachable code warnings 19477 * [369]12180 Inline optimization fails for variadic function 19478 * [370]12340 loop unroller + gcse produces wrong code 19479 19480 C++ compiler and library 19481 19482 * [371]3907 nested template parameter collides with member name 19483 * [372]5293 confusing message when binding a temporary to a reference 19484 * [373]5296 [DR115] Pointers to functions and to template functions 19485 behave differently in deduction 19486 * [374]7939 ICE on function template specialization 19487 * [375]8656 Unable to assign function with __attribute__ and pointer 19488 return type to an appropriate variable 19489 * [376]10147 Confusing error message for invalid template function 19490 argument 19491 * [377]11400 std::search_n() makes assumptions about Size parameter 19492 * [378]11409 issues with using declarations, overloading, and 19493 built-in functions 19494 * [379]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 19495 multiple bits in mask 19496 * [380]11786 operator() call on variable in other namespace not 19497 recognized 19498 * [381]11867 static_cast ignores ambiguity 19499 * [382]11928 bug with conversion operators that are typedefs 19500 * [383]12114 Uninitialized memory accessed in dtor 19501 * [384]12163 static_cast + explicit constructor regression 19502 * [385]12181 Wrong code with comma operator and c++ 19503 * [386]12236 regparm and fastcall messes up parameters 19504 * [387]12266 incorrect instantiation of unneeded template during 19505 overload resolution 19506 * [388]12296 istream::peek() doesn't set eofbit 19507 * [389]12298 [sjlj exceptions] Stack unwind destroys 19508 not-yet-constructed object 19509 * [390]12369 ICE with templates and friends 19510 * [391]12337 apparently infinite loop in g++ 19511 * [392]12344 stdcall attribute ignored if function returns a pointer 19512 * [393]12451 missing(late) class forward declaration in cxxabi.h 19513 * [394]12486 g++ accepts invalid use of a qualified name 19514 19515 x86 specific (Intel/AMD) 19516 19517 * [395]8869 [x86 MMX] ICE with const variable optimization and MMX 19518 builtins 19519 * [396]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 19520 -O2 19521 * [397]11689 g++3.3 emits un-assembleable code for k6 architecture 19522 * [398]12116 [k6] Invalid assembly output values with X-MAME code 19523 * [399]12070 ICE converting between double and long double with 19524 -msoft-float 19525 19526 ia64-specific 19527 19528 * [400]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 19529 * [401]11535 __builtin_return_address may not work on ia64 19530 * [402]11693 [ia64] ICE in gen_nop_type 19531 * [403]12224 [ia64] Thread-local storage doesn't work 19532 19533 PowerPC-specific 19534 19535 * [404]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 19536 kernel 19537 * [405]11319 loop miscompiled on ppc32 19538 * [406]11949 ICE Compiler segfault with ffmpeg -maltivec code 19539 19540 SPARC-specific 19541 19542 * [407]11662 wrong code for expr. with cast to long long and 19543 exclusive or 19544 * [408]11965 invalid assembler code for a shift < 32 operation 19545 * [409]12301 (c++) stack corruption when a returned expression throws 19546 an exception 19547 19548 Alpha-specific 19549 19550 * [410]11717 [alpha-linux] unrecognizable insn compiling for.c of 19551 kernel 2.4.22-pre8 19552 19553 HPUX-specific 19554 19555 * [411]11313 problem with #pragma weak and static inline functions 19556 * [412]11712 __STDC_EXT__ not defined for C++ by default anymore? 19557 19558 Solaris specific 19559 19560 * [413]12166 Profiled programs crash if PROFDIR is set 19561 19562 Solaris-x86 specific 19563 19564 * [414]12101 i386 Solaris no longer works with GNU as? 19565 19566 Miscellaneous embedded target-specific bugs 19567 19568 * [415]10988 [m32r-elf] wrong blockmove code with -O3 19569 * [416]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 19570 -O2 19571 * [417]11902 [sh4] spec file improperly inserts rpath even when none 19572 needed 19573 * [418]11903 [sh4] -pthread fails to link due to error in spec file 19574 on sh4 19575 __________________________________________________________________ 19576 19577 [419]GCC 3.3.3 19578 19579 Minor features 19580 19581 In addition to the bug fixes documented below, this release contains 19582 few minor features such as: 19583 * Support for --with-sysroot 19584 * Support for automatic detection of executable stacks 19585 * Support for SSE3 instructions 19586 * Support for thread local storage debugging under GDB on S390 19587 19588 Bug Fixes 19589 19590 This section lists the problem reports (PRs) from GCC's bug tracker 19591 that are known to be fixed in the 3.3.3 release. This list might not be 19592 complete (that is, it is possible that some PRs that have been fixed 19593 are not listed here). 19594 19595 Bootstrap failures and issues 19596 19597 * [420]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 19598 * [421]12399 boehm-gc fails (when building a cross compiler): libtool 19599 unable to infer tagged configuration 19600 * [422]13068 mklibgcc.in doesn't handle multi-level multilib 19601 subdirectories properly 19602 19603 Internal compiler errors (multi-platform) 19604 19605 * [423]10060 ICE (stack overflow) on huge file (300k lines) due to 19606 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 19607 * [424]10555 (c++) ICE on undefined template argument 19608 * [425]10706 (c++) ICE in mangle_class_name_for_template 19609 * [426]11496 (fortran) error in flow_loops_find when -funroll-loops 19610 active 19611 * [427]11741 ICE in pre_insert_copy_insn, in gcse.c 19612 * [428]12440 GCC crashes during compilation of quicktime4linux 2.0.0 19613 * [429]12632 (fortran) -fbounds-check ICE 19614 * [430]12712 (c++) ICE on short legit C++ code fragment with gcc 19615 3.3.2 19616 * [431]12726 (c++) ICE (segfault) on trivial code 19617 * [432]12890 (c++) ICE on compilation of class with throwing method 19618 * [433]12900 (c++) ICE in rtl_verify_flow_info_1 19619 * [434]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 19620 correct code with -O2 -fno-force-mem 19621 * [435]13289 (c++) ICE in regenerate_decl_from_template on recursive 19622 template 19623 * [436]13318 ICE: floating point exception in the loop optimizer 19624 * [437]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 19625 except.c 19626 * [438]13574 (c++) invalid array default initializer in class lets 19627 gcc consume all memory and die 19628 * [439]13475 ICE on SIMD variables with partial value initialization 19629 * [440]13797 (c++) ICE on invalid template parameter 19630 * [441]13824 (java) gcj SEGV with simple .java program 19631 19632 C and optimization bugs 19633 19634 * [442]8776 loop invariants are not removed (most likely) 19635 * [443]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 19636 strncmp by memcmp 19637 * [444]11350 undefined labels with -Os -fPIC 19638 * [445]12826 Optimizer removes reference through volatile pointer 19639 * [446]12500 stabs debug info: void no longer a predefined / builtin 19640 type 19641 * [447]12941 builtin-bitops-1.c miscompilation (latent bug) 19642 * [448]12953 tree inliner bug (in inline_forbidden_p) and fix 19643 * [449]13041 linux-2.6/sound/core/oss/rate.c miscompiled 19644 * [450]13507 spurious printf format warning 19645 * [451]13382 Type information for const pointer disappears during 19646 optimization. 19647 * [452]13394 noreturn attribute ignored on recursive invokation 19648 * [453]13400 Compiled code crashes storing to read-only location 19649 * [454]13521 Endless loop in calculate_global_regs_live 19650 19651 C++ compiler and library 19652 19653 Some of the bug fixes in this list were made to implement decisions 19654 that the ISO C++ standards committee has made concerning several defect 19655 reports (DRs). Links in the list below point to detailed discussion of 19656 the relevant defect report. 19657 * [455]2094 unimplemented: use of `ptrmem_cst' in template type 19658 unification 19659 * [456]2294 using declaration confusion 19660 * [457]5050 template instantiation depth exceeds limit: recursion 19661 problem? 19662 * [458]9371 Bad exception handling in 19663 i/ostream::operator>>/<<(streambuf*) 19664 * [459]9546 bad exception handling in ostream members 19665 * [460]10081 basic_ios::_M_cache_locale leaves NULL members in the 19666 face of unknown locales 19667 * [461]10093 [462][DR 61] Setting failbit in exceptions doesn't work 19668 * [463]10095 istream::operator>>(int&) sets ios::badbit when 19669 ios::failbit is set. 19670 * [464]11554 Warning about reordering of initializers doesn't mention 19671 location of constructor 19672 * [465]12297 istream::sentry::sentry() handles eof() incorrectly. 19673 * [466]12352 Exception safety problems in src/localename.cc 19674 * [467]12438 Memory leak in locale::combine() 19675 * [468]12540 Memory leak in locale::locale(const char*) 19676 * [469]12594 DRs [470]60 [TC] and [471]63 [TC] not implemented 19677 * [472]12657 Resolution of [473]DR 292 (WP) still unimplemented 19678 * [474]12696 memory eating infinite loop in diagnostics (error 19679 recovery problem) 19680 * [475]12815 Code compiled with optimization behaves unexpectedly 19681 * [476]12862 Conflicts between typedefs/enums and namespace member 19682 declarations 19683 * [477]12926 Wrong value after assignment in initialize list using 19684 bit-fields 19685 * [478]12967 Resolution of [479]DR 300 [WP] still unimplemented 19686 * [480]12971 Resolution of [481]DR 328 [WP] still unimplemented 19687 * [482]13007 basic_streambuf::pubimbue, imbue wrong 19688 * [483]13009 Implicitly-defined assignment operator writes to wrong 19689 memory 19690 * [484]13057 regparm attribute not applied to destructor 19691 * [485]13070 -Wformat option ignored in g++ 19692 * [486]13081 forward template declarations in <complex> let inlining 19693 fail 19694 * [487]13239 Assertion does not seem to work correctly anymore 19695 * [488]13262 "xxx is private within this context" when initializing a 19696 self-contained template class 19697 * [489]13290 simple typo in concept checking for std::generate_n 19698 * [490]13323 Template code does not compile in presence of typedef 19699 * [491]13369 __verify_grouping (and __add_grouping?) not correct 19700 * [492]13371 infinite loop with packed struct and inlining 19701 * [493]13445 Template argument replacement "dereferences" a typedef 19702 * [494]13461 Fails to access protected-ctor from public constant 19703 * [495]13462 Non-standard-conforming type set::pointer 19704 * [496]13478 gcc uses wrong constructor to initialize a const 19705 reference 19706 * [497]13544 "conflicting types" for enums in different scopes 19707 * [498]13650 string::compare should not (always) use 19708 traits_type::length() 19709 * [499]13683 bogus warning about passing non-PODs through ellipsis 19710 * [500]13688 Derived class is denied access to protected base class 19711 member class 19712 * [501]13774 Member variable cleared in virtual multiple inheritance 19713 class 19714 * [502]13884 Protect sstream.tcc from extern template use 19715 19716 Java compiler and library 19717 19718 * [503]10746 [win32] garbage collection crash in GCJ 19719 19720 Objective-C compiler and library 19721 19722 * [504]11433 Crash due to dereferencing null pointer when querying 19723 protocol 19724 19725 Fortran compiler and library 19726 19727 * [505]12633 logical expression gives incorrect result with 19728 -fugly-logint option 19729 * [506]13037 [gcse-lm] g77 generates incorrect code 19730 * [507]13213 Hex constant problem when compiling with -fugly-logint 19731 and -ftypeless-boz 19732 19733 x86-specific (Intel/AMD) 19734 19735 * [508]4490 ICE with -m128bit-long-double 19736 * [509]12292 [x86_64] ICE: RTL check: expected code `const_int', have 19737 `reg' in make_field_assignment, in combine.c 19738 * [510]12441 ICE: can't find a register to spill 19739 * [511]12943 array static-init failure under -fpic, -fPIC 19740 * [512]13608 Incorrect code with -O3 -ffast-math 19741 19742 PowerPC-specific 19743 19744 * [513]11598 testcase gcc.dg/20020118-1.c fails runtime check of 19745 __attribute__((aligned(16))) 19746 * [514]11793 ICE in extract_insn, in recog.c (const_vector's) 19747 * [515]12467 vmsumubm emitted when vmsummbm appropriate (typo in 19748 altivec.md) 19749 * [516]12537 g++ generates writeable text sections 19750 19751 SPARC-specific 19752 19753 * [517]12496 wrong result for __atomic_add(&value, -1) when using -O0 19754 -m64 19755 * [518]12865 mprotect call to make trampoline executable may fail 19756 * [519]13354 ICE in sparc_emit_set_const32 19757 19758 ARM-specific 19759 19760 * [520]10467 [arm] ICE in pre_insert_copy_insn, 19761 19762 ia64-specific 19763 19764 * [521]11226 ICE passing struct arg with two floats 19765 * [522]11227 ICE for _Complex float, _Complex long double args 19766 * [523]12644 GCC 3.3.2 fails to compile glibc on ia64 19767 * [524]13149 build gcc-3.3.2 1305 error:unrecognizable insn 19768 * Various fixes for libunwind 19769 19770 Alpha-specific 19771 19772 * [525]12654 Incorrect comparison code generated for Alpha 19773 * [526]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 19774 * [527]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 19775 19776 HPPA-specific 19777 19778 * [528]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 19779 * [529]12158 [hppa] compilation does not terminate at -O1 19780 19781 S390-specific 19782 19783 * [530]11992 Wrong built-in code for memcmp with length 1<<24: only 19784 (1<<24)-1 possible for CLCL-Instruction 19785 19786 SH-specific 19787 19788 * [531]9365 segfault in gen_far_branch (config/sh/sh.c) 19789 * [532]10392 optimizer generates faulty array indexing 19790 * [533]11322 SH profiler outputs multiple definitions of symbol 19791 * [534]13069 gcc/config/sh/rtems.h broken 19792 * [535]13302 Putting a va_list in a struct causes seg fault 19793 * [536]13585 Incorrect optimization of call to sfunc 19794 * Fix inappropriately exported libgcc functions from the shared 19795 library 19796 19797 Other embedded target specific 19798 19799 * [537]8916 [mcore] unsigned char assign gets hosed. 19800 * [538]11576 [h8300] ICE in change_address_1, in emit-rtl.c 19801 * [539]13122 [h8300] local variable gets corrupted by function call 19802 when -fomit-frame-pointer is given 19803 * [540]13256 [cris] strict_low_part mistreated in delay slots 19804 * [541]13373 [mcore] optimization with -frerun-cse-after-loop 19805 -fexpensive-optimizations produces wrong code on mcore 19806 19807 GNU HURD-specific 19808 19809 * [542]12561 gcc/config/t-gnu needs updating to work with 19810 --with-sysroot 19811 19812 Tru64 Unix specific 19813 19814 * [543]6243 testsuite fails almost all tests due to no libintl in 19815 LD_LIBRARY_PATH during test. 19816 * [544]11397 weak aliases broken on Tru64 UNIX 19817 19818 AIX-specific 19819 19820 * [545]12505 build failure due to defines of uchar in cpphash.h and 19821 sys/types.h 19822 * [546]13150 WEAK symbols not exported by collect2 19823 19824 IRIX-specific 19825 19826 * [547]12666 fixincludes problem on IRIX 6.5.19m 19827 19828 Solaris-specific 19829 19830 * [548]12969 Including sys/byteorder.h breaks configure checks 19831 19832 Testsuite problems (compiler is not affected) 19833 19834 * [549]10819 testsuite creates CR+LF on compiler version lines in 19835 test summary files 19836 * [550]11612 abi_check not finding correct libgcc_s.so.1 19837 19838 Miscellaneous 19839 19840 * [551]13211 using -###, incorrect warnings about unused linker file 19841 are produced 19842 __________________________________________________________________ 19843 19844 [552]GCC 3.3.4 19845 19846 This is the [553]list of problem reports (PRs) from GCC's bug tracking 19847 system that are known to be fixed in the 3.3.4 release. This list might 19848 not be complete (that is, it is possible that some PRs that have been 19849 fixed are not listed here). 19850 __________________________________________________________________ 19851 19852 [554]GCC 3.3.5 19853 19854 This is the [555]list of problem reports (PRs) from GCC's bug tracking 19855 system that are known to be fixed in the 3.3.5 release. This list might 19856 not be complete (that is, it is possible that some PRs that have been 19857 fixed are not listed here). 19858 __________________________________________________________________ 19859 19860 [556]GCC 3.3.6 19861 19862 This is the [557]list of problem reports (PRs) from GCC's bug tracking 19863 system that are known to be fixed in the 3.3.6 release. This list might 19864 not be complete (that is, it is possible that some PRs that have been 19865 fixed are not listed here). 19866 19867 19868 For questions related to the use of GCC, please consult these web 19869 pages and the [558]GCC manuals. If that fails, the 19870 [559]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 19871 web pages and the development of GCC are welcome on our developer 19872 list at [560]gcc (a] gcc.gnu.org. All of [561]our lists have public 19873 archives. 19874 19875 Copyright (C) [562]Free Software Foundation, Inc. Verbatim copying and 19876 distribution of this entire article is permitted in any medium, 19877 provided this notice is preserved. 19878 19879 These pages are [563]maintained by the GCC team. Last modified 19880 2025-06-05. 19881 19882 References 19883 19884 1. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 19885 2. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 19886 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 19887 4. https://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 19888 5. https://gcc.gnu.org/news/dfa.html 19889 6. https://gcc.gnu.org/projects/c-status.html 19890 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 19891 8. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 19892 9. https://gcc.gnu.org/PR10140 19893 10. https://gcc.gnu.org/PR10198 19894 11. https://gcc.gnu.org/PR10338 19895 12. https://gcc.gnu.org/PR3581 19896 13. https://gcc.gnu.org/PR4382 19897 14. https://gcc.gnu.org/PR5533 19898 15. https://gcc.gnu.org/PR6387 19899 16. https://gcc.gnu.org/PR6412 19900 17. https://gcc.gnu.org/PR6620 19901 18. https://gcc.gnu.org/PR6663 19902 19. https://gcc.gnu.org/PR7068 19903 20. https://gcc.gnu.org/PR7083 19904 21. https://gcc.gnu.org/PR7647 19905 22. https://gcc.gnu.org/PR7675 19906 23. https://gcc.gnu.org/PR7718 19907 24. https://gcc.gnu.org/PR8116 19908 25. https://gcc.gnu.org/PR8358 19909 26. https://gcc.gnu.org/PR8511 19910 27. https://gcc.gnu.org/PR8564 19911 28. https://gcc.gnu.org/PR8660 19912 29. https://gcc.gnu.org/PR8766 19913 30. https://gcc.gnu.org/PR8803 19914 31. https://gcc.gnu.org/PR8846 19915 32. https://gcc.gnu.org/PR8906 19916 33. https://gcc.gnu.org/PR9216 19917 34. https://gcc.gnu.org/PR9261 19918 35. https://gcc.gnu.org/PR9263 19919 36. https://gcc.gnu.org/PR9429 19920 37. https://gcc.gnu.org/PR9516 19921 38. https://gcc.gnu.org/PR9600 19922 39. https://gcc.gnu.org/PR9629 19923 40. https://gcc.gnu.org/PR9672 19924 41. https://gcc.gnu.org/PR9749 19925 42. https://gcc.gnu.org/PR9794 19926 43. https://gcc.gnu.org/PR9829 19927 44. https://gcc.gnu.org/PR9916 19928 45. https://gcc.gnu.org/PR9936 19929 46. https://gcc.gnu.org/PR10262 19930 47. https://gcc.gnu.org/PR10278 19931 48. https://gcc.gnu.org/PR10446 19932 49. https://gcc.gnu.org/PR10451 19933 50. https://gcc.gnu.org/PR10506 19934 51. https://gcc.gnu.org/PR10549 19935 52. https://gcc.gnu.org/PR2001 19936 53. https://gcc.gnu.org/PR2391 19937 54. https://gcc.gnu.org/PR2960 19938 55. https://gcc.gnu.org/PR4046 19939 56. https://gcc.gnu.org/PR6405 19940 57. https://gcc.gnu.org/PR6798 19941 58. https://gcc.gnu.org/PR6871 19942 59. https://gcc.gnu.org/PR6909 19943 60. https://gcc.gnu.org/PR7189 19944 61. https://gcc.gnu.org/PR7642 19945 62. https://gcc.gnu.org/PR8634 19946 63. https://gcc.gnu.org/PR8750 19947 64. https://gcc.gnu.org/PR2161 19948 65. https://gcc.gnu.org/PR4319 19949 66. https://gcc.gnu.org/PR8602 19950 67. https://gcc.gnu.org/PR9177 19951 68. https://gcc.gnu.org/PR9853 19952 69. https://gcc.gnu.org/PR45 19953 70. https://gcc.gnu.org/PR3784 19954 71. https://gcc.gnu.org/PR764 19955 72. https://gcc.gnu.org/PR5116 19956 73. https://gcc.gnu.org/PR2862 19957 74. https://gcc.gnu.org/PR3663 19958 75. https://gcc.gnu.org/PR3797 19959 76. https://gcc.gnu.org/PR3948 19960 77. https://gcc.gnu.org/PR4137 19961 78. https://gcc.gnu.org/PR4361 19962 79. https://gcc.gnu.org/PR4802 19963 80. https://gcc.gnu.org/PR5837 19964 81. https://gcc.gnu.org/PR4803 19965 82. https://gcc.gnu.org/PR5094 19966 83. https://gcc.gnu.org/PR5730 19967 84. https://gcc.gnu.org/PR6713 19968 85. https://gcc.gnu.org/PR7015 19969 86. https://gcc.gnu.org/PR7086 19970 87. https://gcc.gnu.org/PR7099 19971 88. https://gcc.gnu.org/PR7247 19972 89. https://gcc.gnu.org/PR7441 19973 90. https://gcc.gnu.org/PR7768 19974 91. https://gcc.gnu.org/PR7804 19975 92. https://gcc.gnu.org/PR8099 19976 93. https://gcc.gnu.org/PR8117 19977 94. https://gcc.gnu.org/PR8205 19978 95. https://gcc.gnu.org/PR8645 19979 96. https://gcc.gnu.org/PR8724 19980 97. https://gcc.gnu.org/PR8805 19981 98. https://gcc.gnu.org/PR8691 19982 99. https://gcc.gnu.org/PR8700 19983 100. https://gcc.gnu.org/PR8724 19984 101. https://gcc.gnu.org/PR8949 19985 102. https://gcc.gnu.org/PR9016 19986 103. https://gcc.gnu.org/PR9053 19987 104. https://gcc.gnu.org/PR9152 19988 105. https://gcc.gnu.org/PR9182 19989 106. https://gcc.gnu.org/PR9297 19990 107. https://gcc.gnu.org/PR9318 19991 108. https://gcc.gnu.org/PR9320 19992 109. https://gcc.gnu.org/PR9400 19993 110. https://gcc.gnu.org/PR9424 19994 111. https://gcc.gnu.org/PR9425 19995 112. https://gcc.gnu.org/PR9439 19996 113. https://gcc.gnu.org/PR9474 19997 114. https://gcc.gnu.org/PR9548 19998 115. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 19999 116. https://gcc.gnu.org/PR9555 20000 117. https://gcc.gnu.org/PR9561 20001 118. https://gcc.gnu.org/PR9563 20002 119. https://gcc.gnu.org/PR9582 20003 120. https://gcc.gnu.org/PR9622 20004 121. https://gcc.gnu.org/PR9683 20005 122. https://gcc.gnu.org/PR9791 20006 123. https://gcc.gnu.org/PR9817 20007 124. https://gcc.gnu.org/PR9825 20008 125. https://gcc.gnu.org/PR9826 20009 126. https://gcc.gnu.org/PR9924 20010 127. https://gcc.gnu.org/PR9946 20011 128. https://gcc.gnu.org/PR9964 20012 129. https://gcc.gnu.org/PR9988 20013 130. https://gcc.gnu.org/PR10033 20014 131. https://gcc.gnu.org/PR10097 20015 132. https://gcc.gnu.org/PR10132 20016 133. https://gcc.gnu.org/PR10180 20017 134. https://gcc.gnu.org/PR10199 20018 135. https://gcc.gnu.org/PR10300 20019 136. https://gcc.gnu.org/PR10427 20020 137. https://gcc.gnu.org/PR10503 20021 138. https://gcc.gnu.org/PR5956 20022 139. https://gcc.gnu.org/PR1832 20023 140. https://gcc.gnu.org/PR3924 20024 141. https://gcc.gnu.org/PR5634 20025 142. https://gcc.gnu.org/PR6367 20026 143. https://gcc.gnu.org/PR6491 20027 144. https://gcc.gnu.org/PR6742 20028 145. https://gcc.gnu.org/PR7113 20029 146. https://gcc.gnu.org/PR7236 20030 147. https://gcc.gnu.org/PR7278 20031 148. https://gcc.gnu.org/PR7384 20032 149. https://gcc.gnu.org/PR7388 20033 150. https://gcc.gnu.org/PR8587 20034 151. https://gcc.gnu.org/PR9038 20035 152. https://gcc.gnu.org/PR10197 20036 153. https://gcc.gnu.org/PR6005 20037 154. https://gcc.gnu.org/PR6389 20038 155. https://gcc.gnu.org/PR6576 20039 156. https://gcc.gnu.org/PR6652 20040 157. https://gcc.gnu.org/PR7060 20041 158. https://gcc.gnu.org/PR7073 20042 159. https://gcc.gnu.org/PR7180 20043 160. https://gcc.gnu.org/PR7416 20044 161. https://gcc.gnu.org/PR7570 20045 162. https://gcc.gnu.org/PR7578 20046 163. https://gcc.gnu.org/PR7611 20047 164. https://gcc.gnu.org/PR7709 20048 165. https://gcc.gnu.org/PR7766 20049 166. https://gcc.gnu.org/PR7785 20050 167. https://gcc.gnu.org/PR7786 20051 168. https://gcc.gnu.org/PR8142 20052 169. https://gcc.gnu.org/PR8234 20053 170. https://gcc.gnu.org/PR8415 20054 171. https://gcc.gnu.org/PR8481 20055 172. https://gcc.gnu.org/PR8593 20056 173. https://gcc.gnu.org/PR8759 20057 174. https://gcc.gnu.org/PR8997 20058 175. https://gcc.gnu.org/PR9253 20059 176. https://gcc.gnu.org/PR9254 20060 177. https://gcc.gnu.org/PR9271 20061 178. https://gcc.gnu.org/PR6767 20062 179. https://gcc.gnu.org/PR9911 20063 180. https://gcc.gnu.org/PR10020 20064 181. https://gcc.gnu.org/PR10546 20065 182. https://gcc.gnu.org/PR7029 20066 183. https://gcc.gnu.org/PR2903 20067 184. https://gcc.gnu.org/PR7873 20068 185. https://gcc.gnu.org/PR7680 20069 186. https://gcc.gnu.org/PR8705 20070 187. https://gcc.gnu.org/PR9986 20071 188. https://gcc.gnu.org/PR10056 20072 189. https://gcc.gnu.org/PR6744 20073 190. https://gcc.gnu.org/PR7361 20074 191. https://gcc.gnu.org/PR9496 20075 192. https://gcc.gnu.org/PR7067 20076 193. https://gcc.gnu.org/PR8480 20077 194. https://gcc.gnu.org/PR8784 20078 195. https://gcc.gnu.org/PR10315 20079 196. https://gcc.gnu.org/PR10267 20080 197. https://gcc.gnu.org/PR7916 20081 198. https://gcc.gnu.org/PR7926 20082 199. https://gcc.gnu.org/PR8555 20083 200. https://gcc.gnu.org/PR8994 20084 201. https://gcc.gnu.org/PR9426 20085 202. https://gcc.gnu.org/PR9806 20086 203. https://gcc.gnu.org/PR10077 20087 204. https://gcc.gnu.org/PR10233 20088 205. https://gcc.gnu.org/PR10286 20089 206. https://gcc.gnu.org/PR10308 20090 207. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.1 20091 208. https://gcc.gnu.org/PR11272 20092 209. https://gcc.gnu.org/PR5754 20093 210. https://gcc.gnu.org/PR6597 20094 211. https://gcc.gnu.org/PR6949 20095 212. https://gcc.gnu.org/PR7053 20096 213. https://gcc.gnu.org/PR8164 20097 214. https://gcc.gnu.org/PR8384 20098 215. https://gcc.gnu.org/PR9559 20099 216. https://gcc.gnu.org/PR9649 20100 217. https://gcc.gnu.org/PR9864 20101 218. https://gcc.gnu.org/PR10432 20102 219. https://gcc.gnu.org/PR10475 20103 220. https://gcc.gnu.org/PR10635 20104 221. https://gcc.gnu.org/PR10661 20105 222. https://gcc.gnu.org/PR10700 20106 223. https://gcc.gnu.org/PR10712 20107 224. https://gcc.gnu.org/PR10796 20108 225. https://gcc.gnu.org/PR10890 20109 226. https://gcc.gnu.org/PR10939 20110 227. https://gcc.gnu.org/PR10956 20111 228. https://gcc.gnu.org/PR11041 20112 229. https://gcc.gnu.org/PR11059 20113 230. https://gcc.gnu.org/PR11083 20114 231. https://gcc.gnu.org/PR11105 20115 232. https://gcc.gnu.org/PR11149 20116 233. https://gcc.gnu.org/PR11228 20117 234. https://gcc.gnu.org/PR11282 20118 235. https://gcc.gnu.org/PR11301 20119 236. https://gcc.gnu.org/PR11308 20120 237. https://gcc.gnu.org/PR11473 20121 238. https://gcc.gnu.org/PR11503 20122 239. https://gcc.gnu.org/PR11513 20123 240. https://gcc.gnu.org/PR11198 20124 241. https://gcc.gnu.org/PR11304 20125 242. https://gcc.gnu.org/PR11381 20126 243. https://gcc.gnu.org/PR11536 20127 244. https://gcc.gnu.org/PR11557 20128 245. https://gcc.gnu.org/PR5897 20129 246. https://gcc.gnu.org/PR11279 20130 247. https://gcc.gnu.org/PR11022 20131 248. https://gcc.gnu.org/PR2330 20132 249. https://gcc.gnu.org/PR5388 20133 250. https://gcc.gnu.org/PR5390 20134 251. https://gcc.gnu.org/PR7877 20135 252. https://gcc.gnu.org/PR9393 20136 253. https://gcc.gnu.org/PR10032 20137 254. https://gcc.gnu.org/PR10468 20138 255. https://gcc.gnu.org/PR10527 20139 256. https://gcc.gnu.org/PR10679 20140 257. https://gcc.gnu.org/PR10682 20141 258. https://gcc.gnu.org/PR10689 20142 259. https://gcc.gnu.org/PR10845 20143 260. https://gcc.gnu.org/PR10849 20144 261. https://gcc.gnu.org/PR10888 20145 262. https://gcc.gnu.org/PR10929 20146 263. https://gcc.gnu.org/PR10931 20147 264. https://gcc.gnu.org/PR10940 20148 265. https://gcc.gnu.org/PR10968 20149 266. https://gcc.gnu.org/PR10990 20150 267. https://gcc.gnu.org/PR11039 20151 268. https://gcc.gnu.org/PR11062 20152 269. https://gcc.gnu.org/PR11095 20153 270. https://gcc.gnu.org/PR11098 20154 271. https://gcc.gnu.org/PR11137 20155 272. https://gcc.gnu.org/PR11154 20156 273. https://gcc.gnu.org/PR11329 20157 274. https://gcc.gnu.org/PR11332 20158 275. https://gcc.gnu.org/PR11431 20159 276. https://gcc.gnu.org/PR11528 20160 277. https://gcc.gnu.org/PR11546 20161 278. https://gcc.gnu.org/PR11567 20162 279. https://gcc.gnu.org/PR11645 20163 280. https://gcc.gnu.org/PR5179 20164 281. https://gcc.gnu.org/PR8204 20165 282. https://gcc.gnu.org/PR10838 20166 283. https://gcc.gnu.org/PR10886 20167 284. https://gcc.gnu.org/PR11349 20168 285. https://gcc.gnu.org/PR4823 20169 286. https://gcc.gnu.org/PR8878 20170 287. https://gcc.gnu.org/PR9815 20171 288. https://gcc.gnu.org/PR10402 20172 289. https://gcc.gnu.org/PR10504 20173 290. https://gcc.gnu.org/PR10673 20174 291. https://gcc.gnu.org/PR11044 20175 292. https://gcc.gnu.org/PR11089 20176 293. https://gcc.gnu.org/PR11420 20177 294. https://gcc.gnu.org/PR9362 20178 295. https://gcc.gnu.org/PR10142 20179 296. https://gcc.gnu.org/PR10663 20180 297. https://gcc.gnu.org/PR10835 20181 298. https://gcc.gnu.org/PR10876 20182 299. https://gcc.gnu.org/PR10955 20183 300. https://gcc.gnu.org/PR11018 20184 301. https://gcc.gnu.org/PR11556 20185 302. https://gcc.gnu.org/PR10907 20186 303. https://gcc.gnu.org/PR11320 20187 304. https://gcc.gnu.org/PR11599 20188 305. https://gcc.gnu.org/PR9745 20189 306. https://gcc.gnu.org/PR10871 20190 307. https://gcc.gnu.org/PR11440 20191 308. https://gcc.gnu.org/PR7594 20192 309. https://gcc.gnu.org/PR10557 20193 310. https://gcc.gnu.org/PR11054 20194 311. https://gcc.gnu.org/PR10834 20195 312. https://gcc.gnu.org/PR10842 20196 313. https://gcc.gnu.org/PR11052 20197 314. https://gcc.gnu.org/PR11183 20198 315. https://gcc.gnu.org/PR11084 20199 316. https://gcc.gnu.org/PR10331 20200 317. https://gcc.gnu.org/PR10413 20201 318. https://gcc.gnu.org/PR11096 20202 319. https://gcc.gnu.org/PR2873 20203 320. https://gcc.gnu.org/PR3163 20204 321. https://gcc.gnu.org/PR5287 20205 322. https://gcc.gnu.org/PR10148 20206 323. https://gcc.gnu.org/PR8787 20207 324. https://gcc.gnu.org/PR10900 20208 325. https://gcc.gnu.org/PR1607 20209 326. https://gcc.gnu.org/PR4252 20210 327. https://gcc.gnu.org/PR4490 20211 328. https://gcc.gnu.org/PR10355 20212 329. https://gcc.gnu.org/PR10726 20213 330. https://gcc.gnu.org/PR10805 20214 331. https://gcc.gnu.org/PR10815 20215 332. https://gcc.gnu.org/PR10877 20216 333. https://gcc.gnu.org/PR11280 20217 334. https://gcc.gnu.org/PR11466 20218 335. https://gcc.gnu.org/PR10737 20219 336. https://gcc.gnu.org/PR10810 20220 337. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.2 20221 338. https://gcc.gnu.org/PR8336 20222 339. https://gcc.gnu.org/PR9330 20223 340. https://gcc.gnu.org/PR9631 20224 341. https://gcc.gnu.org/PR9877 20225 342. https://gcc.gnu.org/PR11687 20226 343. https://gcc.gnu.org/PR12263 20227 344. https://gcc.gnu.org/PR12490 20228 345. https://gcc.gnu.org/PR7277 20229 346. https://gcc.gnu.org/PR7939 20230 347. https://gcc.gnu.org/PR11063 20231 348. https://gcc.gnu.org/PR11207 20232 349. https://gcc.gnu.org/PR11522 20233 350. https://gcc.gnu.org/PR11595 20234 351. https://gcc.gnu.org/PR11646 20235 352. https://gcc.gnu.org/PR11665 20236 353. https://gcc.gnu.org/PR11852 20237 354. https://gcc.gnu.org/PR11878 20238 355. https://gcc.gnu.org/PR11883 20239 356. https://gcc.gnu.org/PR11991 20240 357. https://gcc.gnu.org/PR12146 20241 358. https://gcc.gnu.org/PR12215 20242 359. https://gcc.gnu.org/PR12369 20243 360. https://gcc.gnu.org/PR12446 20244 361. https://gcc.gnu.org/PR12510 20245 362. https://gcc.gnu.org/PR12544 20246 363. https://gcc.gnu.org/PR9862 20247 364. https://gcc.gnu.org/PR10962 20248 365. https://gcc.gnu.org/PR11370 20249 366. https://gcc.gnu.org/PR11637 20250 367. https://gcc.gnu.org/PR11885 20251 368. https://gcc.gnu.org/PR12082 20252 369. https://gcc.gnu.org/PR12180 20253 370. https://gcc.gnu.org/PR12340 20254 371. https://gcc.gnu.org/PR3907 20255 372. https://gcc.gnu.org/PR5293 20256 373. https://gcc.gnu.org/PR5296 20257 374. https://gcc.gnu.org/PR7939 20258 375. https://gcc.gnu.org/PR8656 20259 376. https://gcc.gnu.org/PR10147 20260 377. https://gcc.gnu.org/PR11400 20261 378. https://gcc.gnu.org/PR11409 20262 379. https://gcc.gnu.org/PR11740 20263 380. https://gcc.gnu.org/PR11786 20264 381. https://gcc.gnu.org/PR11867 20265 382. https://gcc.gnu.org/PR11928 20266 383. https://gcc.gnu.org/PR12114 20267 384. https://gcc.gnu.org/PR12163 20268 385. https://gcc.gnu.org/PR12181 20269 386. https://gcc.gnu.org/PR12236 20270 387. https://gcc.gnu.org/PR12266 20271 388. https://gcc.gnu.org/PR12296 20272 389. https://gcc.gnu.org/PR12298 20273 390. https://gcc.gnu.org/PR12369 20274 391. https://gcc.gnu.org/PR12337 20275 392. https://gcc.gnu.org/PR12344 20276 393. https://gcc.gnu.org/PR12451 20277 394. https://gcc.gnu.org/PR12486 20278 395. https://gcc.gnu.org/PR8869 20279 396. https://gcc.gnu.org/PR9786 20280 397. https://gcc.gnu.org/PR11689 20281 398. https://gcc.gnu.org/PR12116 20282 399. https://gcc.gnu.org/PR12070 20283 400. https://gcc.gnu.org/PR11184 20284 401. https://gcc.gnu.org/PR11535 20285 402. https://gcc.gnu.org/PR11693 20286 403. https://gcc.gnu.org/PR12224 20287 404. https://gcc.gnu.org/PR11087 20288 405. https://gcc.gnu.org/PR11319 20289 406. https://gcc.gnu.org/PR11949 20290 407. https://gcc.gnu.org/PR11662 20291 408. https://gcc.gnu.org/PR11965 20292 409. https://gcc.gnu.org/PR12301 20293 410. https://gcc.gnu.org/PR11717 20294 411. https://gcc.gnu.org/PR11313 20295 412. https://gcc.gnu.org/PR11712 20296 413. https://gcc.gnu.org/PR12166 20297 414. https://gcc.gnu.org/PR12101 20298 415. https://gcc.gnu.org/PR10988 20299 416. https://gcc.gnu.org/PR11805 20300 417. https://gcc.gnu.org/PR11902 20301 418. https://gcc.gnu.org/PR11903 20302 419. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.3 20303 420. https://gcc.gnu.org/PR11890 20304 421. https://gcc.gnu.org/PR12399 20305 422. https://gcc.gnu.org/PR13068 20306 423. https://gcc.gnu.org/PR10060 20307 424. https://gcc.gnu.org/PR10555 20308 425. https://gcc.gnu.org/PR10706 20309 426. https://gcc.gnu.org/PR11496 20310 427. https://gcc.gnu.org/PR11741 20311 428. https://gcc.gnu.org/PR12440 20312 429. https://gcc.gnu.org/PR12632 20313 430. https://gcc.gnu.org/PR12712 20314 431. https://gcc.gnu.org/PR12726 20315 432. https://gcc.gnu.org/PR12890 20316 433. https://gcc.gnu.org/PR12900 20317 434. https://gcc.gnu.org/PR13060 20318 435. https://gcc.gnu.org/PR13289 20319 436. https://gcc.gnu.org/PR13318 20320 437. https://gcc.gnu.org/PR13392 20321 438. https://gcc.gnu.org/PR13574 20322 439. https://gcc.gnu.org/PR13475 20323 440. https://gcc.gnu.org/PR13797 20324 441. https://gcc.gnu.org/PR13824 20325 442. https://gcc.gnu.org/PR8776 20326 443. https://gcc.gnu.org/PR10339 20327 444. https://gcc.gnu.org/PR11350 20328 445. https://gcc.gnu.org/PR12826 20329 446. https://gcc.gnu.org/PR12500 20330 447. https://gcc.gnu.org/PR12941 20331 448. https://gcc.gnu.org/PR12953 20332 449. https://gcc.gnu.org/PR13041 20333 450. https://gcc.gnu.org/PR13507 20334 451. https://gcc.gnu.org/PR13382 20335 452. https://gcc.gnu.org/PR13394 20336 453. https://gcc.gnu.org/PR13400 20337 454. https://gcc.gnu.org/PR13521 20338 455. https://gcc.gnu.org/PR2094 20339 456. https://gcc.gnu.org/PR2294 20340 457. https://gcc.gnu.org/PR5050 20341 458. https://gcc.gnu.org/PR9371 20342 459. https://gcc.gnu.org/PR9546 20343 460. https://gcc.gnu.org/PR10081 20344 461. https://gcc.gnu.org/PR10093 20345 462. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 20346 463. https://gcc.gnu.org/PR10095 20347 464. https://gcc.gnu.org/PR11554 20348 465. https://gcc.gnu.org/PR12297 20349 466. https://gcc.gnu.org/PR12352 20350 467. https://gcc.gnu.org/PR12438 20351 468. https://gcc.gnu.org/PR12540 20352 469. https://gcc.gnu.org/PR12594 20353 470. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 20354 471. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 20355 472. https://gcc.gnu.org/PR12657 20356 473. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 20357 474. https://gcc.gnu.org/PR12696 20358 475. https://gcc.gnu.org/PR12815 20359 476. https://gcc.gnu.org/PR12862 20360 477. https://gcc.gnu.org/PR12926 20361 478. https://gcc.gnu.org/PR12967 20362 479. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 20363 480. https://gcc.gnu.org/PR12971 20364 481. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 20365 482. https://gcc.gnu.org/PR13007 20366 483. https://gcc.gnu.org/PR13009 20367 484. https://gcc.gnu.org/PR13057 20368 485. https://gcc.gnu.org/PR13070 20369 486. https://gcc.gnu.org/PR13081 20370 487. https://gcc.gnu.org/PR13239 20371 488. https://gcc.gnu.org/PR13262 20372 489. https://gcc.gnu.org/PR13290 20373 490. https://gcc.gnu.org/PR13323 20374 491. https://gcc.gnu.org/PR13369 20375 492. https://gcc.gnu.org/PR13371 20376 493. https://gcc.gnu.org/PR13445 20377 494. https://gcc.gnu.org/PR13461 20378 495. https://gcc.gnu.org/PR13462 20379 496. https://gcc.gnu.org/PR13478 20380 497. https://gcc.gnu.org/PR13544 20381 498. https://gcc.gnu.org/PR13650 20382 499. https://gcc.gnu.org/PR13683 20383 500. https://gcc.gnu.org/PR13688 20384 501. https://gcc.gnu.org/PR13774 20385 502. https://gcc.gnu.org/PR13884 20386 503. https://gcc.gnu.org/PR10746 20387 504. https://gcc.gnu.org/PR11433 20388 505. https://gcc.gnu.org/PR12633 20389 506. https://gcc.gnu.org/PR13037 20390 507. https://gcc.gnu.org/PR13213 20391 508. https://gcc.gnu.org/PR4490 20392 509. https://gcc.gnu.org/PR12292 20393 510. https://gcc.gnu.org/PR12441 20394 511. https://gcc.gnu.org/PR12943 20395 512. https://gcc.gnu.org/PR13608 20396 513. https://gcc.gnu.org/PR11598 20397 514. https://gcc.gnu.org/PR11793 20398 515. https://gcc.gnu.org/PR12467 20399 516. https://gcc.gnu.org/PR12537 20400 517. https://gcc.gnu.org/PR12496 20401 518. https://gcc.gnu.org/PR12865 20402 519. https://gcc.gnu.org/PR13354 20403 520. https://gcc.gnu.org/PR10467 20404 521. https://gcc.gnu.org/PR11226 20405 522. https://gcc.gnu.org/PR11227 20406 523. https://gcc.gnu.org/PR12644 20407 524. https://gcc.gnu.org/PR13149 20408 525. https://gcc.gnu.org/PR12654 20409 526. https://gcc.gnu.org/PR12965 20410 527. https://gcc.gnu.org/PR13031 20411 528. https://gcc.gnu.org/PR11634 20412 529. https://gcc.gnu.org/PR12158 20413 530. https://gcc.gnu.org/PR11992 20414 531. https://gcc.gnu.org/PR9365 20415 532. https://gcc.gnu.org/PR10392 20416 533. https://gcc.gnu.org/PR11322 20417 534. https://gcc.gnu.org/PR13069 20418 535. https://gcc.gnu.org/PR13302 20419 536. https://gcc.gnu.org/PR13585 20420 537. https://gcc.gnu.org/PR8916 20421 538. https://gcc.gnu.org/PR11576 20422 539. https://gcc.gnu.org/PR13122 20423 540. https://gcc.gnu.org/PR13256 20424 541. https://gcc.gnu.org/PR13373 20425 542. https://gcc.gnu.org/PR12561 20426 543. https://gcc.gnu.org/PR6243 20427 544. https://gcc.gnu.org/PR11397 20428 545. https://gcc.gnu.org/PR12505 20429 546. https://gcc.gnu.org/PR13150 20430 547. https://gcc.gnu.org/PR12666 20431 548. https://gcc.gnu.org/PR12969 20432 549. https://gcc.gnu.org/PR10819 20433 550. https://gcc.gnu.org/PR11612 20434 551. https://gcc.gnu.org/PR13211 20435 552. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.4 20436 553. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 20437 554. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.5 20438 555. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 20439 556. https://gcc.gnu.org/gcc-3.3/changes.html#GCC3.3.6 20440 557. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 20441 558. https://gcc.gnu.org/onlinedocs/ 20442 559. mailto:gcc-help (a] gcc.gnu.org 20443 560. mailto:gcc (a] gcc.gnu.org 20444 561. https://gcc.gnu.org/lists.html 20445 562. https://www.fsf.org/ 20446 563. https://gcc.gnu.org/about.html 20447 ====================================================================== 20448 http://gcc.gnu.org/gcc-3.2/index.html 20449 20450 GCC 3.2 Release Series 20451 20452 (This release series is no longer supported.) 20453 20454 April 25, 2003 20455 20456 The [1]GNU project and the GCC developers are pleased to announce the 20457 release of GCC 3.2.3. 20458 20459 The purpose of the GCC 3.2 release series is to provide a stable 20460 platform for OS distributors to use building their next releases. A 20461 primary objective was to stabilize the C++ ABI; we believe that the 20462 interface to the compiler and the C++ standard library are now 20463 relatively stable. 20464 20465 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 20466 interoperate with code compiled by GCC 3.1.1 or earlier. 20467 20468 Please refer to our [2]detailed list of news, caveats, and bug-fixes 20469 for further information. 20470 20471 Release History 20472 20473 GCC 3.2.3 20474 April 25, 2003 ([3]changes) 20475 20476 GCC 3.2.2 20477 February 5, 2003 ([4]changes) 20478 20479 GCC 3.2.1 20480 November 19, 2002 ([5]changes) 20481 20482 GCC 3.2 20483 August 14, 2002 ([6]changes) 20484 20485 References and Acknowledgements 20486 20487 GCC used to stand for the GNU C Compiler, but since the compiler 20488 supports several other languages aside from C, it now stands for the 20489 GNU Compiler Collection. 20490 20491 A list of [7]successful builds is updated as new information becomes 20492 available. 20493 20494 The GCC developers would like to thank the numerous people that have 20495 contributed new features, improvements, bug fixes, and other changes as 20496 well as test results to GCC. This [8]amazing group of volunteers is 20497 what makes GCC successful. 20498 20499 For additional information about GCC please refer to the [9]GCC project 20500 web site or contact the [10]GCC development mailing list. 20501 20502 To obtain GCC please use [11]our mirror sites, or our CVS server. 20503 20504 20505 For questions related to the use of GCC, please consult these web 20506 pages and the [12]GCC manuals. If that fails, the 20507 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20508 web pages and the development of GCC are welcome on our developer 20509 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 20510 archives. 20511 20512 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 20513 distribution of this entire article is permitted in any medium, 20514 provided this notice is preserved. 20515 20516 These pages are [17]maintained by the GCC team. Last modified 20517 2022-10-26. 20518 20519 References 20520 20521 1. http://www.gnu.org/ 20522 2. https://gcc.gnu.org/gcc-3.2/changes.html 20523 3. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 20524 4. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 20525 5. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 20526 6. https://gcc.gnu.org/gcc-3.2/changes.html#3.2 20527 7. https://gcc.gnu.org/gcc-3.2/buildstat.html 20528 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20529 9. https://gcc.gnu.org/index.html 20530 10. mailto:gcc (a] gcc.gnu.org 20531 11. https://gcc.gnu.org/mirrors.html 20532 12. https://gcc.gnu.org/onlinedocs/ 20533 13. mailto:gcc-help (a] gcc.gnu.org 20534 14. mailto:gcc (a] gcc.gnu.org 20535 15. https://gcc.gnu.org/lists.html 20536 16. https://www.fsf.org/ 20537 17. https://gcc.gnu.org/about.html 20538 ====================================================================== 20539 http://gcc.gnu.org/gcc-3.2/changes.html 20540 20541 GCC 3.2 Release Series 20542 Changes, New Features, and Fixes 20543 20544 The latest release in the 3.2 release series is [1]GCC 3.2.3. 20545 20546 Caveats and New Features 20547 20548 Caveats 20549 20550 * The C++ compiler does not correctly zero-initialize 20551 pointers-to-data members. You must explicitly initialize them. For 20552 example: int S::*m(0); will work, but depending on 20553 default-initialization to zero will not work. This bug cannot be 20554 fixed in GCC 3.2 without inducing unacceptable risks. It will be 20555 fixed in GCC 3.3. 20556 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 20557 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 20558 a number of C++ ABI fixes which make its C++ compiler generate 20559 binary code which is incompatible with the C++ compilers found in 20560 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 20561 20562 Frontend Enhancements 20563 20564 C/C++/Objective-C 20565 20566 * The method of constructing the list of directories to be searched 20567 for header files has been revised. If a directory named by a -I 20568 option is a standard system include directory, the option is 20569 ignored to ensure that the default search order for system 20570 directories and the special treatment of system header files are 20571 not defeated. 20572 * The C and Objective-C compilers no longer accept the "Naming Types" 20573 extension (typedef foo = bar); it was already unavailable in C++. 20574 Code which uses it will need to be changed to use the "typeof" 20575 extension instead: typedef typeof(bar) foo. (We have removed this 20576 extension without a period of deprecation because it has caused the 20577 compiler to crash since version 3.0 and no one noticed until very 20578 recently. Thus we conclude it is not in widespread use.) 20579 20580 C++ 20581 20582 * GCC 3.2 fixed several differences between the C++ ABI implemented 20583 in GCC and the multi-vendor standard, but more have been found 20584 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 20585 code which is affected by these bugs. We will fix these bugs in 20586 some future release, once we are confident that all have been 20587 found; until then, it is our intention to make changes to the ABI 20588 only if they are necessary for correct compilation of C++, as 20589 opposed to conformance to the ABI documents. 20590 * For details on how to build an ABI compliant compiler for GNU/Linux 20591 systems, check the [3]common C++ ABI page. 20592 20593 New Targets and Target Specific Improvements 20594 20595 IA-32 20596 20597 * Fixed a number of bugs in SSE and MMX intrinsics. 20598 * Fixed common compiler crashes with SSE instruction set enabled 20599 (implied by -march=pentium3, pentium4, athlon-xp) 20600 * __m128 and __m128i is not 128bit aligned when used in structures. 20601 20602 x86-64 20603 20604 * A bug whereby the compiler could generate bad code for bzero has 20605 been fixed. 20606 * ABI fixes (implying ABI incompatibilities with previous version in 20607 some corner cases) 20608 * Fixed prefetch code generation 20609 __________________________________________________________________ 20610 20611 [4]GCC 3.2.3 20612 20613 3.2.3 is a bug fix release only; there are no new features that were 20614 not present in GCC 3.2.2. 20615 20616 Bug Fixes 20617 20618 This section lists the problem reports (PRs) from GCC's bug tracking 20619 system that are known to be fixed in the 3.2.3 release. This list might 20620 not be complete (that is, it is possible that some PRs that have been 20621 fixed are not listed here), and some of the titles have been changed to 20622 make them more clear. 20623 20624 Internal Compiler Errors (multi-platform) 20625 20626 * [5]3782: (c++) -quiet -fstats produces a segmentation fault in 20627 cc1plus 20628 * [6]6440: (c++) template specializations cause ICE 20629 * [7]7050: (c++) ICE on: (i ? get_string() : throw) 20630 * [8]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 20631 * [9]7982: (c++) ICE due to infinite recursion (using STL set) 20632 * [10]8068: exceedingly high (infinite) memory usage 20633 * [11]8178: ICE with __builtin_ffs 20634 * [12]8396: ICE in copy_to_mode_reg, in explow.c 20635 * [13]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 20636 * [14]9768: ICE when optimizing inline code at -O2 20637 * [15]9798: (c++) Infinite recursion (segfault) in 20638 cp/decl.c:push_using_directive with recursive using directives 20639 * [16]9799: mismatching structure initializer with nested flexible 20640 array member: ICE 20641 * [17]9928: ICE on duplicate enum declaration 20642 * [18]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 20643 sparc, alpha) 20644 * [19]10352: ICE in find_reloads_toplev 20645 * [20]10336: ICE with -Wunreachable-code 20646 20647 C/optimizer bugs: 20648 20649 * [21]8224: Incorrect joining of signed and unsigned division 20650 * [22]8613: -O2 produces wrong code with builtin strlen and 20651 postincrements 20652 * [23]8828: gcc reports some code is unreachable when it is not 20653 * [24]9226: GCSE breaking argument passing 20654 * [25]9853: miscompilation of non-constant structure initializer 20655 * [26]9797: C99-style struct initializers are miscompiled 20656 * [27]9967: Some standard C function calls should not be replaced 20657 when optimizing for size 20658 * [28]10116: ce2: invalid merge of join_bb in the context of switch 20659 statements 20660 * [29]10171: wrong code for inlined function 20661 * [30]10175: -Wunreachable-code doesn't work for single lines 20662 20663 C++ compiler and library: 20664 20665 * [31]8316: Confusing diagnostic for code that misuses conversion 20666 operators 20667 * [32]9169: filebuf output fails if codecvt<>::out returns noconv 20668 * [33]9420: incomplete type incorrectly reported 20669 * [34]9459: typeof in return type specification of template not 20670 supported 20671 * [35]9507: filebuf::open handles ios_base::ate incorrectly 20672 * [36]9538: Out-of-bounds memory access in streambuf::sputbackc 20673 * [37]9602: Total confusion about template/friend/virtual/abstract 20674 * [38]9993: destructor not called for local object created within and 20675 returned from infinite loop 20676 * [39]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 20677 system 20678 20679 Java compiler and library: 20680 20681 * [40]9652: libgcj build fails on irix6.5.1[78] 20682 * [41]10144: gas on solaris complains about bad .stabs lines for 20683 java, native as unaffected 20684 20685 x86-specific (Intel/AMD): 20686 20687 * [42]8746: gcc miscompiles Linux kernel ppa driver on x86 20688 * [43]9888: -mcpu=k6 -Os produces out of range loop instructions 20689 * [44]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 20690 failed 20691 * [45]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 20692 failed 20693 20694 SPARC-specific: 20695 20696 * [46]7784: [Sparc] ICE in extract_insn, in recog.c 20697 * [47]7796: sparc extra failure with -m64 on execute/930921-1.c in 20698 unroll.c 20699 * [48]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 20700 * [49]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 20701 execute/loop-2d.c 20702 * [50]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 20703 * [51]9414: Scheduling bug on Ultrasparc 20704 * [52]10067: GCC-3.2.2 outputs invalid asm on sparc64 20705 20706 m68k-specific: 20707 20708 * [53]7248: broken "inclusive or" code 20709 * [54]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 20710 20711 PowerPC-specific: 20712 20713 * [55]9732: Wrong code with -O2 -fPIC 20714 * [56]10073: ICE: powerpc cannot split insn 20715 20716 Alpha-specific: 20717 20718 * [57]7702: optimization problem on a DEC alpha under OSF1 20719 * [58]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 20720 20721 HP-specific: 20722 20723 * [59]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 20724 * [60]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 20725 (missing symbol) 20726 * [61]10271: Floating point args don't get reloaded across function 20727 calls with -O2 20728 20729 MIPS specific: 20730 20731 * [62]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 20732 compile/920501-4.c 20733 20734 CRIS specific: 20735 20736 * [63]10377: gcc-3.2.2 creates bad assembler code for cris 20737 20738 Miscellaneous and minor bugs: 20739 20740 * [64]6955: collect2 says "core dumped" when there is no core 20741 __________________________________________________________________ 20742 20743 [65]GCC 3.2.2 20744 20745 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 20746 install by means of the DESTDIR variable. Parts of the GCC tree have 20747 featured that support long before, but now it is available even from 20748 the top level. 20749 20750 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 20751 features that were not present in GCC 3.2.1. 20752 20753 Bug Fixes 20754 20755 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 20756 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 20757 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 20758 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 20759 change, and thus restores ABI-compatibility with previous releases 20760 (except GCC 3.2.1) on these platforms. 20761 20762 This section lists the problem reports (PRs) from GCC's bug tracking 20763 system that are known to be fixed in the 3.2.2 release. This list might 20764 not be complete (that is, it is possible that some PRs that have been 20765 fixed are not listed here) and some of the titles have been changed to 20766 make them more clear. 20767 20768 Internal Compiler Errors (multi-platform) 20769 20770 * [66]5919: (c++) ICE when passing variable array to template 20771 function 20772 * [67]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 20773 * [68]7507: ICE with -O2 when address of called function is a 20774 complicated expression 20775 * [69]7622: ICE with nested inline functions if function's address is 20776 taken 20777 * [70]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 20778 [71]9258) 20779 * [72]8031: (c++) ICE in code comparing typeids and casting from 20780 virtual base 20781 * [73]8275: ICE in simplify_subreg 20782 * [74]8332: (c++) builtin strlen/template interaction causes ICE 20783 * [75]8372: (c++) ICE on explicit call of destructor 20784 * [76]8439: (c, not c++) empty struct causes ICE 20785 * [77]8442: (c++) ICE with nested template classes 20786 * [78]8518: ICE when compiling mplayer ("extern inline" issue) 20787 * [79]8615: (c++) ICE with out-of-range character constant template 20788 argument 20789 * [80]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 20790 * [81]8799: (c++) ICE: error reporting routines re-entered 20791 * [82]9328: (c++) ICE with typeof(X) for overloaded X 20792 * [83]9465: (preprocessor) cpp -traditional ICE on null bytes 20793 20794 C++ (compiler and library) bugs 20795 20796 * [84]47: scoping in nested classes is broken 20797 * [85]6745: problems with iostream rdbuf() member function 20798 * [86]8214: conversion from const char* const to char* sometimes 20799 accepted illegally 20800 * [87]8493: builtin strlen and overload resolution (same bug as 20801 [88]8332) 20802 * [89]8503: strange behaviour of function types 20803 * [90]8727: compiler confused by inheritance from an anonymous struct 20804 * [91]7445: poor performance of std::locale::classic() in 20805 multi-threaded applications 20806 * [92]8230: mishandling of overflow in vector<T>::resize 20807 * [93]8399: sync_with_stdio(false) breaks unformatted input 20808 * [94]8662: illegal access of private member of unnamed class is 20809 accepted 20810 * [95]8707: "make distclean" fails in libstdc++-v3 directory 20811 * [96]8708: __USE_MALLOC doesn't work 20812 * [97]8790: Use of non-thread-safe strtok in src/localename.cc 20813 * [98]8887: Bug in date formats with --enable-clocale=generic 20814 * [99]9076: Call Frame Instructions are not handled correctly during 20815 unwind operation 20816 * [100]9151: std::setprecision limited to 16 digits when outputting a 20817 double to a stream 20818 * [101]9168: codecvt<char, char, mbstate_t> overwrites output buffers 20819 * [102]9269: libstdc++ headers: explicit specialization of function 20820 must precede its first use 20821 * [103]9322: return value of basic_streambuf<>::getloc affected by 20822 locale::global 20823 * [104]9433: segfault in runtime support for dynamic_cast 20824 20825 C and optimizer bugs 20826 20827 * [105]8032: GCC incorrectly initializes static structs that have 20828 flexible arrays 20829 * [106]8639: simple arithmetic expression broken 20830 * [107]8794: optimization improperly eliminates certain expressions 20831 * [108]8832: traditional "asm volatile" code is illegally optimized 20832 * [109]8988: loop optimizer bug: with -O2, code is generated that 20833 segfaults (found on i386, bug present for all platforms) 20834 * [110]9492: structure copy clobbers subsequent stores to structure 20835 20836 Objective-C bugs 20837 20838 * [111]9267: Objective-C parser won't build with newer bison versions 20839 (e.g. 1.875) 20840 20841 Ada bugs 20842 20843 * [112]8344: Ada build problem due to conflict between gcc/final.o, 20844 gcc/ada/final.o 20845 20846 Preprocessor bugs 20847 20848 * [113]8524: _Pragma within macros is improperly expanded 20849 * [114]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 20850 -fshort-wchar 20851 20852 ARM-specific 20853 20854 * [115]9090: arm ICE with >= -O2; regression from gcc-2.95 20855 20856 x86-specific (Intel/AMD) 20857 20858 * [116]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 20859 * [117]8599: loop unroll bug with -march=k6-3 20860 * [118]9506: ABI breakage in structure return (affects BSD and 20861 Cygwin, but not GNU/Linux) 20862 20863 FreeBSD 5.0 specific 20864 20865 * [119]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 20866 20867 RTEMS-specific 20868 20869 * [120]9292: hppa1.1-rtems configurery problems 20870 * [121]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 20871 * [122]9295: [mips-rtems] config/mips/rtems.h init/fini issue 20872 * [123]9296: gthr-rtems regression 20873 * [124]9316: powerpc-rtems: extending multilibs 20874 20875 HP-PA specific 20876 20877 * [125]9493: ICE with -O2 when building a simple function 20878 20879 Documentation 20880 20881 * [126]7341: hyperlink to gcov in GCC documentation doesn't work 20882 * [127]8947: Please add a warning about "-malign-double" in docs 20883 * [128]7448, [129]8882: typo cleanups 20884 __________________________________________________________________ 20885 20886 [130]GCC 3.2.1 20887 20888 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 20889 generates code that is known not to be binary-compatible with the 20890 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 20891 in the distribution, for details. 20892 20893 This release also removes an old GCC extension, "naming types", and the 20894 documentation now directs users to use a different GCC extension, 20895 __typeof__, instead. The feature had evidently been broken for a while. 20896 20897 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 20898 the new warning there are no new features that were not present in GCC 20899 3.2. 20900 20901 In addition, the previous fix for [131]PR 7445 (poor performance of 20902 std::locale::classic() in multi-threaded applications) was reverted 20903 ("unfixed"), because the "fix" was not thread-safe. 20904 20905 Bug Fixes 20906 20907 This section lists the problem reports (PRs) from GCC's bug tracking 20908 system that are known to be fixed in the 3.2.1 release. This list might 20909 not be complete (that is, it is possible that some PRs that have been 20910 fixed are not listed here). As you can see, the number of bug fixes is 20911 quite large, so it is strongly recommended that users of earlier GCC 20912 3.x releases upgrade to GCC 3.2.1. 20913 20914 Internal Compiler Errors (multi-platform) 20915 20916 * [132]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 20917 * [133]5661: (c++) ICE instantiating template on array of unknown 20918 size (bad code) 20919 * [134]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 20920 64-bit platforms 20921 * [135]6994: (c++) ICE in find_function_data 20922 * [136]7150: preprocessor: GCC -dM -E gives an ICE 20923 * [137]7160: ICE when optimizing branches without a return value 20924 * [138]7228: (c++) ICE when using member template and template 20925 function 20926 * [139]7266: (c++) ICE with -pedantic on missing typename 20927 * [140]7353: ICE from use of "Naming Types" extension, see above 20928 * [141]7411: ICE in instantiate_virtual_regs_1, in function.c 20929 * [142]7478: (c++) ICE on static_cast inside template 20930 * [143]7526: preprocessor core dump when _Pragma implies #pragma 20931 dependency 20932 * [144]7721: (c++) ICE on simple (but incorrect) template ([145]7803 20933 is a duplicate) 20934 * [146]7754: (c++) ICE on union with template parameter 20935 * [147]7788: (c++) redeclaring a definition as an incomplete class 20936 causes ICE 20937 * [148]8031: (c++) ICE in comptypes, in cp/typeck.c 20938 * [149]8055: preprocessor dies with SIG11 when building FreeBSD 20939 kernel 20940 * [150]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 20941 variables 20942 * [151]8134: (c++) ICE in force_store_init_value on legal code 20943 * [152]8149: (c++) ICE on incomplete type 20944 * [153]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 20945 initialization 20946 20947 C++ (compiler and library) bugs 20948 20949 * [154]5607: No pointer adjustment in covariant return types 20950 * [155]6579: Infinite loop with statement expressions in member 20951 initialization 20952 * [156]6803: Default copy constructor bug in GCC 3.1 20953 * [157]7176: g++ confused by friend and static member with same name 20954 * [158]7188: Segfault with template class and recursive (incorrect) 20955 initializer list 20956 * [159]7306: Regression: GCC 3.x fails to compile code with virtual 20957 inheritance if a method has a variable number of arguments 20958 * [160]7461: ctype<char>::classic_table() returns offset array on 20959 Cygwin 20960 * [161]7524: f(const float arg[3]) fails 20961 * [162]7584: Erroneous ambiguous base error on using declaration 20962 * [163]7676: Member template overloading problem 20963 * [164]7679: infinite loop when a right parenthesis is missing 20964 * [165]7811: default locale not taken from environment 20965 * [166]7961: compare( char *) implemented incorrectly in 20966 basic_string<> 20967 * [167]8071: basic_ostream::operator<<(streambuf*) loops forever if 20968 streambuf::underflow() leaves gptr() NULL (dups: [168]8127, 20969 [169]6745) 20970 * [170]8096: deque::at() throws std::range_error instead of 20971 std::out_of_range 20972 * [171]8127: cout << cin.rdbuf() infinite loop 20973 * [172]8218: Excessively large memory consumed for classes with large 20974 array members 20975 * [173]8287: GCC 3.2: Destructor called for non-constructed local 20976 object 20977 * [174]8347: empty vector range used in string construction causes 20978 core dump 20979 * [175]8348: fail() flag is set in istringstream when eof() flag is 20980 set 20981 * [176]8391: regression: infinite loop in cp/decl2.c(finish_file) 20982 20983 C and optimizer bugs 20984 20985 * [177]6627: -fno-align-functions doesn't seem to disable function 20986 alignment 20987 * [178]6631: life_analysis misoptimizes code to initialize fields of 20988 a structure 20989 * [179]7102: unsigned char division results in floating exception 20990 * [180]7120: Run once loop should *always* be unrolled 20991 (pessimization) 20992 * [181]7209: Bug involving array referencing and ?: operator 20993 * [182]7515: invalid inlining of global function with -O3 20994 * [183]7814: incorrect scheduling for glibc-2.2.92 strcpy test 20995 * [184]8467: bug in sibling call optimization 20996 20997 Preprocessor bugs 20998 20999 * [185]4890: incorrect line markers from the traditional preprocessor 21000 * [186]7357: -M option omits system headers files (making it the same 21001 as -MM) 21002 * [187]7358: Changes to Sun's make Dependencies 21003 * [188]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 21004 C headers 21005 * [189]7862: Interrupting GCC -MD removes .d file but not .o 21006 * [190]8190: Failed compilation deletes -MD dependency file 21007 * [191]8524: _Pragma within macro is improperly expanded 21008 21009 x86 specific (Intel/AMD) 21010 21011 * [192]5351: (i686-only) function pass-by-value structure copy 21012 corrupts stack ([193]7591 is a duplicate) 21013 * [194]6845, [195]7034, [196]7124, [197]7174: ICE's with 21014 -march=pentium3/pentium2/athlon (these are all the same underlying 21015 bug, in MMX register use) 21016 * [198]7134, [199]7375, [200]7390: ICE with -march=athlon (maybe same 21017 as above?) 21018 * [201]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 21019 * [202]6981: wrong code in 64-bit manipulation on x86 21020 * [203]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 21021 macro 21022 * [204]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 21023 intrinsics are broken 21024 * [205]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 21025 -march=pentium4 21026 * [206]7693: Typo in i386 mmintrin.h header 21027 * [207]7723: ICE - Pentium3 sse - GCC 3.2 21028 * [208]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 21029 * [209]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 21030 21031 PowerPC specific 21032 21033 * [210]5967: GCC bug when profiling nested functions on powerpc 21034 * [211]6984: wrong code generated with -O2, -O3, -Os for do-while 21035 loop on PowerPC 21036 * [212]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 21037 * [213]7130: miscompiled code for GCC-3.1 on 21038 powerpc-unknown-linux-gnu with -funroll-all-loops 21039 * [214]7133: PowerPC ICE: unrecognizable insn 21040 * [215]7380: ICE in extract_insn, at recog.c:2148 21041 * [216]8252: ICE on Altivec code with optimization turned on 21042 * [217]8451: Altivec ICE in GCC 3.2 21043 21044 HP/PA specific 21045 21046 * [218]7250: __ashrdi3 returns wrong value on 32 bit hppa 21047 21048 SPARC specific 21049 21050 * [219]6668: when using --disable-multilib, libgcc_s.so is installed 21051 in the wrong place on sparc-solaris 21052 * [220]7151: ICE when compiling for UltraSPARC 21053 * [221]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 21054 double and -O1 21055 * [222]7842: [REGRESSION] SPARC code gen bug 21056 21057 ARM specific 21058 21059 * [223]7856: [arm] invalid offset in constant pool reference 21060 * [224]7967: optimization produces wrong code (ARM) 21061 21062 Alpha specific 21063 21064 * [225]7374: __builtin_fabsl broken on alpha 21065 21066 IBM s390 specific 21067 21068 * [226]7370: ICE in fixup_var_refs_1 on s390x 21069 * [227]7409: loop optimization bug on s390x-linux-gnu 21070 * [228]8232: s390x: ICE when using bcmp with int length argument 21071 21072 SCO specific 21073 21074 * [229]7623: SCO OpenServer build fails with machmode.def: undefined 21075 symbol: BITS_PER_UNIT 21076 21077 m68k/Coldfire specific 21078 21079 * [230]8314: crtbegin, crtend need to be multilib'ed for this 21080 platform 21081 21082 Documentation 21083 21084 * [231]761: Document some undocumented options 21085 * [232]5610: Fix documentation about invoking SSE instructions 21086 (-mfpmath=sse) 21087 * [233]7484: List -Wmissing-declarations as C-only option 21088 * [234]7531: -mcmodel not documented for x86-64 21089 * [235]8120: Update documentation of bad use of ## 21090 __________________________________________________________________ 21091 21092 [236]GCC 3.2 21093 21094 3.2 is a small bug fix release, but there is a change to the 21095 application binary interface (ABI), hence the change to the second part 21096 of the version number. 21097 21098 The main purpose of the 3.2 release is to correct a couple of problems 21099 in the C++ ABI, with the intention of providing a stable interface 21100 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 21101 21102 Bug Fixes 21103 21104 C++ 21105 21106 * [237]7320: g++ 3.2 relocation problem 21107 * [238]7470: vtable: virtual function pointers not in declaration 21108 order 21109 21110 libstdc++ 21111 21112 * [239]6410: Trouble with non-ASCII monetary symbols and wchar_t 21113 * [240]6503, [241]6642, [242]7186: Problems with comparing or 21114 subtracting various types of const and non-const iterators 21115 * [243]7216: ambiguity with basic_iostream::traits_type 21116 * [244]7220: problem with basic_istream::ignore(0,delimiter) 21117 * [245]7222: locale::operator==() doesn't work on std::locale("") 21118 * [246]7286: placement operator delete issue 21119 * [247]7442: cxxabi.h does not match the C++ ABI 21120 * [248]7445: poor performance of std::locale::classic() in 21121 multi-threaded applications 21122 21123 x86-64 specific 21124 21125 * [249]7291: off-by-one in generated inline bzero code for x86-64 21126 21127 21128 For questions related to the use of GCC, please consult these web 21129 pages and the [250]GCC manuals. If that fails, the 21130 [251]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21131 web pages and the development of GCC are welcome on our developer 21132 list at [252]gcc (a] gcc.gnu.org. All of [253]our lists have public 21133 archives. 21134 21135 Copyright (C) [254]Free Software Foundation, Inc. Verbatim copying and 21136 distribution of this entire article is permitted in any medium, 21137 provided this notice is preserved. 21138 21139 These pages are [255]maintained by the GCC team. Last modified 21140 2025-04-08. 21141 21142 References 21143 21144 1. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 21145 2. https://gcc.gnu.org/gcc-3.1/changes.html 21146 3. https://gcc.gnu.org/gcc-3.2/c++-abi.html 21147 4. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.3 21148 5. https://gcc.gnu.org/PR3782 21149 6. https://gcc.gnu.org/PR6440 21150 7. https://gcc.gnu.org/PR7050 21151 8. https://gcc.gnu.org/PR7741 21152 9. https://gcc.gnu.org/PR7982 21153 10. https://gcc.gnu.org/PR8068 21154 11. https://gcc.gnu.org/PR8178 21155 12. https://gcc.gnu.org/PR8396 21156 13. https://gcc.gnu.org/PR8674 21157 14. https://gcc.gnu.org/PR9768 21158 15. https://gcc.gnu.org/PR9798 21159 16. https://gcc.gnu.org/PR9799 21160 17. https://gcc.gnu.org/PR9928 21161 18. https://gcc.gnu.org/PR10114 21162 19. https://gcc.gnu.org/PR10352 21163 20. https://gcc.gnu.org/PR10336 21164 21. https://gcc.gnu.org/PR8224 21165 22. https://gcc.gnu.org/PR8613 21166 23. https://gcc.gnu.org/PR8828 21167 24. https://gcc.gnu.org/PR9226 21168 25. https://gcc.gnu.org/PR9853 21169 26. https://gcc.gnu.org/PR9797 21170 27. https://gcc.gnu.org/PR9967 21171 28. https://gcc.gnu.org/PR10116 21172 29. https://gcc.gnu.org/PR10171 21173 30. https://gcc.gnu.org/PR10175 21174 31. https://gcc.gnu.org/PR8316 21175 32. https://gcc.gnu.org/PR9169 21176 33. https://gcc.gnu.org/PR9420 21177 34. https://gcc.gnu.org/PR9459 21178 35. https://gcc.gnu.org/PR9507 21179 36. https://gcc.gnu.org/PR9538 21180 37. https://gcc.gnu.org/PR9602 21181 38. https://gcc.gnu.org/PR9993 21182 39. https://gcc.gnu.org/PR10167 21183 40. https://gcc.gnu.org/PR9652 21184 41. https://gcc.gnu.org/PR10144 21185 42. https://gcc.gnu.org/PR8746 21186 43. https://gcc.gnu.org/PR9888 21187 44. https://gcc.gnu.org/PR9638 21188 45. https://gcc.gnu.org/PR9954 21189 46. https://gcc.gnu.org/PR7784 21190 47. https://gcc.gnu.org/PR7796 21191 48. https://gcc.gnu.org/PR8281 21192 49. https://gcc.gnu.org/PR8366 21193 50. https://gcc.gnu.org/PR8726 21194 51. https://gcc.gnu.org/PR9414 21195 52. https://gcc.gnu.org/PR10067 21196 53. https://gcc.gnu.org/PR7248 21197 54. https://gcc.gnu.org/PR8343 21198 55. https://gcc.gnu.org/PR9732 21199 56. https://gcc.gnu.org/PR10073 21200 57. https://gcc.gnu.org/PR7702 21201 58. https://gcc.gnu.org/PR9671 21202 59. https://gcc.gnu.org/PR8694 21203 60. https://gcc.gnu.org/PR9953 21204 61. https://gcc.gnu.org/PR10271 21205 62. https://gcc.gnu.org/PR6362 21206 63. https://gcc.gnu.org/PR10377 21207 64. https://gcc.gnu.org/PR6955 21208 65. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.2 21209 66. https://gcc.gnu.org/PR5919 21210 67. https://gcc.gnu.org/PR7129 21211 68. https://gcc.gnu.org/PR7507 21212 69. https://gcc.gnu.org/PR7622 21213 70. https://gcc.gnu.org/PR7681 21214 71. https://gcc.gnu.org/PR9528 21215 72. https://gcc.gnu.org/PR8031 21216 73. https://gcc.gnu.org/PR8275 21217 74. https://gcc.gnu.org/PR8332 21218 75. https://gcc.gnu.org/PR8372 21219 76. https://gcc.gnu.org/PR8439 21220 77. https://gcc.gnu.org/PR8442 21221 78. https://gcc.gnu.org/PR8518 21222 79. https://gcc.gnu.org/PR8615 21223 80. https://gcc.gnu.org/PR8663 21224 81. https://gcc.gnu.org/PR8799 21225 82. https://gcc.gnu.org/PR9328 21226 83. https://gcc.gnu.org/PR9465 21227 84. https://gcc.gnu.org/PR47 21228 85. https://gcc.gnu.org/PR6745 21229 86. https://gcc.gnu.org/PR8214 21230 87. https://gcc.gnu.org/PR8493 21231 88. https://gcc.gnu.org/PR8332 21232 89. https://gcc.gnu.org/PR8503 21233 90. https://gcc.gnu.org/PR8727 21234 91. https://gcc.gnu.org/PR7445 21235 92. https://gcc.gnu.org/PR8230 21236 93. https://gcc.gnu.org/PR8399 21237 94. https://gcc.gnu.org/PR8662 21238 95. https://gcc.gnu.org/PR8707 21239 96. https://gcc.gnu.org/PR8708 21240 97. https://gcc.gnu.org/PR8790 21241 98. https://gcc.gnu.org/PR8887 21242 99. https://gcc.gnu.org/PR9076 21243 100. https://gcc.gnu.org/PR9151 21244 101. https://gcc.gnu.org/PR9168 21245 102. https://gcc.gnu.org/PR9269 21246 103. https://gcc.gnu.org/PR9322 21247 104. https://gcc.gnu.org/PR9433 21248 105. https://gcc.gnu.org/PR8032 21249 106. https://gcc.gnu.org/PR8639 21250 107. https://gcc.gnu.org/PR8794 21251 108. https://gcc.gnu.org/PR8832 21252 109. https://gcc.gnu.org/PR8988 21253 110. https://gcc.gnu.org/PR9492 21254 111. https://gcc.gnu.org/PR9267 21255 112. https://gcc.gnu.org/PR8344 21256 113. https://gcc.gnu.org/PR8524 21257 114. https://gcc.gnu.org/PR8880 21258 115. https://gcc.gnu.org/PR9090 21259 116. https://gcc.gnu.org/PR8588 21260 117. https://gcc.gnu.org/PR8599 21261 118. https://gcc.gnu.org/PR9506 21262 119. https://gcc.gnu.org/PR9484 21263 120. https://gcc.gnu.org/PR9292 21264 121. https://gcc.gnu.org/PR9293 21265 122. https://gcc.gnu.org/PR9295 21266 123. https://gcc.gnu.org/PR9296 21267 124. https://gcc.gnu.org/PR9316 21268 125. https://gcc.gnu.org/PR9493 21269 126. https://gcc.gnu.org/PR7341 21270 127. https://gcc.gnu.org/PR8947 21271 128. https://gcc.gnu.org/PR7448 21272 129. https://gcc.gnu.org/PR8882 21273 130. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2.1 21274 131. https://gcc.gnu.org/PR7445 21275 132. https://gcc.gnu.org/PR2521 21276 133. https://gcc.gnu.org/PR5661 21277 134. https://gcc.gnu.org/PR6419 21278 135. https://gcc.gnu.org/PR6994 21279 136. https://gcc.gnu.org/PR7150 21280 137. https://gcc.gnu.org/PR7160 21281 138. https://gcc.gnu.org/PR7228 21282 139. https://gcc.gnu.org/PR7266 21283 140. https://gcc.gnu.org/PR7353 21284 141. https://gcc.gnu.org/PR7411 21285 142. https://gcc.gnu.org/PR7478 21286 143. https://gcc.gnu.org/PR7526 21287 144. https://gcc.gnu.org/PR7721 21288 145. https://gcc.gnu.org/PR7803 21289 146. https://gcc.gnu.org/PR7754 21290 147. https://gcc.gnu.org/PR7788 21291 148. https://gcc.gnu.org/PR8031 21292 149. https://gcc.gnu.org/PR8055 21293 150. https://gcc.gnu.org/PR8067 21294 151. https://gcc.gnu.org/PR8134 21295 152. https://gcc.gnu.org/PR8149 21296 153. https://gcc.gnu.org/PR8160 21297 154. https://gcc.gnu.org/PR5607 21298 155. https://gcc.gnu.org/PR6579 21299 156. https://gcc.gnu.org/PR6803 21300 157. https://gcc.gnu.org/PR7176 21301 158. https://gcc.gnu.org/PR7188 21302 159. https://gcc.gnu.org/PR7306 21303 160. https://gcc.gnu.org/PR7461 21304 161. https://gcc.gnu.org/PR7524 21305 162. https://gcc.gnu.org/PR7584 21306 163. https://gcc.gnu.org/PR7676 21307 164. https://gcc.gnu.org/PR7679 21308 165. https://gcc.gnu.org/PR7811 21309 166. https://gcc.gnu.org/PR7961 21310 167. https://gcc.gnu.org/PR8071 21311 168. https://gcc.gnu.org/PR8127 21312 169. https://gcc.gnu.org/PR6745 21313 170. https://gcc.gnu.org/PR8096 21314 171. https://gcc.gnu.org/PR8127 21315 172. https://gcc.gnu.org/PR8218 21316 173. https://gcc.gnu.org/PR8287 21317 174. https://gcc.gnu.org/PR8347 21318 175. https://gcc.gnu.org/PR8348 21319 176. https://gcc.gnu.org/PR8391 21320 177. https://gcc.gnu.org/PR6627 21321 178. https://gcc.gnu.org/PR6631 21322 179. https://gcc.gnu.org/PR7102 21323 180. https://gcc.gnu.org/PR7120 21324 181. https://gcc.gnu.org/PR7209 21325 182. https://gcc.gnu.org/PR7515 21326 183. https://gcc.gnu.org/PR7814 21327 184. https://gcc.gnu.org/PR8467 21328 185. https://gcc.gnu.org/PR4890 21329 186. https://gcc.gnu.org/PR7357 21330 187. https://gcc.gnu.org/PR7358 21331 188. https://gcc.gnu.org/PR7602 21332 189. https://gcc.gnu.org/PR7862 21333 190. https://gcc.gnu.org/PR8190 21334 191. https://gcc.gnu.org/PR8524 21335 192. https://gcc.gnu.org/PR5351 21336 193. https://gcc.gnu.org/PR7591 21337 194. https://gcc.gnu.org/PR6845 21338 195. https://gcc.gnu.org/PR7034 21339 196. https://gcc.gnu.org/PR7124 21340 197. https://gcc.gnu.org/PR7174 21341 198. https://gcc.gnu.org/PR7134 21342 199. https://gcc.gnu.org/PR7375 21343 200. https://gcc.gnu.org/PR7390 21344 201. https://gcc.gnu.org/PR6890 21345 202. https://gcc.gnu.org/PR6981 21346 203. https://gcc.gnu.org/PR7242 21347 204. https://gcc.gnu.org/PR7396 21348 205. https://gcc.gnu.org/PR7630 21349 206. https://gcc.gnu.org/PR7693 21350 207. https://gcc.gnu.org/PR7723 21351 208. https://gcc.gnu.org/PR7951 21352 209. https://gcc.gnu.org/PR8146 21353 210. https://gcc.gnu.org/PR5967 21354 211. https://gcc.gnu.org/PR6984 21355 212. https://gcc.gnu.org/PR7114 21356 213. https://gcc.gnu.org/PR7130 21357 214. https://gcc.gnu.org/PR7133 21358 215. https://gcc.gnu.org/PR7380 21359 216. https://gcc.gnu.org/PR8252 21360 217. https://gcc.gnu.org/PR8451 21361 218. https://gcc.gnu.org/PR7250 21362 219. https://gcc.gnu.org/PR6668 21363 220. https://gcc.gnu.org/PR7151 21364 221. https://gcc.gnu.org/PR7335 21365 222. https://gcc.gnu.org/PR7842 21366 223. https://gcc.gnu.org/PR7856 21367 224. https://gcc.gnu.org/PR7967 21368 225. https://gcc.gnu.org/PR7374 21369 226. https://gcc.gnu.org/PR7370 21370 227. https://gcc.gnu.org/PR7409 21371 228. https://gcc.gnu.org/PR8232 21372 229. https://gcc.gnu.org/PR7623 21373 230. https://gcc.gnu.org/PR8314 21374 231. https://gcc.gnu.org/PR761 21375 232. https://gcc.gnu.org/PR5610 21376 233. https://gcc.gnu.org/PR7484 21377 234. https://gcc.gnu.org/PR7531 21378 235. https://gcc.gnu.org/PR8120 21379 236. https://gcc.gnu.org/gcc-3.2/changes.html#GCC3.2 21380 237. https://gcc.gnu.org/PR7320 21381 238. https://gcc.gnu.org/PR7470 21382 239. https://gcc.gnu.org/PR6410 21383 240. https://gcc.gnu.org/PR6503 21384 241. https://gcc.gnu.org/PR6642 21385 242. https://gcc.gnu.org/PR7186 21386 243. https://gcc.gnu.org/PR7216 21387 244. https://gcc.gnu.org/PR7220 21388 245. https://gcc.gnu.org/PR7222 21389 246. https://gcc.gnu.org/PR7286 21390 247. https://gcc.gnu.org/PR7442 21391 248. https://gcc.gnu.org/PR7445 21392 249. https://gcc.gnu.org/PR7291 21393 250. https://gcc.gnu.org/onlinedocs/ 21394 251. mailto:gcc-help (a] gcc.gnu.org 21395 252. mailto:gcc (a] gcc.gnu.org 21396 253. https://gcc.gnu.org/lists.html 21397 254. https://www.fsf.org/ 21398 255. https://gcc.gnu.org/about.html 21399 ====================================================================== 21400 http://gcc.gnu.org/gcc-3.1/index.html 21401 21402 GCC 3.1 21403 21404 (This release series is no longer supported.) 21405 21406 July 27, 2002 21407 21408 The GNU project and the GCC developers are pleased to announce the 21409 release of GCC 3.1.1. 21410 21411 The links below still apply to GCC 3.1.1. 21412 21413 May 15, 2002 21414 21415 The [1]GNU project and the GCC developers are pleased to announce the 21416 release of GCC 3.1. 21417 21418 GCC used to stand for the GNU C Compiler, but since the compiler 21419 supports several other languages aside from C, it now stands for the 21420 GNU Compiler Collection. 21421 21422 A list of [2]successful builds is updated as new information becomes 21423 available. 21424 21425 The GCC developers would like to thank the numerous people that have 21426 contributed [3]new features, improvements, bug fixes, and other changes 21427 as well as test results to GCC. This [4]amazing group of volunteers is 21428 what makes GCC successful. 21429 21430 For additional information about GCC please refer to the [5]GCC project 21431 web site or contact the [6]GCC development mailing list. 21432 21433 To obtain GCC please use [7]our mirror sites, or our CVS server. 21434 __________________________________________________________________ 21435 21436 21437 For questions related to the use of GCC, please consult these web 21438 pages and the [8]GCC manuals. If that fails, the 21439 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21440 web pages and the development of GCC are welcome on our developer 21441 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 21442 archives. 21443 21444 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 21445 distribution of this entire article is permitted in any medium, 21446 provided this notice is preserved. 21447 21448 These pages are [13]maintained by the GCC team. Last modified 21449 2024-07-04. 21450 21451 References 21452 21453 1. https://www.gnu.org/ 21454 2. https://gcc.gnu.org/gcc-3.1/buildstat.html 21455 3. https://gcc.gnu.org/gcc-3.1/changes.html 21456 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 21457 5. https://gcc.gnu.org/index.html 21458 6. mailto:gcc (a] gcc.gnu.org 21459 7. https://gcc.gnu.org/mirrors.html 21460 8. https://gcc.gnu.org/onlinedocs/ 21461 9. mailto:gcc-help (a] gcc.gnu.org 21462 10. mailto:gcc (a] gcc.gnu.org 21463 11. https://gcc.gnu.org/lists.html 21464 12. https://www.fsf.org/ 21465 13. https://gcc.gnu.org/about.html 21466 ====================================================================== 21467 http://gcc.gnu.org/gcc-3.1/changes.html 21468 21469 GCC 3.1 Release Series 21470 Changes, New Features, and Fixes 21471 21472 Additional changes in GCC 3.1.1 21473 21474 * A bug related to how structures and unions are returned has been 21475 fixed for powerpc-*-netbsd*. 21476 * An important bug in the implementation of -fprefetch-loop-arrays 21477 has been fixed. Previously the optimization prefetched random 21478 blocks of memory for most targets except for i386. 21479 * The Java compiler now compiles Java programs much faster and also 21480 works with parallel make. 21481 * Nested functions have been fixed for mips*-*-netbsd*. 21482 * Some missing floating point support routines have beed added for 21483 mips*-*-netbsd*. 21484 * This [1]message gives additional information about the bugs fixed 21485 in this release. 21486 21487 Caveats 21488 21489 * The -traditional C compiler option has been deprecated and will be 21490 removed in GCC 3.3. (It remains possible to preprocess non-C code 21491 with the traditional preprocessor.) 21492 * The default debugging format for most ELF platforms (including 21493 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 21494 from stabs to DWARF2. This requires GDB 5.1.1 or later. 21495 21496 General Optimizer Improvements 21497 21498 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 21499 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 21500 for profile driven optimizations. 21501 Options -fprofile-arcs and -fbranch-probabilities can now be used 21502 to improve speed of the generated code by profiling the actual 21503 program behaviour on typical runs. In the absence of profile info 21504 the compiler attempts to guess the profile statically. 21505 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 21506 monitor performance of the generated code. 21507 According to the SPECInt2000 results on an AMD Athlon CPU, the code 21508 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 21509 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 21510 is about 2.1% faster compared to 2.95.3. Tests were done using the 21511 -O2 -march=athlon command-line options. 21512 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 21513 infrastructure developed by CodeSourcery, LLC for the C++ front 21514 end, so that it is now used in the C front end too. Inlining 21515 functions as trees exposes them earlier to the compiler, giving it 21516 more opportunities for optimization. 21517 * Support for data prefetching instructions has been added to the GCC 21518 back end and several targets. A new __builtin_prefetch intrinsic is 21519 available to explicitly insert prefetch instructions and 21520 experimental support for loop array prefetching has been added (see 21521 -fprefetch-loop-array documentation). 21522 * Support for emitting debugging information for macros has been 21523 added for DWARF2. It is activated using -g3. 21524 21525 New Languages and Language specific improvements 21526 21527 C/C++ 21528 21529 * A few more [4]ISO C99 features. 21530 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 21531 * The preprocessor's symbol table has been merged with the symbol 21532 table of the C, C++ and Objective-C front ends. 21533 * The preprocessor consumes less memory than the preprocessor in GCC 21534 3.0, often significantly so. On normal input files, it typically 21535 consumes less memory than pre-3.0 cccp-based GCC, too. 21536 21537 C++ 21538 21539 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 21540 was a workaround to allow std compliant code to work with the 21541 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 21542 * The C++ ABI has been fixed so that void (A::*)() const is mangled 21543 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 21544 affects pointer to cv-qualified member function types. 21545 * The C++ ABI has been changed to correctly handle this code: 21546 struct A { 21547 void operator delete[] (void *, size_t); 21548 }; 21549 21550 struct B : public A { 21551 }; 21552 21553 new B[10]; 21554 21555 The amount of storage allocated for the array will be greater than 21556 it was in 3.0, in order to store the number of elements in the 21557 array, so that the correct size can be passed to operator delete[] 21558 when the array is deleted. Previously, the value passed to operator 21559 delete[] was unpredictable. 21560 This change will only affect code that declares a two-argument 21561 operator delete[] with a second parameter of type size_t in a base 21562 class, and does not override that definition in a derived class. 21563 * The C++ ABI has been changed so that: 21564 struct A { 21565 void operator delete[] (void *, size_t); 21566 void operator delete[] (void *); 21567 }; 21568 21569 does not cause unnecessary storage to be allocated when an array of 21570 A objects is allocated. 21571 This change will only affect code that declares both of these forms 21572 of operator delete[], and declared the two-argument form before the 21573 one-argument form. 21574 * The C++ ABI has been changed so that when a parameter is passed by 21575 value, any cleanup for that parameter is performed in the caller, 21576 as specified by the ia64 C++ ABI, rather than the called function 21577 as before. As a result, classes with a non-trivial destructor but a 21578 trivial copy constructor will be passed and returned by invisible 21579 reference, rather than by bitwise copy as before. 21580 * G++ now supports the "named return value optimization": for code 21581 like 21582 A f () { 21583 A a; 21584 ... 21585 return a; 21586 } 21587 21588 G++ will allocate a in the return value slot, so that the return 21589 becomes a no-op. For this to work, all return statements in the 21590 function must return the same variable. 21591 * Improvements to the C++ library are listed in [5]the libstdc++-v3 21592 FAQ. 21593 21594 Objective-C 21595 21596 * Annoying linker warnings (due to incorrect code being generated) 21597 have been fixed. 21598 * If a class method cannot be found, the compiler no longer issues a 21599 warning if a corresponding instance method exists in the root 21600 class. 21601 * Forward @protocol declarations have been fixed. 21602 * Loading of categories has been fixed in certain situations (GNU run 21603 time only). 21604 * The class lookup in the run-time library has been rewritten so that 21605 class method dispatch is more than twice as fast as it used to be 21606 (GNU run time only). 21607 21608 Java 21609 21610 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 21611 javax.transaction. 21612 * Property files and other system resources can be compiled into 21613 executables which use libgcj using the new gcj --resource feature. 21614 * libgcj has been ported to more platforms. In particular there is 21615 now a mostly-functional mingw32 (Windows) target port. 21616 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 21617 Java code can now be called from a C/C++ application. 21618 * gcj can now use builtin functions for certain known methods, for 21619 instance Math.cos. 21620 * gcj can now automatically remove redundant array-store checks in 21621 some common cases. 21622 * The --no-store-checks optimization option was added. This can be 21623 used to omit runtime store checks for code which is known not to 21624 throw ArrayStoreException 21625 * The following third party interface standards were added to libgcj: 21626 org.w3c.dom and org.xml.sax. 21627 * java.security has been merged with GNU Classpath. The new package 21628 is now JDK 1.2 compliant, and much more complete. 21629 * A bytecode verifier was added to the libgcj interpreter. 21630 * java.lang.Character was rewritten to comply with the Unicode 3.0 21631 standard, and improve performance. 21632 * Partial support for many more locales was added to libgcj. 21633 * Socket timeouts have been implemented. 21634 * libgcj has been merged into a single shared library. There are no 21635 longer separate shared libraries for the garbage collector and 21636 zlib. 21637 * Several performance improvements were made to gcj and libgcj: 21638 + Hash synchronization (thin locks) 21639 + A special allocation path for finalizer-free objects 21640 + Thread-local allocation 21641 + Parallel GC, and other GC tweaks 21642 21643 Fortran 21644 21645 Fortran improvements are listed in [6]the Fortran documentation. 21646 21647 Ada 21648 21649 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 21650 tools. The GNAT compiler fully implements the Ada language as defined 21651 by the ISO/IEC 8652 standard. 21652 21653 Please note that the integration of the Ada front end is still work in 21654 progress. 21655 21656 New Targets and Target Specific Improvements 21657 21658 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 21659 architecture used in new editions of Donald E. Knuth's The Art of 21660 Computer Programming. 21661 * Axis Communications has contributed its port to the CRIS CPU 21662 architecture, used in the ETRAX system-on-a-chip series. 21663 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 21664 SH5 64-bit RISC microprocessor architecture, extending the existing 21665 SH port. 21666 * UltraSPARC is fully supported in 64-bit mode. The option -m64 21667 enables it. 21668 * For compatibility with the Sun compiler #pragma redefine_extname 21669 has been implemented on Solaris. 21670 * The x86 back end has had some noticeable work done to it. 21671 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 21672 Jaeger have contributed a port to the AMD x86-64 architecture. 21673 For more information on x86-64 see http://www.x86-64.org. 21674 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 21675 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 21676 enable the respective instruction sets. Intel C++ compatible 21677 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 21678 will be added in next major release. 21679 + Following those improvements, targets for Pentium MMX, K6-2, 21680 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 21681 added. Refer to the documentation on -march= and -mcpu= 21682 options for details. 21683 + For those targets that support it, -mfpmath=sse will cause the 21684 compiler to generate SSE/SSE2 instructions for floating point 21685 math instead of x87 instructions. Usually, this will lead to 21686 quicker code -- especially on the Pentium 4. Note that only 21687 scalar floating point instructions are used and GCC does not 21688 exploit SIMD features yet. 21689 + Prefetch support has been added to the Pentium III, Pentium 4, 21690 K6-2, K6-3, and Athlon series. 21691 + Code generated for floating point to integer conversions has 21692 been improved leading to better performance of many 3D 21693 applications. 21694 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 21695 * C++ support for AIX has been improved. 21696 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 21697 PowerPC port supporting the AltiVec programming model (SIMD). The 21698 support, though presently useful, is experimental and is expected 21699 to stabilize for 3.2. The support is written to conform to 21700 Motorola's AltiVec specs. See -maltivec. 21701 21702 [8]Obsolete Systems 21703 21704 Support for a number of older systems has been declared obsolete in GCC 21705 3.1. Unless there is activity to revive them, the next release of GCC 21706 will have their sources permanently removed. 21707 21708 All configurations of the following processor architectures have been 21709 declared obsolete: 21710 * MIL-STD-1750A, 1750a-*-* 21711 * AMD A29k, a29k-*-* 21712 * Convex, c*-convex-* 21713 * Clipper, clipper-*-* 21714 * Elxsi, elxsi-*-* 21715 * Intel i860, i860-*-* 21716 * Sun picoJava, pj-*-* and pjl-*-* 21717 * Western Electric 32000, we32k-*-* 21718 21719 Most configurations of the following processor architectures have been 21720 declared obsolete, but we are preserving a few systems which may have 21721 active developers. It is unlikely that the remaining systems will 21722 survive much longer unless we see definite signs of port activity. 21723 * Motorola 88000 except 21724 + Generic a.out, m88k-*-aout* 21725 + Generic SVR4, m88k-*-sysv4 21726 + OpenBSD, m88k-*-openbsd* 21727 * NS32k except 21728 + NetBSD, ns32k-*-netbsd* 21729 + OpenBSD, ns32k-*-openbsd*. 21730 * ROMP except 21731 + OpenBSD, romp-*-openbsd*. 21732 21733 Finally, only some configurations of these processor architectures are 21734 being obsoleted. 21735 * Alpha: 21736 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 21737 alpha*-*-osf[45], are still supported.) 21738 * ARM: 21739 + RISCiX, arm-*-riscix*. 21740 * i386: 21741 + 386BSD, i?86-*-bsd* 21742 + Chorus, i?86-*-chorusos* 21743 + DG/UX, i?86-*-dgux* 21744 + FreeBSD 1.x, i?86-*-freebsd1.* 21745 + IBM AIX, i?86-*-aix* 21746 + ISC UNIX, i?86-*-isc* 21747 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 21748 + NEXTstep, i?86-next-* 21749 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 21750 + RTEMS/coff, i?86-*-rtemscoff* 21751 + RTEMS/go32, i?86-go32-rtems* 21752 + Sequent/BSD, i?86-sequent-bsd* 21753 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 21754 i?86-sequent-sysv3* 21755 + SunOS, i?86-*-sunos* 21756 * Motorola 68000: 21757 + Altos, m68[k0]*-altos-* 21758 + Apollo, m68[k0]*-apollo-* 21759 + Apple A/UX, m68[k0]*-apple-* 21760 + Bull, m68[k0]*-bull-* 21761 + Convergent, m68[k0]*-convergent-* 21762 + Generic SVR3, m68[k0]*-*-sysv3* 21763 + ISI, m68[k0]*-isi-* 21764 + LynxOS, m68[k0]*-*-lynxos* 21765 + NEXT, m68[k0]*-next-* 21766 + RTEMS/coff, m68[k0]*-*-rtemscoff* 21767 + Sony, m68[k0]*-sony-* 21768 * MIPS: 21769 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 21770 + Generic BSD, mips-*-bsd* 21771 + Generic System V, mips-*-sysv* 21772 + IRIX before version 5, mips-sgi-irix[1234]* 21773 + RiscOS, mips-*-riscos* 21774 + Sony, mips-sony-* 21775 + Tandem, mips-tandem-* 21776 * SPARC: 21777 + RTEMS/a.out, sparc-*-rtemsaout*. 21778 21779 Documentation improvements 21780 21781 * The old manual ("Using and Porting the GNU Compiler Collection") 21782 has been replaced by a users manual ("Using the GNU Compiler 21783 Collection") and a separate internals reference manual ("GNU 21784 Compiler Collection Internals"). 21785 * More complete and much improved documentation about GCC's internal 21786 representation used by the C and C++ front ends. 21787 * Many cleanups and improvements in general. 21788 21789 21790 For questions related to the use of GCC, please consult these web 21791 pages and the [9]GCC manuals. If that fails, the 21792 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21793 web pages and the development of GCC are welcome on our developer 21794 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 21795 archives. 21796 21797 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 21798 distribution of this entire article is permitted in any medium, 21799 provided this notice is preserved. 21800 21801 These pages are [14]maintained by the GCC team. Last modified 21802 2025-06-05. 21803 21804 References 21805 21806 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 21807 2. https://gcc.gnu.org/news/profiledriven.html 21808 3. https://gcc.gnu.org/benchmarks/ 21809 4. https://gcc.gnu.org/projects/c-status.html 21810 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 21811 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 21812 7. https://www.adacore.com/ 21813 8. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 21814 9. https://gcc.gnu.org/onlinedocs/ 21815 10. mailto:gcc-help (a] gcc.gnu.org 21816 11. mailto:gcc (a] gcc.gnu.org 21817 12. https://gcc.gnu.org/lists.html 21818 13. https://www.fsf.org/ 21819 14. https://gcc.gnu.org/about.html 21820 ====================================================================== 21821 http://gcc.gnu.org/gcc-3.0/index.html 21822 21823 GCC 3.0.4 21824 21825 (This release series is no longer supported.) 21826 21827 February 20, 2002 21828 21829 The [1]GNU project and the GCC developers are pleased to announce the 21830 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 21831 series. 21832 21833 GCC used to stand for the GNU C Compiler, but since the compiler 21834 supports several other languages aside from C, it now stands for the 21835 GNU Compiler Collection. 21836 21837 GCC 3.0.x has several new optimizations, new targets, new languages and 21838 many other new features, relative to GCC 2.95.x. See the [2]new 21839 features page for a more complete list. 21840 21841 A list of [3]successful builds is updated as new information becomes 21842 available. 21843 21844 The GCC developers would like to thank the numerous people that have 21845 contributed new features, test results, bug fixes, etc to GCC. This 21846 [4]amazing group of volunteers is what makes GCC successful. 21847 21848 And finally, we can't in good conscience fail to mention some 21849 [5]caveats to using GCC 3.0.x. 21850 21851 For additional information about GCC please refer to the [6]GCC project 21852 web site or contact the [7]GCC development mailing list. 21853 21854 To obtain GCC please use [8]our mirror sites, or our CVS server. 21855 __________________________________________________________________ 21856 21857 Previous 3.0.x Releases 21858 21859 December 20, 2001: GCC 3.0.3 has been released. 21860 October 25, 2001: GCC 3.0.2 has been released. 21861 August 20, 2001: GCC 3.0.1 has been released. 21862 June 18, 2001: GCC 3.0 has been released. 21863 21864 21865 For questions related to the use of GCC, please consult these web 21866 pages and the [9]GCC manuals. If that fails, the 21867 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21868 web pages and the development of GCC are welcome on our developer 21869 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 21870 archives. 21871 21872 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 21873 distribution of this entire article is permitted in any medium, 21874 provided this notice is preserved. 21875 21876 These pages are [14]maintained by the GCC team. Last modified 21877 2024-07-04. 21878 21879 References 21880 21881 1. https://www.gnu.org/ 21882 2. https://gcc.gnu.org/gcc-3.0/features.html 21883 3. https://gcc.gnu.org/gcc-3.0/buildstat.html 21884 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 21885 5. https://gcc.gnu.org/gcc-3.0/caveats.html 21886 6. https://gcc.gnu.org/index.html 21887 7. mailto:gcc (a] gcc.gnu.org 21888 8. https://gcc.gnu.org/mirrors.html 21889 9. https://gcc.gnu.org/onlinedocs/ 21890 10. mailto:gcc-help (a] gcc.gnu.org 21891 11. mailto:gcc (a] gcc.gnu.org 21892 12. https://gcc.gnu.org/lists.html 21893 13. https://www.fsf.org/ 21894 14. https://gcc.gnu.org/about.html 21895 ====================================================================== 21896 http://gcc.gnu.org/gcc-3.0/features.html 21897 21898 GCC 3.0 New Features 21899 21900 Additional changes in GCC 3.0.4 21901 21902 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 21903 system, which use the ELF object file format, on x86 processors. 21904 * Correct debugging information is generated from functions that have 21905 lines from multiple files (e.g. yacc output). 21906 * A fix for whitespace handling in the -traditional preprocessor, 21907 which can affect Fortran. 21908 * Fixes to the exception handling runtime. 21909 * More fixes for bad code generation in C++. 21910 * A fix for shared library generation under AIX 4.3. 21911 * Documentation updates. 21912 * Port of GCC to Tensilica's Xtensa processor contributed. 21913 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 21914 21915 Additional changes in GCC 3.0.3 21916 21917 * A fix to correct an accidental change to the PowerPC ABI. 21918 * Fixes for bad code generation on a variety of architectures. 21919 * Improvements to the debugging information generated for C++ 21920 classes. 21921 * Fixes for bad code generation in C++. 21922 * A fix to avoid crashes in the C++ demangler. 21923 * A fix to the C++ standard library to avoid buffer overflows. 21924 * Miscellaneous improvements for a variety of architectures. 21925 21926 Additional changes in GCC 3.0.2 21927 21928 * Fixes for bad code generation during loop unrolling. 21929 * Fixes for bad code generation by the sibling call optimization. 21930 * Minor improvements to x86 code generation. 21931 * Implementation of function descriptors in C++ vtables for IA64. 21932 * Numerous minor bug-fixes. 21933 21934 Additional changes in GCC 3.0.1 21935 21936 * C++ fixes for incorrect code-generation. 21937 * Improved cross-compiling support for the C++ standard library. 21938 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 21939 in GCC 3.0. 21940 * Fixes for various exception-handling bugs. 21941 * A port to the S/390 architecture. 21942 21943 General Optimizer Improvements 21944 21945 * [2]Basic block reordering pass. 21946 * New if-conversion pass with support for conditional (predicated) 21947 execution. 21948 * New tail call and sibling call elimination optimizations. 21949 * New register renaming pass. 21950 * New (experimental) [3]static single assignment (SSA) representation 21951 support. 21952 * New dead-code elimination pass implemented using the SSA 21953 representation. 21954 * [4]Global null pointer test elimination. 21955 * [5]Global code hoisting/unification. 21956 * More builtins and optimizations for stdio.h, string.h and old BSD 21957 functions, as well as for ISO C99 functions. 21958 * New builtin __builtin_expect for giving hints to the branch 21959 predictor. 21960 21961 New Languages and Language specific improvements 21962 21963 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 21964 and supported, including the run-time library containing most 21965 common non-GUI Java classes, a bytecode interpreter, and the Boehm 21966 conservative garbage collector. Many bugs have been fixed. GCJ can 21967 compile Java source or Java bytecodes to either native code or Java 21968 class files, and supports native methods written in either the 21969 standard JNI or the more efficient and convenient CNI. 21970 * Here is a [6]partial list of C++ improvements, both new features 21971 and those no longer supported. 21972 * New C++ ABI. On the IA-64 platform GCC is capable of 21973 inter-operating with other IA-64 compilers. 21974 * The new ABI also significantly reduces the size of symbol and debug 21975 information. 21976 * New C++ support library and many C++ bug fixes, vastly improving 21977 our conformance to the ISO C++ standard. 21978 * New [7]inliner for C++. 21979 * Rewritten C preprocessor, integrated into the C, C++ and Objective 21980 C compilers, with very many improvements including ISO C99 support 21981 and [8]improvements to dependency generation. 21982 * Support for more [9]ISO C99 features. 21983 * Many improvements to support for checking calls to format functions 21984 such as printf and scanf, including support for ISO C99 format 21985 features, extensions from the Single Unix Specification and GNU 21986 libc 2.2, checking of strfmon formats and features to assist in 21987 auditing for format string security bugs. 21988 * New warnings for C code that may have undefined semantics because 21989 of violations of sequence point rules in the C standard (such as a 21990 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 21991 * Additional warning option -Wfloat-equal. 21992 * Improvements to -Wtraditional. 21993 * Fortran improvements are listed in [10]the Fortran documentation. 21994 21995 New Targets and Target Specific Improvements 21996 21997 * New x86 back end, generating much improved code. 21998 * Support for a generic i386-elf target contributed. 21999 * New option to emit x86 assembly code using Intel style syntax 22000 (-mintel-syntax). 22001 * HPUX 11 support contributed. 22002 * Improved PowerPC code generation, including scheduled prologue and 22003 epilogue. 22004 * Port of GCC to Intel's IA-64 processor contributed. 22005 * Port of GCC to Motorola's MCore 210 and 340 contributed. 22006 * New unified back-end for Arm, Thumb and StrongArm contributed. 22007 * Port of GCC to Intel's XScale processor contributed. 22008 * Port of GCC to Atmel's AVR microcontrollers contributed. 22009 * Port of GCC to Mitsubishi's D30V processor contributed. 22010 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 22011 processor family) contributed. 22012 * Port of GCC to Fujitsu's FR30 processor contributed. 22013 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 22014 * Port of GCC to Sun's picoJava processor core contributed. 22015 22016 Documentation improvements 22017 22018 * Substantially rewritten and improved C preprocessor manual. 22019 * Many improvements to other documentation. 22020 * Manpages for gcc, cpp and gcov are now generated automatically from 22021 the master Texinfo manual, eliminating the problem of manpages 22022 being out of date. (The generated manpages are only extracts from 22023 the full manual, which is provided in Texinfo form, from which 22024 info, HTML, other formats and a printed manual can be generated.) 22025 * Generated info files are included in the release tarballs alongside 22026 their Texinfo sources, avoiding problems on some platforms with 22027 building makeinfo as part of the GCC distribution. 22028 22029 Other significant improvements 22030 22031 * Garbage collection used internally by the compiler for most memory 22032 allocation instead of obstacks. 22033 * Lengauer and Tarjan algorithm used for computing dominators in the 22034 CFG. This algorithm can be significantly faster and more space 22035 efficient than our older algorithm. 22036 * gccbug script provided to assist in submitting bug reports to our 22037 bug tracking system. (Bug reports previously submitted directly to 22038 our mailing lists, for which you received no bug tracking number, 22039 should be submitted again using gccbug if you can reproduce the 22040 problem with GCC 3.0.) 22041 * The internal libgcc library is [11]built as a shared library on 22042 systems that support it. 22043 * Extensive testsuite included with GCC, with many new tests. In 22044 addition to tests for GCC bugs that have been fixed, many tests 22045 have been added for language features, compiler warnings and 22046 builtin functions. 22047 * Additional language-independent warning options -Wpacked, -Wpadded, 22048 -Wunreachable-code and -Wdisabled-optimization. 22049 * Target-independent options -falign-functions, -falign-loops and 22050 -falign-jumps. 22051 22052 Plus a great many bug fixes and almost all the [12]features found in 22053 GCC 2.95. 22054 22055 22056 For questions related to the use of GCC, please consult these web 22057 pages and the [13]GCC manuals. If that fails, the 22058 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22059 web pages and the development of GCC are welcome on our developer 22060 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 22061 archives. 22062 22063 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 22064 distribution of this entire article is permitted in any medium, 22065 provided this notice is preserved. 22066 22067 These pages are [18]maintained by the GCC team. Last modified 22068 2025-06-05. 22069 22070 References 22071 22072 1. http://www.netbsd.org/ 22073 2. https://gcc.gnu.org/news/reorder.html 22074 3. https://gcc.gnu.org/news/ssa.html 22075 4. https://gcc.gnu.org/news/null.html 22076 5. https://gcc.gnu.org/news/unify.html 22077 6. https://gcc.gnu.org/gcc-3.0/c++features.html 22078 7. https://gcc.gnu.org/news/inlining.html 22079 8. https://gcc.gnu.org/news/dependencies.html 22080 9. https://gcc.gnu.org/projects/c-status.html 22081 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 22082 11. https://gcc.gnu.org/gcc-3.0/libgcc.html 22083 12. https://gcc.gnu.org/gcc-2.95/features.html 22084 13. https://gcc.gnu.org/onlinedocs/ 22085 14. mailto:gcc-help (a] gcc.gnu.org 22086 15. mailto:gcc (a] gcc.gnu.org 22087 16. https://gcc.gnu.org/lists.html 22088 17. https://www.fsf.org/ 22089 18. https://gcc.gnu.org/about.html 22090 ====================================================================== 22091 http://gcc.gnu.org/gcc-3.0/caveats.html 22092 22093 GCC 3.0 Caveats 22094 22095 * -fstrict-aliasing is now part of -O2 and higher optimization 22096 levels. This allows the compiler to assume the strictest aliasing 22097 rules applicable to the language being compiled. For C and C++, 22098 this activates optimizations based on the type of expressions. This 22099 optimization may thus break old, non-compliant code. 22100 * Enumerations are now properly promoted to int in function 22101 parameters and function returns. Normally this change is not 22102 visible, but when using -fshort-enums this is an ABI change. 22103 * The undocumented extension that allowed C programs to have a label 22104 at the end of a compound statement has been deprecated and may be 22105 removed in a future version. Programs that now generate a warning 22106 about this may be fixed by adding a null statement (a single 22107 semicolon) after the label. 22108 * The poorly documented extension that allowed string constants in C, 22109 C++ and Objective C to contain unescaped newlines has been 22110 deprecated and may be removed in a future version. Programs using 22111 this extension may be fixed in several ways: the bare newline may 22112 be replaced by \n, or preceded by \n\, or string concatenation may 22113 be used with the bare newline preceded by \n" and " placed at the 22114 start of the next line. 22115 * The Chill compiler is not included in GCC 3.0, because of the lack 22116 of a volunteer to convert it to use garbage collection. 22117 * Certain non-standard iostream methods from earlier versions of 22118 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 22119 ostream::form, and istream::gets. 22120 * The new C++ ABI is not yet fully supported by current (as of 22121 2001-07-01) releases and development versions of GDB, or any 22122 earlier versions. There is a problem setting breakpoints by line 22123 number, and other related issues that have been fixed in GCC 3.0 22124 but not yet handled in GDB: 22125 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 22126 22127 22128 For questions related to the use of GCC, please consult these web 22129 pages and the [2]GCC manuals. If that fails, the 22130 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22131 web pages and the development of GCC are welcome on our developer 22132 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 22133 22134 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 22135 distribution of this entire article is permitted in any medium, 22136 provided this notice is preserved. 22137 22138 These pages are [7]maintained by the GCC team. Last modified 22139 2022-10-26. 22140 22141 References 22142 22143 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 22144 2. https://gcc.gnu.org/onlinedocs/ 22145 3. mailto:gcc-help (a] gcc.gnu.org 22146 4. mailto:gcc (a] gcc.gnu.org 22147 5. https://gcc.gnu.org/lists.html 22148 6. https://www.fsf.org/ 22149 7. https://gcc.gnu.org/about.html 22150 ====================================================================== 22151 http://gcc.gnu.org/gcc-2.95/index.html 22152 22153 GCC 2.95 22154 22155 (This release series is no longer supported.) 22156 22157 March 16, 2001: The GNU project and the GCC developers are pleased to 22158 announce the release of GCC version 2.95.3. 22159 22160 Release History 22161 22162 GCC 2.95.3 22163 March 16, 2001 22164 22165 GCC 2.95.2 22166 October 27, 1999 22167 22168 GCC 2.95.1 22169 August 19, 1999 22170 22171 GCC 2.95 22172 July 31, 1999. This is the first release of GCC since the April 22173 1999 GCC/EGCS reunification and includes nearly a year's worth 22174 of new development and bugfixes. 22175 22176 References and Acknowledgements 22177 22178 GCC used to stand for the GNU C Compiler, but since the compiler 22179 supports several other languages aside from C, it now stands for the 22180 GNU Compiler Collection. 22181 22182 The whole suite has been extensively [1]regression tested and 22183 [2]package tested. It should be reliable and suitable for widespread 22184 use. 22185 22186 The compiler has several new optimizations, new targets, new languages 22187 and other new features. See the [3]new features page for a more 22188 complete list of new features found in the GCC 2.95 releases. 22189 22190 The sources include installation instructions in both HTML and 22191 plaintext forms in the install directory in the distribution. However, 22192 the most up to date installation instructions and [4]build/test status 22193 are on the web pages. We will update those pages as new information 22194 becomes available. 22195 22196 The GCC developers would like to thank the numerous people that have 22197 contributed new features, test results, bugfixes, etc to GCC. This 22198 [5]amazing group of volunteers is what makes GCC successful. 22199 22200 And finally, we can't in good conscience fail to mention some 22201 [6]caveats to using GCC 2.95. 22202 22203 Download GCC 2.95 from one of our many [7]mirror sites. 22204 22205 For additional information about GCC please see the [8]GCC project web 22206 server or contact the [9]GCC development mailing list. 22207 22208 22209 For questions related to the use of GCC, please consult these web 22210 pages and the [10]GCC manuals. If that fails, the 22211 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22212 web pages and the development of GCC are welcome on our developer 22213 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 22214 archives. 22215 22216 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 22217 distribution of this entire article is permitted in any medium, 22218 provided this notice is preserved. 22219 22220 These pages are [15]maintained by the GCC team. Last modified 22221 2022-10-26. 22222 22223 References 22224 22225 1. https://gcc.gnu.org/gcc-2.95/regress.html 22226 2. https://gcc.gnu.org/gcc-2.95/othertest.html 22227 3. https://gcc.gnu.org/gcc-2.95/features.html 22228 4. https://gcc.gnu.org/gcc-2.95/buildstat.html 22229 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 22230 6. https://gcc.gnu.org/gcc-2.95/caveats.html 22231 7. https://gcc.gnu.org/mirrors.html 22232 8. https://gcc.gnu.org/index.html 22233 9. mailto:gcc (a] gcc.gnu.org 22234 10. https://gcc.gnu.org/onlinedocs/ 22235 11. mailto:gcc-help (a] gcc.gnu.org 22236 12. mailto:gcc (a] gcc.gnu.org 22237 13. https://gcc.gnu.org/lists.html 22238 14. https://www.fsf.org/ 22239 15. https://gcc.gnu.org/about.html 22240 ====================================================================== 22241 http://gcc.gnu.org/gcc-2.95/features.html 22242 22243 GCC 2.95 New Features 22244 22245 * General Optimizer Improvements: 22246 + [1]Localized register spilling to improve speed and code 22247 density especially on small register class machines. 22248 + [2]Global CSE using lazy code motion algorithms. 22249 + [3]Improved global constant/copy propagation. 22250 + [4]Improved control flow graph analysis and manipulation. 22251 + [5]Local dead store elimination. 22252 + [6]Memory Load hoisting/store sinking in loops. 22253 + [7]Type based alias analysis is enabled by default. Note this 22254 feature will expose bugs in the Linux kernel. Please refer to 22255 the FAQ (as shipped with GCC 2.95) for additional information 22256 on this issue. 22257 + Major revamp of GIV detection, combination and simplification 22258 to improve loop performance. 22259 + Major improvements to register allocation and reloading. 22260 * New Languages and Language specific improvements 22261 + [8]Many C++ improvements. 22262 + [9]Many Fortran improvements. 22263 + [10]Java front-end has been integrated. A [11]runtime library 22264 is available separately. 22265 + [12]ISO C99 support 22266 + [13]Chill front-end and runtime has been integrated. 22267 + Boehm garbage collector support in libobjc. 22268 + More support for various pragmas which appear in vendor 22269 include files 22270 * New Targets and Target Specific Improvements 22271 + [14]SPARC back end rewrite. 22272 + -mschedule=8000 will optimize code for PA8000 class 22273 processors; -mpa-risc-2-0 will generate code for PA2.0 22274 processors 22275 + Various micro-optimizations for the ia32 port. K6 22276 optimizations 22277 + Compiler will attempt to align doubles in the stack on the 22278 ia32 port 22279 + Alpha EV6 support 22280 + PowerPC 750 22281 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 22282 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 22283 + c3x, c4x 22284 + HyperSPARC 22285 + SparcLite86x 22286 + sh4 22287 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 22288 arm-linux) 22289 + vxWorks targets include support for vxWorks threads 22290 + StrongARM 110 and ARM9 support added. ARM Scheduling 22291 parameters rewritten. 22292 + Various changes to the MIPS port to avoid assembler macros, 22293 which in turn improves performance 22294 + Various performance improvements to the i960 port. 22295 + Major rewrite of ns32k port 22296 * Other significant improvements 22297 + [15]Ability to dump cfg information and display it using vcg. 22298 + The new faster scheme for fixing vendor header files is 22299 enabled by default. 22300 + Experimental internationalization support. 22301 + multibyte character support 22302 + Some compile-time speedups for pathological problems 22303 + Better support for complex types 22304 * Plus the usual mountain of bugfixes 22305 * Core compiler is based on the gcc2 development tree from Sept 30, 22306 1998, so we have all of the [16]features found in GCC 2.8. 22307 22308 Additional Changes in GCC 2.95.1 22309 22310 * Generic bugfixes and improvements 22311 + Various documentation fixes related to the GCC/EGCS merger. 22312 + Fix memory management bug which could lead to spurious aborts, 22313 core dumps or random parsing errors in the compiler. 22314 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 22315 support. 22316 + Fix infinite loop in the CSE optimizer. 22317 + Avoid undefined behavior in compiler FP emulation code 22318 + Fix install problem when prefix is overridden on the make 22319 install command. 22320 + Fix problem with unwanted installation of assert.h on some 22321 systems. 22322 + Fix problem with finding the wrong assembler in a single tree 22323 build. 22324 + Avoid increasing the known alignment of a register that is 22325 already known to be a pointer. 22326 * Platform specific bugfixes and improvements 22327 + Codegen bugfix for prologue/epilogue for cpu32 target. 22328 + Fix long long code generation bug for the Coldfire target. 22329 + Fix various aborts in the SH compiler. 22330 + Fix bugs in libgcc support library for the SH. 22331 + Fix alpha ev6 code generation bug. 22332 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 22333 AIX platforms. 22334 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 22335 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 22336 targets. 22337 + Fix weak symbol handling for rs6000/ppc svr4 targets. 22338 + Fix various problems with 64bit code generation for the 22339 rs6000/ppc port. 22340 + Fix codegen bug which caused tetex to be mis-compiled on the 22341 x86. 22342 + Fix compiler abort in new cfg code exposed by x86 port. 22343 + Fix out of range array reference in code convert flat 22344 registers to the x87 stacked FP register file. 22345 + Fix minor vxworks configuration bug. 22346 + Fix return type of bsearch for SunOS 4.x. 22347 * Language & Runtime specific fixes. 22348 + The G++ signature extension has been deprecated. It will be 22349 removed in the next major release of G++. Use of signatures 22350 will result in a warning from the compiler. 22351 + Several bugs relating to templates and namespaces were fixed. 22352 + A bug that caused crashes when combining templates with -g on 22353 DWARF1 platforms was fixed. 22354 + Pointers-to-members, virtual functions, and multiple 22355 inheritance should now work together correctly. 22356 + Some code-generation bugs relating to function try blocks were 22357 fixed. 22358 + G++ is a little bit more lenient with certain archaic 22359 constructs than in GCC 2.95. 22360 + Fix to prevent shared library version #s from bring truncated 22361 to 1 digit 22362 + Fix missing std:: in the libstdc++ library. 22363 + Fix stream locking problems in libio. 22364 + Fix problem in java compiler driver. 22365 22366 Additional Changes in GCC 2.95.2 22367 22368 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 22369 the optimizations performed by -fstrict-aliasing are valid according to 22370 the C and C++ standards, the optimization have caused some problems, 22371 particularly with old non-conforming code. 22372 22373 The GCC developers are experimenting with ways to warn users about code 22374 which violates the C/C++ standards, but those warnings are not ready 22375 for widespread use at this time. Rather than wait for those warnings 22376 the GCC developers have chosen to disable -fstrict-aliasing by default 22377 for the GCC 2.95.2 release. 22378 22379 We strongly encourage developers to find and fix code which violates 22380 the C/C++ standards as -fstrict-aliasing may be enabled by default in 22381 future releases. Use the option -fstrict-aliasing to re-enable these 22382 optimizations. 22383 * Generic bugfixes and improvements 22384 + Fix incorrectly optimized memory reference in global common 22385 subexpression elimination (GCSE) optimization pass. 22386 + Fix code generation bug in regmove.c in which it could 22387 incorrectly change a "const" value. 22388 + Fix bug in optimization of conditionals involving volatile 22389 memory references. 22390 + Avoid over-allocation of stack space for some procedures. 22391 + Fixed bug in the compiler which caused incorrect optimization 22392 of an obscure series of bit manipulations, shifts and 22393 arithmetic. 22394 + Fixed register allocator bug which caused teTeX to be 22395 mis-compiled on SPARC targets. 22396 + Avoid incorrect optimization of degenerate case statements for 22397 certain targets such as the ARM. 22398 + Fix out of range memory reference in the jump optimizer. 22399 + Avoid dereferencing null pointer in fix-header. 22400 + Fix test for GCC specific features so that it is possible to 22401 bootstrap with gcc-2.6.2 and older versions of GCC. 22402 + Fix typo in scheduler which could potentially cause out of 22403 range memory accesses. 22404 + Avoid incorrect loop reversal which caused incorrect code for 22405 certain loops on PowerPC targets. 22406 + Avoid incorrect optimization of switch statements on certain 22407 targets (for example the ARM). 22408 * Platform specific bugfixes and improvements 22409 + Work around bug in Sun V5.0 compilers which caused bootstrap 22410 comparison failures on SPARC targets. 22411 + Fix SPARC back end bug which caused aborts in final.c. 22412 + Fix sparc-hal-solaris2* configuration fragments. 22413 + Fix bug in sparc block profiling. 22414 + Fix obscure code generation bug for the PARISC targets. 22415 + Define __STDC_EXT__ for HPUX configurations. 22416 + Various POWERPC64 code generation bugfixes. 22417 + Fix abort for PPC targets using ELF (ex GNU/Linux). 22418 + Fix collect2 problems for AIX targets. 22419 + Correct handling of .file directive for PPC targets. 22420 + Fix bug in fix_trunc x86 patterns. 22421 + Fix x86 port to correctly pop the FP stack for functions that 22422 return structures in memory. 22423 + Fix minor bug in strlen x86 pattern. 22424 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 22425 + Fix template repository code to handle leading underscore in 22426 mangled names. 22427 + Fix weak/weak alias support for OpenBSD. 22428 + GNU/Linux for the ARM has C++ compatible include files. 22429 * Language & Runtime specific fixes. 22430 + Fix handling of constructor attribute in the C front-end which 22431 caused problems building the Chill runtime library on some 22432 targets. 22433 + Fix minor problem merging type qualifiers in the C front-end. 22434 + Fix aliasing bug for pointers and references (C/C++). 22435 + Fix incorrect "non-constant initializer bug" when -traditional 22436 or -fwritable-strings is enabled. 22437 + Fix build error for Chill front-end on SunOS. 22438 + Do not complain about duplicate instantiations when using 22439 -frepo (C++). 22440 + Fix array bounds handling in C++ front-end which caused 22441 problems with dwarf debugging information in some 22442 circumstances. 22443 + Fix minor namespace problem. 22444 + Fix problem linking java programs. 22445 22446 Additional Changes in GCC 2.95.3 22447 22448 * Generic bugfixes and improvements 22449 + Fix numerous problems that caused incorrect optimization in 22450 the register reloading code. 22451 + Fix numerous problems that caused incorrect optimization in 22452 the loop optimizer. 22453 + Fix aborts in the functions build_insn_chain and scan_loops 22454 under some circumstances. 22455 + Fix an alias analysis bug. 22456 + Fix an infinite compilation bug in the combiner. 22457 + A few problems with complex number support have been fixed. 22458 + It is no longer possible for gcc to act as a fork bomb when 22459 installed incorrectly. 22460 + The -fpack-struct option should be recognized now. 22461 + Fixed a bug that caused incorrect code to be generated due to 22462 a lost stack adjustment. 22463 * Platform specific bugfixes and improvements 22464 + Support building ARM toolchains hosted on Windows. 22465 + Fix attribute calculations in ARM toolchains. 22466 + arm-linux support has been improved. 22467 + Fix a PIC failure on sparc targets. 22468 + On ix86 targets, the regparm attribute should now work 22469 reliably. 22470 + Several updates for the h8300 port. 22471 + Fix problem building libio with glibc 2.2. 22472 22473 22474 For questions related to the use of GCC, please consult these web 22475 pages and the [17]GCC manuals. If that fails, the 22476 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22477 web pages and the development of GCC are welcome on our developer 22478 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 22479 archives. 22480 22481 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 22482 distribution of this entire article is permitted in any medium, 22483 provided this notice is preserved. 22484 22485 These pages are [22]maintained by the GCC team. Last modified 22486 2025-06-05. 22487 22488 References 22489 22490 1. https://gcc.gnu.org/news/spill.html 22491 2. https://gcc.gnu.org/news/lcm.html 22492 3. https://gcc.gnu.org/news/cprop.html 22493 4. https://gcc.gnu.org/news/cfg.html 22494 5. https://gcc.gnu.org/news/dse.html 22495 6. https://gcc.gnu.org/news/hoist.html 22496 7. https://gcc.gnu.org/news/alias.html 22497 8. https://gcc.gnu.org/gcc-2.95/c++features.html 22498 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 22499 10. https://gcc.gnu.org/news/gcj-announce.txt 22500 11. https://gcc.gnu.org/news/javaannounce.html 22501 12. https://gcc.gnu.org/projects/c-status.html 22502 13. https://gcc.gnu.org/news/chill.html 22503 14. https://gcc.gnu.org/news/sparc.html 22504 15. https://gcc.gnu.org/news/egcs-vcg.html 22505 16. https://gcc.gnu.org/egcs-1.0/features-2.8.html 22506 17. https://gcc.gnu.org/onlinedocs/ 22507 18. mailto:gcc-help (a] gcc.gnu.org 22508 19. mailto:gcc (a] gcc.gnu.org 22509 20. https://gcc.gnu.org/lists.html 22510 21. https://www.fsf.org/ 22511 22. https://gcc.gnu.org/about.html 22512 ====================================================================== 22513 http://gcc.gnu.org/gcc-2.95/caveats.html 22514 22515 GCC 2.95 Caveats 22516 22517 * GCC 2.95 will issue an error for invalid asm statements that had 22518 been silently accepted by earlier versions of the compiler. This is 22519 particularly noticeable when compiling older versions of the Linux 22520 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 22521 for more information on this issue. 22522 * GCC 2.95 implements type based alias analysis to disambiguate 22523 memory references. Some programs, particularly the Linux kernel 22524 violate ANSI/ISO aliasing rules and therefore may not operate 22525 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 22526 shipped with GCC 2.95) for more information on this issue. 22527 * GCC 2.95 has a known bug in its handling of complex variables for 22528 64bit targets. Instead of silently generating incorrect code, GCC 22529 2.95 will issue a fatal error for situations it can not handle. 22530 This primarily affects the Fortran community as Fortran makes more 22531 use of complex variables than C or C++. 22532 * GCC 2.95 has an integrated libstdc++, but does not have an 22533 integrated libg++. Furthermore old libg++ releases will not work 22534 with GCC 2.95. You can retrieve a recent copy of libg++ from the 22535 [1]GCC ftp server. 22536 Note most C++ programs only need libstdc++. 22537 * Exception handling may not work with shared libraries, particularly 22538 on alphas, hppas, rs6000/powerpc and mips based platforms. 22539 Exception handling is known to work on x86 GNU/Linux platforms with 22540 shared libraries. 22541 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 22542 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 22543 or EGCS 1.1. As a result it may be necessary to fix C++ code before 22544 it will compile with GCC 2.95. 22545 * G++ is also converting toward the ISO C++ standard; as a result 22546 code which was previously valid (and thus accepted by other 22547 compilers and older versions of g++) may no longer be accepted. The 22548 flag -fpermissive may allow some non-conforming code to compile 22549 with GCC 2.95. 22550 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 22551 1.1.x, EGCS 1.0.x or GCC 2.8.x. 22552 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 22553 between Sept 30, 1998 and April 30, 1999 (the official end of the 22554 GCC 2.8 project). Future GCC releases will include all the changes 22555 from the defunct GCC 2.8 sources. 22556 22557 22558 For questions related to the use of GCC, please consult these web 22559 pages and the [2]GCC manuals. If that fails, the 22560 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22561 web pages and the development of GCC are welcome on our developer 22562 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 22563 22564 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 22565 distribution of this entire article is permitted in any medium, 22566 provided this notice is preserved. 22567 22568 These pages are [7]maintained by the GCC team. Last modified 22569 2022-10-26. 22570 22571 References 22572 22573 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 22574 2. https://gcc.gnu.org/onlinedocs/ 22575 3. mailto:gcc-help (a] gcc.gnu.org 22576 4. mailto:gcc (a] gcc.gnu.org 22577 5. https://gcc.gnu.org/lists.html 22578 6. https://www.fsf.org/ 22579 7. https://gcc.gnu.org/about.html 22580 ====================================================================== 22581 http://gcc.gnu.org/egcs-1.1/index.html 22582 22583 EGCS 1.1 22584 22585 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 22586 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 22587 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 22588 22589 EGCS is a free software project to further the development of the GNU 22590 compilers using an open development environment. 22591 22592 EGCS 1.1 is a major new release of the EGCS compiler system. It has 22593 been [1]extensively tested and is believed to be stable and suitable 22594 for widespread use. 22595 22596 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 22597 development sources; it contains all of the new features found in GCC 22598 2.8.1 as well as all new development from GCC up to June 6, 1998. 22599 22600 EGCS 1.1 also contains many improvements and features not found in GCC 22601 or in older versions of EGCS: 22602 * Global common subexpression elimination and global constant/copy 22603 propagation (aka [2]gcse) 22604 * Ongoing improvements to the [3]alias analysis support to allow for 22605 better optimizations throughout the compiler. 22606 * Vastly improved [4]C++ compiler and integrated C++ runtime 22607 libraries. 22608 * Fixes for the /tmp symlink race security problems. 22609 * New targets including mips16, arm-thumb and 64-bit PowerPC. 22610 * Improvements to GNU Fortran (g77) compiler and runtime library made 22611 since g77 version 0.5.23. 22612 22613 See the [5]new features page for a more complete list of new features 22614 found in EGCS 1.1 releases. 22615 22616 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 22617 1.1: 22618 * General improvements and fixes 22619 + Avoid some stack overflows when compiling large functions. 22620 + Avoid incorrect loop invariant code motions. 22621 + Fix some core dumps on Linux kernel code. 22622 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 22623 + Fix code generation problem in gcse. 22624 + Various documentation related fixes. 22625 * g++/libstdc++ improvements and fixes 22626 + MT safe EH fix for setjmp/longjmp based exception handling. 22627 + Fix a few bad interactions between optimization and exception 22628 handling. 22629 + Fixes for demangling of template names starting with "__". 22630 + Fix a bug that would fail to run destructors in some cases 22631 with -O2. 22632 + Fix 'new' of classes with virtual bases. 22633 + Fix crash building Qt on the Alpha. 22634 + Fix failure compiling WIFEXITED macro on GNU/Linux. 22635 + Fix some -frepo failures. 22636 * g77 and libf2c improvements and fixes 22637 + Various documentation fixes. 22638 + Avoid compiler crash on RAND intrinsic. 22639 + Fix minor bugs in makefiles exposed by BSD make programs. 22640 + Define _XOPEN_SOURCE for libI77 build to avoid potential 22641 problems on some 64-bit systems. 22642 + Fix problem with implicit endfile on rewind. 22643 + Fix spurious recursive I/O errors. 22644 * platform specific improvements and fixes 22645 + Match all versions of UnixWare7. 22646 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 22647 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 22648 from unsigned ints to double precision floats. 22649 + Fix ARM ABI issue with NetBSD. 22650 + Fix a few arm code generation bugs. 22651 + Fixincludes will fix additional broken SCO OpenServer header 22652 files. 22653 + Fix a m68k back end bug which caused invalid offsets in reg+d 22654 addresses. 22655 + Fix problems with 64bit AIX 4.3 support. 22656 + Fix handling of long longs for varargs/stdarg functions on the 22657 ppc. 22658 + Minor fixes to CPP predefines for Windows. 22659 + Fix code generation problems with gpr<->fpr copies for 64bit 22660 ppc. 22661 + Fix a few coldfire code generation bugs. 22662 + Fix some more header file problems on SunOS 4.x. 22663 + Fix assert.h handling for RTEMS. 22664 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 22665 + Fix x86 compiler abort in reg-stack pass. 22666 + Fix cygwin/windows problem with section attributes. 22667 + Fix Alpha code generation problem exposed by SMP Linux 22668 kernels. 22669 + Fix typo in m68k 32->64bit integer conversion. 22670 + Make sure target libraries build with -fPIC for PPC & Alpha 22671 targets. 22672 22673 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 22674 1.1.1: 22675 * General improvements and fixes 22676 + Fix bug in loop optimizer which caused the SPARC (and 22677 potentially other) ports to segfault. 22678 + Fix infinite recursion in alias analysis and combiner code. 22679 + Fix bug in regclass preferencing. 22680 + Fix incorrect loop reversal which caused incorrect code to be 22681 generated for several targets. 22682 + Fix return value for builtin memcpy. 22683 + Reduce compile time for certain loops which exposed quadratic 22684 behavior in the loop optimizer. 22685 + Fix bug which caused volatile memory to be written multiple 22686 times when only one write was needed/desired. 22687 + Fix compiler abort in caller-save.c 22688 + Fix combiner bug which caused incorrect code generation for 22689 certain division by constant operations. 22690 + Fix incorrect code generation due to a bug in range check 22691 optimizations. 22692 + Fix incorrect code generation due to mis-handling of clobbered 22693 values in CSE. 22694 + Fix compiler abort/segfault due to incorrect register 22695 splitting when unrolling loops. 22696 + Fix code generation involving autoincremented addresses with 22697 ternary operators. 22698 + Work around bug in the scheduler which caused qt to be 22699 mis-compiled on some platforms. 22700 + Fix code generation problems with -fshort-enums. 22701 + Tighten security for temporary files. 22702 + Improve compile time for codes which make heavy use of 22703 overloaded functions. 22704 + Fix multiply defined constructor/destructor symbol problems. 22705 + Avoid setting bogus RPATH environment variable during 22706 bootstrap. 22707 + Avoid GNU-make dependencies in the texinfo subdir. 22708 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 22709 --enable-cpp=<dirname> can be used to specify an additional 22710 install directory for the cpp wrapper script. 22711 + Fix CSE bug which caused incorrect label-label refs to appear 22712 on some platforms. 22713 + Avoid linking in EH routines from libgcc if they are not 22714 needed. 22715 + Avoid obscure bug in aliasing code. 22716 + Fix bug in weak symbol handling. 22717 * Platform-specific improvements and fixes 22718 + Fix detection of PPro/PII on Unixware 7. 22719 + Fix compiler segfault when building spec99 and other programs 22720 for SPARC targets. 22721 + Fix code-generation bugs for integer and floating point 22722 conditional move instructions on the PPro/PII. 22723 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 22724 + Fix build failure for the arc port. 22725 + Fix floating point format configuration for i?86-gnu port. 22726 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 22727 threads are enabled. 22728 + Fix coldfire code generation bugs. 22729 + Fix "unrecognized insn" problems for Alpha and PPC ports. 22730 + Fix h8/300 code generation problem with floating point values 22731 in memory. 22732 + Fix unrecognized insn problems for the m68k port. 22733 + Fix namespace-pollution problem for the x86 port. 22734 + Fix problems with old assembler on x86 NeXT systems. 22735 + Fix PIC code-generation problems for the SPARC port. 22736 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 22737 + Fix minor ISO namespace violation in Alpha varargs/stdarg 22738 support. 22739 + Fix incorrect "braf" instruction usage for the SH port. 22740 + Fix minor bug in va-sh which prevented its use with -ansi. 22741 + Fix problems recognizing and supporting FreeBSD. 22742 + Handle OpenBSD systems correctly. 22743 + Minor fixincludes fix for Digital UNIX 4.0B. 22744 + Fix problems with ctors/dtors in SCO shared libraries. 22745 + Abort instead of generating incorrect code for PPro/PII 22746 floating point conditional moves. 22747 + Avoid multiply defined symbols on GNU/Linux systems using 22748 libc-5.4.xx. 22749 + Fix abort in alpha compiler. 22750 * Fortran-specific fixes 22751 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 22752 is in the documented, non-Y2K-compliant range of 0-99, instead 22753 of being returned as 100 in the year 2000. 22754 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 22755 milliseconds value properly in Values(8). 22756 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 22757 information properly in SArray(7). 22758 22759 Each release includes installation instructions in both HTML and 22760 plaintext forms (see the INSTALL directory in the toplevel directory of 22761 the distribution). However, we also keep the most up to date 22762 installation instructions and [6]build/test status on our web page. We 22763 will update those pages as new information becomes available. 22764 22765 The EGCS project would like to thank the numerous people that have 22766 contributed new features, test results, bugfixes, etc. This [7]amazing 22767 group of volunteers is what makes EGCS successful. 22768 22769 And finally, we can't in good conscience fail to mention some 22770 [8]caveats to using EGCS 1.1. 22771 22772 Download EGCS from egcs.cygnus.com (USA California). 22773 22774 The EGCS 1.1 release is also available on many mirror sites. 22775 [9]Goto mirror list to find a closer site. 22776 22777 22778 For questions related to the use of GCC, please consult these web 22779 pages and the [10]GCC manuals. If that fails, the 22780 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22781 web pages and the development of GCC are welcome on our developer 22782 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 22783 archives. 22784 22785 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 22786 distribution of this entire article is permitted in any medium, 22787 provided this notice is preserved. 22788 22789 These pages are [15]maintained by the GCC team. Last modified 22790 2024-05-18. 22791 22792 References 22793 22794 1. https://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 22795 2. https://gcc.gnu.org/news/gcse.html 22796 3. https://gcc.gnu.org/news/alias.html 22797 4. https://gcc.gnu.org/egcs-1.1/c++features.html 22798 5. https://gcc.gnu.org/egcs-1.1/features.html 22799 6. https://gcc.gnu.org/egcs-1.1/buildstat.html 22800 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 22801 8. https://gcc.gnu.org/egcs-1.1/caveats.html 22802 9. https://gcc.gnu.org/mirrors.html 22803 10. https://gcc.gnu.org/onlinedocs/ 22804 11. mailto:gcc-help (a] gcc.gnu.org 22805 12. mailto:gcc (a] gcc.gnu.org 22806 13. https://gcc.gnu.org/lists.html 22807 14. https://www.fsf.org/ 22808 15. https://gcc.gnu.org/about.html 22809 ====================================================================== 22810 http://gcc.gnu.org/egcs-1.1/features.html 22811 22812 EGCS 1.1 new features 22813 22814 * Integrated GNU Fortran (g77) compiler and runtime library with 22815 improvements, based on g77 version 0.5.23. 22816 * Vast improvements in the C++ compiler; so many they have [1]page of 22817 their own! 22818 * Compiler implements [2]global common subexpression elimination and 22819 global copy/constant propagation. 22820 * More major improvements in the [3]alias analysis code. 22821 * More major improvements in the exception handling code to improve 22822 performance, lower static overhead and provide the infrastructure 22823 for future improvements. 22824 * The infamous /tmp symlink race security problems have been fixed. 22825 * The regmove optimization pass has been nearly completely rewritten 22826 to improve performance of generated code. 22827 * The compiler now recomputes register usage information before local 22828 register allocation. By providing more accurate information to the 22829 priority based allocator, we get better register allocation. 22830 * The register reloading phase of the compiler optimizes spill code 22831 much better than in previous releases. 22832 * Some bad interactions between the register allocator and 22833 instruction scheduler have been fixed, resulting in much better 22834 code for certain programs. Additionally, we have tuned the 22835 scheduler in various ways to improve performance of generated code 22836 for some architectures. 22837 * The compiler's branch shortening algorithms have been significantly 22838 improved to work better on targets which align jump targets. 22839 * The compiler now supports -Os to prefer optimizing for code space 22840 over optimizing for code speed. 22841 * The compiler will now totally eliminate library calls which compute 22842 constant values. This primarily helps targets with no integer 22843 div/mul support and targets without floating point support. 22844 * The compiler now supports an extensive "--help" option. 22845 * cpplib has been greatly improved and may be suitable for limited 22846 use. 22847 * Memory footprint for the compiler has been significantly reduced 22848 for some pathological cases. 22849 * The time to build EGCS has been improved for certain targets 22850 (particularly the alpha and mips platforms). 22851 * Many infrastructure improvements throughout the compiler, plus the 22852 usual mountain of bugfixes and minor improvements. 22853 * Target dependent improvements: 22854 + SPARC port now includes V8 plus and V9 support as well as 22855 performance tuning for Ultra class machines. The SPARC port 22856 now uses the Haifa scheduler. 22857 + Alpha port has been tuned for the EV6 processor and has an 22858 optimized expansion of memcpy/bzero. The Alpha port now uses 22859 the Haifa scheduler. 22860 + RS6000/PowerPC: support for the Power64 architecture and AIX 22861 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 22862 + x86: Alignment of static store data and jump targets is per 22863 Intel recommendations now. Various improvements throughout the 22864 x86 port to improve performance on Pentium processors 22865 (including improved epilogue sequences for Pentium chips and 22866 back end improvements which should help register allocation on 22867 all x86 variants. Conditional move support has been fixed and 22868 enabled for PPro processors. The x86 port also better supports 22869 64bit operations now. Unixware 7, a System V Release 5 target, 22870 is now supported and SCO OpenServer targets can support GAS. 22871 + MIPS has improved multiply/multiply-add support and now 22872 includes mips16 ISA support. 22873 + M68k has many micro-optimizations and Coldfire fixes. 22874 * Core compiler is based on the GCC development tree from June 9, 22875 1998, so we have all of the [4]features found in GCC 2.8. 22876 22877 22878 For questions related to the use of GCC, please consult these web 22879 pages and the [5]GCC manuals. If that fails, the 22880 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22881 web pages and the development of GCC are welcome on our developer 22882 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 22883 22884 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 22885 distribution of this entire article is permitted in any medium, 22886 provided this notice is preserved. 22887 22888 These pages are [10]maintained by the GCC team. Last modified 22889 2023-09-02. 22890 22891 References 22892 22893 1. https://gcc.gnu.org/egcs-1.1/c++features.html 22894 2. https://gcc.gnu.org/news/gcse.html 22895 3. https://gcc.gnu.org/news/alias.html 22896 4. https://gcc.gnu.org/egcs-1.0/features-2.8.html 22897 5. https://gcc.gnu.org/onlinedocs/ 22898 6. mailto:gcc-help (a] gcc.gnu.org 22899 7. mailto:gcc (a] gcc.gnu.org 22900 8. https://gcc.gnu.org/lists.html 22901 9. https://www.fsf.org/ 22902 10. https://gcc.gnu.org/about.html 22903 ====================================================================== 22904 http://gcc.gnu.org/egcs-1.1/caveats.html 22905 22906 EGCS 1.1 Caveats 22907 22908 * EGCS has an integrated libstdc++, but does not have an integrated 22909 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 22910 Lu has made a libg++-2.8.1.2 snapshot available which may work with 22911 EGCS. 22912 Note most C++ programs only need libstdc++. 22913 * Exception handling may not work with shared libraries, particularly 22914 on alphas, hppas, rs6000/powerpc and mips based platforms. 22915 Exception handling is known to work on x86-linux platforms with 22916 shared libraries. 22917 * Some versions of the Linux kernel have bugs which prevent them from 22918 being compiled or from running when compiled by EGCS. See the FAQ 22919 (as shipped with EGCS 1.1) for additional information. 22920 * In general, EGCS is more rigorous about rejecting invalid C++ code 22921 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 22922 a result it may be necessary to fix C++ code before it will compile 22923 with EGCS. 22924 * G++ is also converting toward the ISO C++ standard; as a result 22925 code which was previously valid (and thus accepted by other 22926 compilers and older versions of g++) may no longer be accepted. 22927 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 22928 or GCC 2.8.x due to changes necessary to support thread safe 22929 exception handling. 22930 22931 22932 For questions related to the use of GCC, please consult these web 22933 pages and the [1]GCC manuals. If that fails, the 22934 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22935 web pages and the development of GCC are welcome on our developer 22936 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 22937 22938 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 22939 distribution of this entire article is permitted in any medium, 22940 provided this notice is preserved. 22941 22942 These pages are [6]maintained by the GCC team. Last modified 22943 2022-10-26. 22944 22945 References 22946 22947 1. https://gcc.gnu.org/onlinedocs/ 22948 2. mailto:gcc-help (a] gcc.gnu.org 22949 3. mailto:gcc (a] gcc.gnu.org 22950 4. https://gcc.gnu.org/lists.html 22951 5. https://www.fsf.org/ 22952 6. https://gcc.gnu.org/about.html 22953 ====================================================================== 22954 http://gcc.gnu.org/egcs-1.0/index.html 22955 22956 EGCS 1.0 22957 22958 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 22959 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 22960 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 22961 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 22962 22963 EGCS is a collaborative effort involving several groups of hackers 22964 using an open development model to accelerate development and testing 22965 of GNU compilers and runtime libraries. 22966 22967 An important goal of EGCS is to allow wide scale testing of 22968 experimental features and optimizations; therefore, EGCS contains some 22969 features and optimizations which are still under development. However, 22970 EGCS has been carefully tested and should be comparable in quality to 22971 most GCC releases. 22972 22973 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 22974 development sources; it contains nearly all of the new features found 22975 in GCC 2.8. 22976 22977 EGCS 1.0 also contains many improvements and features not found in GCC 22978 2.7 and even the GCC 2.8 series (which was released after the original 22979 EGCS 1.0 release). 22980 * Integrated C++ runtime libraries, including support for most major 22981 GNU/Linux systems! 22982 * The integrated libstdc++ library includes a verbatim copy of SGI's 22983 STL release. 22984 * Integrated GNU Fortran compiler. 22985 * New instruction scheduler. 22986 * New alias analysis code. 22987 22988 See the [1]new features page for a more complete list of new features. 22989 22990 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 22991 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 22992 EGCS 1.0 release: 22993 * Add support for Red Hat 5.0 Linux and better support for Linux 22994 systems using glibc2. 22995 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 22996 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 22997 fix these problems. 22998 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 22999 handling interfaces. 23000 To avoid future compatibility problems, we strongly urge anyone who 23001 is planning on distributing shared libraries that contain C++ code 23002 to upgrade to EGCS 1.0.1 first. 23003 Soon after EGCS 1.0 was released, the GCC developers made some 23004 incompatible changes in libgcc's exception handling interfaces. 23005 These changes were needed to solve problems on some platforms. This 23006 means that GCC 2.8.0, when released, will not be seamlessly 23007 compatible with shared libraries built by EGCS 1.0. The reason is 23008 that the libgcc.a in GCC 2.8.0 will not contain a function needed 23009 by the old interface. 23010 The result of this is that there may be compatibility problems with 23011 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 23012 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 23013 and libgcc.a has the support routines for both the old and the new 23014 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 23015 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 23016 The maintainers of GCC 2.x have decided against including seamless 23017 support for the old interface in 2.8.0, since it was never 23018 "official", so to avoid future compatibility problems we recommend 23019 against distributing any shared libraries built by EGCS 1.0 that 23020 contain C++ code (upgrade to 1.0.1 and use that). 23021 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 23022 The x86 changes fix code generation errors exposed when building 23023 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 23024 The hppa change fixes a compiler abort when configured for use with 23025 RTEMS. 23026 The MIPS changes fix problems with the definition of LONG_MAX on 23027 newer systems, allow for command line selection of the target ABI, 23028 and fix one code generation problem. 23029 The rs6000/ppc change fixes some problems with passing structures 23030 to varargs/stdarg functions. 23031 * A few machine independent bugfixes, mostly to fix code generation 23032 errors when building Linux kernels or glibc. 23033 * Fix a few critical exception handling and template bugs in the C++ 23034 compiler. 23035 * Fix Fortran namelist bug on alphas. 23036 * Fix build problems on x86-solaris systems. 23037 23038 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 23039 serious problems in EGCS 1.0.1. 23040 * General improvements and fixes 23041 + Memory consumption significantly reduced, especially for 23042 templates and inline functions. 23043 + Fix various problems with glibc2.1. 23044 + Fix loop optimization bug exposed by rs6000/ppc port. 23045 + Fix to avoid potential code generation problems in jump.c. 23046 + Fix some undefined symbol problems in dwarf1 debug support. 23047 * g++/libstdc++ improvements and fixes 23048 + libstdc++ in the EGCS release has been updated and should be 23049 link compatible with libstdc++-2.8. 23050 + Various fixes in libio/libstdc++ to work better on GNU/Linux 23051 systems. 23052 + Fix problems with duplicate symbols on systems that do not 23053 support weak symbols. 23054 + Memory corruption bug and undefined symbols in bastring have 23055 been fixed. 23056 + Various exception handling fixes. 23057 + Fix compiler abort for very long thunk names. 23058 * g77 improvements and fixes 23059 + Fix compiler crash for omitted bound in Fortran CASE 23060 statement. 23061 + Add missing entries to g77 lang-options. 23062 + Fix problem with -fpedantic in the g77 compiler. 23063 + Fix "backspace" problem with g77 on alphas. 23064 + Fix x86 back end problem with Fortran literals and -fpic. 23065 + Fix some of the problems with negative subscripts for g77 on 23066 alphas. 23067 + Fixes for Fortran builds on cygwin32/mingw32. 23068 * platform specific improvements and fixes 23069 + Fix long double problems on x86 (exposed by glibc). 23070 + x86 ports define i386 again to keep imake happy. 23071 + Fix exception handling support on NetBSD ports. 23072 + Several changes to collect2 to fix many problems with AIX. 23073 + Define __ELF__ for GNU/Linux on rs6000. 23074 + Fix -mcall-linux problem on GNU/Linux on rs6000. 23075 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 23076 + Allow autoconf to select a proper install problem on AIX 3.1. 23077 + m68k port support includes -mcpu32 option as well as cpu32 23078 multilibs. 23079 + Fix stdarg bug for irix6. 23080 + Allow EGCS to build on irix5 without the gnu assembler. 23081 + Fix problem with static linking on sco5. 23082 + Fix bootstrap on sco5 with native compiler. 23083 + Fix for abort building newlib on H8 target. 23084 + Fix fixincludes handling of math.h on SunOS. 23085 + Minor fix for Motorola 3300 m68k systems. 23086 23087 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 23088 problems reported by Red Hat for builds of Red Hat 5.1. 23089 * Generic bugfixes: 23090 + Fix a typo in the libio library which resulted in incorrect 23091 behavior of istream::get. 23092 + Fix the Fortran negative array index problem. 23093 + Fix a major problem with the ObjC runtime thread support 23094 exposed by glibc2. 23095 + Reduce memory consumption of the Haifa scheduler. 23096 * Target specific bugfixes: 23097 + Fix one x86 floating point code generation bug exposed by 23098 glibc2 builds. 23099 + Fix one x86 internal compiler error exposed by glibc2 builds. 23100 + Fix profiling bugs on the Alpha. 23101 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 23102 + Fix rs6000/ppc bug when converting values from integer types 23103 to floating point types. 23104 23105 The EGCS 1.0 releases include installation instructions in both HTML 23106 and plaintext forms (see the INSTALL directory in the toplevel 23107 directory of the distribution). However, we also keep the most up to 23108 date installation instructions and [2]build/test status on our web 23109 page. We will update those pages as new information becomes available. 23110 23111 And, we can't in good conscience fail to mention some [3]caveats to 23112 using EGCS. 23113 23114 Update: Big thanks to Stanford for providing a high speed link for 23115 downloading EGCS (go.cygnus.com)! 23116 23117 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 23118 (USA California -- High speed link provided by Stanford). 23119 23120 The EGCS 1.0 release is also available many mirror sites. 23121 [4]Goto mirror list to find a closer site 23122 23123 We'd like to thank the numerous people that have contributed new 23124 features, test results, bugfixes, etc. Unfortunately, they're far too 23125 numerous to mention by name. 23126 23127 23128 For questions related to the use of GCC, please consult these web 23129 pages and the [5]GCC manuals. If that fails, the 23130 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23131 web pages and the development of GCC are welcome on our developer 23132 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 23133 23134 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 23135 distribution of this entire article is permitted in any medium, 23136 provided this notice is preserved. 23137 23138 These pages are [10]maintained by the GCC team. Last modified 23139 2023-09-02. 23140 23141 References 23142 23143 1. https://gcc.gnu.org/egcs-1.0/features.html 23144 2. https://gcc.gnu.org/egcs-1.0/buildstat.html 23145 3. https://gcc.gnu.org/egcs-1.0/caveats.html 23146 4. https://gcc.gnu.org/mirrors.html 23147 5. https://gcc.gnu.org/onlinedocs/ 23148 6. mailto:gcc-help (a] gcc.gnu.org 23149 7. mailto:gcc (a] gcc.gnu.org 23150 8. https://gcc.gnu.org/lists.html 23151 9. https://www.fsf.org/ 23152 10. https://gcc.gnu.org/about.html 23153 ====================================================================== 23154 http://gcc.gnu.org/egcs-1.0/features.html 23155 23156 EGCS 1.0 features 23157 23158 * Core compiler is based on the gcc2 development tree from Aug 2, 23159 1997, so we have most of the [1]features found in GCC 2.8. 23160 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 23161 * Vast improvements in the C++ compiler; so many they have [2]page of 23162 their own! 23163 * Integrated C++ runtime libraries, including support for most major 23164 GNU/Linux systems! 23165 * New instruction scheduler from IBM Haifa which includes support for 23166 function wide instruction scheduling as well as superscalar 23167 scheduling. 23168 * Significantly improved alias analysis code. 23169 * Improved register allocation for two address machines. 23170 * Significant code generation improvements for Fortran code on 23171 Alphas. 23172 * Various optimizations from the g77 project as well as improved loop 23173 optimizations. 23174 * Dwarf2 debug format support for some targets. 23175 * egcs libstdc++ includes the SGI STL implementation without changes. 23176 * As a result of these and other changes, egcs libstc++ is not binary 23177 compatible with previous releases of libstdc++. 23178 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 23179 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 23180 1.1), Support for RTEMS on several embedded targets, Support for 23181 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 23182 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 23183 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 23184 * RS6000/PowerPC ports generate code which can run on all 23185 RS6000/PowerPC variants by default. 23186 * -mcpu= and -march= switches for the x86 port to allow better 23187 control over how the x86 port generates code. 23188 * Includes the template repository patch (aka repo patch); note the 23189 new template code makes repo obsolete for ELF systems using gnu-ld 23190 such as GNU/Linux. 23191 * Plus the usual assortment of bugfixes and improvements. 23192 23193 23194 For questions related to the use of GCC, please consult these web 23195 pages and the [3]GCC manuals. If that fails, the 23196 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23197 web pages and the development of GCC are welcome on our developer 23198 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 23199 23200 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 23201 distribution of this entire article is permitted in any medium, 23202 provided this notice is preserved. 23203 23204 These pages are [8]maintained by the GCC team. Last modified 23205 2022-10-26. 23206 23207 References 23208 23209 1. https://gcc.gnu.org/egcs-1.0/features-2.8.html 23210 2. https://gcc.gnu.org/egcs-1.0/c++features.html 23211 3. https://gcc.gnu.org/onlinedocs/ 23212 4. mailto:gcc-help (a] gcc.gnu.org 23213 5. mailto:gcc (a] gcc.gnu.org 23214 6. https://gcc.gnu.org/lists.html 23215 7. https://www.fsf.org/ 23216 8. https://gcc.gnu.org/about.html 23217 ====================================================================== 23218 http://gcc.gnu.org/egcs-1.0/caveats.html 23219 23220 EGCS 1.0 Caveats 23221 23222 * EGCS has an integrated libstdc++, but does not have an integrated 23223 libg++. Furthermore old libg++ releases will not work with egc; HJ 23224 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 23225 Note most C++ programs only need libstdc++. 23226 * Note that using -pedantic or -Wreturn-type can cause an explosion 23227 in the amount of memory needed for template-heavy C++ code, such as 23228 code that uses STL. Also note that -Wall includes -Wreturn-type, so 23229 if you use -Wall you will need to specify -Wno-return-type to turn 23230 it off. 23231 * Exception handling may not work with shared libraries, particularly 23232 on alphas, hppas, and mips based platforms. Exception handling is 23233 known to work on x86-linux platforms with shared libraries. 23234 * Some versions of the Linux kernel have bugs which prevent them from 23235 being compiled or from running when compiled by EGCS. See the FAQ 23236 (as shipped with EGCS 1.0) for additional information. 23237 * In general, EGCS is more rigorous about rejecting invalid C++ code 23238 or deprecated C++ constructs than G++ 2.7. As a result it may be 23239 necessary to fix C++ code before it will compile with EGCS. 23240 * G++ is also aggressively tracking the C++ standard; as a result 23241 code which was previously valid (and thus accepted by other 23242 compilers and older versions of G++) may no longer be accepted. 23243 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 23244 1.0.x and later releases should work with Red Hat Linux 5.0. 23245 23246 23247 For questions related to the use of GCC, please consult these web 23248 pages and the [1]GCC manuals. If that fails, the 23249 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 23250 web pages and the development of GCC are welcome on our developer 23251 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 23252 23253 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 23254 distribution of this entire article is permitted in any medium, 23255 provided this notice is preserved. 23256 23257 These pages are [6]maintained by the GCC team. Last modified 23258 2022-10-26. 23259 23260 References 23261 23262 1. https://gcc.gnu.org/onlinedocs/ 23263 2. mailto:gcc-help (a] gcc.gnu.org 23264 3. mailto:gcc (a] gcc.gnu.org 23265 4. https://gcc.gnu.org/lists.html 23266 5. https://www.fsf.org/ 23267 6. https://gcc.gnu.org/about.html 23268 ====================================================================== 23269
Indexes created Tue Oct 27 00:25:24 UTC 2026