NEWS revision 1.18
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 GCC 12 Release Series 10 11 Jun 20, 2024 12 13 The GCC developers are pleased to announce the release of GCC 12.4. 14 15 This release is a bug-fix release, containing fixes for regressions in 16 GCC 12.3 relative to previous releases of GCC. 17 18 Release History 19 20 GCC 12.4 21 Jun 20, 2024 ([1]changes, [2]documentation) 22 23 GCC 12.3 24 May 8, 2023 ([3]changes, [4]documentation) 25 26 GCC 12.2 27 Aug 19, 2022 ([5]changes, [6]documentation) 28 29 GCC 12.1 30 May 6, 2022 ([7]changes, [8]documentation) 31 32 References and Acknowledgements 33 34 GCC used to stand for the GNU C Compiler, but since the compiler 35 supports several other languages aside from C, it now stands for the 36 GNU Compiler Collection. 37 38 The GCC developers would like to thank the numerous people that have 39 contributed new features, improvements, bug fixes, and other changes as 40 well as test results to GCC. This [9]amazing group of volunteers is 41 what makes GCC successful. 42 43 For additional information about GCC please refer to the [10]GCC 44 project web site or contact the [11]GCC development mailing list. 45 46 To obtain GCC please use [12]our mirror sites or [13]our version 47 control system. 48 49 50 For questions related to the use of GCC, please consult these web 51 pages and the [14]GCC manuals. If that fails, the 52 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 53 web pages and the development of GCC are welcome on our developer 54 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 55 archives. 56 57 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 58 distribution of this entire article is permitted in any medium, 59 provided this notice is preserved. 60 61 These pages are [19]maintained by the GCC team. Last modified 62 2024-06-20. 63 64 References 65 66 1. https://gcc.gnu.org/gcc-12/changes.html 67 2. https://gcc.gnu.org/onlinedocs/12.4.0/ 68 3. https://gcc.gnu.org/gcc-12/changes.html 69 4. https://gcc.gnu.org/onlinedocs/12.3.0/ 70 5. https://gcc.gnu.org/gcc-12/changes.html 71 6. https://gcc.gnu.org/onlinedocs/12.2.0/ 72 7. https://gcc.gnu.org/gcc-12/changes.html 73 8. https://gcc.gnu.org/onlinedocs/12.1.0/ 74 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Contributors.html 75 10. https://gcc.gnu.org/index.html 76 11. mailto:gcc (a] gcc.gnu.org 77 12. https://gcc.gnu.org/mirrors.html 78 13. https://gcc.gnu.org/git.html 79 14. https://gcc.gnu.org/onlinedocs/ 80 15. mailto:gcc-help (a] gcc.gnu.org 81 16. mailto:gcc (a] gcc.gnu.org 82 17. https://gcc.gnu.org/lists.html 83 18. https://www.fsf.org/ 84 19. https://gcc.gnu.org/about.html 85 ====================================================================== 86 http://gcc.gnu.org/gcc-12/changes.html 87 GCC 12 Release Series 88 Changes, New Features, and Fixes 89 90 This page is a "brief" summary of some of the huge number of 91 improvements in GCC 12. You may also want to check out our [1]Porting 92 to GCC 12 page and the [2]full GCC documentation. 93 94 Caveats 95 96 * An ABI incompatibility between C and C++ when passing or returning 97 by value certain aggregates containing zero width bit-fields has 98 been discovered on various targets. As mentioned in [3]PR102024, 99 since the [4]PR42217 fix in GCC 4.5 the C++ front-end has been 100 removing zero width bit-fields from the internal representation of 101 the aggregates after the layout of those aggregates, but the C 102 front-end kept them, so passing e.g. struct S { float a; int : 0; 103 float b; } or struct T { float c; int : 0; } by value could differ 104 between C and C++. Starting with GCC 12 the C++ front-end no longer 105 removes those bit-fields from the internal representation and per 106 clarified psABI some targets have been changed, so that they either 107 ignore those bit-fields in the argument passing by value decisions 108 in both C and C++, or they always take them into account. x86-64, 109 ARM and AArch64 will always ignore them (so there is a C ABI 110 incompatibility between GCC 11 and earlier with GCC 12 or later), 111 PowerPC64 ELFv2 and S/390 always take them into account (so there 112 is a C++ ABI incompatibility, GCC 4.4 and earlier compatible with 113 GCC 12 or later, incompatible with GCC 4.5 through GCC 11). RISC-V 114 has changed the handling of these already starting with GCC 10. As 115 the ABI requires, MIPS takes them into account handling function 116 return values so there is a C++ ABI incompatibility with GCC 4.5 117 through 11. For function arguments on MIPS, refer to [5]the MIPS 118 specific entry. GCC 12 on the above targets will report such 119 incompatibilities as warnings or other diagnostics unless 120 -Wno-psabi is used. 121 * C: Computed gotos require a pointer type now. 122 * C++: Two non-standard std::pair constructors have been deprecated. 123 These allowed the use of an rvalue and a literal 0 to construct a 124 pair containing a move-only type and a pointer. The nullptr keyword 125 should be used to initialize the pointer member instead of a 126 literal 0, as this is portable to other C++ implementations. 127 * The configuration option --enable-libstdcxx-allocator no longer 128 supports the bitmap, mt, and pool arguments. Those configurations 129 had been broken for some time. 130 * D: Building and bootstrapping GDC, the D compiler, now requires a 131 working GDC compiler (GCC version 9.1 or later) and D runtime 132 library, libphobos, as the D front end is written in D. On some 133 targets, libphobos isn't enabled by default, but compiles and works 134 if --enable-libphobos is used. Other targets may require a more 135 recent minimum version of GCC to bootstrap. Specifics are 136 documented for affected targets in the [6]manual for installing 137 GCC. 138 * Fortran: OpenMP code using the omp_lib.h include file can no longer 139 be compiled with -std=f95 but now requires at least -std=f2003. 140 Alternatively, use the omp_lib module, which still supports 141 -std=f95 and is recommended to be used instead in general. 142 * OpenMP offloading to Intel MIC has been deprecated and will be 143 removed in a future release. 144 * The cr16 target with the cr16-*-* configuration has been obsoleted 145 and will be removed in a future release. 146 * The hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11* configurations 147 targeting 32-bit PA-RISC with HP-UX have been obsoleted and will be 148 removed in a future release. 149 * The m32c*-*-rtems* configuration has been obsoleted and will be 150 removed in a future release. 151 * The support for the m32r-*-linux*, m32rle-*-linux*, 152 m68k*-*-openbsd* and vax-*-openbsd* configurations has been 153 removed. 154 * STABS: Support for emitting the STABS debugging format is 155 deprecated and will be removed in the next release. All ports now 156 default to emit DWARF (version 2 or later) debugging info or are 157 obsoleted. 158 * The optimization level -Ofast now implies 159 -fno-semantic-interposition. 160 161 General Improvements 162 163 * Vectorization is enabled at -O2 which is now equivalent to what 164 would have been -O2 -ftree-vectorize -fvect-cost-model=very-cheap 165 in the past. Note that the default vectorizer cost model has been 166 changed; it used to behave as if -fvect-cost-model=cheap had been 167 specified. 168 * GCC now supports the [7]ShadowCallStack sanitizer, which can be 169 enabled using the command-line option 170 [8]-fsanitize=shadow-call-stack. This sanitizer currently only 171 works on AArch64 targets and it requires an environment in which 172 all code has been compiled with -ffixed-r18. Its primary initial 173 user is the Linux kernel. 174 175 New Languages and Language specific improvements 176 177 * OpenMP 178 + OpenMP 5.0 support has been extended: The close map modifier 179 and the affinity clause are now supported. In addition, 180 Fortran gained the following features which were available in 181 C and C++ before: declare variant is now available, depobj, 182 mutexinoutset and iterator can now also be used with the 183 depend clause, defaultmap has been updated for OpenMP 5.0, and 184 the loop directive and combined directives involving the 185 master directive have been added. 186 + The following OpenMP 5.1 features have been added: support for 187 expressing OpenMP directives as C++ 11 attributes, the masked 188 and scope constructs, the nothing and error directives, and 189 using primary with the proc_bind clause and OMP_PROC_BIND 190 environment variable, the reproducible and unconstrained 191 modifiers to the order clause, and, for C/C++ only, the align 192 and allocator modifiers to the allocate clause and the atomic 193 extensions are now available. The OMP_PLACE environment 194 variable supports the OpenMP 5.1 features. In addition, the 195 OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT environment variables 196 and their associated API routines are now supported as well as 197 the memory-allocation routines added for Fortran and extended 198 for C/C++ in OpenMP 5.1. In Fortran code, strictly structured 199 blocks can be used. 200 + The [9]OpenMP Implementation Status can be found in the 201 libgomp manual. 202 * Version 2.6 of the [10]OpenACC specification continues to be 203 maintained and improved in the C, C++ and Fortran compilers. See 204 the [11]implementation status section on the OpenACC wiki page and 205 the [12]run-time library documentation for further information. In 206 addition to general performance tuning and bug fixing, new features 207 include: 208 + OpenACC worker parallelism for [13]AMD GPUs (already for a 209 long time supported for [14]Nvidia GPUs). 210 + Data privatization/sharing at the OpenACC gang level. 211 + Considerable improvements for the experimental OpenACC 212 'kernels' decomposition ([15]--param 213 openacc-kernels=decompose). 214 + A new warning flag [16]-Wopenacc-parallelism to warn about 215 potentially suboptimal choices related to OpenACC parallelism. 216 * The offload target code generation for OpenMP and OpenACC can now 217 be better adjusted using the new [17]-foffload-options= flag and 218 the pre-existing but now documented [18]-foffload= flag. 219 220 Ada 221 222 * Ada 2022 223 + Added the -gnat2022 flag to indicate strict Ada 2022 224 compliance. The old -gnat2020 flag is now deprecated. 225 + Support for Big Numbers (Annex G) has seen continuous 226 improvements. It is now considered complete. It is compatible 227 with SPARK, i.e. can be used from SPARK code. 228 + Continuous improvements to the Ada 2022 standard since GCC 11. 229 + Greatly improved compile time support. More functions can now 230 have the with Static aspect and can be used in more contexts. 231 * Ada 2022 extensions. The use of the -gnatX flag is necessary to 232 access these features as they are not considered stable or 233 standard. 234 + Fixed lower bound for unconstrained arrays. 235 o type Matrix is array (Natural range 0 .. <>, Natural 236 range 0 .. <>) of Integer; is now valid. 237 o Subtypes can also specify a lower bound: subtype String_1 238 is String (1 .. <>);. Boundaries from slices will "slide" 239 to the correct lower bound of the subtype. 240 + Generalized Object.Operand notation. The following code is now 241 valid V.Add_Element(42);, with V being a vector, for example. 242 + Additional when constructs. Keywords return, goto and raise 243 can now use when in addition to the existing exit when. The 244 following expression is therefore now valid raise 245 Constraint_Error with "Element is null" when Element = null; 246 + Pattern matching 247 o The case statement has been extended to cover records and 248 arrays as well as finer grained casing on scalar types. 249 In the future it is expected to provide more compile time 250 guarantees when accessing discriminated fields. Case 251 exhaustion is supported for pattern matching. An example 252 would be 253 type Sign is (Neg, Zero, Pos); 254 255 function Multiply (S1, S2 : Sign) return Sign is 256 (case (S1, S2) is 257 when (Neg, Neg) | (Pos, Pos) => Pos, 258 when (Zero, <>) | (<>, Zero) => Zero, 259 when (Neg, Pos) | (Pos, Neg) => Neg); 260 261 * gnatfind and gnatxref, which were already deprecated, have been 262 removed. 263 * Greatly expanded code covered by contracts. Thanks to this work, 264 there are now several Ada standard libraries fully proven in SPARK 265 which means they have no runtime nor logical errors. They are 266 mostly numeric and string handling libraries. 267 * Enable return-slot optimization for Pure functions. 268 * General optimizations, improvements and additions to the standard 269 library. Performance, correctness and in some cases stability was 270 improved. Memory pools have also seen some minor enhancements. 271 * Improvements to embedded-RTOS targets such as RTEMS, VxWorks and 272 QNX. Older targets were removed or cleaned. 273 * Added some [19]hardening features. 274 275 C family 276 277 * Support for __builtin_shufflevector compatible with the clang 278 language extension was added. 279 * Support for attribute unavailable was added. 280 * A new built-in function, __builtin_assoc_barrier, was added. It can 281 be used to inhibit re-association of floating-point expressions. 282 * Support for __builtin_dynamic_object_size compatible with the clang 283 language extension was added. 284 * New warnings: 285 + [20]-Wbidi-chars warns about potentially misleading UTF-8 286 bidirectional control characters. The default is 287 -Wbidi-chars=unpaired ([21]PR103026) 288 + [22]-Warray-compare warns about comparisons between two 289 operands of array type ([23]PR97573) 290 * Enhancements to existing warnings: 291 + [24]-Wattributes has been extended so that it's possible to 292 use -Wno-attributes=ns::attr or -Wno-attributes=ns:: to 293 suppress warnings about unknown scoped attributes (in C++11 294 and C2X). Similarly, #pragma GCC diagnostic ignored_attributes 295 "vendor::attr" can be used to achieve the same effect 296 ([25]PR101940) 297 298 C 299 300 * Some new features from the upcoming C2X revision of the ISO C 301 standard are supported with -std=c2x and -std=gnu2x. Some of these 302 features are also supported as extensions when compiling for older 303 language versions. In addition to the features listed, some 304 features previously supported as extensions and now added to the C 305 standard are enabled by default in C2X mode and not diagnosed with 306 -std=c2x -Wpedantic. 307 + Digit separators (as in C++) are supported for C2X. 308 + The #elifdef and #elifndef preprocessing directives are now 309 supported. 310 + The printf and scanf format checking with [26]-Wformat now 311 supports the %b format specified by C2X for binary integers, 312 and the %B format recommended by C2X for printf. 313 314 C++ 315 316 * Several C++23 features have been implemented: 317 + [27]P1938R3, if consteval ([28]PR100974) 318 + [29]P0849R8, auto(x): decay-copy in the language 319 ([30]PR103049) 320 + [31]P2242R3, Non-literal variables (and labels and gotos) in 321 constexpr functions ([32]PR102612) 322 + [33]P2334R1, Support for preprocessing directives elifdef and 323 elifndef ([34]PR102616) 324 + [35]P2360R0, Extend init-statement to allow alias-declaration 325 ([36]PR102617) 326 + [37]P2128R6, Multidimensional subscript operator 327 + [38]DR 2397, auto specifier for pointers and references to 328 arrays ([39]PR100975) 329 * Several C++ Defect Reports have been resolved, e.g.: 330 + [40]DR 960, Covariant functions and lvalue/rvalue references 331 + [41]DR 1227, Mixing immediate and non-immediate contexts in 332 deduction failure 333 + [42]DR 1315, Restrictions on non-type template arguments in 334 partial specializations 335 + [43]DR 2082, Referring to parameters in unevaluated operands 336 of default arguments 337 + [44]DR 2351, void{} 338 + [45]DR 2374, Overly permissive specification of enum 339 direct-list-initialization 340 + [46]DR 2397, auto specifier for pointers and references to 341 arrays 342 + [47]DR 2446, Questionable type-dependency of concept-ids 343 * New command-line option -fimplicit-constexpr can be used to make 344 inline functions implicitly constexpr ([48]git) 345 * New command-line option -ffold-simple-inlines can be used to fold 346 calls to certain trivial inline functions (currently std::move, 347 std::forward, std::addressof and std::as_const). In contrast to 348 inlining such calls, folding means that no intermediate code or 349 debug information will be generated for them; this minimizes the 350 abstraction penalty incurred for using these functions versus using 351 the fundamental operations from which they're defined (e.g. 352 std::move versus static_cast). This flag is enabled by default when 353 -fno-inline is not active. 354 * Deduction guides can be declared at class scope ([49]PR79501) 355 * [50]-Wuninitialized warns about using uninitialized variables in 356 member initializer lists ([51]PR19808) 357 * [52]-Wint-in-bool-context is now disabled when instantiating a 358 template ([53]git) 359 * Stricter checking of attributes on friend declarations: if a friend 360 declaration has an attribute, that declaration must be a 361 definition. Moreover, a C++11 attribute cannot appear in the middle 362 of the decl-specifier-seq. ([54]PR99032) 363 * New warning options for C++ language mismatches: 364 -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions, 365 -Wc++20-extensions, and -Wc++23-extensions. They are enabled by 366 default and can be used to control existing pedwarns about 367 occurrences of new C++ constructs in code using an old C++ standard 368 dialect. 369 * New warning [55]-Wmissing-requires warns about missing requires 370 ([56]git) 371 * The existing std::is_constant_evaluated in if warning was extended 372 to warn in more cases ([57]PR100995) 373 * [58]-Waddress has been enhanced so that it now warns about, for 374 instance, comparing the address of a nonstatic member function to 375 null ([59]PR102103) 376 * Errors about narrowing are no longer hidden if they occur in system 377 headers 378 * Ordered comparison of null pointers is now rejected ([60]PR99701) 379 * Anonymous structs with bases are now rejected ([61]git) 380 * The compiler rejects taking the address of an immediate member 381 function ([62]PR102753) 382 * The compiler has support for C++20 383 __cpp_lib_is_pointer_interconvertible and 384 __cpp_lib_is_layout_compatible to help the C++ library implement 385 [63]P0466, Layout-compatibility and Pointer-interconvertibility 386 Traits ([64]PR101539) 387 * Memory usage of constraint subsumption has been improved 388 ([65]PR100828) 389 * constinit thread_local variables are optimized better 390 ([66]PR101786) 391 * Support for C++17 std::hardware_destructive_interference_size was 392 added, along with the [67]-Winterference-size warning ([68]git) 393 * Many bugs in the CTAD handling have been fixed ([69]PR101344, 394 [70]PR101883, [71]PR89062, [72]PR101233, [73]PR88252, [74]PR86439, 395 [75]PR98832, [76]PR102933 ...) 396 * Two-stage name lookup for dependent operator expressions has been 397 corrected ([77]PR51577) 398 * Several issues with constrained variable templates have been fixed 399 ([78]PR98486) 400 * The compiler performs less instantiating when doing speculative 401 constant evaluation ([79]git) 402 * Various diagnostic improvements; e.g., a more precise caret 403 location for pointer-to-member expressions 404 * The new -fconstexpr-fp-except flag allows IEC559 floating point 405 exceptions in constant-expressions. 406 407 Runtime Library (libstdc++) 408 409 * Improved experimental C++20 support, including: 410 + std::vector, std::basic_string, std::optional, and 411 std::variant can be used in constexpr functions. 412 + std::make_shared for arrays with default initialization, and 413 std::atomic<std::shared_ptr<T>>. 414 + Layout-compatibility and pointer-interconvertibility traits. 415 * Improved experimental C++23 support, including: 416 + Monadic operations for std::optional. 417 + std::expected 418 + std::move_only_function 419 + <spanstream> 420 + std::basic_string::resize_and_overwrite 421 + std::unique_ptr can be used in constexpr functions. 422 + <stacktrace> (not built by default, requires linking to an 423 extra library). 424 + <stdatomic.h> 425 + std::invoke_r 426 + constexpr std::type_info::operator== 427 428 D 429 430 * New features: 431 + Support for the D programming language has been updated to 432 version 2.100.1 of the language and run-time library. Full 433 changelog for this release and previous releases can be found 434 on the [80]dlang.org website. 435 + On supported targets, the __traits(compiles) expression can 436 now be used to determine whether a target-specific built-in is 437 available without error during CTFE ([81]PR101127). 438 + Functions annotated with pragma(inline, true) are now compiled 439 into every module where they are used from ([82]PR106563). 440 + Partial support for directly importing C99 sources into a D 441 compilation ([83]ImportC) has been added to the language. A 442 notable missing feature is support for preprocessing C 443 imports, which can be worked around by preprocessing all C 444 sources used for importing ahead of time. 445 * New language options: 446 + -fcheck=, enables or disables the code generation of specific 447 run-time contract checks. 448 + -fcheckaction=, controls the run-time behavior on an assert, 449 array bounds check, or final switch contract failure. The 450 default is -fcheckaction=throw. 451 + -fdump-c++-spec=, dumps all compiled extern(C++) declarations 452 as C++ code to the given file. The supplementary option 453 -fdump-c++-spec-verbose turns on emission of comments for 454 ignored declarations in the generated spec. 455 + -fextern-std=, controls which C++ standard extern(C++) 456 declarations are compiled to be compatible with. The default 457 is -fextern-std=c++17. 458 + -fpreview=, added to enable upcoming D language features in 459 the compiler. 460 + -frevert=, added to revert D language changes to support older 461 D codebases that need more time to transition. 462 + -fsave-mixins=, saves mixins expanded at compile-time to a 463 file. 464 * Deprecated and removed features: 465 + The -Wtemplates compiler switch has been removed, as it had 466 been superceded by -ftransition=templates, which more 467 accurately reports on which templates have been instantiated. 468 + The -ftransition=dip25 and -ftransition=dip1000 compiler 469 switches have been renamed to -fpreview=dip25 and 470 -fpreview=dip1000. 471 472 Fortran 473 474 * WG5/N1942, "TS 29113 Further Interoperability of Fortran with C", 475 is now fully supported. In addition to implementing previously 476 missing functionality, such as support for character arguments of 477 length greater than one in functions marked bind(c) and gaps in the 478 handling for assumed-rank arrays, numerous other bugs have been 479 fixed, and an extensive set of new conformance test cases has been 480 added. 481 * GCC 12 now uses OPERATION as the name of the function to the 482 CO_REDUCE intrinsic for the pairwise reduction, thus conforming to 483 the Fortran 2018 standard. Previous versions used OPERATOR which 484 conforms to TS 18508. 485 * On POWER systems which support it, the -mabi=ieeelongdouble option 486 now selects the IEEE 128-bit floating point format for 487 REAL(KIND=16). R16_IBM and R16_IEEE have been added to the 488 -fconvert option, the CONVERT specifier of the OPEN statement and 489 the GFORTRAN_CONVERT_UNIT environment variable. 490 491 Go 492 493 * GCC 12 provides a complete implementation of the Go 1.18 user 494 packages. 495 * Although Go 1.18 includes support for generic programming, that 496 support is not yet available in GCC. 497 498 libgccjit 499 500 * The libgccjit API gained 30 new entry points: 501 + 17 new "reflection" entry points for querying functions and 502 types ([84]LIBGCCJIT_ABI_16) 503 + [85]gcc_jit_lvalue_set_tls_model for supporting thread-local 504 variables ([86]LIBGCCJIT_ABI_17) 505 + [87]gcc_jit_lvalue_set_link_section for setting the link 506 section of global variables, analogous to 507 [88]__attribute__((section(".section"))) 508 ([89]LIBGCCJIT_ABI_18) 509 + 4 new entry points for initializing global variables and 510 creating constructors for rvalues ([90]LIBGCCJIT_ABI_19) 511 + Support for sized integer types, including 128-bit integers 512 and helper functions for such types ([91]LIBGCCJIT_ABI_20) 513 + [92]gcc_jit_context_new_bitcast for reinterpreting the bits of 514 an rvalue as a different type ([93]LIBGCCJIT_ABI_21) 515 + [94]gcc_jit_lvalue_set_register_name for setting a specific 516 register for a variable ([95]LIBGCCJIT_ABI_22) 517 + [96]gcc_jit_context_set_bool_print_errors_to_stderr 518 ([97]LIBGCCJIT_ABI_23) 519 + 2 new entry points for setting the alignment of a variable 520 ([98]LIBGCCJIT_ABI_24) 521 * libgccjit has gained support for the use of various atomic builtins 522 ([99]PR96066, [100]PR96067) 523 * [101]gcc_jit_context_new_cast is now able to handle truncation and 524 extension between different integer types ([102]PR95498) 525 526 New Targets and Target Specific Improvements 527 528 AArch64 & arm 529 530 * Newer revisions of the Arm Architecture are supported as arguments 531 to the -march option: armv8.7-a, armv8.8-a, armv9-a. 532 * The Arm Cortex-A510 CPU is now supported through the cortex-a510 533 argument to the -mcpu and -mtune options. 534 * GCC can now auto-vectorize operations performing sign-differing 535 dot-product operations, taking advantage of instructions in the 536 Advanced SIMD (AArch64/AArch32) and SVE (AArch64) instruction sets. 537 538 AArch64 539 540 * A number of new CPUs are supported through the -mcpu and -mtune 541 options (GCC identifiers in parentheses). 542 + Ampere-1 (ampere1). 543 + Arm Cortex-A710 (cortex-a710). 544 + Arm Cortex-X2 (cortex-x2). 545 * The 64-byte atomic load/store intrinsics to accelerator memory from 546 the [103]2020 Arm Architecture extensions are supported through the 547 +ls64 option extension. 548 * Initial code generation support is supported for hardware 549 instructions used to accelerate the memcpy,memmove and memset 550 standard functions. These instructions can be generated when 551 compiling with the +mopsoption extension. 552 * The ACLE Advanced SIMD intrinsics accessible through the arm_neon.h 553 header have been significantly reimplemented and generate 554 higher-performing code than previous GCC versions. 555 * The option -mtune=neoverse-512tvb is added to tune for Arm Neoverse 556 cores that have a total vector bandwidth of 512 bits. Please refer 557 to the documentation for more details. 558 559 AMD Radeon (GCN) 560 561 * Debug experience with ROCGDB has been improved. 562 * Support for the type __int128_t/integer(kind=16) was added. 563 * For offloading, the limitation of using only one wavefront per 564 compute unit (CU) has been lifted. Up to 40 workgroups per CU and 565 16 wavefronts per workgroup are supported (up to a limit of 40 566 wavefronts in total, per CU). Additionally, the number of used 567 wavefronts and workgroups was tuned for performance. 568 569 arm 570 571 * Support is added for accessing the stack canary value via the TLS 572 register through the -fstack-protector-guard=tls and 573 -mstack-protector-guard-offset= options. This intended for use in 574 Linux kernel development. Please refer to the documentation for 575 more details. 576 577 BPF 578 579 * Support for CO-RE (compile-once, run-everywhere) has been added to 580 the BPF back end. CO-RE allows to compile portable BPF programs 581 that are able to run among different versions of the Linux kernel. 582 583 IA-32/x86-64 584 585 * New ISA extension support for Intel AVX512-FP16 was added. 586 AVX512-FP16 intrinsics are available via the -mavx512fp16 compiler 587 switch. 588 * For both C and C++ the _Float16 type is supported on x86 systems 589 with SSE2 enabled. Without {-mavx512fp16}, all operations will be 590 emulated in software and float instructions. 591 * Mitigation against straight line speculation (SLS) for function 592 return and indirect jump is supported via 593 -mharden-sls=[none|all|return|indirect-jmp]. 594 * Add CS prefix to call and jmp to indirect thunk with branch target 595 in r8-r15 registers via -mindirect-branch-cs-prefix. 596 * Always use global offset table (GOT) to access external data and 597 function symbols when the new -mno-direct-extern-access 598 command-line option is specified. 599 600 LoongArch 601 602 * Support for the LoongArch architecture instruction set has been 603 added. 604 * The Loongson CPU codename LA464 and LoongArch 64-bit generic CPU 605 codename loongarch64 are supported through the -march= and -mtune= 606 options (GCC identifiers in parentheses). 607 + Loongson LA464 core (la464). 608 + LoongArch 64-bit generic core (loongarch64). 609 610 MIPS 611 612 * The ABI passing arguments containing zero-width fields (for 613 example, C/C++ zero-width bit-fields, GNU C/C++ zero-length arrays, 614 and GNU C empty structs) has changed. Now a zero-width field will 615 not prevent an aligned 64-bit floating-point field next to it from 616 being passed through FPR. This is compatible with LLVM, but 617 incompatible with previous GCC releases. GCC 12 on MIPS will report 618 such incompatibilities as an inform unless -Wno-psabi is used. 619 * The ABI returning values containing C++17 empty bases has changed. 620 Now an empty base will not prevent an aggregate containing only one 621 or two floating-point fields from being returned through FPR. This 622 is compatible with GCC 6 and earlier, but incompatible with GCC 7 623 through 11. GCC 12 on MIPS will report such incompatibilities as an 624 inform unless -Wno-psabi is used. 625 626 NVPTX 627 628 * The -march flag has been added. The -misa flag is now considered an 629 alias of the -march flag. 630 * Support for PTX ISA target architectures sm_53, sm_70, sm_75 and 631 sm_80 has been added. These can be specified using the -march flag. 632 * The default PTX ISA target architecture has been set back to sm_30, 633 to fix support for sm_30 boards. 634 * The -march-map flag has been added. The -march-map value will be 635 mapped to an valid -march flag value. For instance, 636 -march-map=sm_50 maps to -march=sm_35. This can be used to specify 637 that generated code is to be executed on a board with at least some 638 specific compute capability, without having to know the valid 639 values for the -march flag. 640 * The -mptx flag has been added to specify the PTX ISA version for 641 the generated code; permitted values are 3.1 (matches previous GCC 642 versions), 6.0, 6.3, and 7.0. If not specified, the used version is 643 the minimal version required for -march but at least 6.0. 644 * An mptx-3.1 multilib was added. This allows using older drivers 645 which do not support PTX ISA version 6.0. 646 * The new __PTX_SM__ predefined macro allows code to check the PTX 647 ISA target architecture being targeted by the compiler. 648 * The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__ 649 predefined macros allows code to check the PTX ISA version being 650 targeted by the compiler. 651 652 PowerPC / PowerPC64 / RS6000 653 654 * The internal implementation of Power's target-specific built-in 655 functions has been rewritten to be easier and less error-prone to 656 maintain. Every attempt has been made to ensure that the new 657 behavior matches the old behavior, but inevitably some bugs can be 658 expected. Please [104]report any problems. 659 * The built-in functions __builtin_get_texasr, __builtin_get_texasru, 660 __builtin_get_tfhar, __builtin_get_tfiar, __builtin_set_texasr, 661 __builtin_set_texasru, __builtin_set_tfhar, and __builtin_set_tfiar 662 now behave as documented in all supported configurations. On prior 663 releases, the arguments and return values of these functions were 664 treated as unsigned long long instead of as unsigned long, when the 665 options -m32 -mpowerpc64 were in effect. 666 * The overloaded built-in functions vec_cntlz_lsbb and vec_cnttz_lsbb 667 now behave as documented. On prior releases, these built-in 668 functions had incorrect semantics on little-endian targets. 669 670 PRU 671 672 * The [105]__regio_symbol variable qualifier has been added. It 673 allows easier access in C programs to the __R30 and __R31 CPU I/O 674 registers. 675 676 RISC-V 677 678 * Default ISA spec version was bump to 20191213, more detail see this 679 [106]announcement 680 * New ISA extension support for zba, zbb, zbc, zbs was added. 681 * New ISA extension support for vector and scalar crypto was added, 682 only support architecture testing macro and -march= parsing. 683 * The option -mtune=thead-c906 is added to tune for T-HEAD c906 684 cores. 685 * libstdc++ no longer attempts to detect built-in atomics. 686 Distributions that have out-of-tree workarounds for -latomic should 687 check their ABIs again. 688 689 Operating Systems 690 691 Improvements to Static Analyzer 692 693 * The analyzer has gained a 694 [107]-Wanalyzer-use-of-uninitialized-value warning, similar to 695 [108]-Wuninitialized and [109]-Wmaybe-uninitialized, but based on 696 an interprocedural path-sensitive analysis ([110]PR95006). 697 Such warnings are not disabled by the new 698 [111]-ftrivial-auto-var-init (see below), as the latter is 699 considered a mitigation option. 700 * [112]-Wanalyzer-write-to-const and 701 [113]-Wanalyzer-write-to-string-literal will now check for 702 [114]__attribute__ ((access, ....)) on calls to externally-defined 703 functions, and complain about read-only regions pointed to by 704 arguments marked with a write_only or read_write attribute 705 ([115]PR104793). 706 * The analyzer's "taint" mode, activated by 707 [116]-fanalyzer-checker=taint (in addition to [117]-fanalyzer), has 708 gained four new taint-based warnings: 709 + [118]-Wanalyzer-tainted-allocation-size for e.g. 710 attacker-controlled malloc and alloca, 711 + [119]-Wanalyzer-tainted-divisor for detecting where an 712 attacker can inject a divide-by-zero, 713 + [120]-Wanalyzer-tainted-offset for attacker-controlled pointer 714 offsets, 715 + [121]-Wanalyzer-tainted-size for attacker-controlled values 716 being used as a size parameter to calls to memset or to 717 functions marked with [122]__attribute__ ((access, ....)). 718 The existing [123]-Wanalyzer-tainted-array-index has been reworded 719 to talk about "attacker-controlled" rather than "tainted" values, 720 for consistency with the new warnings. 721 A new [124]__attribute__ ((tainted_args)) has been added to the C 722 and C++ front ends, usable on functions, and on function pointer 723 callback fields in structs. The analyzer's taint mode will treat 724 all parameters and buffers pointed to by parameters of such 725 functions as being attacker-controlled, such as for annotating 726 system calls in an operating system kernel as being an "attack 727 surface". 728 * The analyzer now respects [125]__attribute__((const)): it will 729 treat such functions as returning the same value when given the 730 same inputs ([126]PR104434), and as having no side effects 731 ([127]PR104576). 732 * The analyzer is now able to split its analysis into multiple 733 execution paths in places where there isn't a split in the control 734 flow graph. For example, it now handles realloc calls by splitting 735 the execution path into three possible outcomes for the call: 736 + failure, returning NULL 737 + success, growing the buffer in-place without moving it 738 + success, allocating a new buffer, copying the content of the 739 old buffer to it, and freeing the old buffer 740 * The analyzer's interprocedural path exploration logic is now able 741 to track calls through function pointers. 742 * The analyzer now makes the assumption that if we know PTR is 743 non-NULL, then (PTR + OFFSET) is also non-NULL. This isn't strictly 744 true, but eliminates false positives in practice ([128]PR101962). 745 * The analyzer has gained some initial support for inline assembler 746 code. This is extremely limited, and is purely to help suppress 747 false positives when analyzing the Linux kernel, which makes heavy 748 use of inline assembler ([129]PR101570). 749 * The way the analyzer tracks the state of memory along an execution 750 path has been improved in various ways for GCC 12: 751 + An optimization for representing bulk updates to memory (e.g. 752 zero fills) has been removed as it never worked well. In GCC 753 12 it has been replaced with a simpler and more accurate 754 approach, eliminating many false positives ([130]PR95006). 755 + Various optimizations have been added, speeding up the 756 analysis on a particularly problematic source file from 4 757 minutes down to 17 seconds ([131]PR104943, [132]PR104954, and 758 [133]PR104955). 759 + The analyzer now tracks the sizes of dynamically-allocated 760 regions, both on the heap (via malloc etc) and stack (via 761 alloca), though none of the analyzer warnings make use of this 762 yet in GCC 12. 763 * The analyzer's handling of switch statements has been rewritten, 764 fixing various bugs. 765 766 Other significant improvements 767 768 Eliminating uninitialized variables 769 770 * GCC can now [134]initialize all stack variables implicitly, 771 including padding. This is intended to eliminate all classes of 772 uninitialized stack variable flaws. Lack of explicit initialization 773 will still warn when [135]-Wuninitialized is active. For best 774 debugging, use of the new command-line option 775 [136]-ftrivial-auto-var-init=pattern can be used to fill variables 776 with a repeated 0xFE pattern, which tends to illuminate many bugs 777 (e.g. pointers receive invalid addresses, sizes and indices are 778 very large). For best production results, the new command-line 779 option [137]-ftrivial-auto-var-init=zero can be used to fill 780 variables with 0x00, which tends to provide a safer state for bugs 781 (e.g. pointers are NULL, strings are NUL filled, and sizes and 782 indices are 0). 783 784 Debugging formats 785 786 * GCC can now generate debugging information in [138]CTF, a 787 lightweight debugging format that provides information about C 788 types and the association between functions and data symbols and 789 types. This format is designed to be embedded in ELF files and to 790 be very compact and simple. A new command-line option -gctf enables 791 the generation of CTF. 792 * GCC can now generate debugging information in BTF. This is a 793 debugging format mainly used in BPF programs and the Linux kernel. 794 The compiler can generate BTF for any target, when enabled with the 795 command-line option -gbtf 796 797 GCC 12.1 798 799 This is the [139]list of problem reports (PRs) from GCC's bug tracking 800 system that are known to be fixed in the 12.1 release. This list might 801 not be complete (that is, it is possible that some PRs that have been 802 fixed are not listed here). 803 804 GCC 12.2 805 806 This is the [140]list of problem reports (PRs) from GCC's bug tracking 807 system that are known to be fixed in the 12.2 release. This list might 808 not be complete (that is, it is possible that some PRs that have been 809 fixed are not listed here). 810 811 Target Specific Changes 812 813 LoongArch 814 815 * The default setting of -m[check|no-check]-zero-division is changed 816 for optimized code. Now -mno-check-zero-division is the default for 817 all optimization levels but -O0 and -Og. The old behavior can be 818 obtained by explicitly passing -mcheck-zero-division to GCC. 819 820 GCC 12.3 821 822 Target Specific Changes 823 824 x86-64 825 826 * GCC now supports AMD CPUs based on the znver4 core via 827 -march=znver4. The switch makes GCC consider using 512-bit vectors 828 when auto-vectorizing. 829 830 This is the [141]list of problem reports (PRs) from GCC's bug tracking 831 system that are known to be fixed in the 12.3 release. This list might 832 not be complete (that is, it is possible that some PRs that have been 833 fixed are not listed here). 834 835 GCC 12.4 836 837 This is the [142]list of problem reports (PRs) from GCC's bug tracking 838 system that are known to be fixed in the 12.4 release. This list might 839 not be complete (that is, it is possible that some PRs that have been 840 fixed are not listed here). 841 842 843 For questions related to the use of GCC, please consult these web 844 pages and the [143]GCC manuals. If that fails, the 845 [144]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 846 web pages and the development of GCC are welcome on our developer 847 list at [145]gcc (a] gcc.gnu.org. All of [146]our lists have public 848 archives. 849 850 Copyright (C) [147]Free Software Foundation, Inc. Verbatim copying and 851 distribution of this entire article is permitted in any medium, 852 provided this notice is preserved. 853 854 These pages are [148]maintained by the GCC team. Last modified 855 2024-06-20. 856 857 References 858 859 1. https://gcc.gnu.org/gcc-12/porting_to.html 860 2. https://gcc.gnu.org/onlinedocs/index.html#current 861 3. https://gcc.gnu.org/PR102024 862 4. https://gcc.gnu.org/PR42217 863 5. https://gcc.gnu.org/gcc-12/changes.html#mips_zero_width_fields 864 6. https://gcc.gnu.org/install/specific.html 865 7. https://clang.llvm.org/docs/ShadowCallStack.html 866 8. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Instrumentation-Options.html#index-fsanitize_003dshadow-call-stack 867 9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/OpenMP-Implementation-Status.html 868 10. https://www.openacc.org/ 869 11. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-12 870 12. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/Enabling-OpenACC.html 871 13. https://gcc.gnu.org/gcc-12/changes.html#amdgcn 872 14. https://gcc.gnu.org/gcc-12/changes.html#nvptx 873 15. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-param 874 16. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wopenacc-parallelism 875 17. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload-options 876 18. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload 877 19. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gnat_rm/Security-Hardening-Features.html#Security-Hardening-Features 878 20. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wbidi-chars 879 21. https://gcc.gnu.org/PR103026 880 22. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Warray-compare 881 23. https://gcc.gnu.org/PR97573 882 24. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wattributes 883 25. https://gcc.gnu.org/PR101940 884 26. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wformat 885 27. https://wg21.link/p1938 886 28. https://gcc.gnu.org/PR100974 887 29. https://wg21.link/p0849 888 30. https://gcc.gnu.org/PR103049 889 31. https://wg21.link/p2242 890 32. https://gcc.gnu.org/PR102612 891 33. https://wg21.link/p2334 892 34. https://gcc.gnu.org/PR102616 893 35. https://wg21.link/p2360 894 36. https://gcc.gnu.org/PR102617 895 37. https://wg21.link/p2128 896 38. https://wg21.link/cwg2397 897 39. https://gcc.gnu.org/PR100975 898 40. https://wg21.link/cwg960 899 41. https://wg21.link/cwg1227 900 42. https://wg21.link/cwg1315 901 43. https://wg21.link/cwg2082 902 44. https://wg21.link/cwg2351 903 45. https://wg21.link/cwg2374 904 46. https://wg21.link/cwg2397 905 47. https://wg21.link/cwg2446 906 48. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=87c2080b 907 49. https://gcc.gnu.org/PR79501 908 50. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 909 51. https://gcc.gnu.org/PR19808 910 52. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wint-in-bool-context 911 53. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a2b12bc 912 54. https://gcc.gnu.org/PR99032 913 55. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmissing-requires 914 56. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=e18e56c7 915 57. https://gcc.gnu.org/PR100995 916 58. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Waddress 917 59. https://gcc.gnu.org/PR102103 918 60. https://gcc.gnu.org/PR99701 919 61. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3ead06c1 920 62. https://gcc.gnu.org/PR102753 921 63. https://wg21.link/p0466 922 64. https://gcc.gnu.org/PR101539 923 65. https://gcc.gnu.org/PR100828 924 66. https://gcc.gnu.org/PR101786 925 67. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Winterference-size 926 68. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=76b75018 927 69. https://gcc.gnu.org/PR101344 928 70. https://gcc.gnu.org/PR101883 929 71. https://gcc.gnu.org/PR89062 930 72. https://gcc.gnu.org/PR101233 931 73. https://gcc.gnu.org/PR88252 932 74. https://gcc.gnu.org/PR86439 933 75. https://gcc.gnu.org/PR98832 934 76. https://gcc.gnu.org/PR102933 935 77. https://gcc.gnu.org/PR51577 936 78. https://gcc.gnu.org/PR98486 937 79. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=1595fe44 938 80. https://dlang.org/changelog/2.100.1.html 939 81. https://gcc.gnu.org/PR101127 940 82. https://gcc.gnu.org/PR106563 941 83. https://dlang.org/spec/importc.html 942 84. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-16 943 85. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_tls_model 944 86. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-17 945 87. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_link_section 946 88. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Variable-Attributes.html#index-section-variable-attribute 947 89. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-18 948 90. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-19 949 91. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-20 950 92. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_bitcast 951 93. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-21 952 94. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_register_name 953 95. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-22 954 96. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/contexts.html#c.gcc_jit_context_set_bool_print_errors_to_stderr 955 97. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-23 956 98. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-24 957 99. https://gcc.gnu.org/PR96066 958 100. https://gcc.gnu.org/PR96067 959 101. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_cast 960 102. https://gcc.gnu.org/PR95498 961 103. https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/arm-a-profile-architecture-developments-2020 962 104. https://gcc.gnu.org/bugs/ 963 105. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Named-Address-Spaces.html#PRU-Named-Address-Spaces 964 106. https://groups.google.com/a/groups.riscv.org/g/sw-dev/c/aE1ZeHHCYf4 965 107. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-use-of-uninitialized-value 966 108. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 967 109. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 968 110. https://gcc.gnu.org/PR95006 969 111. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 970 112. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 971 113. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 972 114. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 973 115. https://gcc.gnu.org/PR104793 974 116. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker 975 117. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer 976 118. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size 977 119. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor 978 120. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset 979 121. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size 980 122. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html 981 123. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index 982 124. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-tainted_005fargs-function-attribute 983 125. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-const-function-attribute 984 126. https://gcc.gnu.org/PR104434 985 127. https://gcc.gnu.org/PR104576 986 128. https://gcc.gnu.org/PR101962 987 129. https://gcc.gnu.org/PR101570 988 130. https://gcc.gnu.org/PR95006 989 131. https://gcc.gnu.org/PR104943 990 132. https://gcc.gnu.org/PR104954 991 133. https://gcc.gnu.org/PR104955 992 134. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 993 135. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized 994 136. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 995 137. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init 996 138. https://ctfstd.org/ 997 139. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.0 998 140. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.2 999 141. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.3 1000 142. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.4 1001 143. https://gcc.gnu.org/onlinedocs/ 1002 144. mailto:gcc-help (a] gcc.gnu.org 1003 145. mailto:gcc (a] gcc.gnu.org 1004 146. https://gcc.gnu.org/lists.html 1005 147. https://www.fsf.org/ 1006 148. https://gcc.gnu.org/about.html 1007 ====================================================================== 1008 http://gcc.gnu.org/gcc-11/index.html 1009 GCC 11 Release Series 1010 1011 May 29, 2023 1012 1013 The GCC developers are pleased to announce the release of GCC 11.4. 1014 1015 This release is a bug-fix release, containing fixes for regressions in 1016 GCC 11.3 relative to previous releases of GCC. 1017 1018 Release History 1019 1020 GCC 11.4 1021 May 29, 2023 ([1]changes, [2]documentation) 1022 1023 GCC 11.3 1024 April 21, 2022 ([3]changes, [4]documentation) 1025 1026 GCC 11.2 1027 July 28, 2021 ([5]changes, [6]documentation) 1028 1029 GCC 11.1 1030 April 27, 2021 ([7]changes, [8]documentation) 1031 1032 References and Acknowledgements 1033 1034 GCC used to stand for the GNU C Compiler, but since the compiler 1035 supports several other languages aside from C, it now stands for the 1036 GNU Compiler Collection. 1037 1038 The GCC developers would like to thank the numerous people that have 1039 contributed new features, improvements, bug fixes, and other changes as 1040 well as test results to GCC. This [9]amazing group of volunteers is 1041 what makes GCC successful. 1042 1043 For additional information about GCC please refer to the [10]GCC 1044 project web site or contact the [11]GCC development mailing list. 1045 1046 To obtain GCC please use [12]our mirror sites or [13]our version 1047 control system. 1048 1049 1050 For questions related to the use of GCC, please consult these web 1051 pages and the [14]GCC manuals. If that fails, the 1052 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1053 web pages and the development of GCC are welcome on our developer 1054 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 1055 archives. 1056 1057 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 1058 distribution of this entire article is permitted in any medium, 1059 provided this notice is preserved. 1060 1061 These pages are [19]maintained by the GCC team. Last modified 1062 2024-05-30. 1063 1064 References 1065 1066 1. https://gcc.gnu.org/gcc-11/changes.html 1067 2. https://gcc.gnu.org/onlinedocs/11.4.0/ 1068 3. https://gcc.gnu.org/gcc-11/changes.html 1069 4. https://gcc.gnu.org/onlinedocs/11.3.0/ 1070 5. https://gcc.gnu.org/gcc-11/changes.html 1071 6. https://gcc.gnu.org/onlinedocs/11.2.0/ 1072 7. https://gcc.gnu.org/gcc-11/changes.html 1073 8. https://gcc.gnu.org/onlinedocs/11.1.0/ 1074 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html 1075 10. https://gcc.gnu.org/index.html 1076 11. mailto:gcc (a] gcc.gnu.org 1077 12. https://gcc.gnu.org/mirrors.html 1078 13. https://gcc.gnu.org/git.html 1079 14. https://gcc.gnu.org/onlinedocs/ 1080 15. mailto:gcc-help (a] gcc.gnu.org 1081 16. mailto:gcc (a] gcc.gnu.org 1082 17. https://gcc.gnu.org/lists.html 1083 18. https://www.fsf.org/ 1084 19. https://gcc.gnu.org/about.html 1085 ====================================================================== 1086 http://gcc.gnu.org/gcc-11/changes.html 1087 GCC 11 Release Series 1088 Changes, New Features, and Fixes 1089 1090 This page is a "brief" summary of some of the huge number of 1091 improvements in GCC 11. You may also want to check out our [1]Porting 1092 to GCC 11 page and the [2]full GCC documentation. 1093 1094 Caveats 1095 1096 * The default mode for C++ is now -std=gnu++17 instead of 1097 -std=gnu++14. Note that [3]C++17 changes to template template 1098 parameter matching can be disabled independently of other features 1099 with -fno-new-ttp-matching. 1100 * When building GCC itself, the host compiler must now support C++11, 1101 rather than C++98. In particular bootstrapping GCC 11 using an 1102 older version of GCC requires a binary of GCC 4.8 or later, rather 1103 than of GCC 3.4 or later as was the case for bootstrapping GCC 10. 1104 * Naming and location of auxiliary and dump output files changed. If 1105 you compile multiple input files in a single command, if you enable 1106 Link Time Optimization, or if you use -dumpbase, -dumpdir, 1107 -save-temps=*, and you expect any file other than the primary 1108 output file(s) to be created as a side effect, watch out for 1109 improvements and a few surprises. See [4]the patch, particularly 1110 its textual description, for more details about the changes. 1111 * -gsplit-dwarf no longer enables debug info generation on its own 1112 but requires a separate -g for this. 1113 * The libstdc++ configure option --enable-cheaders=c_std is 1114 deprecated and will be removed in a future release. It should be 1115 possible to use --enable-cheaders=c_global (the default) with no 1116 change in behaviour. 1117 * The front end for compiling BRIG format of Heterogeneous System 1118 Architecture Intermediate Language (HSAIL) has been deprecated and 1119 will likely be removed in a future release. 1120 * Some short options of the gcov tool have been renamed: -i to -j and 1121 -j to -H. 1122 1123 General Improvements 1124 1125 * [5]ThreadSanitizer improvements to support alternative runtimes and 1126 environments. The [6]Linux Kernel Concurrency Sanitizer (KCSAN) is 1127 now supported. 1128 + Add --param tsan-distinguish-volatile to optionally emit 1129 instrumentation distinguishing volatile accesses. 1130 + Add --param tsan-instrument-func-entry-exit to optionally 1131 control if function entries and exits should be instrumented. 1132 * In previous releases of GCC, the "column numbers" emitted in 1133 diagnostics were actually a count of bytes from the start of the 1134 source line. This could be problematic, both because of: 1135 + multibyte characters (requiring more than one byte to encode), 1136 and 1137 + multicolumn characters (requiring more than one column to 1138 display in a monospace font) 1139 For example, the character ("GREEK SMALL LETTER PI (U+03C0)") 1140 occupies one column, and its UTF-8 encoding requires two bytes; the 1141 character ("SLIGHTLY SMILING FACE (U+1F642)") occupies two 1142 columns, and its UTF-8 encoding requires four bytes. 1143 In GCC 11 the column numbers default to being column numbers, 1144 respecting multi-column characters. The old behavior can be 1145 restored using a new option [7]-fdiagnostics-column-unit=byte. 1146 There is also a new option [8]-fdiagnostics-column-origin=, 1147 allowing the pre-existing default of the left-hand column being 1148 column 1 to be overridden if desired (e.g. for 0-based columns). 1149 The output of [9]-fdiagnostics-format=json has been extended to 1150 supply both byte counts and column numbers for all source 1151 locations. 1152 Additionally, in previous releases of GCC, tab characters in the 1153 source would be emitted verbatim when quoting source code, but be 1154 prefixed with whitespace or line number information, leading to 1155 misalignments in the resulting output when compared with the actual 1156 source. Tab characters are now printed as an appropriate number of 1157 spaces, using the [10]-ftabstop option (which defaults to 8 spaces 1158 per tab stop). 1159 * Introduce [11]Hardware-assisted AddressSanitizer support. This 1160 sanitizer currently only works for the AArch64 target. It helps 1161 debug address problems similarly to [12]AddressSanitizer but is 1162 based on partial hardware assistance and provides probabilistic 1163 protection to use less RAM at run time. [13]Hardware-assisted 1164 AddressSanitizer is not production-ready for user space, and is 1165 provided mainly for use compiling the Linux Kernel. 1166 To use this sanitizer the command line arguments are: 1167 + -fsanitize=hwaddress to instrument userspace code. 1168 + -fsanitize=kernel-hwaddress to instrument kernel code. 1169 * For targets that produce DWARF debugging information GCC now 1170 defaults to [14]DWARF version 5 (with the exception of VxWorks and 1171 Darwin/Mac OS X which default to version 2 and AIX which defaults 1172 to version 4). This can produce up to 25% more compact debug 1173 information compared to earlier versions. 1174 To take full advantage of DWARF version 5 GCC needs to be built 1175 against binutils version 2.35.2 or higher. When GCC is built 1176 against earlier versions of binutils GCC will still emit DWARF 1177 version 5 for most debuginfo data, but will generate version 4 1178 debug line tables (even when explicitly given -gdwarf-5). 1179 The following debug information consumers can process DWARF version 1180 5: 1181 + GDB 8.0, or higher 1182 + valgrind 3.17.0 1183 + elfutils 0.172, or higher (for use with systemtap, 1184 dwarves/pahole, perf and libabigail) 1185 + dwz 0.14 1186 Programs embedding libbacktrace are urged to upgrade to the version 1187 shipping with GCC 11. 1188 To make GCC 11 generate an older DWARF version use -g together with 1189 -gdwarf-2, -gdwarf-3 or -gdwarf-4. 1190 * Vectorizer improvements: 1191 + The straight-line code vectorizer now considers the whole 1192 function when vectorizing and can handle opportunities 1193 crossing CFG merges and backedges. 1194 * A series of conditional expressions that compare the same variable 1195 can be transformed into a switch statement if each of them contains 1196 a comparison expression. Example: 1197 int IsHTMLWhitespace(int aChar) { 1198 return aChar == 0x0009 || aChar == 0x000A || 1199 aChar == 0x000C || aChar == 0x000D || 1200 aChar == 0x0020; 1201 } 1202 1203 This statement can be transformed into a switch statement and then 1204 expanded into a bit-test. 1205 * New command-line options: 1206 + [15]-fbit-tests, enabled by default, can be used to enable or 1207 disable switch expansion using bit-tests. 1208 * Inter-procedural optimization improvements: 1209 + A new IPA-modref pass was added to track side effects of 1210 function calls and improve precision of points-to-analysis. 1211 The pass can be controlled by the [16]-fipa-modref option. 1212 + The identical code folding pass (controlled by [17]-fipa-icf) 1213 was significantly improved to increase the number of unified 1214 functions and to reduce compile-time memory use. 1215 + IPA-CP (Interprocedural constant propagation) heuristics 1216 improved its estimation of potential usefulness of known loop 1217 bounds and strides by taking the estimated frequency of these 1218 loops into account. 1219 * Link-time optimization improvements: 1220 + The LTO bytecode format was optimized for smaller object files 1221 and faster streaming. 1222 + Memory allocation of the linking stage was improved to reduce 1223 peak memory use. 1224 * Profile driven optimization improvements: 1225 + Using [18]-fprofile-values, was improved by tracking more 1226 target values for e.g. indirect calls. 1227 + GCOV data file format outputs smaller files by representing 1228 zero counters in a more compact way. 1229 1230 New Languages and Language specific improvements 1231 1232 * GCC 11 adds support for non-rectangular loop nests in OpenMP 1233 constructs and the allocator routines of [19]OpenMP 5.0, including 1234 initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD 1235 environment variable and the active-levels routines are now 1236 supported. For C/C++, the declare variant and map support has been 1237 extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP 1238 5.0 support has been extended, including the following features 1239 which were before only available in C and C++: order(concurrent), 1240 device_type, memorder-clauses for flush, lastprivate with 1241 conditional modifier, atomic construct and reduction clause 1242 extensions of OpenMP 5.0, if clause with simd and cancel modifiers, 1243 target data without map clause, and limited support for the 1244 requires construct. 1245 * Version 2.6 of the [20]OpenACC specification continues to be 1246 maintained and improved in the C, C++ and Fortran compilers. See 1247 the [21]implementation status section on the OpenACC wiki page and 1248 the [22]run-time library documentation for further information. 1249 1250 C family 1251 1252 * New attributes: 1253 + The [23]no_stack_protector attribute has been added to mark 1254 functions which should not be instrumented with stack 1255 protection (-fstack-protector). 1256 + The existing [24]malloc attribute has been extended so that it 1257 can be used to identify allocator/deallocator API pairs. A 1258 pair of new [25]-Wmismatched-dealloc and 1259 [26]-Wmismatched-new-delete warnings will complain about 1260 mismatched calls, and [27]-Wfree-nonheap-object about 1261 deallocation calls with pointers not obtained from allocation 1262 functions. Additionally, the static analyzer will use these 1263 attributes when checking for leaks, double-frees, 1264 use-after-frees, and similar issues. 1265 * New warnings: 1266 + [28]-Wmismatched-dealloc, enabled by default, warns about 1267 calls to deallocation functions with pointers returned from 1268 mismatched allocation functions. 1269 + [29]-Wsizeof-array-div, enabled by -Wall, warns about 1270 divisions of two sizeof operators when the first one is 1271 applied to an array and the divisor does not equal the size of 1272 the array element. 1273 + [30]-Wstringop-overread, enabled by default, warns about calls 1274 to string functions reading past the end of the arrays passed 1275 to them as arguments. In prior GCC releases most instances of 1276 his warning are diagnosed by -Wstringop-overflow. 1277 + [31]-Wtsan, enabled by default, warns about unsupported 1278 features in ThreadSanitizer (currently 1279 std::atomic_thread_fence). 1280 * Enhancements to existing warnings: 1281 + [32]-Wfree-nonheap-object detects many more instances of calls 1282 to deallocation functions with pointers not returned from a 1283 dynamic memory allocation function. 1284 + [33]-Wmaybe-uninitialized diagnoses passing pointers or 1285 references to uninitialized memory to functions taking 1286 const-qualified arguments. 1287 + [34]-Wuninitialized detects reads from uninitialized 1288 dynamically allocated memory. 1289 * For ELF targets that support the GNU or FreeBSD OSABIs, the used 1290 attribute will now save the symbol declaration it is applied to 1291 from linker garbage collection. 1292 To support this behavior, used symbols that have not been placed in 1293 specific sections (e.g. with the section attribute, or the 1294 -f{function,data}-sections options) will be placed in new, unique 1295 sections. 1296 This functionality requires Binutils version 2.36 or later. 1297 1298 C 1299 1300 * Several new features from the upcoming C2X revision of the ISO C 1301 standard are supported with -std=c2x and -std=gnu2x. Some of these 1302 features are also supported as extensions when compiling for older 1303 language versions. In addition to the features listed, some 1304 features previously supported as extensions and now added to the C 1305 standard are enabled by default in C2X mode and not diagnosed with 1306 -std=c2x -Wpedantic. 1307 + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>. 1308 + As in C++, function definitions no longer need to give names 1309 for unused function parameters. 1310 + The expansions of the true and false macros in <stdbool.h> 1311 have changed so that they have type bool. 1312 + The [[nodiscard]] standard attribute is now supported. 1313 + The __has_c_attribute preprocessor operator is now supported. 1314 + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN, 1315 DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros 1316 for _FloatN, _FloatNx and _DecimalN types, are provided in 1317 <float.h>. There are also corresponding built-in functions 1318 __builtin_nansdN for decimal signaling NaNs. 1319 + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and 1320 LDBL_IS_IEC_60559 are provided in <float.h>. 1321 + The feature test macro __STDC_WANT_IEC_60559_EXT__ is 1322 supported by <float.h>. 1323 + Labels may appear before declarations and at the end of a 1324 compound statement. 1325 * New warnings: 1326 + [35]-Warray-parameter, enabled by -Wall, warns about 1327 redeclarations of functions with ordinary array arguments 1328 declared using inconsistent forms. The warning also enables 1329 the detection of the likely out of bounds accesses in calls to 1330 such functions with smaller arrays. 1331 + [36]-Wvla-parameter, enabled by -Wall, warns redeclarations of 1332 functions with variable length array arguments declared using 1333 inconsistent forms or with mismatched bounds. The warning also 1334 enables the detection of the likely out of bounds accesses in 1335 calls to such functions with smaller arrays. 1336 1337 C++ 1338 1339 * The default mode has been changed to -std=gnu++17. 1340 * Several C++20 features have been implemented: 1341 + the compiler now supports consteval virtual functions 1342 + P2082R1, Fixing CTAD for aggregates 1343 + P0593R6, Pseudo-destructors end object lifetimes 1344 + P1907R1, Inconsistencies with non-type template parameters 1345 (complete implementation) 1346 + P1975R0, Fixing the wording of parenthesized 1347 aggregate-initialization 1348 + P1009R2, Array size deduction in new-expressions 1349 + P1099R5, using enum 1350 + Modules, Requires -fmodules-ts and some aspects are 1351 incomplete. Refer to [37]C++ 20 Status 1352 * The C++ front end has experimental support for some of the upcoming 1353 C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b 1354 or -std=gnu++2b flags, including 1355 + P0330R8, Literal Suffix for (signed) size_t. 1356 For a full list of new features, see [38]the C++ status page. 1357 * Several C++ Defect Reports have been resolved, e.g.: 1358 + DR 625, Use of auto as a template-argument 1359 + DR 1512, Pointer comparison vs qualification conversions 1360 + DR 1722, Should lambda to function pointer conversion function 1361 be noexcept? 1362 + DR 1914, Duplicate standard attributes 1363 + DR 2032, Default template-arguments of variable templates 1364 + DR 2289, Uniqueness of decomposition declaration names 1365 + DR 2237, Can a template-id name a constructor? 1366 + DR 2303, Partial ordering and recursive variadic inheritance 1367 + DR 2369, Ordering between constraints and substitution 1368 + DR 2450, braced-init-list as a template-argument 1369 * G++ now performs better access checking in templates ([39]PR41437). 1370 * reinterpret_casts in constexpr evaluation are now checked more 1371 completely ([40]PR95307). 1372 * The diagnostic for static_assert has been improved: the compiler 1373 now shows the expression including its template arguments (if there 1374 were any), and can point to the failing clause if the condition 1375 comprised of any logical AND operators ([41]PR97518). 1376 * New warnings: 1377 + [42]-Wctad-maybe-unsupported, disabled by default, warns about 1378 performing class template argument deduction on a type with no 1379 deduction guides. 1380 + [43]-Wrange-loop-construct, enabled by -Wall, warns when a 1381 range-based for-loop is creating unnecessary and expensive 1382 copies. 1383 + [44]-Wdeprecated-enum-enum-conversion, enabled by default in 1384 C++20, warns about deprecated arithmetic conversions on 1385 operands of enumeration types, as outlined in 1386 [depr.arith.conv.enum]. 1387 + [45]-Wdeprecated-enum-float-conversion, enabled by default in 1388 C++20, warns about deprecated arithmetic conversions on 1389 operands where one is of enumeration type and the other is of 1390 a floating-point type, as outlined in [depr.arith.conv.enum]. 1391 + [46]-Wmismatched-new-delete, enabled by -Wall, warns about 1392 calls to C++ operator delete with pointers returned from 1393 mismatched forms of operator new or from other mismatched 1394 allocation functions. 1395 + [47]-Wvexing-parse, enabled by default, warns about the most 1396 vexing parse rule: the cases when a declaration looks like a 1397 variable definition, but the C++ language requires it to be 1398 interpreted as a function declaration. 1399 * Enhancements to existing warnings: 1400 + [48]-Wnonnull considers the implicit this argument of every 1401 C++ nonstatic member function to have been implicitly declared 1402 with attribute nonnull and triggers warnings for calls where 1403 the pointer is null. 1404 1405 Runtime Library (libstdc++) 1406 1407 * Improved C++17 support, including: 1408 + std::from_chars and std::to_chars for floating-point types. 1409 * Improved experimental C++20 support, including: 1410 + Calendar additions to <chrono>. Thanks to Cassio Neri for 1411 optimizations. 1412 + std::bit_cast 1413 + std::source_location 1414 + Atomic wait and notify operations. 1415 + <barrier>, <latch>, and <semaphore> 1416 + <syncstream> 1417 + Efficient access to basic_stringbuf's buffer. 1418 + Heterogeneous lookup in unordered containers. 1419 * Experimental C++23 support, including: 1420 + contains member functions for strings, thanks to Paul Fee. 1421 + std::to_underlying, std::is_scoped_enum 1422 * Experimental support for Data-Parallel Types (simd) from the 1423 Parallelism 2 TS, thanks to Matthias Kretz. 1424 * Faster std::uniform_int_distribution, thanks to Daniel Lemire. 1425 1426 D 1427 1428 * New features: 1429 + A new bottom type typeof(*null) has been added to represent 1430 run-time errors and non-terminating functions. This also 1431 introduces a new standard alias for the type named noreturn, 1432 and is implicitly imported into every module. 1433 + Printf-like and scanf-like functions are now detected by 1434 prefixing them with pragma(printf) for printf-like functions 1435 or pragma(scanf) for scanf-like functions. 1436 + The __traits() expression now supports the extensions 1437 isDeprecated, isDisabled, isFuture, isModule, isPackage, 1438 child, isReturnOnStack, isZeroInit, getTargetInfo, 1439 getLocation, hasPostblit, isCopyable, getVisibility, and 1440 totype. 1441 + An expression-based contract syntax has been added to the 1442 language. 1443 + Function literals can now return a value by reference with the 1444 ref keyword. 1445 + A new syntax is available to declare aliases to function types 1446 using the alias syntax based on the assignment operator. 1447 + New types __c_complex_float, __c_complex_double, 1448 __c_complex_real, and __c_wchar_t have been added for 1449 interfacing with C and C++ code, and are available from the 1450 core.stdc.config module. 1451 + User-defined attributes can now be used to annotate enum 1452 members, alias declarations, and function parameters. 1453 + Template alias parameters can now be instantiated with basic 1454 types such as int or void function(). 1455 + The mixin construct can now be used as types in the form 1456 mixin(string) var. 1457 + The mixin construct can now take an argument list, same as 1458 pragma(msg). 1459 * New intrinsics: 1460 + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror 1461 have been added. 1462 + Byte swap intrinsic core.bitop.byteswap for swapping bytes in 1463 a 2-byte ushort has been added. 1464 + Math intrinsics available from core.math now have overloads 1465 for float and double types. 1466 + Volatile intrinsics core.volatile.volatileLoad and 1467 core.volatile.volatileStore have been moved from the 1468 core.bitop module. 1469 * New attributes: 1470 + The following GCC attributes are now recognized and available 1471 from the gcc.attributes module with short-hand aliases for 1472 convenience: 1473 o @attribute("alloc_size", arguments) or 1474 @alloc_size(arguments). 1475 o @attribute("always_inline") or @always_inline. 1476 o @attribute("used") or @used. 1477 o @attribute("optimize", arguments) or 1478 @optimize(arguments). 1479 o @attribute("cold") or @cold. 1480 o @attribute("noplt") or @noplt. 1481 o @attribute("target_clones", arguments) or 1482 @target_clones(arguments). 1483 o @attribute("no_icf") or @no_icf. 1484 o @attribute("noipa") or @noipa. 1485 o @attribute("symver", arguments) or @symver(arguments). 1486 + New aliases have been added to gcc.attributes for 1487 compatibility with ldc.attributes. 1488 o The @allocSize(arguments) attribute is the same as 1489 @alloc_size(arguments), but uses a 0-based index for 1490 function arguments. 1491 o The @assumeUsed attribute is an alias for 1492 @attribute("used"). 1493 o The @fastmath attribute is an alias for 1494 @optimize("Ofast"). 1495 o The @naked attribute is an alias for @attribute("naked"). 1496 This attribute may not be available on all targets. 1497 o The @restrict attribute has been added to specify that a 1498 function parameter is to be restrict-qualified in the C99 1499 sense of the term. 1500 o The @optStrategy(strategy) attribute is an alias for 1501 @optimize("O0") when the strategy is "none", otherwise 1502 @optimize("Os") for the "optsize" and "minsize" 1503 strategies. 1504 o The @polly attribute is an alias for 1505 @optimize("loop-parallelize-all"). 1506 o The @section(name) attribute is an alias for 1507 @attribute("section", name). 1508 o The @target(arguments) attribute is an alias for 1509 attribute("target", arguments). 1510 o The @weak attribute is an alias for @attribute("weak"). 1511 * New language options: 1512 + -fweak-templates, added to control whether declarations that 1513 can be defined in multiple objects should be emitted as weak 1514 symbols. The default is to emit all symbols with extern 1515 linkage as weak, unless the target lacks support for weak 1516 symbols. 1517 + -Wdeprecated, this option is now enabled by default. 1518 + -Wextra, this option now turns on all warnings that are not 1519 part of the core D language front-end - -Waddress, 1520 -Wcast-result, -Wunknown-pragmas. 1521 + -Wvarargs, added to turn on warnings about questionable usage 1522 of the va_start intrinsic. 1523 * Deprecated and removed features: 1524 + Compiler-recognized attributes are now made available from the 1525 gcc.attributes module, the former module gcc.attribute has 1526 been deprecated and will be removed in a future release. 1527 + The @attribute("alias") attribute has been removed, as it had 1528 been superseded by pragma(mangle). 1529 + The @attribute("forceinline") attribute has been removed and 1530 renamed to @always_inline. 1531 + __vector types that are not supported in hardware are now 1532 rejected at compile-time. Previously all vector types were 1533 accepted by the compiler and emulated when target support was 1534 absent. 1535 + The extern(Pascal) linkage attribute has been removed. 1536 + The deprecation phase for -ftransition=import and 1537 -ftransition=checkimports is finished. These switches no 1538 longer have an effect and are now removed. Symbols that are 1539 not visible in a particular scope will no longer be found by 1540 the compiler. 1541 + It is now an error to use private variables selectively 1542 imported from other modules. Due to a bug, some imported 1543 private members were visible from other modules, violating the 1544 specification. 1545 + The -fweak compiler switch has been removed, as it existed 1546 only for testing. 1547 1548 Fortran 1549 1550 * Added DEPRECATED to !GCC$'s attributes directive. 1551 1552 Go 1553 1554 * GCC 11 provides a complete implementation of the Go 1.16.3 user 1555 packages. 1556 1557 libgccjit 1558 1559 * libgccjit was marked as merely "Alpha" quality when [49]originally 1560 added in GCC 5. Given that we have maintained [50]API and ABI 1561 compatibility since then and it is in use by various projects, we 1562 have removed that caveat. 1563 * libgccjit can now be built for MinGW 1564 * The libgccjit API gained 10 new entry points: 1565 + [51]gcc_jit_global_set_initializer 1566 + 9 entrypoints for [52]directly embedding asm statements into a 1567 compile, analogous to inline asm in the C front end 1568 1569 New Targets and Target Specific Improvements 1570 1571 AArch64 & arm 1572 1573 * A number of new CPUs are supported through arguments to the -mcpu 1574 and -mtune options in both the arm and aarch64 back ends (GCC 1575 identifiers in parentheses): 1576 + Arm Cortex-A78 (cortex-a78). 1577 + Arm Cortex-A78AE (cortex-a78ae). 1578 + Arm Cortex-A78C (cortex-a78c). 1579 + Arm Cortex-X1 (cortex-x1). 1580 + Arm Neoverse V1 (neoverse-v1). 1581 + Arm Neoverse N2 (neoverse-n2). 1582 * GCC can now auto-vectorize operations performing addition, 1583 subtraction, multiplication and the accumulate/subtract variants on 1584 complex numbers, taking advantage of the Advanced SIMD instructions 1585 in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64) 1586 and MVE (AArch32 M-profile) instruction sets. 1587 1588 AArch64 1589 1590 * In addition to the above, the following AArch64-only CPUs are now 1591 supported: 1592 + Fujitsu A64FX (a64fx). 1593 + Arm Cortex-R82 (cortex-r82). 1594 * The AArch64 Armv8-R architecture is now supported through the 1595 -march=armv8-r option. 1596 * Mitigation against the [53]Straight-line Speculation vulnerability 1597 is supported with the -mharden-sls= option. Please refer to the 1598 documentation for usage instructions. 1599 * The availability of Advanced SIMD intrinsics available through the 1600 arm_neon.h header is improved and GCC 11 supports the full set of 1601 intrinsics defined by ACLE Q3 2020. 1602 1603 AMD Radeon (GCN) 1604 1605 * Initial support for gfx908 GPUs has been added. 1606 1607 arm 1608 1609 * Initial auto-vectorization is now available when targeting the MVE 1610 instruction set. 1611 * GCC can now make use of the Low Overhead Branch instruction in 1612 Armv8.1-M to optimize loop counters and checks. 1613 * The -mcpu=cortex-m55 option now supports the extensions +nomve and 1614 +nomve.fp to control generation of MVE and MVE floating-point 1615 instructions. 1616 1617 IA-32/x86-64 1618 1619 * New ISA extension support for Intel TSXLDTRK was added to GCC. 1620 TSXLDTRK intrinsics are available via the -mtsxldtrk compiler 1621 switch. 1622 * New ISA extension support for Intel SERIALIZE was added to GCC. 1623 SERIALIZE intrinsics are available via the -mserialize compiler 1624 switch. 1625 * New ISA extension support for Intel HRESET was added to GCC. HRESET 1626 intrinsics are available via the -mhreset compiler switch. 1627 * New ISA extension support for Intel UINTR was added to GCC. UINTR 1628 intrinsics are available via the -muintr compiler switch. 1629 * New ISA extension support for Intel KEYLOCKER was added to GCC. 1630 KEYLOCKER intrinsics are available via the -mkeylocker compiler 1631 switch. 1632 * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16 1633 was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are 1634 available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler 1635 switches. 1636 * New ISA extension support for Intel AVX-VNNI was added to GCC. 1637 AVX-VNNI intrinsics are available via the -mavxvnni compiler 1638 switch. 1639 * GCC now supports the Intel CPU named Sapphire Rapids through 1640 -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B, 1641 AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG, 1642 TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA 1643 extensions. 1644 * GCC now supports the Intel CPU named Alderlake through 1645 -march=alderlake. The switch enables the CLDEMOTE, PTWRITE, 1646 WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions. 1647 * GCC now supports the Intel CPU named Rocketlake through 1648 -march=rocketlake. Rocket Lake is based on Icelake client and minus 1649 SGX. 1650 * GCC now supports AMD CPUs based on the znver3 core via 1651 -march=znver3. 1652 * GCC now supports micro-architecture levels defined in the x86-64 1653 psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4. 1654 1655 Nios II 1656 1657 * The options -mcustom-insn=N no longer produce compiler warnings if 1658 the custom instruction is not generated due to missing optimization 1659 options such as -fno-math-errno, -ffinite-math-only, or 1660 -funsafe-math-optimizations. These warnings were not consistently 1661 emitted for all custom instructions. 1662 * The -mcustom-fpu-cfg=fph2 option has been added to enable the 1663 custom instructions supported by the Nios II Floating Point 1664 Hardware 2 Component. 1665 1666 NVPTX 1667 1668 * The -misa default has changed from sm_30 to sm_35. 1669 * The -m32 compiler switch has been removed. 1670 * The -msoft-stack-reserve-local format has been fixed. Previously, 1671 it accepted -msoft-stack-reserve-local<n>. It now accepts 1672 -msoft-stack-reserve-local=<n>. 1673 1674 S/390, System z, IBM Z Systems 1675 1676 * The behavior when compiling with -fexcess-precision=standard (e.g., 1677 implied by -std=c99) on s390(x) targets can now be controlled at 1678 configure time with the flag --enable-s390-excess-float-precision. 1679 When enabled, GCC will maintain previous behavior and evaluate 1680 float expressions in double precision, which aligns with the 1681 definition of float_t as double. With the flag disabled, GCC will 1682 always evaluate float expressions in single precision. In native 1683 builds and cross compiles that have target libc headers, GCC will 1684 by default match the definition of float_t in the installed glibc. 1685 1686 RISC-V 1687 1688 * Support address sanitizer for RISC-V. 1689 * Support big-endian for RISC-V, thanks to Marcus Comstedt. 1690 * Implement new style of architecture extension test macros: each 1691 architecture extension has a corresponding feature test macro, 1692 which can be used to test its existence and version information. 1693 * Legacy architecture extension test macros like __riscv_atomic are 1694 deprecated, but will still be supported for at least 2 release 1695 cycles. 1696 * Support IFUNC for riscv*-*-linux*. 1697 * Add new option -misa-spec=* to control ISA spec version. This 1698 controls the default version of each extensions. It defaults to 1699 2.2. 1700 * Introduce the --with-multilib-generator configure time option. This 1701 allows for flexible config multi-lib settings. Its syntax is the 1702 same as RISC-V's multilib-generator. 1703 * Extend the sytax for multilib-generator, support expansion operator 1704 * to reduce the complexity of complicated multi-lib re-use rules. 1705 * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the 1706 pipeline model and architecture extensions, like -mtune=* plus 1707 -march=*. 1708 * Support for TLS stack protector canary access, thanks to Cooper Qu. 1709 * Support __builtin_thread_pointer for RISC-V. 1710 * Introduce shorten_memrefs optimization, which reduces the code size 1711 for memory access, thanks to Craig Blackmore. 1712 1713 Operating Systems 1714 1715 AIX 1716 1717 * GCC for AIX can be built as a 64-bit application and the runtime is 1718 built as FAT libraries containing both 32-bit and 64-bit objects. 1719 * Support AIX Vector Extended ABI with -mabi=vec-extabi. 1720 * Thread-Local uninitiated data placed in local common section. 1721 * Use thread-safe access in ctype. 1722 * Link with libc128.a when long-double-128 enabled. 1723 1724 Improvements to Static Analyzer 1725 1726 * The implementation of how program state is tracked within 1727 [54]-fanalyzer has been completely rewritten for GCC 11, fixing 1728 numerous bugs, and allowing for the analyzer to scale to larger C 1729 source files. 1730 * The analysis of allocations and deallocations has been generalized 1731 beyond malloc and free. 1732 + As preliminary work towards eventually supporting C++, the 1733 malloc/free checking will also check new/delete and 1734 new[]/delete[]. However, C++ is not yet properly supported by 1735 [55]-fanalyzer (for example, exception-handling is 1736 unimplemented). 1737 + As noted above, the existing [56]malloc attribute has been 1738 extended so that it can be used to identify 1739 allocator/deallocator API pairs. The analyzer will use these 1740 attributes when checking for leaks, double-frees, 1741 use-after-frees, and similar issues. 1742 + A new [57]-Wanalyzer-mismatching-deallocation warning has been 1743 added, covering such mismatches as using scalar delete rather 1744 vector delete[]. 1745 * The analyzer has gained warnings 1746 [58]-Wanalyzer-shift-count-negative, 1747 [59]-Wanalyzer-shift-count-overflow, [60]-Wanalyzer-write-to-const, 1748 and [61]-Wanalyzer-write-to-string-literal, all enabled by default 1749 when [62]-fanalyzer is enabled. 1750 * The analyzer can now be extended by GCC plugins, allowing for 1751 domain-specific path-sensitive warnings. An example of using a 1752 [63]GCC plugin to check for misuses of CPython's global interpreter 1753 lock can be seen in the test suite 1754 * The analyzer has gained new debugging options 1755 [64]-fdump-analyzer-json and [65]-fno-analyzer-feasibility. 1756 1757 Other significant improvements 1758 1759 * GCC has gained a new environment variable 1760 [66]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to 1761 request machine-readable fix-it hints without needing to adjust 1762 build flags. 1763 1764 GCC 11.1 1765 1766 This is the [67]list of problem reports (PRs) from GCC's bug tracking 1767 system that are known to be fixed in the 11.1 release. This list might 1768 not be complete (that is, it is possible that some PRs that have been 1769 fixed are not listed here). 1770 1771 GCC 11.2 1772 1773 This is the [68]list of problem reports (PRs) from GCC's bug tracking 1774 system that are known to be fixed in the 11.2 release. This list might 1775 not be complete (that is, it is possible that some PRs that have been 1776 fixed are not listed here). 1777 1778 GCC 11.3 1779 1780 This is the [69]list of problem reports (PRs) from GCC's bug tracking 1781 system that are known to be fixed in the 11.3 release. This list might 1782 not be complete (that is, it is possible that some PRs that have been 1783 fixed are not listed here). 1784 1785 GCC 11.4 1786 1787 This is the [70]list of problem reports (PRs) from GCC's bug tracking 1788 system that are known to be fixed in the 11.4 release. This list might 1789 not be complete (that is, it is possible that some PRs that have been 1790 fixed are not listed here). 1791 1792 Target Specific Changes 1793 1794 x86-64 1795 1796 * The x86-64 ABI of passing and returning structure with a 64-bit 1797 single precision vector changed in GCC 11.1 when 3DNOW is disabled. 1798 Disabling 3DNOW no longer changes how they are passed nor returned. 1799 This ABI change is now diagnosed with -Wpsabi. 1800 * Mitigation against straight line speculation (SLS) for function 1801 return and indirect jump is supported via 1802 -mharden-sls=[none|all|return|indirect-jmp]. 1803 * Add CS prefix to call and jmp to indirect thunk with branch target 1804 in r8-r15 registers via -mindirect-branch-cs-prefix. 1805 1806 1807 For questions related to the use of GCC, please consult these web 1808 pages and the [71]GCC manuals. If that fails, the 1809 [72]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1810 web pages and the development of GCC are welcome on our developer 1811 list at [73]gcc (a] gcc.gnu.org. All of [74]our lists have public 1812 archives. 1813 1814 Copyright (C) [75]Free Software Foundation, Inc. Verbatim copying and 1815 distribution of this entire article is permitted in any medium, 1816 provided this notice is preserved. 1817 1818 These pages are [76]maintained by the GCC team. Last modified 1819 2024-06-14. 1820 1821 References 1822 1823 1. https://gcc.gnu.org/gcc-11/porting_to.html 1824 2. https://gcc.gnu.org/onlinedocs/index.html#current 1825 3. https://wg21.link/p0522r0 1826 4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html 1827 5. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual 1828 6. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html 1829 7. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit 1830 8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin 1831 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1832 10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop 1833 11. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 1834 12. https://github.com/google/sanitizers/wiki/AddressSanitizer 1835 13. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 1836 14. https://dwarfstd.org/doc/DWARF5.pdf 1837 15. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests 1838 16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref 1839 17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf 1840 18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values 1841 19. https://www.openmp.org/specifications/ 1842 20. https://www.openacc.org/ 1843 21. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-11 1844 22. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html 1845 23. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute 1846 24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 1847 25. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 1848 26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 1849 27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 1850 28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 1851 29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div 1852 30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread 1853 31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan 1854 32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 1855 33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 1856 34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized 1857 35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter 1858 36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter 1859 37. https://gcc.gnu.org/projects/cxx-status.html#cxx20 1860 38. https://gcc.gnu.org/projects/cxx-status.html#cxx23 1861 39. https://gcc.gnu.org/PR41437 1862 40. https://gcc.gnu.org/PR95307 1863 41. https://gcc.gnu.org/PR97518 1864 42. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported 1865 43. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct 1866 44. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion 1867 45. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion 1868 46. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 1869 47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse 1870 48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull 1871 49. https://gcc.gnu.org/gcc-5/changes.html#jit 1872 50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html 1873 51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer 1874 52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html 1875 53. https://developer.arm.com/documentation/102587/0102/Straight-line-speculation-frequently-asked-questions 1876 54. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1877 55. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1878 56. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 1879 57. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation 1880 58. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative 1881 59. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow 1882 60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 1883 61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 1884 62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 1885 63. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825 1886 64. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json 1887 65. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility 1888 66. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT 1889 67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0 1890 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2 1891 69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.3 1892 70. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.4 1893 71. https://gcc.gnu.org/onlinedocs/ 1894 72. mailto:gcc-help (a] gcc.gnu.org 1895 73. mailto:gcc (a] gcc.gnu.org 1896 74. https://gcc.gnu.org/lists.html 1897 75. https://www.fsf.org/ 1898 76. https://gcc.gnu.org/about.html 1899 ====================================================================== 1900 http://gcc.gnu.org/gcc-10/index.html 1901 GCC 10 Release Series 1902 1903 (This release series is no longer supported.) 1904 1905 July, 7, 2023 1906 1907 The [1]GNU project and the GCC developers are pleased to announce the 1908 release of GCC 10.5. 1909 1910 This release is a bug-fix release, containing fixes for regressions in 1911 GCC 10.4 relative to previous releases of GCC. 1912 1913 Release History 1914 1915 GCC 10.5 1916 July 7, 2023 ([2]changes, [3]documentation) 1917 1918 GCC 10.4 1919 June 28, 2022 ([4]changes, [5]documentation) 1920 1921 GCC 10.3 1922 April 8, 2021 ([6]changes, [7]documentation) 1923 1924 GCC 10.2 1925 July 23, 2020 ([8]changes, [9]documentation) 1926 1927 GCC 10.1 1928 May 7, 2020 ([10]changes, [11]documentation) 1929 1930 References and Acknowledgements 1931 1932 GCC used to stand for the GNU C Compiler, but since the compiler 1933 supports several other languages aside from C, it now stands for the 1934 GNU Compiler Collection. 1935 1936 The GCC developers would like to thank the numerous people that have 1937 contributed new features, improvements, bug fixes, and other changes as 1938 well as test results to GCC. This [12]amazing group of volunteers is 1939 what makes GCC successful. 1940 1941 For additional information about GCC please refer to the [13]GCC 1942 project web site or contact the [14]GCC development mailing list. 1943 1944 To obtain GCC please use [15]our mirror sites or [16]our version 1945 control system. 1946 1947 1948 For questions related to the use of GCC, please consult these web 1949 pages and the [17]GCC manuals. If that fails, the 1950 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1951 web pages and the development of GCC are welcome on our developer 1952 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 1953 archives. 1954 1955 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 1956 distribution of this entire article is permitted in any medium, 1957 provided this notice is preserved. 1958 1959 These pages are [22]maintained by the GCC team. Last modified 1960 2024-05-30. 1961 1962 References 1963 1964 1. http://www.gnu.org/ 1965 2. https://gcc.gnu.org/gcc-10/changes.html 1966 3. https://gcc.gnu.org/onlinedocs/10.5.0/ 1967 4. https://gcc.gnu.org/gcc-10/changes.html 1968 5. https://gcc.gnu.org/onlinedocs/10.4.0/ 1969 6. https://gcc.gnu.org/gcc-10/changes.html 1970 7. https://gcc.gnu.org/onlinedocs/10.3.0/ 1971 8. https://gcc.gnu.org/gcc-10/changes.html 1972 9. https://gcc.gnu.org/onlinedocs/10.2.0/ 1973 10. https://gcc.gnu.org/gcc-10/changes.html 1974 11. https://gcc.gnu.org/onlinedocs/10.1.0/ 1975 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html 1976 13. https://gcc.gnu.org/index.html 1977 14. mailto:gcc (a] gcc.gnu.org 1978 15. https://gcc.gnu.org/mirrors.html 1979 16. https://gcc.gnu.org/git.html 1980 17. https://gcc.gnu.org/onlinedocs/ 1981 18. mailto:gcc-help (a] gcc.gnu.org 1982 19. mailto:gcc (a] gcc.gnu.org 1983 20. https://gcc.gnu.org/lists.html 1984 21. https://www.fsf.org/ 1985 22. https://gcc.gnu.org/about.html 1986 ====================================================================== 1987 http://gcc.gnu.org/gcc-10/changes.html 1988 GCC 10 Release Series 1989 Changes, New Features, and Fixes 1990 1991 This page is a "brief" summary of some of the huge number of 1992 improvements in GCC 10. You may also want to check out our [1]Porting 1993 to GCC 10 page and the [2]full GCC documentation. 1994 1995 Caveats 1996 1997 * An ABI incompatibility between C++14 and C++17 has been fixed. On 1998 some targets a class with a zero-sized subobject would be passed 1999 incorrectly when compiled as C++17 or C++20. See the [3]C++ notes 2000 below for more details. 2001 * The deprecated Profile Mode and array_allocator extensions have 2002 been removed from libstdc++. 2003 * The non-standard std::__is_nullptr_t type trait is deprecated and 2004 will be removed from libstdc++ in a future release. The standard 2005 trait std::is_null_pointer should be instead. 2006 * The minimum version of the [4]MPFR library required for building 2007 GCC has been increased to version 3.1.0 (released 2011-10-03). 2008 * The automatic template instantiation at link time (-frepo) has been 2009 removed. 2010 * The --param allow-store-data-races internal parameter has been 2011 removed in favor of a new official option -fallow-store-data-races. 2012 While default behavior is unchanged and the new option allows to 2013 correctly maintain a per compilation unit setting across link-time 2014 optimization, alteration of the default via --param 2015 allow-store-data-races will now be diagnosed and build systems have 2016 to be adjusted accordingly. 2017 * Offloading to Heterogeneous System Architecture Intermediate 2018 Language (HSAIL) has been deprecated and will likely be removed in 2019 a future release. 2020 * The type of the std::iterator base class of 2021 std::istreambuf_iterator was changed in C++98 mode to be consistent 2022 with C++11 and later standards. See the [5]libstdc++ notes below 2023 for more details. 2024 2025 General Improvements 2026 2027 * New built-in functions: 2028 + The [6]__has_builtin built-in preprocessor operator can be 2029 used to query support for built-in functions provided by GCC 2030 and other compilers that support it. 2031 + __builtin_roundeven for the corresponding function from 2032 ISO/IEC TS 18661. 2033 * New command-line options: 2034 + [7]-fallocation-dce removes unneeded pairs of new and delete 2035 operators. 2036 + [8]-fprofile-partial-training can now be used to inform the 2037 compiler that code paths not covered by the training run 2038 should not be optimized for size. 2039 + [9]-fprofile-reproducible controls level of reproducibility of 2040 profile gathered by [10]-fprofile-generate. This makes it 2041 possible to rebuild program with same outcome which is useful, 2042 for example, for distribution packages. 2043 + [11]-fprofile-prefix-path can be used in combination with 2044 -fprofile-generate=profile_dir and -fprofile-use=profile_dir 2045 to inform GCC where the base directory of build source tree is 2046 in case it differs between instrumentation and optimized 2047 builds. 2048 + [12]-fanalyzer enables a new static analysis pass and 2049 associated warnings. This pass performs a time-consuming 2050 exploration of paths through the code in the hope of detecting 2051 various common errors, such as double-free bugs. This option 2052 should be regarded as experimental in this release. In 2053 particular, analysis of non-C code is unlikely to work. 2054 * Inter-procedural optimization improvements: 2055 + The inter-procedural scalar replacement of aggregates 2056 (IPA-SRA) pass was re-implemented to work at link-time and can 2057 now also remove computing and returning unused return values. 2058 + [13]-finline-functions is now enabled at -O2 and was retuned 2059 for better code size versus runtime performance trade-offs. 2060 Inliner heuristics was also significantly sped up to avoid 2061 negative impact to -flto -O2 compile times. 2062 + Inliner heuristics and function cloning can now use 2063 value-range information to predict effectivity of individual 2064 transformations. 2065 + During link-time optimization the C++ One Definition Rule is 2066 used to increase precision of type based alias analysis. 2067 * Link-time optimization improvements: 2068 + A new binary [14]lto-dump has been added. It dumps various 2069 information about LTO bytecode object files. 2070 + The parallel phase of the LTO can automatically detect a 2071 running make's jobserver or fall back to number of available 2072 cores. 2073 + The LTO bytecode can be compressed with the [15]zstd 2074 algorithm. The configure script automatically detects zstd 2075 support. 2076 + Most --param values can now be specified at translation unit 2077 granularity. This includes all parameters controlling the 2078 inliner and other inter-procedural optimizations. Unlike 2079 earlier releases, GCC 10 will ignore parameters controlling 2080 optimizations specified at link-time and apply parameters 2081 specified at compile-time in the same manner as done for 2082 optimization flags. 2083 * Profile driven optimization improvements: 2084 + Profile maintenance during compilation and hot/cold code 2085 partitioning have been improved. 2086 + Using [16]-fprofile-values, an instrumented binary can track 2087 multiple values (up to 4) for e.g. indirect calls and provide 2088 more precise profile information. 2089 2090 New Languages and Language-Specific Improvements 2091 2092 * Version 2.6 of the [17]OpenACC specification is now supported by 2093 the C, C++ and Fortran compilers. See the [18]implementation status 2094 section on the OpenACC wiki page and the [19]run-time library 2095 documentation for further information. 2096 * GCC 10 adds a number of newly implemented [20]OpenMP 5.0 features 2097 such as conditional lastprivate clause, scan and loop directives, 2098 order(concurrent) and use_device_addr clauses support, if clause on 2099 simd construct, and partial support for the declare variant 2100 directive, getting closer to full support of the OpenMP 5.0 2101 standard. 2102 * OpenMP and OpenACC now support [21]offloading to AMD Radeon (GCN) 2103 GPUs; supported are the third-generation Fiji (fiji) and the 2104 fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906). 2105 2106 C family 2107 2108 * New attributes: 2109 + The access function and type attribute has been added to 2110 describe how a function accesses objects passed to it by 2111 pointer or reference, and to associate such arguments with 2112 integer arguments denoting the objects' sizes. The attribute 2113 is used to enable the detection of invalid accesses by 2114 user-defined functions, such as those diagnosed by 2115 -Wstringop-overflow. 2116 + The symver attribute can be used to bind symbols to specific 2117 version nodes on ELF platforms. This is preferred to using 2118 inline assembly with GNU as symver directive because the 2119 latter is not compatible with link-time optimizations. 2120 * New warnings: 2121 + [22]-Wstring-compare, enabled by -Wextra, warns about equality 2122 and inequality expressions between zero and the result of a 2123 call to either strcmp and strncmp that evaluate to a constant 2124 as a result of the length of one argument being greater than 2125 the size of the array pointed to by the other. 2126 + [23]-Wzero-length-bounds, enabled by -Warray-bounds, warns 2127 about accesses to elements of zero-length arrays that might 2128 overlap other members of the same object. 2129 * Enhancements to existing warnings: 2130 + [24]-Warray-bounds detects more out-of-bounds accesses to 2131 member arrays as well as accesses to elements of zero-length 2132 arrays. 2133 + [25]-Wformat-overflow makes full use of string length 2134 information computed by the strlen optimization pass. 2135 + [26]-Wrestrict detects overlapping accesses to dynamically 2136 allocated objects. 2137 + [27]-Wreturn-local-addr diagnoses more instances of return 2138 statements returning addresses of automatic variables. 2139 + [28]-Wstringop-overflow detects more out-of-bounds stores to 2140 member arrays including zero-length arrays, dynamically 2141 allocated objects and variable length arrays, as well as more 2142 instances of reads of unterminated character arrays by string 2143 built-in functions. The warning also detects out-of-bounds 2144 accesses by calls to user-defined functions declared with the 2145 new attribute access. 2146 + [29]-Warith-conversion re-enables warnings from -Wconversion, 2147 -Wfloat-conversion, and -Wsign-conversion that are now off by 2148 default for an expression where the result of an arithmetic 2149 operation will not fit in the target type due to promotion, 2150 but the operands of the expression do fit in the target type. 2151 * Extended characters in identifiers may now be specified directly in 2152 the input encoding (UTF-8, by default), in addition to the UCN 2153 syntax (\uNNNN or \UNNNNNNNN) that is already supported: 2154 2155 static const int = 3; 2156 int get_nave_pi() { 2157 return ; 2158 } 2159 2160 C 2161 2162 * Several new features from the upcoming C2X revision of the ISO C 2163 standard are supported with -std=c2x and -std=gnu2x. Some of these 2164 features are also supported as extensions when compiling for older 2165 language versions. In addition to the features listed, some 2166 features previously supported as extensions and now added to the C 2167 standard are enabled by default in C2X mode and not diagnosed with 2168 -std=c2x -Wpedantic. 2169 + The [[]] attribute syntax is supported, as in C++. Existing 2170 attributes can be used with this syntax in forms such as 2171 [[gnu::const]]. The standard attributes [[deprecated]], 2172 [[fallthrough]] and [[maybe_unused]] are supported. 2173 + UTF-8 character constants using the u8'' syntax are supported. 2174 + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and 2175 LDBL_NORM_MAX. 2176 + When decimal floating-point arithmetic is supported, <float.h> 2177 defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and 2178 DEC128_TRUE_MIN, in addition to the macros that were 2179 previously only defined if __STDC_WANT_DEC_FP__ was defined 2180 before including <float.h>. 2181 + In C2X mode, empty parentheses in a function definition give 2182 that function a type with a prototype for subsequent calls; 2183 other old-style function definitions are diagnosed by default 2184 in C2X mode. 2185 + The strftime format checking supports the %OB and %Ob formats. 2186 + In C2X mode, -fno-fp-int-builtin-inexact is enabled by 2187 default. 2188 * GCC now defaults to -fno-common. As a result, global variable 2189 accesses are more efficient on various targets. In C, global 2190 variables with multiple tentative definitions now result in linker 2191 errors. With -fcommon such definitions are silently merged during 2192 linking. 2193 2194 C++ 2195 2196 * Several C++20 features have been implemented: 2197 + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, 2198 P0848R3, P1616R1, P1452R2 2199 + P1668R1, Permitting Unevaluated inline-assembly in constexpr 2200 Functions 2201 + P1161R3, Deprecate a[b,c] 2202 + P0848R3, Conditionally Trivial Special Member Functions 2203 + P1091R3, Extending structured bindings 2204 + P1143R2, Adding the constinit keyword 2205 + P1152R4, Deprecating volatile 2206 + P0388R4, Permit conversions to arrays of unknown bound 2207 + P0784R7, constexpr new 2208 + P1301R4, [[nodiscard("with reason")]] 2209 + P1814R0, class template argument deduction for alias templates 2210 + P1816R0, class template argument deduction for aggregates 2211 + P0960R3, Parenthesized initialization of aggregates 2212 + P1331R2, Allow trivial default initialization in constexpr 2213 contexts 2214 + P1327R1, Allowing dynamic_cast and polymorphic typeid in 2215 constexpr contexts 2216 + P0912R5, Coroutines (requires -fcoroutines) 2217 * Several C++ Defect Reports have been resolved, e.g.: 2218 + DR 1560, lvalue-to-rvalue conversion in ?: 2219 + DR 1813, __is_standard_layout for a class with repeated bases 2220 + DR 2094, volatile scalars are trivially copyable, 2221 + DR 2096, constraints on literal unions 2222 + DR 2413, typename in conversion-function-ids 2223 + DR 2352, Similar types and reference binding 2224 + DR 1601, Promotion of enumeration with fixed underlying type 2225 + DR 330, Qualification conversions and pointers to arrays of 2226 pointers 2227 + DR 1307, Overload resolution based on size of array 2228 initializer-list 2229 + DR 1710, Missing template keyword in class-or-decltype 2230 * New warnings: 2231 + [30]-Wmismatched-tags, disabled by default, warns about 2232 declarations of structs, classes, and class templates and 2233 their specializations with a class-key that does not match 2234 either the definition or the first declaration if no 2235 definition is provided. The option is provided to ease 2236 portability to Windows-based compilers. 2237 + [31]-Wredundant-tags, disabled by default, warns about 2238 redundant class-key and enum-key in contexts where the key can 2239 be eliminated without causing an syntactic ambiguity. 2240 * G++ can now detect modifying constant objects in constexpr 2241 evaluation (which is undefined behavior). 2242 * G++ no longer emits bogus -Wsign-conversion warnings with explicit 2243 casts. 2244 * Narrowing is now detected in more contexts (e.g., case values). 2245 * Memory consumption of the compiler has been reduced in constexpr 2246 evaluation. 2247 * The noexcept-specifier is now properly treated as a complete-class 2248 context as per [class.mem]. 2249 * The attribute deprecated can now be used on namespaces too. 2250 * The ABI of passing and returning certain C++ classes by value 2251 changed on several targets in GCC 10, including [32]AArch64, 2252 [33]ARM, [34]PowerPC ELFv2, [35]S/390 and [36]Itanium. These 2253 changes affect classes with a zero-sized subobject (an empty base 2254 class, or data member with the [[no_unique_address]] attribute) 2255 where all other non-static data members have the same type (this is 2256 called a "homogeneous aggregate" in some ABI specifications, or if 2257 there is only one such member, a "single element"). In -std=c++17 2258 and -std=c++20 modes, classes with an empty base class were not 2259 considered to have a single element or to be a homogeneous 2260 aggregate, and so could be passed differently (in the wrong 2261 registers or at the wrong stack address). This could make code 2262 compiled with -std=c++17 and -std=c++14 ABI incompatible. This has 2263 been corrected and the empty bases are ignored in those ABI 2264 decisions, so functions compiled with -std=c++14 and -std=c++17 are 2265 now ABI compatible again. Example: struct empty {}; struct S : 2266 empty { float f; }; void f(S);. Similarly, in classes containing 2267 non-static data members with empty class types using the C++20 2268 [[no_unique_address]] attribute, those members weren't ignored in 2269 the ABI argument passing decisions as they should be. Both of these 2270 ABI changes are now diagnosed with -Wpsabi. 2271 2272 Runtime Library (libstdc++) 2273 2274 * Improved experimental C++2a support, including: 2275 + Library concepts in <concepts> and <iterator>. 2276 + Constrained algorithms in <ranges>, <algorithm>, and <memory> 2277 (thanks to Patrick Palka). 2278 + New algorithms shift_left and shift_right (thanks to Patrick 2279 Palka). 2280 + std::span (thanks to JeanHeyd Meneide). 2281 + Three-way comparisons in <compare> and throughout the library. 2282 + Constexpr support in <algorithm> and elsewhere (thanks to 2283 Edward Smith-Rowland). 2284 + <stop_token> and std::jthread (thanks to Thomas Rodgers). 2285 + std::atomic_ref and std::atomic<floating point>. 2286 + Integer comparison functions (cmp_equal, cmp_less etc.). 2287 + std::ssize, std::to_array. 2288 + std::construct_at, std::destroy, constexpr std::allocator. 2289 + Mathematical constants in <numbers>. 2290 * Support for RDSEED in std::random_device. 2291 * Reduced header dependencies, leading to faster compilation for some 2292 code. 2293 * The std::iterator base class of std::istreambuf_iterator was 2294 changed in C++98 mode to be consistent with C++11 and later 2295 standards. This is expected to have no noticeable effect except in 2296 the unlikely case of a class which has potentially overlapping 2297 subobjects of type std::istreambuf_iterator<C> and another iterator 2298 type with a std::iterator<input_iterator_tag, C, ...> base class. 2299 The layout of such a type might change when compiled as C++98. 2300 [37]Bug 92285 has more details and concrete examples. 2301 2302 D 2303 2304 * Support for static foreach has been implemented. 2305 * Aliases can now be created directly from any __traits that return 2306 symbols or tuples. Previously, an AliasSeq was necessary in order 2307 to alias their return. 2308 * It is now possible to detect the language ABI specified for a 2309 struct, class, or interface using __traits(getLinkage, ...). 2310 * Support for core.math.toPrec intrinsics has been added. These 2311 intrinsics guarantee the rounding to specific floating-point 2312 precisions at specified points in the code. 2313 * Support for pragma(inline) has been implemented. Previously the 2314 pragma was recognized, but had no effect on the compilation. 2315 * Optional parentheses in asm operands are deprecated and will be 2316 removed in a future release. 2317 * All content imported files are now included in the make dependency 2318 list when compiling with -M. 2319 * Compiler recognized attributes provided by the gcc.attribute module 2320 will now take effect when applied to function prototypes as well as 2321 when applied to full function declarations. 2322 * Added a --enable-libphobos-checking configure option to control 2323 whether run-time checks are compiled into the D runtime library. 2324 * Added a --with-libphobos-druntime-only configure option to indicate 2325 whether to build only the core D runtime library, or both the core 2326 and standard libraries into libphobos. 2327 2328 Fortran 2329 2330 * use_device_addr of version 5.0 of the [38]OpenMP specification is 2331 now supported. Note that otherwise OpenMP 4.5 is partially 2332 supported in the Fortran compiler; the largest missing item is 2333 structure element mapping. 2334 * The default buffer size for I/O using unformatted files has been 2335 increased to 1048576. The buffer size for can now be set at runtime 2336 via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and 2337 GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted 2338 files, respectively. 2339 * Mismatches between actual and dummy argument lists in a single file 2340 are now rejected with an error. Use the new option 2341 -fallow-argument-mismatch to turn these errors into warnings; this 2342 option is implied with -std=legacy. -Wargument-mismatch has been 2343 removed. 2344 * The handling of a BOZ literal constant has been reworked to provide 2345 better conformance to the Fortran 2008 and 2018 standards. In these 2346 Fortran standards, a BOZ literal constant is a typeless and 2347 kindless entity. As a part of the rework, documented and 2348 undocumented extensions to the Fortran standard now emit errors 2349 during compilation. Some of these extensions are permitted with the 2350 -fallow-invalid-boz option, which degrades the error to a warning 2351 and the code is compiled as with older gfortran. 2352 * At any optimization level except-Os, gfortran now uses inline 2353 packing for arguments instead of calling a library routine. If the 2354 source contains a large number of arguments that need to be 2355 repacked, code size or time for compilation can become excessive. 2356 If that is the case, -fno-inline-arg-packing can be used to disable 2357 inline argument packing. 2358 * Legacy extensions: 2359 + For formatted input/output, if the explicit widths after the 2360 data-edit descriptors I, F and G have been omitted, default 2361 widths are used. 2362 + A blank format item at the end of a format specification, i.e. 2363 nothing following the final comma, is allowed. Use the option 2364 -fdec-blank-format-item; this option is implied with -fdec. 2365 + The existing support for AUTOMATIC and STATIC attributes has 2366 been extended to allow variables with the AUTOMATIC attribute 2367 to be used in EQUIVALENCE statements. Use -fdec-static; this 2368 option is implied by -fdec. 2369 + Allow character literals in assignments and DATA statements 2370 for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. 2371 Use the option -fdec-char-conversions; this option is implied 2372 with -fdec. 2373 + DEC comparisons, i.e. allow Hollerith constants to be used in 2374 comparisons with INTEGER, REAL, COMPLEX and CHARACTER 2375 expressions. Use the option -fdec. 2376 * Character type names in errors and warnings now include len in 2377 addition to kind; * is used for assumed length. The kind is omitted 2378 if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4). 2379 * CO_BROADCAST now supports derived type variables including objects 2380 with allocatable components. In this case, the optional arguments 2381 STAT= and ERRMSG= are currently ignored. 2382 * The handling of module and submodule names has been reworked to 2383 allow the full 63-character length mandated by the standard. 2384 Previously symbol names were truncated if the combined length of 2385 module, submodule, and function name exceeded 126 characters. This 2386 change therefore breaks the ABI, but only for cases where this 126 2387 character limit was exceeded. 2388 2389 Go 2390 2391 * GCC 10 provides a complete implementation of the Go 1.14.6 user 2392 packages. 2393 2394 libgccjit 2395 2396 * The libgccjit API gained four new entry points: 2397 + [39]gcc_jit_version_major, [40]gcc_jit_version_minor, and 2398 [41]gcc_jit_version_patchlevel for programmatically checking 2399 the libgccjit version from client code, and 2400 + [42]gcc_jit_context_new_bitfield 2401 2402 New Targets and Target Specific Improvements 2403 2404 AArch64 & arm 2405 2406 * The AArch64 and arm ports now support condition flag output 2407 constraints in inline assembly, as indicated by the 2408 __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for 2409 A32 and T32 targets. Please refer to the documentation for more 2410 details. 2411 2412 AArch64 2413 2414 * There have been several improvements related to the Scalable Vector 2415 Extension (SVE): 2416 + The SVE ACLE types and intrinsics are now supported. They can 2417 be accessed using the header file arm_sve.h. 2418 + It is now possible to create fixed-length SVE types using the 2419 arm_sve_vector_bits attribute. For example: 2420 #if __ARM_FEATURE_SVE_BITS==512 2421 typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512))); 2422 typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512))); 2423 #endif 2424 + -mlow-precision-div, -mlow-precision-sqrt and 2425 -mlow-precision-recip-sqrt now work for SVE. 2426 + -msve-vector-bits=128 now generates vector-length-specific 2427 code for little-endian targets. It continues to generate 2428 vector-length-agnostic code for big-endian targets, just as 2429 previous releases did for all targets. 2430 + The vectorizer is now able to use extending loads and 2431 truncating stores, including gather loads and scatter stores. 2432 + The vectorizer now compares the cost of vectorizing with SVE 2433 and vectorizing with Advanced SIMD and tries to pick the best 2434 one. Previously it would always use SVE if possible. 2435 + If a vector loop uses Advanced SIMD rather than SVE, the 2436 vectorizer now considers using SVE to vectorize the left-over 2437 elements (the scalar tail or epilog). 2438 + Besides these specific points, there have been many general 2439 improvements to the way that the vectorizer uses SVE. 2440 * The -mbranch-protection=pac-ret option now accepts the optional 2441 argument +b-key extension to perform return address signing with 2442 the B-key instead of the A-key. 2443 * The option -moutline-atomics has been added to aid deployment of 2444 the Large System Extensions (LSE) on GNU/Linux systems built with a 2445 baseline architecture targeting Armv8-A. When the option is 2446 specified code is emitted to detect the presence of LSE 2447 instructions at runtime and use them for standard atomic 2448 operations. For more information please refer to the documentation. 2449 * The Transactional Memory Extension is now supported through ACLE 2450 intrinsics. It can be enabled through the +tme option extension 2451 (for example, -march=armv8.5-a+tme). 2452 * A number of features from Armv8.5-A are now supported through ACLE 2453 intrinsics. These include: 2454 + The random number instructions that can be enabled through the 2455 (already present in GCC 9.1) +rng option extension. 2456 + Floating-point intrinsics to round to integer instructions 2457 from Armv8.5-A when targeting -march=armv8.5-a or later. 2458 + Memory Tagging Extension intrinsics enabled through the 2459 +memtag option extension. 2460 * Similarly, the following Armv8.6-A features are now supported 2461 through ACLE intrinsics: 2462 + The bfloat16 extension. This extension is enabled 2463 automatically when Armv8.6-A is selected (such as by 2464 -march=armv8.6-a). It can also be enabled for Armv8.2-A and 2465 later using the +bf16 option extension. 2466 + The Matrix Multiply extension. This extension is split into 2467 three parts, one for each supported data type: 2468 o Support for 8-bit integer matrix multiply instructions. 2469 This extension is enabled automatically when Armv8.6-A is 2470 selected. It can also be enabled for Armv8.2-A and later 2471 using the +i8mm option extension. 2472 o Support for 32-bit floating-point matrix multiply 2473 instructions. This extension can be enabled using the 2474 +f32mm option extension, which also has the effect of 2475 enabling SVE. 2476 o Support for 64-bit floating-point matrix multiply 2477 instructions. This extension can be enabled using the 2478 +f64mm option extension, which likewise has the effect of 2479 enabling SVE. 2480 * SVE2 is now supported through ACLE intrinsics and (to a limited 2481 extent) through autovectorization. It can be enabled through the 2482 +sve2 option extension (for example, -march=armv8.5-a+sve2). 2483 Additional extensions can be enabled through +sve2-sm4, +sve2-aes, 2484 +sve2-sha3 and +sve2-bitperm. 2485 * Support has been added for the following processors (GCC 2486 identifiers in parentheses): 2487 + Arm Cortex-A77 (cortex-a77). 2488 + Arm Cortex-A76AE (cortex-a76ae). 2489 + Arm Cortex-A65 (cortex-a65). 2490 + Arm Cortex-A65AE (cortex-a65ae). 2491 + Arm Cortex-A34 (cortex-a34). 2492 + Marvell ThunderX3 (thunderx3t110). 2493 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2494 options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as 2495 arguments to the equivalent target attributes and pragmas. 2496 2497 arm 2498 2499 * Support for the FDPIC ABI has been added. It uses 64-bit function 2500 descriptors to represent pointers to functions, and enables code 2501 sharing on MMU-less systems. The corresponding target triple is 2502 arm-uclinuxfdpiceabi, and the C library is uclibc-ng. 2503 * Support has been added for the Arm EABI on NetBSD through the 2504 arm*-*-netbsdelf-*eabi* triplet. 2505 * The handling of 64-bit integer operations has been significantly 2506 reworked and improved leading to improved performance and reduced 2507 stack usage when using 64-bit integral data types. The option 2508 -mneon-for-64bits is now deprecated and will be removed in a future 2509 release. 2510 * Support has been added for the following processors (GCC 2511 identifiers in parentheses): 2512 + Arm Cortex-A77 (cortex-a77). 2513 + Arm Cortex-A76AE (cortex-a76ae). 2514 + Arm Cortex-M35P (cortex-m35p). 2515 + Arm Cortex-M55 (cortex-m55). 2516 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2517 options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p. 2518 * Support has been extended for the ACLE [43]data-processing 2519 intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit 2520 multiplication and other related intrinsics aimed at DSP algorithm 2521 optimization. 2522 * Support for -mpure-code in Thumb-1 (v6m) has been added: this 2523 M-profile feature is no longer restricted to targets with MOVT. For 2524 example, -mcpu=cortex-m0 now supports this option. 2525 * Support for the [44]Armv8.1-M Mainline Architecture has been added. 2526 + Armv8.1-M Mainline can be enabled by using the 2527 -march=armv8.1-m.main command-line option. 2528 * Support for the [45]MVE beta ACLE intrinsics has been added. These 2529 intrinsics can be enabled by including the arm_mve.h header file 2530 and passing the +mve or +mve.fp option extensions (for example: 2531 -march=armv8.1-m.main+mve). 2532 * Support for the Custom Datapath Extension beta ACLE [46]intrinsics 2533 has been added. 2534 * Support for Armv8.1-M Mainline Security Extensions architecture has 2535 been added. The -mcmse option, when used in combination with an 2536 Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main 2537 -mcmse), now leads to the generation of improved code sequences 2538 when changing security states. 2539 2540 AMD Radeon (GCN) 2541 2542 * Code generation and in particular vectorization support have been 2543 much improved. 2544 2545 ARC 2546 2547 * The interrupt service routine functions save all used registers, 2548 including extension registers and auxiliary registers used by Zero 2549 Overhead Loops. 2550 * Improve code size by using multiple short instructions instead of a 2551 single long mov or ior instruction when its long immediate constant 2552 is known. 2553 * Fix usage of the accumulator register for ARC600. 2554 * Fix issues with uncached attribute. 2555 * Remove -mq-class option. 2556 * Improve 64-bit integer addition and subtraction operations. 2557 2558 AVR 2559 2560 * Support for the XMEGA-like devices 2561 2562 ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, 2563 ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, 2564 ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, 2565 ATmega4808, ATmega4809 2566 has been added. 2567 * A new command-line option -nodevicespecs has been added. It allows 2568 to provide a custom device-specs file by means of 2569 2570 avr-gcc -nodevicespecs -specs=my-spec-file <options> 2571 and without the need to provide options -B and -mmcu=. See [47]AVR 2572 command-line options for details. This feature is also available in 2573 GCC 9.3+ and GCC 8.4+. 2574 * New command-line options -mdouble=[32,64] and -mlong-double=[32,64] 2575 have been added. They allow to choose the size (in bits) of the 2576 double and long double types, respectively. Whether or not the 2577 mentioned layouts are available, whether the options act as a 2578 multilib option, and the default for either option are controlled 2579 by the new [48]AVR configure options --with-double= and 2580 --with-long-double=. 2581 * A new configure option --with-libf7= has been added. It controls to 2582 which level avr-libgcc provides 64-bit floating point support by 2583 means of [49]Libf7. 2584 * A new configure option --with-double-comparison= has been added. 2585 It's unlikely you need to set this option by hand. 2586 2587 IA-32/x86-64 2588 2589 * Support to expand __builtin_roundeven into the appropriate SSE 4.1 2590 instruction has been added. 2591 * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD 2592 intrinsics are available via the -menqcmd compiler switch. 2593 * GCC now supports the Intel CPU named Cooperlake through 2594 -march=cooperlake. The switch enables the AVX512BF16 ISA 2595 extensions. 2596 * GCC now supports the Intel CPU named Tigerlake through 2597 -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B 2598 AVX512VP2INTERSECT ISA extensions. 2599 2600 MIPS 2601 2602 * The mips*-*-linux* targets now mark object files with appropriate 2603 GNU-stack note, facilitating use of non-executable stack hardening 2604 on GNU/Linux. The soft-float targets have this feature enabled by 2605 default, while for hard-float targets it is required for GCC to be 2606 configured with --with-glibc-version=2.31 against glibc 2.31 or 2607 later. 2608 2609 PowerPC / PowerPC64 / RS6000 2610 2611 * Many vector builtins have been listed as deprecated in the 2612 [50]64-Bit ELF V2 ABI Specification for quite a number of years. 2613 The vector builtins listed in Tables A.8 through A.10 are now 2614 deprecated for GCC 10, and will likely be removed from support in 2615 GCC 11. Note that this does not result in any loss of function. 2616 These deprecated builtins generally provide somewhat nonsensical 2617 argument lists (for example, mixing signed, unsigned, and bool 2618 vector arguments arbitrarily), or are duplicate builtins that are 2619 inconsistent with the expected naming scheme. We expect that this 2620 will be unlikely to affect much if any code, and any required code 2621 changes will be trivial. 2622 2623 PRU 2624 2625 * A new back end targeting TI PRU I/O processors has been contributed 2626 to GCC. 2627 2628 RISC-V 2629 2630 * The riscv*-*-* targets now require GNU binutils version 2.30 or 2631 later, to support new assembly instructions produced by GCC. 2632 2633 V850 2634 2635 * The ABI for V850 nested functions has been changed. Previously the 2636 V850 port used %r20 for the static chain pointer, now the port uses 2637 %r19. This corrects a long standing latent bug in the v850 port 2638 where a call to a nested function would unexpectedly change the 2639 value in %r20. 2640 2641 Operating Systems 2642 2643 Improvements for plugin authors 2644 2645 * GCC diagnostics can now have a chain of events associated with 2646 them, describing a path through the code that triggers the problem. 2647 These can be printed by the diagnostics subsystem in various ways, 2648 controlled by the [51]-fdiagnostics-path-format option, or captured 2649 in JSON form via [52]-fdiagnostics-format=json. 2650 * GCC diagnostics can now be associated with [53]CWE weakness 2651 identifiers, which will appear on the standard error stream, and in 2652 the JSON output from [54]-fdiagnostics-format=json. 2653 2654 Other significant improvements 2655 2656 * To allow inline expansion of both memcpy and memmove, the existing 2657 movmem instruction patterns used for non-overlapping memory copies 2658 have been renamed to cpymem. The movmem name is now used for 2659 overlapping memory moves, consistent with the library functions 2660 memcpy and memmove. 2661 * For many releases, when GCC emits a warning it prints the option 2662 controlling that warning. As of GCC 10, that option text is now a 2663 clickable hyperlink for the documentation of that option (assuming 2664 a [55]sufficiently capable terminal). This behavior can be 2665 controlled via a new [56]-fdiagnostics-urls option (along with 2666 various environment variables and heuristics documented with that 2667 option). 2668 2669 GCC 10.1 2670 2671 This is the [57]list of problem reports (PRs) from GCC's bug tracking 2672 system that are known to be fixed in the 10.1 release. This list might 2673 not be complete (that is, it is possible that some PRs that have been 2674 fixed are not listed here). 2675 2676 GCC 10.2 2677 2678 This is the [58]list of problem reports (PRs) from GCC's bug tracking 2679 system that are known to be fixed in the 10.2 release. This list might 2680 not be complete (that is, it is possible that some PRs that have been 2681 fixed are not listed here). 2682 2683 GCC 10.3 2684 2685 This is the [59]list of problem reports (PRs) from GCC's bug tracking 2686 system that are known to be fixed in the 10.3 release. This list might 2687 not be complete (that is, it is possible that some PRs that have been 2688 fixed are not listed here). 2689 2690 Target Specific Changes 2691 2692 AArch64 2693 2694 * A bug with the Random Number intrinsics in the arm_acle.h header 2695 that resulted in an incorrect status result being returned has been 2696 fixed. 2697 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 2698 options are -mcpu=a64fx and -mtune=a64fx respectively. In 2699 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 2700 tunes the code for the A64FX. This includes tuning the SVE code, 2701 although by default the code is still length-agnostic and so works 2702 for all SVE implementations. Adding -msve-vector-bits=512 makes the 2703 code specific to 512-bit SVE. 2704 2705 x86-64 2706 2707 * GCC 10.3 supports AMD CPUs based on the znver3 core via 2708 -march=znver3. 2709 2710 GCC 10.4 2711 2712 This is the [60]list of problem reports (PRs) from GCC's bug tracking 2713 system that are known to be fixed in the 10.4 release. This list might 2714 not be complete (that is, it is possible that some PRs that have been 2715 fixed are not listed here). 2716 2717 Target Specific Changes 2718 2719 x86-64 2720 2721 * The x86-64 ABI of passing and returning structures with a 64-bit 2722 integer vector changed in GCC 10.1 when MMX is disabled. Disabling 2723 MMX no longer changes how they are passed nor returned. This ABI 2724 change is now diagnosed with -Wpsabi. 2725 2726 GCC 10.5 2727 2728 This is the [61]list of problem reports (PRs) from GCC's bug tracking 2729 system that are known to be fixed in the 10.5 release. This list might 2730 not be complete (that is, it is possible that some PRs that have been 2731 fixed are not listed here). 2732 2733 2734 For questions related to the use of GCC, please consult these web 2735 pages and the [62]GCC manuals. If that fails, the 2736 [63]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2737 web pages and the development of GCC are welcome on our developer 2738 list at [64]gcc (a] gcc.gnu.org. All of [65]our lists have public 2739 archives. 2740 2741 Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and 2742 distribution of this entire article is permitted in any medium, 2743 provided this notice is preserved. 2744 2745 These pages are [67]maintained by the GCC team. Last modified 2746 2023-07-07. 2747 2748 References 2749 2750 1. https://gcc.gnu.org/gcc-10/porting_to.html 2751 2. https://gcc.gnu.org/onlinedocs/index.html#current 2752 3. https://gcc.gnu.org/gcc-10/changes.html#empty_base 2753 4. https://www.mpfr.org/ 2754 5. https://gcc.gnu.org/gcc-10/changes.html#iterator_base 2755 6. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin 2756 7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce 2757 8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training 2758 9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible 2759 10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate 2760 11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path 2761 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html 2762 13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions 2763 14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html 2764 15. https://facebook.github.io/zstd/ 2765 16. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values 2766 17. https://www.openacc.org/ 2767 18. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-10 2768 19. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1 2769 20. https://www.openmp.org/specifications/ 2770 21. https://gcc.gnu.org/wiki/Offloading 2771 22. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare 2772 23. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds 2773 24. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds 2774 25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow 2775 26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict 2776 27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr 2777 28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow 2778 29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion 2779 30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags 2780 31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags 2781 32. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383 2782 33. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711 2783 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707 2784 35. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704 2785 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706 2786 37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285 2787 38. https://www.openmp.org/specifications/ 2788 39. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major 2789 40. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor 2790 41. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel 2791 42. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield 2792 43. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics 2793 44. https://developer.arm.com/Architectures/M-Profile Architecture 2794 45. https://developer.arm.com/architectures/instruction-sets/intrinsics/ 2795 46. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension 2796 47. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs 2797 48. https://gcc.gnu.org/install/configure.html#avr 2798 49. https://gcc.gnu.org/wiki/avr-gcc#Libf7 2799 50. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture 2800 51. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format 2801 52. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2802 53. https://cwe.mitre.org/ 2803 54. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2804 55. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda 2805 56. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls 2806 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0 2807 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2 2808 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3 2809 60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4 2810 61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.5 2811 62. https://gcc.gnu.org/onlinedocs/ 2812 63. mailto:gcc-help (a] gcc.gnu.org 2813 64. mailto:gcc (a] gcc.gnu.org 2814 65. https://gcc.gnu.org/lists.html 2815 66. https://www.fsf.org/ 2816 67. https://gcc.gnu.org/about.html 2817 ====================================================================== 2818 http://gcc.gnu.org/gcc-9/index.html 2819 GCC 9 Release Series 2820 2821 (This release series is no longer supported.) 2822 2823 May 27, 2022 2824 2825 The [1]GNU project and the GCC developers are pleased to announce the 2826 release of GCC 9.5. 2827 2828 This release is a bug-fix release, containing fixes for regressions in 2829 GCC 9.4 relative to previous releases of GCC. 2830 2831 Release History 2832 2833 GCC 9.5 2834 May 27, 2022 ([2]changes, [3]documentation) 2835 2836 GCC 9.4 2837 June 1, 2021 ([4]changes, [5]documentation) 2838 2839 GCC 9.3 2840 Mar 12, 2020 ([6]changes, [7]documentation) 2841 2842 GCC 9.2 2843 Aug 12, 2019 ([8]changes, [9]documentation) 2844 2845 GCC 9.1 2846 May 3, 2019 ([10]changes, [11]documentation) 2847 2848 References and Acknowledgements 2849 2850 GCC used to stand for the GNU C Compiler, but since the compiler 2851 supports several other languages aside from C, it now stands for the 2852 GNU Compiler Collection. 2853 2854 The GCC developers would like to thank the numerous people that have 2855 contributed new features, improvements, bug fixes, and other changes as 2856 well as test results to GCC. This [12]amazing group of volunteers is 2857 what makes GCC successful. 2858 2859 For additional information about GCC please refer to the [13]GCC 2860 project web site or contact the [14]GCC development mailing list. 2861 2862 To obtain GCC please use [15]our mirror sites or [16]our version 2863 control system. 2864 2865 2866 For questions related to the use of GCC, please consult these web 2867 pages and the [17]GCC manuals. If that fails, the 2868 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2869 web pages and the development of GCC are welcome on our developer 2870 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 2871 archives. 2872 2873 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 2874 distribution of this entire article is permitted in any medium, 2875 provided this notice is preserved. 2876 2877 These pages are [22]maintained by the GCC team. Last modified 2878 2024-05-30. 2879 2880 References 2881 2882 1. http://www.gnu.org/ 2883 2. https://gcc.gnu.org/gcc-9/changes.html 2884 3. https://gcc.gnu.org/onlinedocs/9.5.0/ 2885 4. https://gcc.gnu.org/gcc-9/changes.html 2886 5. https://gcc.gnu.org/onlinedocs/9.4.0/ 2887 6. https://gcc.gnu.org/gcc-9/changes.html 2888 7. https://gcc.gnu.org/onlinedocs/9.3.0/ 2889 8. https://gcc.gnu.org/gcc-9/changes.html 2890 9. https://gcc.gnu.org/onlinedocs/9.2.0/ 2891 10. https://gcc.gnu.org/gcc-9/changes.html 2892 11. https://gcc.gnu.org/onlinedocs/9.1.0/ 2893 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html 2894 13. https://gcc.gnu.org/index.html 2895 14. mailto:gcc (a] gcc.gnu.org 2896 15. https://gcc.gnu.org/mirrors.html 2897 16. https://gcc.gnu.org/git.html 2898 17. https://gcc.gnu.org/onlinedocs/ 2899 18. mailto:gcc-help (a] gcc.gnu.org 2900 19. mailto:gcc (a] gcc.gnu.org 2901 20. https://gcc.gnu.org/lists.html 2902 21. https://www.fsf.org/ 2903 22. https://gcc.gnu.org/about.html 2904 ====================================================================== 2905 http://gcc.gnu.org/gcc-9/changes.html 2906 GCC 9 Release Series 2907 Changes, New Features, and Fixes 2908 2909 This page is a "brief" summary of some of the huge number of 2910 improvements in GCC 9. You may also want to check out our [1]Porting to 2911 GCC 9 page and the [2]full GCC documentation. 2912 2913 Caveats 2914 2915 * On Arm targets (arm*-*-*), [3]a bug in the implementation of the 2916 procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has 2917 been fixed: a structure containing a bit-field based on a 64-bit 2918 integral type and where no other element in a structure required 2919 64-bit alignment could be passed incorrectly to functions. This is 2920 an ABI change. If the option -Wpsabi is enabled (on by default) the 2921 compiler will emit a diagnostic note for code that might be 2922 affected. 2923 * Support for a number of older systems and recently unmaintained or 2924 untested target ports of GCC has been declared obsolete in GCC 9. 2925 Unless there is activity to revive them, the next release of GCC 2926 will have their sources permanently removed. 2927 The following ports for individual systems on particular 2928 architectures have been obsoleted: 2929 + Solaris 10 (*-*-solaris2.10). Details can be found in the 2930 [4]announcement. 2931 + Cell Broadband Engine SPU (spu*-*-*). Details can be found in 2932 the [5]announcement. 2933 * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause 2934 ABI incompatibilities with object files compiled with other 2935 versions of GCC. If the std::rotate algorithm is called with an 2936 empty range then it might cause a divide-by-zero error (as a SIGFPE 2937 signal) and crash. The change has been reverted for GCC 9.2.0 and 2938 future releases. For more details see [6]Bug 90920. The problem can 2939 be avoided by recompiling any objects that might call std::rotate 2940 with an empty range, so that the GCC 9.1.0 definition of 2941 std::rotate is not used. 2942 * The automatic template instantiation at link time ([7]-frepo) has 2943 been deprecated and will be removed in a future release. 2944 * The --with-default-libstdcxx-abi=gcc4-compatible configure option 2945 is broken in the 9.1 and 9.2 releases, producing a shared library 2946 with missing symbols (see [8]Bug 90361). As a workaround, configure 2947 without that option and build GCC as normal, then edit the 2948 installed <bits/c++config.h> headers to define the 2949 _GLIBCXX_USE_CXX11_ABI macro to 0. 2950 2951 General Improvements 2952 2953 The following GCC command line options have been introduced or 2954 improved. 2955 * All command line options that take a byte-size argument accept 2956 64-bit integers as well as standard SI and IEC suffixes such as kb 2957 and KiB, MB and MiB, or GB and GiB denoting the corresponding 2958 multiples of bytes. See [9]Invoking GCC for more. 2959 * A new option [10]-flive-patching=[inline-only-static|inline-clone] 2960 generates code suitable for live patching. At the same time it 2961 provides multiple-level control over IPA optimizations. See the 2962 user guide for more details. 2963 * A new option, --completion, has been added to provide more fine 2964 option completion in a shell. It is intended to be used by 2965 Bash-completion. 2966 * GCC's diagnostics now print source code with a left margin showing 2967 line numbers, configurable with 2968 [11]-fno-diagnostics-show-line-numbers. 2969 GCC's diagnostics can also now label regions of the source code to 2970 show pertinent information, such as the types within an expression. 2971 $ g++ t.cc 2972 t.cc: In function 'int test(const shape&, const shape&)': 2973 t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl 2974 e>' and 'boxed_value<double>') 2975 14 | return (width(s1) * height(s1) 2976 | ~~~~~~~~~~~~~~~~~~~~~~ 2977 | | 2978 | boxed_value<[...]> 2979 15 | + width(s2) * height(s2)); 2980 | ^ ~~~~~~~~~~~~~~~~~~~~~~ 2981 | | 2982 | boxed_value<[...]> 2983 2984 These labels can be disabled via [12]-fno-diagnostics-show-labels. 2985 * A new option [13]-fdiagnostics-format=json has been introduced for 2986 emitting diagnostics in a machine-readable format. 2987 * The alignment-related options [14]-falign-functions, 2988 [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps 2989 received support for a secondary alignment (e.g. 2990 -falign-loops=n:m:n2:m2). 2991 * New pair of profiling options ([18]-fprofile-filter-files and 2992 [19]-fprofile-exclude-files) has been added. The options help to 2993 filter which source files are instrumented. 2994 * AddressSanitizer generates more compact redzones for automatic 2995 variables. That helps to reduce memory footprint of a sanitized 2996 binary. 2997 * Numerous improvements have been made to the output of 2998 [20]-fopt-info. 2999 Messages are now prefixed with optimized, missed, or note, rather 3000 than the old behavior of all being prefixed with note. 3001 The output from -fopt-info can now contain information on inlining 3002 decisions: 3003 $ g++ -c inline.cc -O2 -fopt-info-inline-all 3004 inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)( 3005 E)) [with T = char**; E = char*]/2. 3006 inline.cc:24:11: optimized: Inlining void foreach(T, T, void (*)(E)) [with T = 3007 char**; E = char*]/2 into int main(int, char**)/1. 3008 inline.cc:19:12: missed: not inlinable: void inline_me(char*)/0 -> int std::pu 3009 ts(const char*)/3, function body not available 3010 inline.cc:13:8: optimized: Inlined void inline_me(char*)/4 into int main(int, c 3011 har**)/1 which now has time 127.363637 and size 11, net change of +0. 3012 Unit growth for small function inlining: 16->16 (0%) 3013 3014 Inlined 2 calls, eliminated 1 functions 3015 3016 3017 The output from the vectorizer has been rationalized so that failed 3018 attempts to vectorize a loop are displayed in the form 3019 [LOOP-LOCATION]: couldn't vectorize this loop 3020 [PROBLEM-LOCATION]: because of [REASON] 3021 3022 rather than an exhaustive log of all decisions made by the 3023 vectorizer. For example: 3024 $ gcc -c v.c -O3 -fopt-info-all-vec 3025 v.c:7:3: missed: couldn't vectorize loop 3026 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 3027 mory"); 3028 v.c:3:6: note: vectorized 0 loops in function. 3029 v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 3030 mory"); 3031 3032 The old behavior can be obtained via a new -internals suboption of 3033 -fopt-info. 3034 * A new option, [21]-fsave-optimization-record has been added, which 3035 writes a SRCFILE.opt-record.json.gz file describing the 3036 optimization decisions made by GCC. This is similar to the output 3037 of -fopt-info, but with additional metadata such as the inlining 3038 chain, and profile information (if available). 3039 * Inter-procedural propagation of stack alignment can now be 3040 controlled by [22]-fipa-stack-alignment. 3041 * Propagation of addressability, readonly, and writeonly flags on 3042 static variables can now be controlled by 3043 [23]-fipa-reference-addressable. 3044 3045 The following built-in functions have been introduced. 3046 * [24]__builtin_expect_with_probability to provide branch prediction 3047 probability hints to the optimizer. 3048 * [25]__builtin_has_attribute determines whether a function, type, or 3049 variable has been declared with some attribute. 3050 * [26]__builtin_speculation_safe_value can be used to help mitigate 3051 against unsafe speculative execution. 3052 3053 The following attributes have been introduced. 3054 * The [27]copy function attribute has been added. The attribute can 3055 also be applied to type definitions and to variable declarations. 3056 3057 A large number of improvements to code generation have been made, 3058 including but not limited to the following. 3059 * Switch expansion has been improved by using a different strategy 3060 (jump table, bit test, decision tree) for a subset of switch cases. 3061 * A linear function expression defined as a switch statement can be 3062 transformed by [28]-ftree-switch-conversion. For example: 3063 3064 int 3065 foo (int how) 3066 { 3067 switch (how) { 3068 case 2: how = 205; break; 3069 case 3: how = 305; break; 3070 case 4: how = 405; break; 3071 case 5: how = 505; break; 3072 case 6: how = 605; break; 3073 } 3074 return how; 3075 } 3076 3077 can be transformed into 100 * how + 5 (for values defined in the 3078 switch statement). 3079 * Inter-procedural optimization improvements: 3080 + Inliner defaults were tuned to better suit modern C++ 3081 codebases, especially when built with link time-optimizations. 3082 New parameters max-inline-insns-small, max-inline-insns-size, 3083 uninlined-function-insns, uninlined-function-time, 3084 uninlined-thunk-insns, and uninlined-thunk-time were added. 3085 + Hot/cold partitioning is now more precise and aggressive. 3086 + Improved scalability for very large translation units 3087 (especially when link-time optimizing large programs). 3088 * Profile driven optimization improvements: 3089 + [29]-fprofile-use now enables [30]-fversion-loops-for-strides, 3090 [31]-floop-interchange, [32]-floop-unroll-and-jam, 3091 [33]-ftree-loop-distribution. 3092 + Streaming of counter histograms was removed, which reduces the 3093 size of profile files. Histograms are computed on the fly with 3094 link-time optimization. 3095 + The parameter hot-bb-count-ws-permille was reduced from 999 to 3096 990 to account for more precise histograms. 3097 * Link-time optimization improvements: 3098 + Types are now simplified prior to streaming resulting in 3099 significant reductions of LTO object file sizes and link-time 3100 memory use as well as improvements of link-time parallelism. 3101 + The default number of partitions (--param lto-partitions) was 3102 increased from 32 to 128 enabling effective use of CPUs with 3103 more than 32 hyperthreads. --param 3104 lto-max-streaming-parallelism can now be used to control the 3105 number of streaming processes. 3106 + Warnings on C++ One Decl Rule violations (-Wodr) are now more 3107 informative and produce fewer redundant results. 3108 Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an 3109 8-core machine was reduced by about 5% compared to GCC 8.3, and the 3110 size of LTO object files by 7%. LTO link time improves by 11% on an 3111 8-core machine and scales significantly better for more parallel 3112 build environments. The serial stage of the link-time optimization 3113 is 28% faster consuming 20% less memory. The parallel stage now 3114 scales to up to 128 partitions rather than 32 and reduces memory 3115 use for every worker by 30%. 3116 3117 The following improvements to the gcov command-line utility have been 3118 made. 3119 * The gcov tool received a new option [34]--use-hotness-colors (-q) 3120 that can provide perf-like coloring of hot functions. 3121 * The gcov tool has changed its intermediate format to a new JSON 3122 format. 3123 3124 New Languages and Language specific improvements 3125 3126 [35]OpenACC support in C, C++, and Fortran continues to be maintained 3127 and improved. Most of the OpenACC 2.5 specification is implemented. See 3128 the [36]implementation status section on the OpenACC wiki page for 3129 further information. 3130 3131 C family 3132 3133 * Version 5.0 of the [37]OpenMP specification is now partially 3134 supported in the C and C++ compilers. For details which features of 3135 OpenMP 5.0 are and which are not supported in the GCC 9 release see 3136 [38]this mail. 3137 * New extensions: 3138 + [39]__builtin_convertvector built-in for vector conversions 3139 has been added. 3140 * New warnings: 3141 + [40]-Waddress-of-packed-member, enabled by default, warns 3142 about an unaligned pointer value from the address of a packed 3143 member of a struct or union. 3144 * Enhancements to existing warnings: 3145 + [41]-Warray-bounds detects more instances of out-of-bounds 3146 indices. 3147 + [42]-Wattribute-alias also detects attribute mismatches 3148 between alias declarations and their targets, in addition to 3149 mismatches between their types. 3150 + [43]-Wformat-overflow and [44]-Wformat-truncation have been 3151 extended to all formatted input/output functions (where 3152 applicable) and enhanced to detect a subset of instances of 3153 reading past the end of unterminated constant character arrays 3154 in %s directives. 3155 + [45]-Wmissing-attributes detects instances of missing function 3156 attributes on declarations of aliases and weak references. 3157 + [46]-Wstringop-truncation also detects a subset of instances 3158 of reading past the end of unterminated constant character 3159 arrays, 3160 * If a macro is used with the wrong argument count, the C and C++ 3161 front ends now show the definition of that macro via a note. 3162 * The spelling corrector now considers transposed letters, and the 3163 threshold for similarity has been tightened, to avoid nonsensical 3164 suggestions. 3165 3166 C 3167 3168 * There is now experimental support for -std=c2x, to select support 3169 for the upcoming C2X revision of the ISO C standard. This standard 3170 is in the early stages of development and the only feature 3171 supported in GCC 9 is _Static_assert with a single argument 3172 (support for _Static_assert with two arguments was added in C11 and 3173 GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU 3174 extensions, and -Wc11-c2x-compat, to warn for uses of features 3175 added in C2X (such warnings are also enabled by use of -Wpedantic 3176 if not using -std=c2x or -std=gnu2x). 3177 * New warnings: 3178 + [47]-Wabsolute-value warns for calls to standard functions 3179 that compute the absolute value of an argument when a more 3180 appropriate standard function is available. For example, 3181 calling abs(3.14) triggers the warning because the appropriate 3182 function to call to compute the absolute value of a double 3183 argument is fabs. The option also triggers warnings when the 3184 argument in a call to such a function has an unsigned type. 3185 This warning can be suppressed with an explicit type cast and 3186 it is also enabled by -Wextra. 3187 3188 C++ 3189 3190 * New warnings: 3191 + [48]-Wdeprecated-copy, implied by -Wextra, warns about the 3192 C++11 deprecation of implicitly declared copy constructor and 3193 assignment operator if one of them is user-provided. 3194 -Wdeprecated-copy-dtor also warns if the destructor is 3195 user-provided, as specified in C++11. 3196 + [49]-Winit-list-lifetime, on by default, warns about uses of 3197 std::initializer_list that are likely to result in a dangling 3198 pointer, such as returning or assigning from a temporary list. 3199 + [50]-Wredundant-move, implied by -Wextra, warns about 3200 redundant calls to std::move. 3201 + [51]-Wpessimizing-move, implied by -Wall, warns when a call to 3202 std::move prevents copy elision. 3203 + [52]-Wclass-conversion, on by default, warns when a conversion 3204 function will never be called due to the type it converts to. 3205 * The C++ front end has experimental support for some of the upcoming 3206 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 3207 including range-based for statements with initializer, default 3208 constructible and assignable stateless lambdas, lambdas in 3209 unevaluated contexts, language support for empty data members, 3210 allowing pack expansion in lambda init-capture, likely and unlikely 3211 attributes, class types in non-type template parameters, allowing 3212 virtual function calls in constant expressions, explicit(bool), 3213 std::is_constant_evaluated, nested inline namespaces, etc. For a 3214 full list of new features, see [53]the C++ status page. 3215 * The C++ front end now preserves source locations for literals, 3216 id-expression, and mem-initializer for longer. For example it is 3217 now able to pin-point the pertinent locations for bad 3218 initializations such as these 3219 $ g++ -c bad-inits.cc 3220 bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization 3221 10 | { 3, json::object }, 3222 | ~~~~~~^~~~~~ 3223 | | 3224 | json 3225 bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f 3226 permissive] 3227 14 | char buffers[3][5] = { "red", "green", "blue" }; 3228 | ^~~~~~~ 3229 bad-inits.cc: In constructor 'X::X()': 3230 bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv 3231 e] 3232 17 | X() : one(42), two(42), three(42) 3233 | ^~ 3234 | | 3235 | int 3236 3237 rather than emitting the error at the final closing parenthesis or 3238 brace. 3239 * Error-reporting of overload resolution has been special-cased to 3240 make the case of a single failed candidate easier to read. For 3241 example: 3242 $ g++ param-type-mismatch.cc 3243 param-type-mismatch.cc: In function 'int test(int, const char*, float)': 3244 param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char* 3245 *' 3246 8 | return foo::member_1 (first, second, third); 3247 | ^~~~~~ 3248 | | 3249 | const char* 3250 param-type-mismatch.cc:3:46: note: initializing argument 2 of 'static int foo: 3251 :member_1(int, const char**, float)' 3252 3 | static int member_1 (int one, const char **two, float three); 3253 | ~~~~~~~~~~~~~^~~ 3254 3255 highlights both the problematic argument, and the parameter that it 3256 can't be converted to. 3257 * Diagnostics involving binary operators now use color to distinguish 3258 the two operands, and label them separately (as per the example of 3259 source labelling above). 3260 * Diagnostics involving function calls now highlight the pertinent 3261 parameter of the declaration in more places. 3262 $ g++ bad-conversion.cc 3263 bad-conversion.cc: In function 'void caller()': 3264 bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*' 3265 9 | callee (0, false, 2); 3266 | ^~~~~ 3267 | | 3268 | bool 3269 bad-conversion.cc:3:19: note: initializing argument 2 of 'void callee(int, voi 3270 d*, int)' 3271 3 | void callee (int, void *, int) 3272 | ^~~~~~ 3273 3274 * The C++ front end's implementation of [54]-Wformat now shows 3275 precise locations within string literals, and underlines the 3276 pertinent arguments at bogus call sites (the C front end has been 3277 doing this since GCC 7). For example: 3278 $ g++ -c bad-printf.cc -Wall 3279 bad-printf.cc: In function 'void print_field(const char*, float, long int, long 3280 int)': 3281 bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type 3282 'int', but argument 3 has type 'long int' [-Wformat=] 3283 6 | printf ("%s: %*ld ", fieldname, column - width, value); 3284 | ~^~~ ~~~~~~~~~~~~~~ 3285 | | | 3286 | int long int 3287 bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b 3288 ut argument 4 has type 'double' [-Wformat=] 3289 6 | printf ("%s: %*ld ", fieldname, column - width, value); 3290 | ~~~^ ~~~~~ 3291 | | | 3292 | long int double 3293 | %*f 3294 3295 * The C++ front end has gained new fix-it hints for forgetting the 3296 return *this; needed by various C++ operators: 3297 $ g++ -c operator.cc 3298 operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr 3299 &)': 3300 operator.cc:7:3: warning: no return statement in function returning non-void [-W 3301 return-type] 3302 6 | m_ptr = other.m_ptr; 3303 +++ |+ return *this; 3304 7 | } 3305 | ^ 3306 3307 for when the compiler needs a typename: 3308 $ g++ -c template.cc 3309 template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i 3310 s a dependent scope 3311 3 | Traits::type type; 3312 | ^~~~~~ 3313 | typename 3314 3315 when trying to use an accessor member as if it were a data member: 3316 $ g++ -c fncall.cc 3317 fncall.cc: In function 'void hangman(const mystring&)': 3318 fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length 3319 () const' (did you forget the '()' ?) 3320 12 | if (str.get_length > 0) 3321 | ~~~~^~~~~~~~~~ 3322 | () 3323 3324 for C++11's scoped enums: 3325 $ g++ -c enums.cc 3326 enums.cc: In function 'void json::test(const json::value&)': 3327 enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js 3328 on::kind::STRING'? 3329 12 | if (v.get_kind () == STRING) 3330 | ^~~~~~ 3331 | json::kind::STRING 3332 enums.cc:3:44: note: 'json::kind::STRING' declared here 3333 3 | enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ }; 3334 | ^~~~~~ 3335 3336 and a tweak to integrate the suggestions about misspelled members 3337 with that for accessors: 3338 $ g++ -c accessor-fixit.cc 3339 accessor-fixit.cc: In function 'int test(t*)': 3340 accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m 3341 ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const') 3342 17 | return ptr->ratio; 3343 | ^~~~~ 3344 | get_ratio() 3345 3346 In addition, various diagnostics in the C++ front-end have been 3347 streamlined by consolidating the suggestion into the initial error, 3348 rather than emitting a follow-up note: 3349 $ g++ typo.cc 3350 typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF 3351 _SIZE'? 3352 5 | uint8_t buf[BUFSIZE]; 3353 | ^~~~~~~ 3354 | BUF_SIZE 3355 3356 Runtime Library (libstdc++) 3357 3358 * Improved support for C++17, including: 3359 + The C++17 implementation is no longer experimental. 3360 + Parallel algorithms and <execution> (requires [55]Thread 3361 Building Blocks 2018 or newer). 3362 + <memory_resource>. 3363 + Using the types and functions in <filesystem> does not require 3364 linking with -lstdc++fs now. 3365 * Improved experimental support for C++2a, including: 3366 + Type traits std::remove_cvref, std::unwrap_reference, 3367 std::unwrap_decay_ref, std::is_nothrow_convertible, and 3368 std::type_identity. 3369 + Headers <bit> and <version>. 3370 + Uniform container erasure (std::erase_if). 3371 + contains member of maps and sets. 3372 + String prefix and suffix checking (starts_with, ends_with). 3373 + Functions std::midpoint and std::lerp for interpolation. 3374 + std::bind_front. 3375 + std::visit<R>. 3376 + std::assume_aligned. 3377 + Uses-allocator construction utilities. 3378 + std::pmr::polymorphic_allocator<std::byte>. 3379 + Library support for char8_t type. 3380 + Destroying delete. 3381 + std::is_constant_evaluated() function. 3382 * Support for opening file streams with wide character paths on 3383 Windows 3384 * Incomplete support for the C++17 Filesystem library and the 3385 Filesystem TS on Windows. 3386 * Incomplete, experimental support for the Networking TS. 3387 3388 D 3389 3390 * Support for the D programming language has been added to GCC, 3391 implementing version 2.076 of the language and run-time library. 3392 3393 Fortran 3394 3395 * Asynchronous I/O is now fully supported. The program needs to be 3396 linked against the pthreads library to use it, otherwise the I/O is 3397 done synchronously. For systems which do not support POSIX 3398 condition variables, such as AIX, all I/O is still done 3399 synchronously. 3400 * The BACK argument for MINLOC and MAXLOC has been implemented. 3401 * The FINDLOC intrinsic function has been implemented. 3402 * The IS_CONTIGUOUS intrinsic function has been implemented. 3403 * Direct access to the real and imaginary parts of a complex variable 3404 via c%re and c%im has been implemented. 3405 * Type parameter inquiry via str%len and a%kind has been implemented. 3406 * C descriptors and the ISO_Fortran_binding.h source file have been 3407 implemented. 3408 * The MAX and MIN intrinsics are no longer guaranteed to return any 3409 particular value in case one of the arguments is a NaN. Note that 3410 this conforms to the Fortran standard and to what other Fortran 3411 compilers do. If there is a need to handle that case in some 3412 specific way, one needs to explicitly check for NaN's before 3413 calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the 3414 intrinsic module IEEE_ARITHMETIC. 3415 * A new command-line option [56]-fdec-include, set also by the 3416 [57]-fdec option, has been added to increase compatibility with 3417 legacy code. With this option, an INCLUDE directive is also parsed 3418 as a statement, which allows the directive to be spread across 3419 multiple source lines with line continuations. 3420 * A new [58]BUILTIN directive, has been added. The purpose of the 3421 directive is to provide an API between the GCC compiler and the GNU 3422 C Library which would define vector implementations of math 3423 routines. 3424 3425 Go 3426 3427 * GCC 9 provides a complete implementation of the Go 1.12.2 user 3428 packages. 3429 3430 libgccjit 3431 3432 * The libgccjit API gained a new entry point: 3433 [59]gcc_jit_context_add_driver_option. 3434 3435 New Targets and Target Specific Improvements 3436 3437 AArch64 & Arm 3438 3439 * Support has been added for the following processors (GCC 3440 identifiers in parentheses): 3441 + Arm Cortex-A76 (cortex-a76). 3442 + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE 3443 (cortex-a76.cortex-a55). 3444 + Arm Neoverse N1 (neoverse-n1). 3445 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3446 options, for example: -mcpu=cortex-a76 or 3447 -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent 3448 target attributes and pragmas. 3449 * The Armv8.3-A complex number instructions are now supported via 3450 intrinsics when the option -march=armv8.3-a or equivalent is 3451 specified. For the half-precision floating-point variants of these 3452 instructions use the architecture extension flag +fp16, e.g. 3453 -march=armv8.3-a+fp16. 3454 The intrinsics are defined by the ACLE specification. 3455 * The Armv8.5-A architecture is now supported through the 3456 -march=armv8.5-a option. 3457 * The Armv8.5-A architecture also adds some security features that 3458 are optional to all older architecture versions. These are now 3459 supported and only affect the assembler. 3460 + Speculation Barrier instruction through the -march=armv8-a+sb 3461 option. 3462 + Execution and Data Prediction Restriction instructions through 3463 the -march=armv8-a+predres option. 3464 + Speculative Store Bypass Safe instruction through the 3465 -march=armv8-a+ssbs option. This does not require a compiler 3466 option for Arm and thus -march=armv8-a+ssbs is an 3467 AArch64-specific option. 3468 3469 AArch64 specific 3470 3471 * Support has been added for the Arm Neoverse E1 processor 3472 (-mcpu=neoverse-e1). 3473 * The AArch64 port now has support for stack clash protection using 3474 the [60]-fstack-clash-protection option. The probing interval/guard 3475 size can be set by using --param 3476 stack-clash-protection-guard-size=12|16. The value of this 3477 parameter must be in bytes represented as a power of two. The two 3478 supported values for this parameter are 12 (for a 4KiB size, 2^12) 3479 and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and 3480 can be changed at configure time using the flag 3481 --with-stack-clash-protection-guard-size=12|16. 3482 * The option -msign-return-address= has been deprecated. This has 3483 been replaced by the new -mbranch-protection= option. This new 3484 option can now be used to enable the return address signing as well 3485 as the new Branch Target Identification feature of Armv8.5-A 3486 architecture. For more information on the arguments accepted by 3487 this option, please refer to [61]AArch64-Options. 3488 * The following optional extensions to Armv8.5-A architecture are now 3489 supported and only affect the assembler. 3490 + Random Number Generation instructions through the 3491 -march=armv8.5-a+rng option. 3492 + Memory Tagging Extension through the -march=armv8.5-a+memtag 3493 option. 3494 3495 Arm specific 3496 3497 * Support for the deprecated Armv2 and Armv3 architectures and their 3498 variants has been removed. Their corresponding -march values and 3499 the -mcpu options that used these architectures have been removed. 3500 * Support for the Armv5 and Armv5E architectures (which have no known 3501 implementations) has been removed. Note that Armv5T, Armv5TE and 3502 Armv5TEJ architectures remain supported. 3503 * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using 3504 their respective -mcpu options. 3505 3506 AMD GCN 3507 3508 * A new back end targeting AMD GCN GPUs has been contributed to GCC. 3509 The implementation is currently limited to compiling 3510 single-threaded, stand-alone programs. Future versions will add 3511 support for offloading multi-threaded kernels via OpenMP and 3512 OpenACC. The following devices are supported (GCC identifiers in 3513 parentheses): 3514 + Fiji (fiji). 3515 + Vega 10 (gfx900). 3516 3517 ARC 3518 3519 * LRA is now on by default for the ARC target. This can be controlled 3520 by -mlra. 3521 * Add support for frame code-density and branch-and-index 3522 instructions. 3523 3524 C-SKY 3525 3526 * A new back end targeting C-SKY V2 processors has been contributed 3527 to GCC. 3528 3529 IA-32/x86-64 3530 3531 * Support of Intel MPX (Memory Protection Extensions) has been 3532 removed. 3533 * New ISA extension support for Intel PTWRITE was added to GCC. 3534 PTWRITE intrinsics are available via the -mptwrite compiler switch. 3535 * GCC now supports the Intel CPU named Cascade Lake with AVX512 3536 extensions through -march=cascadelake. The switch enables the 3537 following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW, 3538 AVX512DQ, AVX512VNNI. 3539 3540 MIPS 3541 3542 * The Loongson loongson-mmi and loongson-ext extensions have been 3543 split from loongson3a: 3544 + loongson-mmi contains the Loongson MMI (MultiMedia extensions 3545 Instructions). 3546 + loongson-ext contains the Loongson EXT (EXTensions 3547 instructions). 3548 * The Loongson EXT2 (EXTensions R2 instructions) are now supported. 3549 + loongson-ext2 contains the Loongson EXT2 instructions. 3550 Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and 3551 -m[no-]loongson-ext2 enable or disable those extensions. 3552 * Support has been added for the following processors (GCC 3553 identifiers in parentheses): 3554 + Loongson 3A1000 (gs464) which enables loongson-mmi, 3555 loongson-ext by default. 3556 + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi, 3557 loongson-ext, loongson-ext2 by default. 3558 + Loongson 2K1000 (gs264e) which enables loongson-ext, 3559 loongson-ext2, msa by default. 3560 The GCC identifiers can be used as arguments to the -mcpu and 3561 -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments 3562 to the equivalent target attributes and pragmas. 3563 3564 OpenRISC 3565 3566 * A new back end targeting OpenRISC processors has been contributed 3567 to GCC. 3568 3569 S/390, System z, IBM z Systems 3570 3571 * Support for the arch13 architecture has been added. When using the 3572 -march=arch13 option, the compiler will generate code making use of 3573 the new instructions introduced with the vector enhancement 3574 facility 2 and the miscellaneous instruction extension facility 2. 3575 The -mtune=arch13 option enables arch13 specific instruction 3576 scheduling without making use of new instructions. 3577 * Builtins for the new vector instructions have been added and can be 3578 enabled using the -mzvector option. 3579 * Support for ESA architecture machines g5 and g6 is deprecated since 3580 GCC 6.1.0 and has been removed now. 3581 * When compiling with -march=z14 or higher GCC emits alignments hints 3582 on the vector load/store instructions (8 or 16 byte). 3583 * Functions now have a default alignment of 16 bytes. This helps with 3584 branch prediction effects. 3585 * -mfentry is now supported. As well as the mcount mechanism the 3586 __fentry__ is called before the function prologue. However, since 3587 just a single instruction is required to call __fentry__ the call 3588 sequence imposes a smaller overhead than mcount (4 instructions). 3589 The produced code is compatible only with newer glibc versions, 3590 which provide the __fentry__ symbol and do not clobber r0 when 3591 resolving lazily bound functions. -mfentry is only supported when 3592 generating 64-bit code and does not work with nested C functions. 3593 * The -mnop-mcount option can be used to emit NOP instructions 3594 instead of an mcount or fentry call stub. 3595 * With the -mrecord-mcount option a __mcount_loc section is generated 3596 containing pointers to each profiling call stub. This is useful for 3597 automatically patching in and out calls. 3598 3599 Operating Systems 3600 3601 Solaris 3602 3603 * g++ now unconditionally enables large file support when compiling 3604 32-bit code. 3605 * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has 3606 been merged from LLVM. For the moment, this only works for 32-bit 3607 code on both SPARC and x86. 3608 * An initial port of the D runtime library has been completed on 3609 Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC 3610 support is still work-in-progress. 3611 3612 Windows 3613 3614 * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been 3615 fixed. A non-field declaration could cause the current bitfield 3616 allocation unit to be completed, incorrectly placing a following 3617 bitfield into a new allocation unit. The Microsoft ABI is selected 3618 for: 3619 + Mingw targets 3620 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 3621 option is specified, or __attribute__((ms_struct)) is used 3622 + SuperH targets when the -mhitachi option is specified, or 3623 __attribute__((renesas)) is used 3624 3625 Improvements for plugin authors 3626 3627 * GCC's diagnostic subsystem now has a way to logically group 3628 together related diagnostics, auto_diagnostic_group. Such 3629 diagnostics will be nested by the output of 3630 [63]-fdiagnostics-format=json. 3631 * GCC now has a set of [64]user experience guidelines for GCC, with 3632 information and advice on implementing new diagnostics. 3633 3634 Other significant improvements 3635 3636 * GCC's internal "selftest" suite now runs for C++ as well as C (in 3637 debug builds of the compiler). 3638 3639 GCC 9.1 3640 3641 This is the [65]list of problem reports (PRs) from GCC's bug tracking 3642 system that are known to be fixed in the 9.1 release. This list might 3643 not be complete (that is, it is possible that some PRs that have been 3644 fixed are not listed here). 3645 3646 GCC 9.2 3647 3648 This is the [66]list of problem reports (PRs) from GCC's bug tracking 3649 system that are known to be fixed in the 9.2 release. This list might 3650 not be complete (that is, it is possible that some PRs that have been 3651 fixed are not listed here). 3652 3653 GCC 9.3 3654 3655 This is the [67]list of problem reports (PRs) from GCC's bug tracking 3656 system that are known to be fixed in the 9.3 release. This list might 3657 not be complete (that is, it is possible that some PRs that have been 3658 fixed are not listed here). 3659 3660 GCC 9.4 3661 3662 This is the [68]list of problem reports (PRs) from GCC's bug tracking 3663 system that are known to be fixed in the 9.4 release. This list might 3664 not be complete (that is, it is possible that some PRs that have been 3665 fixed are not listed here). 3666 3667 Target Specific Changes 3668 3669 AArch64 3670 3671 * The option -moutline-atomics has been added to aid deployment of 3672 the Large System Extensions (LSE) on GNU/Linux systems built with a 3673 baseline architecture targeting Armv8-A. When the option is 3674 specified code is emitted to detect the presence of LSE 3675 instructions at run time and use them for standard atomic 3676 operations. For more information please refer to the documentation. 3677 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 3678 options are -mcpu=a64fx and -mtune=a64fx respectively. In 3679 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 3680 tunes the code for the A64FX. This includes tuning the SVE code, 3681 although by default the code is still length-agnostic and so works 3682 for all SVE implementations. Adding -msve-vector-bits=512 makes the 3683 code specific to 512-bit SVE. 3684 3685 GCC 9.5 3686 3687 This is the [69]list of problem reports (PRs) from GCC's bug tracking 3688 system that are known to be fixed in the 9.5 release. This list might 3689 not be complete (that is, it is possible that some PRs that have been 3690 fixed are not listed here). 3691 3692 3693 For questions related to the use of GCC, please consult these web 3694 pages and the [70]GCC manuals. If that fails, the 3695 [71]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3696 web pages and the development of GCC are welcome on our developer 3697 list at [72]gcc (a] gcc.gnu.org. All of [73]our lists have public 3698 archives. 3699 3700 Copyright (C) [74]Free Software Foundation, Inc. Verbatim copying and 3701 distribution of this entire article is permitted in any medium, 3702 provided this notice is preserved. 3703 3704 These pages are [75]maintained by the GCC team. Last modified 3705 2024-05-09. 3706 3707 References 3708 3709 1. https://gcc.gnu.org/gcc-9/porting_to.html 3710 2. https://gcc.gnu.org/onlinedocs/index.html#current 3711 3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469 3712 4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html 3713 5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html 3714 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920 3715 7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo 3716 8. https://gcc.gnu.org/PR90361 3717 9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC 3718 10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching 3719 11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers 3720 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels 3721 13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 3722 14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions 3723 15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels 3724 16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops 3725 17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps 3726 18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files 3727 19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files 3728 20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info 3729 21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record 3730 22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment 3731 23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable 3732 24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability 3733 25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1 3734 26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1 3735 27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute 3736 28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion 3737 29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use 3738 30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides 3739 31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange 3740 32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam 3741 33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution 3742 34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov 3743 35. https://www.openacc.org/ 3744 36. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-9 3745 37. https://www.openmp.org/specifications/ 3746 38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html 3747 39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector 3748 40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member 3749 41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds 3750 42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias 3751 43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow 3752 44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation 3753 45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 3754 46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 3755 47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value 3756 48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy 3757 49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime 3758 50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 3759 51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 3760 52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion 3761 53. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 3762 54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat 3763 55. https://github.com/oneapi-src/oneTBB 3764 56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include 3765 57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec 3766 58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive 3767 59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option 3768 60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector 3769 61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options 3770 62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 3771 63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 3772 64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html 3773 65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0 3774 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2 3775 67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3 3776 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4 3777 69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5 3778 70. https://gcc.gnu.org/onlinedocs/ 3779 71. mailto:gcc-help (a] gcc.gnu.org 3780 72. mailto:gcc (a] gcc.gnu.org 3781 73. https://gcc.gnu.org/lists.html 3782 74. https://www.fsf.org/ 3783 75. https://gcc.gnu.org/about.html 3784 ====================================================================== 3785 http://gcc.gnu.org/gcc-8/index.html 3786 GCC 8 Release Series 3787 3788 (This release series is no longer supported.) 3789 3790 May 14, 2021 3791 3792 The [1]GNU project and the GCC developers are pleased to announce the 3793 release of GCC 8.5. 3794 3795 This release is a bug-fix release, containing fixes for regressions in 3796 GCC 8.4 relative to previous releases of GCC. 3797 3798 Release History 3799 3800 GCC 8.5 3801 May 14, 2021 ([2]changes, [3]documentation) 3802 3803 GCC 8.4 3804 Mar 4, 2020 ([4]changes, [5]documentation) 3805 3806 GCC 8.3 3807 Feb 22, 2019 ([6]changes, [7]documentation) 3808 3809 GCC 8.2 3810 Jul 26, 2018 ([8]changes, [9]documentation) 3811 3812 GCC 8.1 3813 May 2, 2018 ([10]changes, [11]documentation) 3814 3815 References and Acknowledgements 3816 3817 GCC used to stand for the GNU C Compiler, but since the compiler 3818 supports several other languages aside from C, it now stands for the 3819 GNU Compiler Collection. 3820 3821 A list of [12]successful builds is updated as new information becomes 3822 available. 3823 3824 The GCC developers would like to thank the numerous people that have 3825 contributed new features, improvements, bug fixes, and other changes as 3826 well as test results to GCC. This [13]amazing group of volunteers is 3827 what makes GCC successful. 3828 3829 For additional information about GCC please refer to the [14]GCC 3830 project web site or contact the [15]GCC development mailing list. 3831 3832 To obtain GCC please use [16]our mirror sites or [17]our version 3833 control system. 3834 3835 3836 For questions related to the use of GCC, please consult these web 3837 pages and the [18]GCC manuals. If that fails, the 3838 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3839 web pages and the development of GCC are welcome on our developer 3840 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3841 archives. 3842 3843 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3844 distribution of this entire article is permitted in any medium, 3845 provided this notice is preserved. 3846 3847 These pages are [23]maintained by the GCC team. Last modified 3848 2024-05-30. 3849 3850 References 3851 3852 1. http://www.gnu.org/ 3853 2. https://gcc.gnu.org/gcc-8/changes.html 3854 3. https://gcc.gnu.org/onlinedocs/8.5.0/ 3855 4. https://gcc.gnu.org/gcc-8/changes.html 3856 5. https://gcc.gnu.org/onlinedocs/8.4.0/ 3857 6. https://gcc.gnu.org/gcc-8/changes.html 3858 7. https://gcc.gnu.org/onlinedocs/8.3.0/ 3859 8. https://gcc.gnu.org/gcc-8/changes.html 3860 9. https://gcc.gnu.org/onlinedocs/8.2.0/ 3861 10. https://gcc.gnu.org/gcc-8/changes.html 3862 11. https://gcc.gnu.org/onlinedocs/8.1.0/ 3863 12. https://gcc.gnu.org/gcc-8/buildstat.html 3864 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 3865 14. https://gcc.gnu.org/index.html 3866 15. mailto:gcc (a] gcc.gnu.org 3867 16. https://gcc.gnu.org/mirrors.html 3868 17. https://gcc.gnu.org/git.html 3869 18. https://gcc.gnu.org/onlinedocs/ 3870 19. mailto:gcc-help (a] gcc.gnu.org 3871 20. mailto:gcc (a] gcc.gnu.org 3872 21. https://gcc.gnu.org/lists.html 3873 22. https://www.fsf.org/ 3874 23. https://gcc.gnu.org/about.html 3875 ====================================================================== 3876 http://gcc.gnu.org/gcc-8/changes.html 3877 GCC 8 Release Series 3878 Changes, New Features, and Fixes 3879 3880 This page is a "brief" summary of some of the huge number of 3881 improvements in GCC 8. You may also want to check out our [1]Porting to 3882 GCC 8 page and the [2]full GCC documentation. 3883 3884 Caveats 3885 3886 * Support for the obsolete SDB/coff debug info format has been 3887 removed. The option -gcoff no longer does anything. 3888 * The Cilk+ extensions to the C and C++ languages have been removed. 3889 * The MPX extensions to the C and C++ languages have been deprecated 3890 and will be removed in a future release. 3891 * The extension allowing arithmetic on std::atomic<void*> and types 3892 like std::atomic<R(*)()> has been deprecated. 3893 * The non-standard C++0x std::copy_exception function was removed. 3894 std::make_exception_ptr should be used instead. 3895 * Support for the powerpc*-*-*spe* target ports which have been 3896 recently unmaintained and untested in GCC has been declared 3897 obsolete in GCC 8 as [3]announced. Unless there is activity to 3898 revive them, the next release of GCC will have their sources 3899 permanently removed. 3900 3901 General Improvements 3902 3903 * Inter-procedural optimization improvements: 3904 + Reworked run-time estimation metrics leading to more realistic 3905 guesses driving inliner and cloning heuristics. 3906 + The ipa-pure-const pass is extended to propagate the malloc 3907 attribute, and the corresponding warning option 3908 -Wsuggest-attribute=malloc emits a diagnostic for functions 3909 which can be annotated with the malloc attribute. 3910 * Profile driven optimization improvements: 3911 + New infrastructure for representing profiles (both statically 3912 guessed and profile feedback) which allows propagation of 3913 additional information about the reliability of the profile. 3914 + A number of improvements in the profile updating code solving 3915 problems found by new verification code. 3916 + Static detection of code which is not executed in a valid run 3917 of the program. This includes paths which trigger undefined 3918 behavior as well as calls to functions declared with the cold 3919 attribute. Newly the noreturn attribute does not imply all 3920 effects of cold to differentiate between exit (which is 3921 noreturn) and abort (which is in addition not executed in 3922 valid runs). 3923 + -freorder-blocks-and-partition, a pass splitting function 3924 bodies into hot and cold regions, is now enabled by default at 3925 -O2 and higher for x86 and x86-64. 3926 * Link-time optimization improvements: 3927 + We have significantly improved debug information on ELF 3928 targets using DWARF by properly preserving language-specific 3929 information. This allows for example the libstdc++ 3930 pretty-printers to work with LTO optimized executables. 3931 * A new option -fcf-protection=[full|branch|return|none] is 3932 introduced to perform code instrumentation to increase program 3933 security by checking that target addresses of control-flow transfer 3934 instructions (such as indirect function call, function return, 3935 indirect jump) are valid. Currently the instrumentation is 3936 supported on x86 GNU/Linux targets only. See the user guide for 3937 further information about the option syntax and section "New 3938 Targets and Target Specific Improvements" for IA-32/x86-64 for more 3939 details. 3940 * The -gcolumn-info option is now enabled by default. It includes 3941 column information in addition to just filenames and line numbers 3942 in DWARF debugging information. 3943 * The polyhedral-based loop nest optimization pass 3944 -floop-nest-optimize has been overhauled. It's still considered 3945 experimental and may not result in any runtime improvements. 3946 * Two new classical loop nest optimization passes have been added. 3947 -floop-unroll-and-jam performs outer loop unrolling and fusing of 3948 the inner loop copies. -floop-interchange exchanges loops in a loop 3949 nest to improve data locality. Both passes are enabled by default 3950 at -O3 and above. 3951 * The classic loop nest optimization pass -ftree-loop-distribution 3952 has been improved and enabled by default at -O3 and above. It 3953 supports loop nest distribution in some restricted scenarios; it 3954 also supports cancellable innermost loop distribution with loop 3955 versioning under run-time alias checks. 3956 * The new option -fstack-clash-protection causes the compiler to 3957 insert probes whenever stack space is allocated statically or 3958 dynamically to reliably detect stack overflows and thus mitigate 3959 the attack vector that relies on jumping over a stack guard page as 3960 provided by the operating system. 3961 * A new pragma GCC unroll has been implemented in the C family of 3962 languages, as well as Fortran and Ada, so as to make it possible 3963 for the user to have a finer-grained control over the loop 3964 unrolling optimization. 3965 * GCC has been enhanced to detect more instances of meaningless or 3966 mutually exclusive attribute specifications and handle such 3967 conflicts more consistently. Mutually exclusive attribute 3968 specifications are ignored with a warning regardless of whether 3969 they appear on the same declaration or on distinct declarations of 3970 the same entity. For example, because the noreturn attribute on the 3971 second declaration below is mutually exclusive with the malloc 3972 attribute on the first, it is ignored and a warning is issued. 3973 > 3974 void* __attribute__ ((malloc)) f (unsigned); 3975 void* __attribute__ ((noreturn)) f (unsigned); 3976 3977 warning: ignoring attribute 'noreturn' because it conflicts with attribute 3978 'malloc' [-Wattributes] 3979 * The gcov tool can distinguish functions that begin on a same line 3980 in a source file. This can be a different template instantiation or 3981 a class constructor: 3982 3983 File 'ins.C' 3984 Lines executed:100.00% of 8 3985 Creating 'ins.C.gcov' 3986 3987 -: 0:Source:ins.C 3988 -: 0:Graph:ins.gcno 3989 -: 0:Data:ins.gcda 3990 -: 0:Runs:1 3991 -: 0:Programs:1 3992 -: 1:template<class T> 3993 -: 2:class Foo 3994 -: 3:{ 3995 -: 4: public: 3996 2: 5: Foo(): b (1000) {} 3997 ------------------ 3998 Foo<char>::Foo(): 3999 1: 5: Foo(): b (1000) {} 4000 ------------------ 4001 Foo<int>::Foo(): 4002 1: 5: Foo(): b (1000) {} 4003 ------------------ 4004 2: 6: void inc () { b++; } 4005 ------------------ 4006 Foo<char>::inc(): 4007 1: 6: void inc () { b++; } 4008 ------------------ 4009 Foo<int>::inc(): 4010 1: 6: void inc () { b++; } 4011 ------------------ 4012 -: 7: 4013 -: 8: private: 4014 -: 9: int b; 4015 -: 10:}; 4016 -: 11: 4017 1: 12:int main(int argc, char **argv) 4018 -: 13:{ 4019 1: 14: Foo<int> a; 4020 1: 15: Foo<char> b; 4021 -: 16: 4022 1: 17: a.inc (); 4023 1: 18: b.inc (); 4024 1: 19:} 4025 4026 * The gcov tool has more accurate numbers for execution of lines in a 4027 source file. 4028 * The gcov tool can use TERM colors to provide more readable output. 4029 * AddressSanitizer gained a new pair of sanitization options, 4030 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 4031 warn about subtraction (or comparison) of pointers that point to a 4032 different memory object: 4033 4034 int 4035 main () 4036 { 4037 /* Heap allocated memory. */ 4038 char *heap1 = (char *)__builtin_malloc (42); 4039 char *heap2 = (char *)__builtin_malloc (42); 4040 if (heap1 > heap2) 4041 return 1; 4042 4043 return 0; 4044 } 4045 4046 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 4047 00050 4048 #0 0x40070f in main /tmp/pointer-compare.c:7 4049 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4050 #2 0x400629 in _start (/tmp/a.out+0x400629) 4051 4052 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 4053 00000003a) 4054 allocated by thread T0 here: 4055 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 4056 _malloc_linux.cc:86 4057 #1 0x4006ea in main /tmp/pointer-compare.c:5 4058 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4059 4060 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 4061 00000007a) 4062 allocated by thread T0 here: 4063 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 4064 _malloc_linux.cc:86 4065 #1 0x4006f8 in main /tmp/pointer-compare.c:6 4066 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 4067 4068 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 4069 4070 * The store merging pass has been enhanced to handle bit-fields and 4071 not just constant stores, but also data copying from adjacent 4072 memory locations into other adjacent memory locations, including 4073 bitwise logical operations on the data. The pass can also handle 4074 byte swapping into memory locations. 4075 * The undefined behavior sanitizer gained two new options included in 4076 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 4077 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 4078 builtins, and -fsanitize=pointer-overflow which performs cheap run 4079 time tests for pointer wrapping. 4080 * A new attribute no_sanitize can be applied to functions to instruct 4081 the compiler not to do sanitization of the options provided as 4082 arguments to the attribute. Acceptable values for no_sanitize match 4083 those acceptable by the -fsanitize command-line option. 4084 4085 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 4086 f () { /* Do something. */; } 4087 4088 New Languages and Language specific improvements 4089 4090 Ada 4091 4092 * For its internal exception handling used on the host for error 4093 recovery in the front-end, the compiler now relies on the native 4094 exception handling mechanism of the host platform, which should be 4095 more efficient than the former mechanism. 4096 4097 BRIG (HSAIL) 4098 4099 In this release cycle, the focus for the BRIGFE was on stabilization 4100 and performance improvements. Also a couple of completely new features 4101 were added. 4102 * Improved support for function and module scope group segment 4103 variables. PRM specs define function and module scope group segment 4104 variables as an experimental feature. However, PRM test suite uses 4105 them. Now group segment is handled by separate book keeping of 4106 module scope and function (kernel) offsets. Each function has a 4107 "frame" in the group segment offset to which is given as an 4108 argument, similar to traditional call stack frame handling. 4109 * Reduce the number of type conversions due to the untyped HSAIL 4110 registers. Instead of always representing the HSAIL's untyped 4111 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 4112 code and builds the register variables as a type used the most when 4113 storing or reading data to/from each register. This reduces the 4114 number of total casts which cannot be always optimized away. 4115 * Support for BRIG_KIND_NONE directives. 4116 * Made -O3 the default optimization level for BRIGFE. 4117 * Fixed illegal addresses generated from address expressions which 4118 refer only to offset 0. 4119 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 4120 mode, the offset is treated as 32-bit unless it's in global, 4121 read-only or kernarg address space. 4122 * Fixed a crash caused sometimes by calls with more than 4 arguments. 4123 * Fixed a mis-execution issue with kernels that have both unexpanded 4124 ID functions and calls to subfunctions. 4125 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 4126 avoid illegal optimizations. 4127 * Ensure per WI copies of private variables are aligned correctly. 4128 * libhsail-rt: Assume the host runtime allocates the work group 4129 memory. 4130 4131 C family 4132 4133 * New command-line options have been added for the C and C++ 4134 compilers: 4135 + [4]-Wmultistatement-macros warns about unsafe macros expanding 4136 to multiple statements used as a body of a statement such as 4137 if, else, while, switch, or for. 4138 + [5]-Wstringop-truncation warns for calls to bounded string 4139 manipulation functions such as strncat, strncpy, and stpncpy 4140 that might either truncate the copied string or leave the 4141 destination unchanged. For example, the following call to 4142 strncat is diagnosed because it appends just three of the four 4143 characters from the source string. 4144 void append (char *buf, size_t bufsize) 4145 { 4146 strncat (buf, ".txt", 3); 4147 } 4148 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 4149 Wstringop-truncation] 4150 Similarly, in the following example, the call to strncpy 4151 specifies the size of the destination buffer as the bound. If 4152 the length of the source string is equal to or greater than 4153 this size the result of the copy will not be NUL-terminated. 4154 Therefore, the call is also diagnosed. To avoid the warning, 4155 specify sizeof buf - 1 as the bound and set the last element 4156 of the buffer to NUL. 4157 void copy (const char *s) 4158 { 4159 char buf[80]; 4160 strncpy (buf, s, sizeof buf); 4161 4162 } 4163 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 4164 tion] 4165 The -Wstringop-truncation option is included in -Wall. 4166 Note that due to GCC bug [6]82944, defining strncat, strncpy, 4167 or stpncpy as a macro in a system header as some 4168 implementations do, suppresses the warning. 4169 + [7]-Wif-not-aligned controls warnings issued in response to 4170 invalid uses of objects declared with attribute 4171 [8]warn_if_not_aligned. 4172 The -Wif-not-aligned option is included in -Wall. 4173 + [9]-Wmissing-attributes warns when a declaration of a function 4174 is missing one or more attributes that a related function is 4175 declared with and whose absence may adversely affect the 4176 correctness or efficiency of generated code. For example, in 4177 C++, the warning is issued when an explicit specialization of 4178 a primary template declared with attribute alloc_align, 4179 alloc_size, assume_aligned, format, format_arg, malloc, or 4180 nonnull is declared without it. Attributes deprecated, error, 4181 and warning suppress the warning. 4182 The -Wmissing-attributes option is included in -Wall. 4183 + [10]-Wpacked-not-aligned warns when a struct or union declared 4184 with attribute packed defines a member with an explicitly 4185 specified alignment greater than 1. Such a member will wind up 4186 under-aligned. For example, a warning will be issued for the 4187 definition of struct A in the following: 4188 struct __attribute__ ((aligned (8))) 4189 S8 { char a[8]; }; 4190 4191 struct __attribute__ ((packed)) A 4192 { 4193 struct S8 s8; 4194 }; 4195 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 4196 The -Wpacked-not-aligned option is included in -Wall. 4197 + -Wcast-function-type warns when a function pointer is cast to 4198 an incompatible function pointer. This warning is enabled by 4199 -Wextra. 4200 + -Wsizeof-pointer-div warns for suspicious divisions of the 4201 size of a pointer by the size of the elements it points to, 4202 which looks like the usual way to compute the array size but 4203 won't work out correctly with pointers. This warning is 4204 enabled by -Wall. 4205 + -Wcast-align=strict warns whenever a pointer is cast such that 4206 the required alignment of the target is increased. For 4207 example, warn if a char * is cast to an int * regardless of 4208 the target machine. 4209 + -fprofile-abs-path creates absolute path names in the .gcno 4210 files. This allows gcov to find the correct sources in 4211 projects where compilations occur with different working 4212 directories. 4213 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 4214 signed integer overflow is now undefined by default at all 4215 optimization levels. Using -fsanitize=signed-integer-overflow is 4216 now the preferred way to audit code, -Wstrict-overflow is 4217 deprecated. 4218 * The [11]-Warray-bounds option has been improved to detect more 4219 instances of out-of-bounds array indices and pointer offsets. For 4220 example, negative or excessive indices into flexible array members 4221 and string literals are detected. 4222 * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to 4223 detect many more instances of overlapping accesses to objects via 4224 restrict-qualified arguments to standard memory and string 4225 manipulation functions such as memcpy and strcpy. For example, the 4226 strcpy call in the function below attempts to truncate the string 4227 by replacing its initial characters with the last four. However, 4228 because the function writes the terminating NUL into a[4], the 4229 copies overlap and the call is diagnosed. 4230 void f (void) 4231 { 4232 char a[] = "abcd1234"; 4233 strcpy (a, a + 4); 4234 4235 } 4236 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 4237 4 [-Wrestrict] 4238 The -Wrestrict option is included in -Wall. 4239 * Several optimizer enhancements have enabled improvements to the 4240 [13]-Wformat-overflow and [14]-Wformat-truncation options. The 4241 warnings detect more instances of buffer overflow and truncation 4242 than in GCC 7 and are better at avoiding certain kinds of false 4243 positives. 4244 * When reporting mismatching argument types at a function call, the C 4245 and C++ compilers now underline both the argument and the pertinent 4246 parameter in the declaration. 4247 $ gcc arg-type-mismatch.cc 4248 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 4249 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 4250 [-fpermissive] 4251 return callee(first, second, third); 4252 ^~~~~~ 4253 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 4254 onst char*, float)' 4255 extern int callee(int one, const char *two, float three); 4256 ~~~~~~~~~~~~^~~ 4257 4258 * When reporting on unrecognized identifiers, the C and C++ compilers 4259 will now emit fix-it hints suggesting #include directives for 4260 various headers in the C and C++ standard libraries. 4261 $ gcc incomplete.c 4262 incomplete.c: In function 'test': 4263 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 4264 return NULL; 4265 ^~~~ 4266 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 4267 t to '#include <stddef.h>'? 4268 incomplete.c:1:1: 4269 +#include <stddef.h> 4270 const char *test(void) 4271 incomplete.c:3:10: 4272 return NULL; 4273 ^~~~ 4274 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 4275 ch function it appears in 4276 4277 $ gcc incomplete.cc 4278 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 4279 std::string s("hello world"); 4280 ^~~~~~ 4281 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 4282 forget to '#include <string>'? 4283 +#include <string> 4284 std::string s("hello world"); 4285 ^~~ 4286 4287 * The C and C++ compilers now use more intuitive locations when 4288 reporting on missing semicolons, and offer fix-it hints: 4289 $ gcc t.c 4290 t.c: In function 'test': 4291 t.c:3:12: error: expected ';' before '}' token 4292 return 42 4293 ^ 4294 ; 4295 } 4296 ~ 4297 4298 * When reporting on missing '}' and ')' tokens, the C and C++ 4299 compilers will now highlight the corresponding '{' and '(' token, 4300 issuing a 'note' if it's on a separate line: 4301 $ gcc unclosed.c 4302 unclosed.c: In function 'log_when_out_of_range': 4303 unclosed.c:12:50: error: expected ')' before '{' token 4304 && (temperature < MIN || temperature > MAX) { 4305 ^~ 4306 ) 4307 unclosed.c:11:6: note: to match this '(' 4308 if (logging_enabled && check_range () 4309 ^ 4310 4311 or highlighting it directly if it's on the same line: 4312 $ gcc unclosed-2.c 4313 unclosed-2.c: In function 'test': 4314 unclosed-2.c:8:45: error: expected ')' before '{' token 4315 if (temperature < MIN || temperature > MAX { 4316 ~ ^~ 4317 ) 4318 4319 They will also emit fix-it hints. 4320 4321 C++ 4322 4323 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 4324 convention, which changes the ABI for some uncommon code: 4325 + Passing an empty class as an argument now takes up no space on 4326 x86_64, as required by the psABI. 4327 + Passing or returning a class with only deleted copy and move 4328 constructors now uses the same calling convention as a class 4329 with a non-trivial copy or move constructor. This only affects 4330 C++17 mode, as in earlier standards passing or returning such 4331 a class was impossible. 4332 + WARNING: In GCC 8.1 the second change mistakenly also affects 4333 classes with a deleted copy constructor and defaulted trivial 4334 move constructor (bug [15]c++/86094). This issue is fixed in 4335 GCC 8.2 (-fabi-version=13). 4336 You can test whether these changes affect your code with -Wabi=11 4337 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 4338 problematic for your project, the GCC 7 ABI can be selected with 4339 -fabi-version=11. 4340 * The value of the C++11 alignof operator has been corrected to match 4341 C _Alignof (minimum alignment) rather than GNU __alignof__ 4342 (preferred alignment); on ia32 targets this means that 4343 alignof(double) is now 4 rather than 8. Code that wants the 4344 preferred alignment should use __alignof__ instead. 4345 * New command-line options have been added for the C++ compiler to 4346 control warnings: 4347 + [16]-Wclass-memaccess warns when objects of non-trivial class 4348 types are manipulated in potentially unsafe ways by raw memory 4349 functions such as memcpy, or realloc. The warning helps detect 4350 calls that bypass user-defined constructors or copy-assignment 4351 operators, corrupt virtual table pointers, data members of 4352 const-qualified types or references, or member pointers. The 4353 warning also detects calls that would bypass access controls 4354 to data members. For example, a call such as: 4355 memcpy (&std::cout, &std::cerr, sizeof std::cout); 4356 results in 4357 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 4358 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 4359 o trivial copy-assignment [-Wclass-memaccess] 4360 The -Wclass-memaccess option is included in -Wall. 4361 * The C++ front end has experimental support for some of the upcoming 4362 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 4363 including designated initializers, default member initializers for 4364 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 4365 lambda [=, this] captures, etc. For a full list of new features, 4366 see [17]the C++ status page. 4367 * When reporting on attempts to access private fields of a class or 4368 struct, the C++ compiler will now offer fix-it hints showing how to 4369 use an accessor function to get at the field in question, if one 4370 exists. 4371 $ gcc accessor.cc 4372 accessor.cc: In function 'void test(foo*)': 4373 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 4374 if (ptr->m_ratio >= 0.5) 4375 ^~~~~~~ 4376 accessor.cc:7:10: note: declared private here 4377 double m_ratio; 4378 ^~~~~~~ 4379 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 4380 foo::get_ratio() const' 4381 if (ptr->m_ratio >= 0.5) 4382 ^~~~~~~ 4383 get_ratio() 4384 4385 * The C++ compiler can now give you a hint if you use a macro before 4386 it was defined (e.g. if you mess up the order of your #include 4387 directives): 4388 $ gcc ordering.cc 4389 ordering.cc:2:24: error: expected ';' at end of member declaration 4390 virtual void clone() const OVERRIDE { } 4391 ^~~~~ 4392 ; 4393 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 4394 virtual void clone() const OVERRIDE { } 4395 ^~~~~~~~ 4396 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 4397 In file included from ordering.cc:5: 4398 c++11-compat.h:2: note: it was later defined here 4399 #define OVERRIDE override 4400 4401 4402 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 4403 you when you can use a static_cast, const_cast, or 4404 reinterpret_cast. 4405 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 4406 old-style-cast-fixits.cc: In function 'void test(void*)': 4407 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 4408 -Wold-style-cast] 4409 foo *f = (foo *)ptr; 4410 ^~~ 4411 ---------- 4412 static_cast<foo *> (ptr) 4413 4414 * When reporting on problems within extern "C" linkage 4415 specifications, the C++ compiler will now display the location of 4416 the start of the extern "C". 4417 $ gcc -c extern-c.cc 4418 extern-c.cc:3:1: error: template with C linkage 4419 template <typename T> void test (void); 4420 ^~~~~~~~ 4421 In file included from extern-c.cc:1: 4422 unclosed.h:1:1: note: 'extern "C"' linkage started here 4423 extern "C" { 4424 ^~~~~~~~~~ 4425 extern-c.cc:3:39: error: expected '}' at end of input 4426 template <typename T> void test (void); 4427 ^ 4428 In file included from extern-c.cc:1: 4429 unclosed.h:1:12: note: to match this '{' 4430 extern "C" { 4431 ^ 4432 4433 * When reporting on mismatching template types, the C++ compiler will 4434 now use color to highlight the mismatching parts of the template, 4435 and will elide the parameters that are common between two 4436 mismatching templates, printing [...] instead: 4437 $ gcc templates.cc 4438 templates.cc: In function 'void test()': 4439 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 4440 e>' to 'vector<int>' 4441 fn_1(vector<double> ()); 4442 ^~~~~~~~~~~~~~~~~ 4443 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 4444 ,double>' to 'map<[...],int>' 4445 fn_2(map<int, double>()); 4446 ^~~~~~~~~~~~~~~~~~ 4447 4448 Those [...] elided parameters can be seen using -fno-elide-type: 4449 $ gcc templates.cc -fno-elide-type 4450 templates.cc: In function 'void test()': 4451 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 4452 e>' to 'vector<int>' 4453 fn_1(vector<double> ()); 4454 ^~~~~~~~~~~~~~~~~ 4455 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 4456 ouble>' to 'map<int,int>' 4457 fn_2(map<int, double>()); 4458 ^~~~~~~~~~~~~~~~~~ 4459 4460 The C++ compiler has also gained an option 4461 -fdiagnostics-show-template-tree which visualizes such mismatching 4462 templates in a hierarchical form: 4463 $ gcc templates-2.cc -fdiagnostics-show-template-tree 4464 templates-2.cc: In function 'void test()': 4465 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 4466 ble>' to 'vector<int>' 4467 vector< 4468 [double != int]> 4469 fn_1(vector<double> ()); 4470 ^~~~~~~~~~~~~~~~~ 4471 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 4472 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 4473 ap<[...],vector<float>>,vector<float>>' 4474 map< 4475 map< 4476 [...], 4477 vector< 4478 [double != float]>>, 4479 vector< 4480 [double != float]>> 4481 fn_2(map<map<int, vector<double>>, vector<double>> ()); 4482 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4483 4484 which again works with -fno-elide-type: 4485 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 4486 templates-2.cc: In function 'void test()': 4487 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 4488 ble>' to 'vector<int>' 4489 vector< 4490 [double != int]> 4491 fn_1(vector<double> ()); 4492 ^~~~~~~~~~~~~~~~~ 4493 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 4494 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 4495 <int,vector<float>>,vector<float>>' 4496 map< 4497 map< 4498 int, 4499 vector< 4500 [double != float]>>, 4501 vector< 4502 [double != float]>> 4503 fn_2(map<map<int, vector<double>>, vector<double>> ()); 4504 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4505 4506 * Flowing off the end of a non-void function is considered 4507 unreachable and may be subject to optimization on that basis. As a 4508 result of this change, -Wreturn-type warnings are enabled by 4509 default for C++. 4510 4511 Runtime Library (libstdc++) 4512 4513 * Improved experimental support for C++17, including the following 4514 features: 4515 + Deduction guides to support class template argument deduction. 4516 + std::filesystem implementation. 4517 + std::char_traits<char> and std::char_traits<wchar_t> are 4518 usable in constant expressions. 4519 + std::to_chars and std::from_chars (for integers only, not for 4520 floating point types). 4521 * Experimental support for C++2a: std::to_address (thanks to Glen 4522 Fernandes) and std::endian. 4523 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 4524 entropy count for the random device, if known (thanks to Xi 4525 Ruoyao). 4526 * Support for std::experimental::source_location. 4527 * AddressSanitizer integration for std::vector, detecting 4528 out-of-range accesses to the unused capacity of a vector. 4529 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 4530 Mathematical Special Functions. 4531 4532 Fortran 4533 4534 * The main version of libfortran has been changed to 5. 4535 * Parameterized derived types, a major feature of Fortran 2003, have 4536 been implemented. 4537 * Partial support is provided for Fortran 2018 teams, which are 4538 hierarchical subsets of images that execute independently of other 4539 image subsets. 4540 * The maximum rank for arrays has been increased to 15, conforming to 4541 the Fortran 2008 standard. 4542 * Transformational intrinsics are now fully supported in 4543 initialization expressions. 4544 * New flag -fc-prototypes to write C prototypes for BIND(C) 4545 procedures and variables. 4546 * If -fmax-stack-var-size is honored if given together with -Ofast, 4547 -fstack-arrays is no longer set in that case. 4548 * New options -fdefault-real-16 and -fdefault-real-10 to control the 4549 default kind of REAL variables. 4550 * A warning is now issued if an array subscript inside a DO loop 4551 could lead to an out-of-bounds-access. The new option 4552 -Wdo-subscript, enabled by -Wextra, warns about this even if the 4553 compiler can not prove that the code will be executed. 4554 * The Fortran front end now attempts to interchange loops if it is 4555 deemed profitable. So far, this is restricted to FORALL and DO 4556 CONCURRENT statements with multiple indices. This behavior be 4557 controlled with the new flag -ffrontend-loop-interchange, which is 4558 enabled with optimization by default. The 4559 -Wfrontend-loop-interchange option warns about such occurrences. 4560 * When an actual argument contains too few elements for a dummy 4561 argument, an error is now issued. The -std=legacy option can be 4562 used to still compile such code. 4563 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 4564 integers, making records larger than 2GiB possible. 4565 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 4566 effect. The record length for preconnected units is now larger than 4567 any practical limit, same as for sequential access units opened 4568 without an explicit RECL= specifier. 4569 * Character variables longer than HUGE(0) elements are now possible 4570 on 64-bit targets. Note that this changes the procedure call ABI 4571 for all procedures with character arguments on 64-bit targets, as 4572 the type of the hidden character length argument has changed. The 4573 hidden character length argument is now of type INTEGER(C_SIZE_T). 4574 * Partial support is provided for Fortran 2018 teams, which are 4575 hierarchical subsets of images that execute independently of other 4576 image subsets. 4577 4578 Go 4579 4580 * GCC 8 provides a complete implementation of the Go 1.10.1 user 4581 packages. 4582 * The garbage collector is now fully concurrent. As before, values 4583 stored on the stack are scanned conservatively, but value stored in 4584 the heap are scanned precisely. 4585 * Escape analysis is fully implemented and enabled by default in the 4586 Go front end. This significantly reduces the number of heap 4587 allocations by allocating values on the stack instead. 4588 4589 libgccjit 4590 4591 The libgccjit API gained four new entry points: 4592 * [18]gcc_jit_type_get_vector and 4593 * [19]gcc_jit_context_new_rvalue_from_vector for working with 4594 vectors, 4595 * [20]gcc_jit_type_get_aligned 4596 * [21]gcc_jit_function_get_address 4597 4598 The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is 4599 now easier-to-read. 4600 4601 New Targets and Target Specific Improvements 4602 4603 AArch64 4604 4605 * The Armv8.4-A architecture is now supported. It can be used by 4606 specifying the -march=armv8.4-a option. 4607 * The Dot Product instructions are now supported as an optional 4608 extension to the Armv8.2-A architecture and newer and are mandatory 4609 on Armv8.4-A. The extension can be used by specifying the +dotprod 4610 architecture extension. E.g. -march=armv8.2-a+dotprod. 4611 * The Armv8-A +crypto extension has now been split into two 4612 extensions for finer grained control: 4613 + +aes which contains the Armv8-A AES crytographic instructions. 4614 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 4615 instructions. 4616 Using +crypto will now enable these two extensions. 4617 * New Armv8.4-A FP16 Floating Point Multiplication Variant 4618 instructions have been added. These instructions are mandatory in 4619 Armv8.4-A but available as an optional extension to Armv8.2-A and 4620 Armv8.3-A. The new extension can be used by specifying the +fp16fml 4621 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 4622 the instructions can be enabled by specifying +fp16. 4623 * New cryptographic instructions have been added as optional 4624 extensions to Armv8.2-A and newer. These instructions can be 4625 enabled with: 4626 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 4627 implies +sha2. 4628 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 4629 * The Scalable Vector Extension (SVE) is now supported as an optional 4630 extension to the Armv8.2-A architecture and newer. This support 4631 includes automatic vectorization with SVE instructions, but it does 4632 not yet include the SVE Arm C Language Extensions (ACLE). It can be 4633 enabled by specifying the +sve architecture extension (for example, 4634 -march=armv8.2-a+sve). By default, the generated code works with 4635 all vector lengths, but it can be made specific to N-bit vectors 4636 using -msve-vector-bits=N. 4637 * Support has been added for the following processors (GCC 4638 identifiers in parentheses): 4639 + Arm Cortex-A75 (cortex-a75). 4640 + Arm Cortex-A55 (cortex-a55). 4641 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 4642 (cortex-a75.cortex-a55). 4643 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4644 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 4645 arguments to the equivalent target attributes and pragmas. 4646 4647 ARC 4648 4649 * Added support for: 4650 + Fast interrupts. 4651 + Naked functions. 4652 + aux variable attributes. 4653 + uncached type qualifier. 4654 + Secure functions via sjli instruction. 4655 * New exception handling implementation. 4656 * Revamped trampoline implementation. 4657 * Refactored small data feature implementation, controlled via the -G 4658 command-line option. 4659 * New support for reduced register set ARC architecture 4660 configurations, controlled via the -mrf16 command-line option. 4661 * Refurbished and improved support for zero overhead loops. 4662 Introduced -mlpc-width command-line option to control the width of 4663 the lp_count register. 4664 4665 ARM 4666 4667 * The -mfpu option now takes a new option setting of -mfpu=auto. When 4668 set to this the floating-point and SIMD settings are derived from 4669 the settings of the -mcpu or -march options. The internal CPU 4670 configurations have been updated with information about the 4671 permitted floating-point configurations supported. See the user 4672 guide for further information about the extended option syntax for 4673 controlling architectural extensions via the -march option. 4674 -mfpu=auto is now the default setting unless the compiler has been 4675 configured with an explicit --with-fpu option. 4676 * The -march and -mcpu options now accept optional extensions to the 4677 architecture or CPU option, allowing the user to enable or disable 4678 any such extensions supported by that architecture or CPU such as 4679 (but not limited to) floating-point and AdvancedSIMD. For example: 4680 the option -mcpu=cortex-a53+nofp will generate code for the 4681 Cortex-A53 processor with no floating-point support. This, in 4682 combination with the new -mfpu=auto option, provides a 4683 straightforward way of specifying a valid build target through a 4684 single -mcpu or -march option. The -mtune option accepts the same 4685 arguments as -mcpu but only the CPU name has an effect on tuning. 4686 The architecture extensions do not have any effect. For details of 4687 what extensions a particular architecture or CPU option supports 4688 please refer to the [23]documentation. 4689 * The -mstructure-size-boundary option has been deprecated and will 4690 be removed in a future release. 4691 * The default link behavior for Armv6 and Armv7-R targets has been 4692 changed to produce BE8 format when generating big-endian images. A 4693 new flag -mbe32 can be used to force the linker to produce legacy 4694 BE32 format images. There is no change of behavior for Armv6-M and 4695 other Armv7 or later targets: these already defaulted to BE8 4696 format. This change brings GCC into alignment with other compilers 4697 for the ARM architecture. 4698 * The Armv8-R architecture is now supported. It can be used by 4699 specifying the -march=armv8-r option. 4700 * The Armv8.3-A architecture is now supported. It can be used by 4701 specifying the -march=armv8.3-a option. 4702 * The Armv8.4-A architecture is now supported. It can be used by 4703 specifying the -march=armv8.4-a option. 4704 * The Dot Product instructions are now supported as an optional 4705 extension to the Armv8.2-A architecture and newer and are mandatory 4706 on Armv8.4-A. The extension can be used by specifying the +dotprod 4707 architecture extension. E.g. -march=armv8.2-a+dotprod. 4708 * Support for setting extensions and architectures using the GCC 4709 target pragma and attribute has been added. It can be used by 4710 specifying #pragma GCC target ("arch=..."), #pragma GCC target 4711 ("+extension"), __attribute__((target("arch=..."))) or 4712 __attribute__((target("+extension"))). 4713 * New Armv8.4-A FP16 Floating Point Multiplication Variant 4714 instructions have been added. These instructions are mandatory in 4715 Armv8.4-A but available as an optional extension to Armv8.2-A and 4716 Armv8.3-A. The new extension can be used by specifying the +fp16fml 4717 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 4718 the instructions can be enabled by specifying +fp16. 4719 * Support has been added for the following processors (GCC 4720 identifiers in parentheses): 4721 + Arm Cortex-A75 (cortex-a75). 4722 + Arm Cortex-A55 (cortex-a55). 4723 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 4724 (cortex-a75.cortex-a55). 4725 + Arm Cortex-R52 for Armv8-R (cortex-r52). 4726 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4727 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 4728 arguments to the equivalent target attributes and pragmas. 4729 4730 AVR 4731 4732 * The AVR port now supports the following XMEGA-like devices: 4733 4734 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 4735 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 4736 ATtiny3214, ATtiny3216, ATtiny3217 4737 The new devices are listed under [24]-mmcu=avrxmega3. 4738 + These devices see flash memory in the RAM address space, so 4739 that features like PROGMEM and __flash are not needed any more 4740 (as opposed to other AVR families for which read-only data 4741 will be located in RAM except special, non-standard features 4742 are used to locate and access such data). This requires that 4743 the compiler is used with Binutils 2.29 or newer so that 4744 [25]read-only data will be located in flash memory. 4745 + A new command-line option -mshort-calls is supported. This 4746 option is used internally for multilib selection of the 4747 avrxmega3 variants. It is not an optimization option. Do not 4748 set it by hand. 4749 * The compiler now generates [26]efficient interrupt service routine 4750 (ISR) prologues and epilogues. This is achieved by using the new 4751 [27]AVR pseudo instruction __gcc_isr which is supported and 4752 resolved by the GNU assembler. 4753 + As the __gcc_isr pseudo-instruction will be resolved by the 4754 assembler, inline assembly is transparent to the process. This 4755 means that when inline assembly uses an instruction like INC 4756 that clobbers the condition code, then the assembler will 4757 detect this and generate an appropriate ISR prologue / 4758 epilogue chunk to save / restore SREG as needed. 4759 + A new command-line option -mno-gas-isr-prologues disables the 4760 generation of the __gcc_isr pseudo instruction. Any non-naked 4761 ISR will save and restore SREG, tmp_reg and zero_reg, no 4762 matter whether the respective register is clobbered or used. 4763 + The feature is turned on per default for all optimization 4764 levels except for -O0 and -Og. It is explicitly enabled by 4765 means of option -mgas-isr-prologues. 4766 + Support has been added for a new [28]AVR function attribute 4767 no_gccisr. It can be used to disable __gcc_isr pseudo 4768 instruction generation for individual ISRs. 4769 + This optimization is only available if GCC is configured with 4770 GNU Binutils 2.29 or newer; or at least with a version of 4771 Binutils that implements feature [29]PR21683. 4772 * The compiler no more saves / restores registers in main; the effect 4773 is the same as if attribute OS_task was specified for main. This 4774 optimization can be switched off by the new command-line option 4775 -mno-main-is-OS_task. 4776 4777 IA-32/x86-64 4778 4779 * The x86 port now supports the naked function attribute. 4780 * Better tuning for znver1 and Intel Core based CPUs. 4781 * Vectorization cost metrics has been reworked leading to significant 4782 improvements on some benchmarks. 4783 * GCC now supports the Intel CPU named Cannonlake through 4784 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 4785 and SHA ISA extensions. 4786 * GCC now supports the Intel CPU named Icelake through 4787 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 4788 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 4789 ISA extensions. 4790 * GCC now supports the Intel Control-flow Enforcement Technology 4791 (CET) extension through -fcf-protection option. 4792 4793 NDS32 4794 4795 * New command-line options -mext-perf, -mext-perf2, and -mext-string 4796 have been added for performance extension instructions. 4797 4798 Nios II 4799 4800 * The Nios II back end has been improved to generate better-optimized 4801 code. Changes include switching to LRA, more accurate cost models, 4802 and more compact code for addressing static variables. 4803 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 4804 added. 4805 * The stack-smashing protection options are now enabled on Nios II. 4806 4807 PA-RISC 4808 4809 * The default call ABI on 32-bit linux has been changed from callee 4810 copies to caller copies. This affects objects larger than eight 4811 bytes passed by value. The goal is to improve compatibility with 4812 x86 and resolve issues with OpenMP. 4813 * Other PA-RISC targets are unchanged. 4814 4815 PowerPC / PowerPC64 / RS6000 4816 4817 * The PowerPC SPE support is split off to a separate powerpcspe port. 4818 The separate port is deprecated and might be removed in a future 4819 release. 4820 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 4821 powerpc*-*-linux*paired*) is deprecated and will be removed in a 4822 future release. 4823 * The Xilinx floating point support (-mxilinx-fpu, 4824 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 4825 release. 4826 * Support for using big-endian AltiVec intrinsics on a little-endian 4827 target (-maltivec=be) is deprecated and will be removed in a future 4828 release. 4829 4830 Tile 4831 4832 * The TILE-Gx port is deprecated and will be removed in a future 4833 release. 4834 4835 Operating Systems 4836 4837 Windows 4838 4839 * GCC on Microsoft Windows can now be configured via 4840 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 4841 specific behavior for GCC itself with regards to supporting the 4842 wildcard character. Prior versions of GCC would follow the 4843 configuration of the MinGW runtime. This behavior can still be 4844 obtained by not using the above options or by using 4845 --enable-mingw-wildcard=platform. 4846 4847 Improvements for plugin authors 4848 4849 * Plugins can now register a callback hook for when comments are 4850 encountered by the C and C++ compilers, e.g. allowing for plugins 4851 to handle documentation markup in code comments. 4852 * The gdbinit support script for debugging GCC now has a 4853 break-on-diagnostic command, providing an easy way to trigger a 4854 breakpoint whenever a diagnostic is emitted. 4855 * The API for creating fix-it hints now supports newlines, and for 4856 emitting mutually incompatible fix-it hints for one diagnostic. 4857 4858 GCC 8.1 4859 4860 This is the [30]list of problem reports (PRs) from GCC's bug tracking 4861 system that are known to be fixed in the 8.1 release. This list might 4862 not be complete (that is, it is possible that some PRs that have been 4863 fixed are not listed here). 4864 4865 GCC 8.2 4866 4867 This is the [31]list of problem reports (PRs) from GCC's bug tracking 4868 system that are known to be fixed in the 8.2 release. This list might 4869 not be complete (that is, it is possible that some PRs that have been 4870 fixed are not listed here). 4871 4872 General Improvements 4873 4874 * Fixed LTO link-time performance problems caused by an overflow in 4875 the partitioning algorithm while building large binaries. 4876 4877 Language Specific Changes 4878 4879 C++ 4880 4881 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 4882 returning of classes with a deleted copy constructor and defaulted 4883 trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces 4884 -fabi-version=13 and makes it the default, ABI incompatibilities 4885 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++ 4886 changes for more details. 4887 4888 Target Specific Changes 4889 4890 IA-32/x86-64 4891 4892 * -mtune=native performance regression [34]PR84413 on Intel Skylake 4893 processors has been fixed. 4894 4895 GCC 8.3 4896 4897 This is the [35]list of problem reports (PRs) from GCC's bug tracking 4898 system that are known to be fixed in the 8.3 release. This list might 4899 not be complete (that is, it is possible that some PRs that have been 4900 fixed are not listed here). 4901 4902 Windows 4903 4904 * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been 4905 fixed. A non-field declaration could cause the current bitfield 4906 allocation unit to be completed, incorrectly placing a following 4907 bitfield into a new allocation unit. The Microsoft ABI is selected 4908 for: 4909 + Mingw targets 4910 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 4911 option is specified, or __attribute__((ms_struct)) is used 4912 + SuperH targets when the -mhitachi option is specified, or 4913 __attribute__((renesas)) is used 4914 GCC 8 introduced additional cases of this defect, but rather than 4915 resolve only those regressions, we decided to resolve all the cases 4916 of this defect in single change. 4917 4918 GCC 8.4 4919 4920 This is the [37]list of problem reports (PRs) from GCC's bug tracking 4921 system that are known to be fixed in the 8.4 release. This list might 4922 not be complete (that is, it is possible that some PRs that have been 4923 fixed are not listed here). 4924 4925 GCC 8.5 4926 4927 This is the [38]list of problem reports (PRs) from GCC's bug tracking 4928 system that are known to be fixed in the 8.5 release. This list might 4929 not be complete (that is, it is possible that some PRs that have been 4930 fixed are not listed here). 4931 4932 Target Specific Changes 4933 4934 AArch64 4935 4936 * The option -moutline-atomics has been added to aid deployment of 4937 the Large System Extensions (LSE) on GNU/Linux systems built with a 4938 baseline architecture targeting Armv8-A. When the option is 4939 specified code is emitted to detect the presence of LSE 4940 instructions at run time and use them for standard atomic 4941 operations. For more information please refer to the documentation. 4942 4943 4944 For questions related to the use of GCC, please consult these web 4945 pages and the [39]GCC manuals. If that fails, the 4946 [40]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4947 web pages and the development of GCC are welcome on our developer 4948 list at [41]gcc (a] gcc.gnu.org. All of [42]our lists have public 4949 archives. 4950 4951 Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and 4952 distribution of this entire article is permitted in any medium, 4953 provided this notice is preserved. 4954 4955 These pages are [44]maintained by the GCC team. Last modified 4956 2023-01-11. 4957 4958 References 4959 4960 1. https://gcc.gnu.org/gcc-8/porting_to.html 4961 2. https://gcc.gnu.org/onlinedocs/index.html#current 4962 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 4963 4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 4964 5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 4965 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 4966 7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 4967 8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 4968 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 4969 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 4970 11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 4971 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 4972 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 4973 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 4974 15. https://gcc.gnu.org/PR86094 4975 16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 4976 17. https://gcc.gnu.org/projects/cxx-status.html#cxx2a 4977 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 4978 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 4979 20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 4980 21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 4981 22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 4982 23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 4983 24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 4984 25. https://sourceware.org/PR21472 4985 26. https://gcc.gnu.org/PR20296 4986 27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 4987 28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 4988 29. https://sourceware.org/PR21683 4989 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 4990 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 4991 32. https://gcc.gnu.org/PR86094 4992 33. https://gcc.gnu.org/gcc-8/changes.html#cxx 4993 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 4994 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 4995 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 4996 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 4997 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5 4998 39. https://gcc.gnu.org/onlinedocs/ 4999 40. mailto:gcc-help (a] gcc.gnu.org 5000 41. mailto:gcc (a] gcc.gnu.org 5001 42. https://gcc.gnu.org/lists.html 5002 43. https://www.fsf.org/ 5003 44. https://gcc.gnu.org/about.html 5004 ====================================================================== 5005 http://gcc.gnu.org/gcc-7/index.html 5006 GCC 7 Release Series 5007 5008 (This release series is no longer supported.) 5009 5010 Nov 14, 2019 5011 5012 The [1]GNU project and the GCC developers are pleased to announce the 5013 release of GCC 7.5. 5014 5015 This release is a bug-fix release, containing fixes for regressions in 5016 GCC 7.4 relative to previous releases of GCC. 5017 5018 Release History 5019 5020 GCC 7.5 5021 Nov 14, 2019 ([2]changes, [3]documentation) 5022 5023 GCC 7.4 5024 Dec 6, 2018 ([4]changes, [5]documentation) 5025 5026 GCC 7.3 5027 Jan 25, 2018 ([6]changes, [7]documentation) 5028 5029 GCC 7.2 5030 Aug 14, 2017 ([8]changes, [9]documentation) 5031 5032 GCC 7.1 5033 May 2, 2017 ([10]changes, [11]documentation) 5034 5035 References and Acknowledgements 5036 5037 GCC used to stand for the GNU C Compiler, but since the compiler 5038 supports several other languages aside from C, it now stands for the 5039 GNU Compiler Collection. 5040 5041 A list of [12]successful builds is updated as new information becomes 5042 available. 5043 5044 The GCC developers would like to thank the numerous people that have 5045 contributed new features, improvements, bug fixes, and other changes as 5046 well as test results to GCC. This [13]amazing group of volunteers is 5047 what makes GCC successful. 5048 5049 For additional information about GCC please refer to the [14]GCC 5050 project web site or contact the [15]GCC development mailing list. 5051 5052 To obtain GCC please use [16]our mirror sites or [17]our version 5053 control system. 5054 5055 5056 For questions related to the use of GCC, please consult these web 5057 pages and the [18]GCC manuals. If that fails, the 5058 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5059 web pages and the development of GCC are welcome on our developer 5060 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 5061 archives. 5062 5063 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5064 distribution of this entire article is permitted in any medium, 5065 provided this notice is preserved. 5066 5067 These pages are [23]maintained by the GCC team. Last modified 5068 2024-05-30. 5069 5070 References 5071 5072 1. http://www.gnu.org/ 5073 2. https://gcc.gnu.org/gcc-7/changes.html 5074 3. https://gcc.gnu.org/onlinedocs/7.5.0/ 5075 4. https://gcc.gnu.org/gcc-7/changes.html 5076 5. https://gcc.gnu.org/onlinedocs/7.4.0/ 5077 6. https://gcc.gnu.org/gcc-7/changes.html 5078 7. https://gcc.gnu.org/onlinedocs/7.3.0/ 5079 8. https://gcc.gnu.org/gcc-7/changes.html 5080 9. https://gcc.gnu.org/onlinedocs/7.2.0/ 5081 10. https://gcc.gnu.org/gcc-7/changes.html 5082 11. https://gcc.gnu.org/onlinedocs/7.1.0/ 5083 12. https://gcc.gnu.org/gcc-7/buildstat.html 5084 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5085 14. https://gcc.gnu.org/index.html 5086 15. mailto:gcc (a] gcc.gnu.org 5087 16. https://gcc.gnu.org/mirrors.html 5088 17. https://gcc.gnu.org/git.html 5089 18. https://gcc.gnu.org/onlinedocs/ 5090 19. mailto:gcc-help (a] gcc.gnu.org 5091 20. mailto:gcc (a] gcc.gnu.org 5092 21. https://gcc.gnu.org/lists.html 5093 22. https://www.fsf.org/ 5094 23. https://gcc.gnu.org/about.html 5095 ====================================================================== 5096 http://gcc.gnu.org/gcc-7/changes.html 5097 GCC 7 Release Series 5098 Changes, New Features, and Fixes 5099 5100 This page is a brief summary of some of the huge number of improvements 5101 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 5102 the [2]full GCC documentation. 5103 5104 Caveats 5105 5106 * GCC now uses [3]LRA (a new local register allocator) by default for 5107 new targets. 5108 * The non-standard C++0x type traits has_trivial_default_constructor, 5109 has_trivial_copy_constructor and has_trivial_copy_assign have been 5110 removed. 5111 * The libstdc++ [4]Profile Mode has been deprecated and will be 5112 removed in a future version. 5113 * The Cilk+ extensions to the C and C++ languages have been 5114 deprecated. 5115 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 5116 affects conformance to the procedure call standard (AAPCS) has been 5117 fixed. The bug affects some C++ code where class objects are passed 5118 by value to functions and could result in incorrect or inconsistent 5119 code being generated. This is an ABI change. If the option -Wpsabi 5120 is enabled (on by default) the compiler will emit a diagnostic note 5121 for code that might be affected. 5122 5123 General Optimizer Improvements 5124 5125 * GCC 7 can determine the return value or range of return values of 5126 some calls to the sprintf family of functions and make it available 5127 to other optimization passes. Some calls to the snprintf function 5128 with a zero size argument can be folded into constants. This 5129 optimization is included in -O1 and can be selectively controlled 5130 by the -fprintf-return-value option. 5131 * A new store merging pass has been added. It merges constant stores 5132 to adjacent memory locations into fewer, wider, stores. It is 5133 enabled by the -fstore-merging option and at the -O2 optimization 5134 level or higher (and -Os). 5135 * A new code hoisting optimization has been added to the partial 5136 redundancy elimination pass. It attempts to move evaluation of 5137 expressions executed on all paths to the function exit as early as 5138 possible. This primarily helps improve code size, but can improve 5139 the speed of the generated code as well. It is enabled by the 5140 -fcode-hoisting option and at the -O2 optimization level or higher 5141 (and -Os). 5142 * A new interprocedural bitwise constant propagation optimization has 5143 been added, which propagates knowledge about which bits of 5144 variables are known to be zero (including pointer alignment 5145 information) across the call graph. It is enabled by the 5146 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 5147 at the -O2 optimization level and higher (and -Os). This 5148 optimization supersedes interprocedural alignment propagation of 5149 GCC 6, and therefore the option -fipa-cp-alignment is now 5150 deprecated and ignored. 5151 * A new interprocedural value range propagation optimization has been 5152 added, which propagates integral range information across the call 5153 graph when variable values can be proven to be within those ranges. 5154 It is enabled by the -fipa-vrp option and at the -O2 optimization 5155 level and higher (and -Os). 5156 * A new loop splitting optimization pass has been added. Certain 5157 loops which contain a condition that is always true on one side of 5158 the iteration space and always false on the other are split into 5159 two loops, such that each of the two new loops iterates on just one 5160 side of the iteration space and the condition does not need to be 5161 checked inside of the loop. It is enabled by the -fsplit-loops 5162 option and at the -O3 optimization level or higher. 5163 * The shrink-wrapping optimization can now separate portions of 5164 prologues and epilogues to improve performance if some of the work 5165 done traditionally by prologues and epilogues is not needed on 5166 certain paths. This is controlled by the -fshrink-wrap-separate 5167 option, enabled by default. It requires target support, which is 5168 currently only implemented in the PowerPC and AArch64 ports. 5169 * AddressSanitizer gained a new sanitization option, 5170 -fsanitize-address-use-after-scope, which enables sanitization of 5171 variables whose address is taken and used after a scope where the 5172 variable is defined: 5173 5174 int 5175 main (int argc, char **argv) 5176 { 5177 char *ptr; 5178 { 5179 char my_char; 5180 ptr = &my_char; 5181 } 5182 5183 *ptr = 123; 5184 return *ptr; 5185 } 5186 5187 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 5188 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 5189 WRITE of size 1 at 0x7fffb8dba990 thread T0 5190 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 5191 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 5192 #2 0x400739 in _start (/tmp/a.out+0x400739) 5193 5194 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 5195 #0 0x40067f in main /tmp/use-after-scope-1.c:3 5196 5197 This frame has 1 object(s): 5198 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 5199 5200 The option is enabled by default with -fsanitize=address and 5201 disabled by default with -fsanitize=kernel-address. Compared to the 5202 LLVM compiler, where the option already exists, the implementation 5203 in the GCC compiler has some improvements and advantages: 5204 + Complex uses of gotos and case labels are properly handled and 5205 should not report any false positive or false negatives. 5206 + C++ temporaries are sanitized. 5207 + Sanitization can handle invalid memory stores that are 5208 optimized out by the LLVM compiler when optimization is 5209 enabled. 5210 * The -fsanitize=signed-integer-overflow suboption of the 5211 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 5212 on arithmetic operations with generic vectors. 5213 * Version 5 of the DWARF debugging information standard is supported 5214 through the -gdwarf-5 option. The DWARF version 4 debugging 5215 information remains the default until consumers of debugging 5216 information are adjusted. 5217 5218 New Languages and Language specific improvements 5219 5220 OpenACC support in C, C++, and Fortran continues to be maintained and 5221 improved. See the [6]OpenACC and [7]Offloading wiki pages for further 5222 information. 5223 5224 Ada 5225 5226 * On mainstream native platforms, Ada programs no longer require the 5227 stack to be made executable in order to run properly. 5228 5229 BRIG (HSAIL) 5230 5231 Support for processing BRIG 1.0 files was added in this release. BRIG 5232 is a binary format for HSAIL (Heterogeneous System Architecture 5233 Intermediate Language). The BRIG front end can be used for implementing 5234 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 5235 GCC-supported targets. An implementation of an HSAIL runtime library, 5236 libhsail-rt is also included. 5237 5238 C family 5239 5240 * New command-line options have been added for the C and C++ 5241 compilers: 5242 + -Wimplicit-fallthrough warns when a switch case falls through. 5243 This warning has five different levels. The compiler is able 5244 to parse a wide range of fallthrough comments, depending on 5245 the level. It also handles control-flow statements, such as 5246 ifs. It's possible to suppress the warning by either adding a 5247 fallthrough comment, or by using a null statement: 5248 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 5249 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 5250 is enabled by -Wextra. 5251 + -Wpointer-compare warns when a pointer is compared with a zero 5252 character constant. Such code is now invalid in C++11 and GCC 5253 rejects it. This warning is enabled by default. 5254 + -Wduplicated-branches warns when an if-else has identical 5255 branches. 5256 + -Wrestrict warns when an argument passed to a 5257 restrict-qualified parameter aliases with another argument. 5258 + -Wmemset-elt-size warns for memset calls, when the first 5259 argument references an array, and the third argument is a 5260 number equal to the number of elements of the array, but not 5261 the size of the array. This warning is enabled by -Wall. 5262 + -Wint-in-bool-context warns about suspicious uses of integer 5263 values where boolean values are expected. This warning is 5264 enabled by -Wall. 5265 + -Wswitch-unreachable warns when a switch statement has 5266 statements between the controlling expression and the first 5267 case label which will never be executed. This warning is 5268 enabled by default. 5269 + -Wexpansion-to-defined warns when defined is used outside #if. 5270 This warning is enabled by -Wextra or -Wpedantic. 5271 + -Wregister warns about uses of the register storage specifier. 5272 In C++17 this keyword has been removed and for C++17 this is a 5273 pedantic warning enabled by default. The warning is not 5274 emitted for the GNU Explicit Register Variables extension. 5275 + -Wvla-larger-than=N warns about unbounded uses of 5276 variable-length arrays, and about bounded uses of 5277 variable-length arrays whose bound can be larger than N bytes. 5278 + -Wduplicate-decl-specifier warns when a declaration has 5279 duplicate const, volatile, restrict or _Atomic specifier. This 5280 warning is enabled by -Wall. 5281 * GCC 6's C and C++ front ends were able to offer suggestions for 5282 misspelled field names: 5283 5284 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 5285 you mean 'color'? 5286 return ptr->colour; 5287 ^~~~~~ 5288 5289 GCC 7 greatly expands the scope of these suggestions. Firstly, it 5290 adds fix-it hints to such suggestions: 5291 5292 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 5293 you mean 'color'? 5294 return ptr->colour; 5295 ^~~~~~ 5296 color 5297 5298 The suggestions now cover many other things, such as misspelled 5299 function names: 5300 5301 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 5302 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 5303 ion] 5304 gtk_widget_showall (w); 5305 ^~~~~~~~~~~~~~~~~~ 5306 gtk_widget_show_all 5307 5308 misspelled macro names and enum values: 5309 5310 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 5311 ction); did you mean 'MAX_ITEMS'? 5312 int array[MAX_ITEM]; 5313 ^~~~~~~~ 5314 MAX_ITEMS 5315 5316 misspelled type names: 5317 5318 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 5319 gned'? 5320 void test (singed char e); 5321 ^~~~~~ 5322 signed 5323 5324 and, in the C front end, named initializers: 5325 5326 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 5327 '? 5328 struct s test = { .colour = 3 }; 5329 ^~~~~~ 5330 color 5331 5332 * The preprocessor can now offer suggestions for misspelled 5333 directives, e.g.: 5334 5335 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 5336 #endfi 5337 ^~~~~ 5338 endif 5339 5340 * Warnings about format strings now underline the pertinent part of 5341 the string, and can offer suggested fixes. In some cases, the 5342 pertinent argument is underlined. 5343 5344 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 5345 nt 3 has type 'int' [-Wformat=] 5346 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 5347 ~^ ~~~~~ 5348 %d 5349 5350 * The new -Wdangling-else command-line option has been split out of 5351 -Wparentheses and warns about dangling else. 5352 * The -Wshadow warning has been split into three variants. 5353 -Wshadow=global warns for any shadowing. This is the default when 5354 using -Wshadow without any argument. -Wshadow=local only warns for 5355 a local variable shadowing another local variable or parameter. 5356 -Wshadow=compatible-local only warns for a local variable shadowing 5357 another local variable or parameter whose type is compatible (in 5358 C++ compatible means that the type of the shadowing variable can be 5359 converted to that of the shadowed variable). 5360 The following example shows the different kinds of shadow warnings: 5361 5362 enum operation { add, count }; 5363 struct container { int nr; }; 5364 5365 int 5366 container_count (struct container c, int count) 5367 { 5368 int r = 0; 5369 for (int count = 0; count > 0; count--) 5370 { 5371 struct container count = c; 5372 r += count.nr; 5373 } 5374 return r; 5375 } 5376 5377 -Wshadow=compatible-local will warn for the parameter being 5378 shadowed with the same type: 5379 5380 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 5381 compatible-local] 5382 for (int count = 0; count > 0; count--) 5383 ^~~~~ 5384 warn-test.c:5:42: note: shadowed declaration is here 5385 container_count (struct container c, int count) 5386 ^~~~~ 5387 5388 -Wshadow=local will warn for the above and for the shadowed 5389 declaration with incompatible type: 5390 5391 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 5392 hadow=local] 5393 struct container count = c; 5394 ^~~~~ 5395 warn-test.c:8:12: note: shadowed declaration is here 5396 for (int count = 0; count > 0; count--) 5397 ^~~~~ 5398 5399 -Wshadow=global will warn for all of the above and the shadowing of 5400 the global declaration: 5401 5402 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 5403 -Wshadow] 5404 container_count (struct container c, int count) 5405 ^~~~~ 5406 warn-test.c:1:23: note: shadowed declaration is here 5407 enum operation { add, count }; 5408 ^~~~~ 5409 5410 * GCC 7 contains a number of enhancements that help detect buffer 5411 overflow and other forms of invalid memory accesses. 5412 + The -Walloc-size-larger-than=size option detects calls to 5413 standard and user-defined memory allocation functions 5414 decorated with attribute alloc_size whose argument exceeds the 5415 specified size (PTRDIFF_MAX by default). The option also 5416 detects arithmetic overflow in the computation of the size in 5417 two-argument allocation functions like calloc where the total 5418 size is the product of the two arguments. Since calls with an 5419 excessive size cannot succeed they are typically the result of 5420 programming errors. Such bugs have been known to be the source 5421 of security vulnerabilities and a target of exploits. 5422 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 5423 For example, the following call to malloc incorrectly tries to 5424 avoid passing a negative argument to the function and instead 5425 ends up unconditionally invoking it with an argument less than 5426 or equal to zero. Since after conversion to the type of the 5427 argument of the function (size_t) a negative argument results 5428 in a value in excess of the maximum PTRDIFF_MAX the call is 5429 diagnosed. 5430 5431 void* f (int n) 5432 { 5433 return malloc (n > 0 ? 0 : n); 5434 } 5435 5436 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 5437 147483647 [-Walloc-size-larger-than=] 5438 5439 + The -Walloc-zero option detects calls to standard and 5440 user-defined memory allocation functions decorated with 5441 attribute alloc_size with a zero argument. -Walloc-zero is not 5442 included in either -Wall or -Wextra and must be explicitly 5443 enabled. 5444 + The -Walloca option detects all calls to the alloca function 5445 in the program. -Walloca is not included in either -Wall or 5446 -Wextra and must be explicitly enabled. 5447 + The -Walloca-larger-than=size option detects calls to the 5448 alloca function whose argument either may exceed the specified 5449 size, or that is not known to be sufficiently constrained to 5450 avoid exceeding it. -Walloca-larger-than is not included in 5451 either -Wall or -Wextra and must be explicitly enabled. 5452 For example, compiling the following snippet with 5453 -Walloca-larger-than=1024 results in a warning because even 5454 though the code appears to call alloca only with sizes of 1kb 5455 and less, since n is signed, a negative value would result in 5456 a call to the function well in excess of the limit. 5457 5458 void f (int n) 5459 { 5460 char *d; 5461 if (n < 1025) 5462 d = alloca (n); 5463 else 5464 d = malloc (n); 5465 5466 } 5467 5468 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 5469 ong unsigned int' [-Walloca-larger-than=] 5470 5471 In contrast, a call to alloca that isn't bounded at all such 5472 as in the following function will elicit the warning below 5473 regardless of the size argument to the option. 5474 5475 void f (size_t n) 5476 { 5477 char *d = alloca (n); 5478 5479 } 5480 5481 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 5482 5483 + The -Wformat-overflow=level option detects certain and likely 5484 buffer overflow in calls to the sprintf family of formatted 5485 output functions. Although the option is enabled even without 5486 optimization it works best with -O2 and higher. 5487 For example, in the following snippet the call to sprintf is 5488 diagnosed because even though its output has been constrained 5489 using the modulo operation it could result in as many as three 5490 bytes if mday were negative. The solution is to either 5491 allocate a larger buffer or make sure the argument is not 5492 negative, for example by changing mday's type to unsigned or 5493 by making the type of the second operand of the modulo 5494 expression unsigned: 100U. 5495 5496 void* f (int mday) 5497 { 5498 char *buf = malloc (3); 5499 sprintf (buf, "%02i", mday % 100); 5500 return buf; 5501 } 5502 5503 warning: 'sprintf may write a terminating nul past the end of the destination [- 5504 Wformat-overflow=] 5505 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 5506 5507 + The -Wformat-truncation=level option detects certain and 5508 likely output truncation in calls to the snprintf family of 5509 formatted output functions. -Wformat-truncation=1 is included 5510 in -Wall and enabled without optimization but works best with 5511 -O2 and higher. 5512 For example, the following function attempts to format an 5513 integer between 0 and 255 in hexadecimal, including the 0x 5514 prefix, into a buffer of four characters. But since the 5515 function must always terminate output by the null character 5516 ('\0') such a buffer is only big enough to fit just one digit 5517 plus the prefix. Therefore the snprintf call is diagnosed. To 5518 avoid the warning either use a bigger buffer or handle the 5519 function's return value which indicates whether or not its 5520 output has been truncated. 5521 5522 void f (unsigned x) 5523 { 5524 char d[4]; 5525 snprintf (d, sizeof d, "%#02x", x & 0xff); 5526 5527 } 5528 5529 warning: 'snprintf' output may be truncated before the last format character [-W 5530 format-truncation=] 5531 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 5532 5533 + The -Wnonnull option has been enhanced to detect a broader set 5534 of cases of passing null pointers to functions that expect a 5535 non-null argument (those decorated with attribute nonnull). By 5536 taking advantage of optimizations the option can detect many 5537 more cases of the problem than in prior GCC versions. 5538 + The -Wstringop-overflow=type option detects buffer overflow in 5539 calls to string handling functions like memcpy and strcpy. The 5540 option relies on [8]Object Size Checking and has an effect 5541 similar to defining the _FORTIFY_SOURCE macro. 5542 -Wstringop-overflow=2 is enabled by default. 5543 For example, in the following snippet, because the call to 5544 strncat specifies a maximum that allows the function to write 5545 past the end of the destination, it is diagnosed. To correct 5546 the problem and avoid the overflow the function should be 5547 called with a size of at most sizeof d - strlen(d) - 1. 5548 5549 void f (const char *fname) 5550 { 5551 char d[8]; 5552 strncpy (d, "/tmp/", sizeof d); 5553 strncat (d, fname, sizeof d); 5554 5555 } 5556 5557 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 5558 ow=] 5559 5560 * The <limits.h> header provided by GCC defines macros such as 5561 INT_WIDTH for the width in bits of integer types, if 5562 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 5563 included. The <stdint.h> header defines such macros as SIZE_WIDTH 5564 and INTMAX_WIDTH for the width of some standard typedef names for 5565 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 5566 before the header is included; note that GCC's implementation of 5567 this header is only used for freestanding compilations, not hosted 5568 compilations, on most systems. These macros come from ISO/IEC TS 5569 18661-1:2014. 5570 * The <float.h> header provided by GCC defines the macro 5571 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 5572 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 5573 included. This represents the number of decimal digits for which 5574 conversions between decimal character strings and binary formats, 5575 in both directions, are correctly rounded, and currently has the 5576 value of UINTMAX_MAX on all systems, reflecting that GCC's 5577 compile-time conversions are correctly rounded for any number of 5578 digits. 5579 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 5580 __builtin_mul_overflow_p built-in functions have been added. These 5581 work similarly to their siblings without the _p suffix, but do not 5582 actually store the result of the arithmetics anywhere, just return 5583 whether the operation would overflow. Calls to these built-ins with 5584 integer constant arguments evaluate to integer constants 5585 expressions. 5586 For example, in the following, c is assigned the result of a * b 5587 only if the multiplication does not overflow, otherwise it is 5588 assigned the value zero. The multiplication is performed at 5589 compile-time and without triggering a -Woverflow warning. 5590 5591 enum { 5592 a = 12345678, 5593 b = 87654321, 5594 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 5595 }; 5596 5597 C 5598 5599 * The C front end now supports type names _FloatN for floating-point 5600 types with IEEE interchange formats and _FloatNx for floating-point 5601 types with IEEE extended formats. These type names come from 5602 ISO/IEC TS 18661-3:2015. 5603 The set of types supported depends on the target for which GCC is 5604 configured. Most targets support _Float32, _Float32x and _Float64. 5605 _Float128 is supported on targets where IEEE binary128 encoding was 5606 already supported as long double or __float128. _Float64x is 5607 supported on targets where a type with either binary128 or Intel 5608 extended precision format is available. 5609 Constants with these types are supported using suffixes fN, FN, fNx 5610 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 5611 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 5612 defined before it is included. 5613 These new types are always distinct from each other and from float, 5614 double and long double, even if they have the same encoding. 5615 Complex types such as _Complex _Float128 are also supported. 5616 Type-generic built-in functions such as __builtin_isinf support the 5617 new types, and the following type-specific built-in functions have 5618 versions (suffixed fN or fNx) for the new types: 5619 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 5620 __builtin_inf, __builtin_nan, __builtin_nans. 5621 * Compilation with -fopenmp is now compatible with the C11 _Atomic 5622 keyword. 5623 5624 C++ 5625 5626 * The C++ front end has experimental support for all of the current 5627 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 5628 constexpr, class template argument deduction, auto template 5629 parameters, and structured bindings. For a full list of new 5630 features, see [9]the C++ status page. 5631 * C++17 support for new of over-aligned types can be enabled in other 5632 modes with the -faligned-new flag. 5633 * The C++17 evaluation order requirements can be selected in other 5634 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 5635 with -fno-strong-eval-order. 5636 * The default semantics of inherited constructors has changed in all 5637 modes, following [10]P0136. Essentially, overload resolution 5638 happens as if calling the inherited constructor directly, and the 5639 compiler fills in construction of the other bases and members as 5640 needed. Most uses should not need any changes. The old behavior can 5641 be restored with -fno-new-inheriting-ctors, or -fabi-version less 5642 than 11. 5643 * The resolution of DR 150 on matching of template template 5644 parameters, allowing default template arguments to make a template 5645 match a parameter, is currently enabled by default in C++17 mode 5646 only. The default can be overridden with -f{no-,}new-ttp-matching. 5647 * The C++ front end will now provide fix-it hints for some missing 5648 semicolons, allowing for automatic fixes by IDEs: 5649 5650 test.cc:4:11: error: expected ';' after class definition 5651 class a {} 5652 ^ 5653 ; 5654 5655 * -Waligned-new has been added to the C++ front end. It warns about 5656 new of type with extended alignment without -faligned-new. 5657 5658 Runtime Library (libstdc++) 5659 5660 * The type of exception thrown by iostreams, std::ios_base::failure, 5661 now uses the [11]cxx11 ABI. 5662 * Experimental support for C++17, including the following new 5663 features: 5664 + std::string_view; 5665 + std::any, std::optional, and std::variant; 5666 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 5667 invoke_result; 5668 + std::is_swappable, and std::is_nothrow_swappable; 5669 + std::apply, and std::make_from_tuple; 5670 + std::void_t, std::bool_constant, std::conjunction, 5671 std::disjunction, and std::negation; 5672 + Variable templates for type traits; 5673 + Mathematical Special Functions; 5674 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 5675 std::chrono::abs; 5676 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 5677 + std::scoped_lock, std::shared_mutex, 5678 std::atomic<T>::is_always_lock_free; 5679 + std::sample, std::default_searcher, std::boyer_moore_searcher 5680 and std::boyer_moore_horspool_searcher; 5681 + Extraction and re-insertion of map and set nodes, try_emplace 5682 members for maps, and functions for accessing containers 5683 std::size, std::empty, and std::data; 5684 + std::shared_ptr support for arrays, 5685 std::shared_ptr<T>::weak_type, 5686 std::enable_shared_from_this<T>::weak_from_this(), and 5687 std::owner_less<void>; 5688 + std::byte; 5689 + std::as_const, std::not_fn, 5690 std::has_unique_object_representations, constexpr 5691 std::addressof. 5692 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 5693 Voutilainen for work on the C++17 support. 5694 * A new power-of-two rehashing policy for use with the _Hashtable 5695 internals, thanks to Franois Dumont. 5696 5697 Fortran 5698 5699 * Support for a number of extensions for compatibility with legacy 5700 code with new flags: 5701 + -fdec-structure Support for DEC STRUCTURE and UNION 5702 + -fdec-intrinsic-ints Support for new integer intrinsics with 5703 B/I/J/K prefixes such as BABS, JIAND... 5704 + -fdec-math Support for additional math intrinsics, including 5705 COTAN and degree-valued trigonometric functions such as TAND, 5706 ASIND... 5707 + -fdec Enable the -fdec-* family of extensions. 5708 * New flag -finit-derived to allow default initialization of 5709 derived-type variables. 5710 * Improved DO loops with step equal to 1 or -1, generates faster code 5711 without a loop preheader. A new warning, -Wundefined-do-loop, warns 5712 when a loop iterates either to HUGE(i) (with step equal to 1), or 5713 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 5714 at run time with -fcheck=do enabled: 5715 5716 program test 5717 implicit none 5718 integer(1) :: i 5719 do i = -HUGE(i)+10, -HUGE(i)-1, -1 5720 print *, i 5721 end do 5722 end program test 5723 5724 At line 8 of file do_check_12.f90 5725 Fortran runtime error: Loop iterates infinitely 5726 5727 * Version 4.5 of the [12]OpenMP specification is now partially 5728 supported in the Fortran compiler; the largest missing item is 5729 structure element mapping. 5730 * User-defined derived-type input/output (UDTIO) is added. 5731 * Derived type coarrays with allocatable and pointer components are 5732 partially supported. 5733 * Non-constant stop codes and error stop codes (Fortran 2015 5734 feature). 5735 * Derived types with allocatable components of recursive type. 5736 * Intrinsic assignment to polymorphic variables. 5737 * Improved submodule support. 5738 * Improved diagnostics (polymorphic results in pure functions). 5739 * Coarray: Support for failed images (Fortan 2015 feature). 5740 5741 Go 5742 5743 * GCC 7 provides a complete implementation of the Go 1.8.1 user 5744 packages. 5745 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 5746 conservative and less concurrent. 5747 * Escape analysis is available for experimental use via the 5748 -fgo-optimize-allocs option. The -fgo-debug-escape prints 5749 information useful for debugging escape analysis choices. 5750 5751 Java (GCJ) 5752 5753 The GCC Java front end and associated libjava runtime library have been 5754 removed from GCC. 5755 5756 libgccjit 5757 5758 The libgccjit API gained support for marking calls as requiring 5759 tail-call optimization via a new entry point: 5760 [13]gcc_jit_rvalue_set_bool_require_tail_call. 5761 5762 libgccjit performs numerous checks at the API boundary, but if these 5763 succeed, it previously ignored errors and other diagnostics emitted 5764 within the core of GCC, and treated the compile of a gcc_jit_context as 5765 having succeeded. As of GCC 7 it now ensures that if any diagnostics 5766 are emitted, they are visible from the libgccjit API, and that the the 5767 context is flagged as having failed. 5768 5769 New Targets and Target Specific Improvements 5770 5771 AArch64 5772 5773 * GCC has been updated to the latest revision of the procedure call 5774 standard (AAPCS64) to provide support for parameter passing when 5775 data types have been over-aligned. 5776 * The ARMv8.3-A architecture is now supported. It can be used by 5777 specifying the -march=armv8.3-a option. 5778 * The option -msign-return-address= is supported to enable return 5779 address protection using ARMv8.3-A Pointer Authentication 5780 Extensions. For more information on the arguments accepted by this 5781 option, please refer to [14]AArch64-Options. 5782 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 5783 Extensions are now supported. They can be used by specifying the 5784 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 5785 Floating-Point Extensions introduce new half-precision data 5786 processing floating-point instructions. 5787 * Support has been added for the following processors (GCC 5788 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 5789 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 5790 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 5791 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 5792 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 5793 The GCC identifiers can be used as arguments to the -mcpu or -mtune 5794 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 5795 arguments to the equivalent target attributes and pragmas. 5796 5797 ARC 5798 5799 * Added support for ARC HS and ARC EM processors. 5800 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 5801 * Added support for NPS400 ARC700 based CPUs. 5802 * Thread Local Storage is now supported by ARC CPUs. 5803 * Fixed errors for ARC600 when using 32x16 multiplier option. 5804 * Fixed PIE for ARC CPUs. 5805 * New CPU templates are supported via multilib. 5806 5807 ARM 5808 5809 * Support for the ARMv5 and ARMv5E architectures has been deprecated 5810 (which have no known implementations) and will be removed in a 5811 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 5812 architectures remain supported. The values armv5 and armv5e of 5813 -march are thus deprecated. 5814 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 5815 Extensions are now supported. They can be used by specifying the 5816 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 5817 Floating-Point Extensions introduce new half-precision data 5818 processing floating-point instructions. 5819 * The ARMv8-M architecture is now supported in its two architecture 5820 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 5821 Floating-Point Extensions. They can be used by specifying the 5822 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 5823 * Support has been added for the following processors (GCC 5824 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 5825 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 5826 identifiers can be used as arguments to the -mcpu or -mtune 5827 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 5828 * A new command-line option -mpure-code has been added. It does not 5829 allow constant data to be placed in code sections. This option is 5830 only available when generating non-PIC code for ARMv7-M targets. 5831 * Support for the ACLE Coprocessor Intrinsics has been added. This 5832 enables the generation of coprocessor instructions through the use 5833 of intrinsics such as cdp, ldc, and others. 5834 * The configure option --with-multilib-list now accepts the value 5835 rmprofile to build multilib libraries for a range of embedded 5836 targets. See our [15]installation instructions for details. 5837 5838 AVR 5839 5840 * On the reduced Tiny cores, the progmem [16]variable attribute is 5841 now properly supported. Respective read-only variables are located 5842 in flash memory in section .progmem.data. No special code is needed 5843 to access such variables; the compiler automatically adds an offset 5844 of 0x4000 to all addresses, which is needed to access variables in 5845 flash memory. As opposed to ordinary cores where it is sufficient 5846 to specify the progmem attribute with definitions, on the reduced 5847 Tiny cores the attribute also has to be specified with (external) 5848 declarations: 5849 5850 extern const int array[] __attribute__((__progmem__)); 5851 5852 int get_value2 (void) 5853 { 5854 /* Access via addresses array + 0x4004 and array + 0x4005. */ 5855 return array[2]; 5856 } 5857 5858 const int* get_address (unsigned idx) 5859 { 5860 /* Returns array + 0x4000 + 2 * idx. */ 5861 return &array[idx]; 5862 } 5863 5864 * A new command-line option -Wmisspelled-isr has been added. It turns 5865 off or turns into errors warnings that are reported for 5866 interrupt service routines (ISRs) which don't follow AVR-LibC's 5867 naming convention of prefixing ISR names with __vector. 5868 * __builtin_avr_nops(n) is a new [17]built-in function that inserts n 5869 NOP instructions into the instruction stream. n must be a value 5870 known at compile time. 5871 5872 IA-32/x86-64 5873 5874 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 5875 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 5876 variable precision (4VNNIW), AVX-512 Vector Population Count 5877 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 5878 been added. 5879 5880 NVPTX 5881 5882 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 5883 See the [18]Offloading Wiki on how to configure it. 5884 5885 PowerPC / PowerPC64 / RS6000 5886 5887 * The PowerPC port now uses LRA by default. 5888 * GCC now diagnoses inline assembly that clobbers register r2. This 5889 has always been invalid code, and is no longer quietly tolerated. 5890 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 5891 enhanced to generate more of the new instructions by default, and 5892 to provide more built-in functions to generate code for other new 5893 instructions. 5894 * The configuration option --enable-gnu-indirect-function is now 5895 enabled by default on PowerPC GNU/Linux builds. 5896 * The PowerPC port will now allow 64-bit and 32-bit integer types to 5897 be allocated to the VSX vector registers (ISA 2.06 and above). In 5898 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 5899 allocated in the vector registers. Previously, only 64-bit integer 5900 types were allowed in the traditional floating point registers. 5901 * New options -mstack-protector-guard=global, 5902 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 5903 -mstack-protector-guard-offset= change how the stack protector gets 5904 the value to use as canary. 5905 5906 S/390, System z, IBM z Systems, IBM Z 5907 5908 * Support for the IBM z14 processor has been added. When using the 5909 -march=z14 option, the compiler will generate code making use of 5910 the new instructions introduced with the vector enhancement 5911 facility and the miscellaneous instruction extension facility 2. 5912 The -mtune=z14 option enables z14 specific instruction scheduling 5913 without making use of new instructions. 5914 * Builtins for the new vector instructions have been added and can be 5915 enabled using the -mzvector option. 5916 5917 RISC-V 5918 5919 * Support for the RISC-V instruction set has been added. 5920 5921 RX 5922 5923 Basic support for atomic built-in function has been added. It is 5924 currently implemented by flipping interrupts off and on as needed. 5925 5926 SH 5927 5928 * Support for SH5/SH64 has been removed. 5929 * Improved utilization of delay slots on SH2A. 5930 * Improved utilization of zero-displacement conditional branches. 5931 * The following deprecated options have been removed 5932 + -mcbranchdi 5933 + -mcmpeqdi 5934 + -minvalid-symbols 5935 + -msoft-atomic 5936 + -mspace 5937 + -madjust-unroll 5938 * Support for the following SH2A instructions has been added 5939 + mov.b @-Rm,R0 5940 + mov.w @-Rm,R0 5941 + mov.l @-Rm,R0 5942 + mov.b R0,@Rn+ 5943 + mov.w R0,@Rn+ 5944 + mov.l R0,@Rn+ 5945 5946 SPARC 5947 5948 * The SPARC port now uses LRA by default. 5949 * Support for the new Subtract-Extended-with-Carry instruction 5950 available in SPARC M7 (Niagara 7) has been added. 5951 5952 Operating Systems 5953 5954 AIX 5955 5956 * Visibility support has been enabled for AIX 7.1 and above. 5957 5958 Fuchsia 5959 5960 * Support has been added for the [19]Fuchsia OS. 5961 5962 RTEMS 5963 5964 * The ABI changes on ARM so that no short enums are used by default. 5965 5966 Other significant improvements 5967 5968 * -fverbose-asm previously emitted information on the meanings of 5969 assembly expressions. This has been extended so that it now also 5970 prints comments showing the source lines that correspond to the 5971 assembly, making it easier to read the generated assembly 5972 (especially with larger functions). For example, given this C 5973 source file: 5974 5975 int test (int n) 5976 { 5977 int i; 5978 int total = 0; 5979 5980 for (i = 0; i < n; i++) 5981 total += i * i; 5982 return total; 5983 } 5984 5985 -fverbose-asm now gives output similar to this for the function 5986 body (when compiling for x86_64, with -Os): 5987 5988 .text 5989 .globl test 5990 .type test, @@function 5991 test: 5992 .LFB0: 5993 .cfi_startproc 5994 # example.c:4: int total = 0; 5995 xorl %eax, %eax # <retval> 5996 # example.c:6: for (i = 0; i < n; i++) 5997 xorl %edx, %edx # i 5998 .L2: 5999 # example.c:6: for (i = 0; i < n; i++) 6000 cmpl %edi, %edx # n, i 6001 jge .L5 #, 6002 # example.c:7: total += i * i; 6003 movl %edx, %ecx # i, tmp92 6004 imull %edx, %ecx # i, tmp92 6005 # example.c:6: for (i = 0; i < n; i++) 6006 incl %edx # i 6007 # example.c:7: total += i * i; 6008 addl %ecx, %eax # tmp92, <retval> 6009 jmp .L2 # 6010 .L5: 6011 # example.c:10: } 6012 ret 6013 .cfi_endproc 6014 6015 * Two new options have been added for printing fix-it hints: 6016 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 6017 emitted in a machine-readable form, suitable for consumption 6018 by IDEs. For example, given: 6019 6020 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 6021 you mean 'color'? 6022 return ptr->colour; 6023 ^~~~~~ 6024 color 6025 6026 it will emit: 6027 6028 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 6029 6030 + -fdiagnostics-generate-patch will print a patch in "unified" 6031 format after any diagnostics are printed, showing the result 6032 of applying all fix-it hints. For the above example it would 6033 emit: 6034 6035 --- spellcheck-fields.cc 6036 +++ spellcheck-fields.cc 6037 @@ -49,5 +49,5 @@ 6038 6039 color get_color(struct s *ptr) 6040 { 6041 - return ptr->colour; 6042 + return ptr->color; 6043 } 6044 6045 * The gcc and g++ driver programs will now provide suggestions for 6046 misspelled arguments to command-line options. 6047 6048 $ gcc -c test.c -ftls-model=global-dinamic 6049 gcc: error: unknown TLS model 'global-dinamic' 6050 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 6051 cal-dynamic local-exec; did you mean 'global-dynamic'? 6052 6053 * The compiler will now provide suggestions for misspelled 6054 parameters. 6055 6056 $ gcc -c test.c --param max-early-inliner-iteration=3 6057 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 6058 x-early-inliner-iterations'? 6059 6060 * Profile-guided optimization (PGO) instrumentation, as well as test 6061 coverage (GCOV), can newly instrument constructors (functions marks 6062 with __attribute__((constructor))), destructors and C++ 6063 constructors (and destructors) of classes that are used as the type 6064 of a global variable. 6065 * A new option -fprofile-update=atomic prevents creation of corrupted 6066 profiles created during an instrumentation run (-fprofile=generate) 6067 of an application. The downside of the option is a speed penalty. 6068 Providing -pthread on the command line selects atomic profile 6069 updating (when supported by the target). 6070 * GCC's already extensive testsuite has gained some new capabilities, 6071 to further improve the reliability of the compiler: 6072 + GCC now has an internal unit-testing API and a suite of tests 6073 for programmatic self-testing of subsystems. 6074 + GCC's C front end has been extended so that it can parse dumps 6075 of GCC's internal representations, allowing for DejaGnu tests 6076 that more directly exercise specific optimization passes. This 6077 covers both the [20]GIMPLE representation (for testing 6078 higher-level optimizations) and the [21]RTL representation, 6079 allowing for more direct testing of lower-level details, such 6080 as register allocation and instruction selection. 6081 6082 GCC 7.1 6083 6084 This is the [22]list of problem reports (PRs) from GCC's bug tracking 6085 system that are known to be fixed in the 7.1 release. This list might 6086 not be complete (that is, it is possible that some PRs that have been 6087 fixed are not listed here). 6088 6089 GCC 7.2 6090 6091 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6092 system that are known to be fixed in the 7.2 release. This list might 6093 not be complete (that is, it is possible that some PRs that have been 6094 fixed are not listed here). 6095 6096 Target Specific Changes 6097 6098 SPARC 6099 6100 * Support for the SPARC M8 processor has been added. 6101 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 6102 around an erratum in LEON3FT processors. 6103 * Use of the Floating-point Multiply Single to Double (FsMULd) 6104 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 6105 options. 6106 6107 Operating Systems 6108 6109 RTEMS 6110 6111 * The Ada run-time support uses now thread-local storage (TLS). 6112 * Support for RISC-V has been added. 6113 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 6114 double has been added. 6115 6116 GCC 7.3 6117 6118 This is the [24]list of problem reports (PRs) from GCC's bug tracking 6119 system that are known to be fixed in the 7.3 release. This list might 6120 not be complete (that is, it is possible that some PRs that have been 6121 fixed are not listed here). 6122 6123 Target Specific Changes 6124 6125 SPARC 6126 6127 * Workarounds for the four [25]LEON3FT errata GRLIB-TN-0010..0013 6128 have been added. Relevant errata are activated by the target 6129 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 6130 6131 Operating Systems 6132 6133 RTEMS 6134 6135 * Support has been added for Epiphany target. 6136 6137 GCC 7.4 6138 6139 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6140 system that are known to be fixed in the 7.4 release. This list might 6141 not be complete (that is, it is possible that some PRs that have been 6142 fixed are not listed here). 6143 6144 GCC 7.5 6145 6146 This is the [27]list of problem reports (PRs) from GCC's bug tracking 6147 system that are known to be fixed in the 7.5 release. This list might 6148 not be complete (that is, it is possible that some PRs that have been 6149 fixed are not listed here). 6150 6151 6152 For questions related to the use of GCC, please consult these web 6153 pages and the [28]GCC manuals. If that fails, the 6154 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6155 web pages and the development of GCC are welcome on our developer 6156 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 6157 archives. 6158 6159 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 6160 distribution of this entire article is permitted in any medium, 6161 provided this notice is preserved. 6162 6163 These pages are [33]maintained by the GCC team. Last modified 6164 2022-10-26. 6165 6166 References 6167 6168 1. https://gcc.gnu.org/gcc-7/porting_to.html 6169 2. https://gcc.gnu.org/onlinedocs/index.html#current 6170 3. https://gcc.gnu.org/wiki/LRAIsDefault 6171 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 6172 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 6173 6. https://gcc.gnu.org/wiki/OpenACC 6174 7. https://gcc.gnu.org/wiki/Offloading 6175 8. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 6176 9. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 6177 10. https://wg21.link/p0136 6178 11. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 6179 12. https://www.openmp.org/specifications/ 6180 13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 6181 14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 6182 15. https://gcc.gnu.org/install/configure.html 6183 16. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 6184 17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 6185 18. https://gcc.gnu.org/wiki/Offloading 6186 19. https://fuchsia.googlesource.com/ 6187 20. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 6188 21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 6189 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 6190 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 6191 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 6192 25. https://www.gaisler.com/index.php/information/app-tech-notes 6193 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 6194 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 6195 28. https://gcc.gnu.org/onlinedocs/ 6196 29. mailto:gcc-help (a] gcc.gnu.org 6197 30. mailto:gcc (a] gcc.gnu.org 6198 31. https://gcc.gnu.org/lists.html 6199 32. https://www.fsf.org/ 6200 33. https://gcc.gnu.org/about.html 6201 ====================================================================== 6202 http://gcc.gnu.org/gcc-6/index.html 6203 GCC 6 Release Series 6204 6205 (This release series is no longer supported.) 6206 6207 October 26, 2018 6208 6209 The [1]GNU project and the GCC developers are pleased to announce the 6210 release of GCC 6.5. 6211 6212 This release is a bug-fix release, containing fixes for regressions in 6213 GCC 6.4 relative to previous releases of GCC. 6214 6215 Release History 6216 6217 GCC 6.5 6218 October 26, 2018 ([2]changes, [3]documentation) 6219 6220 GCC 6.4 6221 July 4, 2017 ([4]changes, [5]documentation) 6222 6223 GCC 6.3 6224 December 21, 2016 ([6]changes, [7]documentation) 6225 6226 GCC 6.2 6227 August 22, 2016 ([8]changes, [9]documentation) 6228 6229 GCC 6.1 6230 April 27, 2016 ([10]changes, [11]documentation) 6231 6232 References and Acknowledgements 6233 6234 GCC used to stand for the GNU C Compiler, but since the compiler 6235 supports several other languages aside from C, it now stands for the 6236 GNU Compiler Collection. 6237 6238 A list of [12]successful builds is updated as new information becomes 6239 available. 6240 6241 The GCC developers would like to thank the numerous people that have 6242 contributed new features, improvements, bug fixes, and other changes as 6243 well as test results to GCC. This [13]amazing group of volunteers is 6244 what makes GCC successful. 6245 6246 For additional information about GCC please refer to the [14]GCC 6247 project web site or contact the [15]GCC development mailing list. 6248 6249 To obtain GCC please use [16]our mirror sites or [17]our version 6250 control system. 6251 6252 6253 For questions related to the use of GCC, please consult these web 6254 pages and the [18]GCC manuals. If that fails, the 6255 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6256 web pages and the development of GCC are welcome on our developer 6257 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6258 archives. 6259 6260 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6261 distribution of this entire article is permitted in any medium, 6262 provided this notice is preserved. 6263 6264 These pages are [23]maintained by the GCC team. Last modified 6265 2024-05-30. 6266 6267 References 6268 6269 1. http://www.gnu.org/ 6270 2. https://gcc.gnu.org/gcc-6/changes.html 6271 3. https://gcc.gnu.org/onlinedocs/6.5.0/ 6272 4. https://gcc.gnu.org/gcc-6/changes.html 6273 5. https://gcc.gnu.org/onlinedocs/6.4.0/ 6274 6. https://gcc.gnu.org/gcc-6/changes.html 6275 7. https://gcc.gnu.org/onlinedocs/6.3.0/ 6276 8. https://gcc.gnu.org/gcc-6/changes.html 6277 9. https://gcc.gnu.org/onlinedocs/6.2.0/ 6278 10. https://gcc.gnu.org/gcc-6/changes.html 6279 11. https://gcc.gnu.org/onlinedocs/6.1.0/ 6280 12. https://gcc.gnu.org/gcc-6/buildstat.html 6281 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6282 14. https://gcc.gnu.org/index.html 6283 15. mailto:gcc (a] gcc.gnu.org 6284 16. https://gcc.gnu.org/mirrors.html 6285 17. https://gcc.gnu.org/git.html 6286 18. https://gcc.gnu.org/onlinedocs/ 6287 19. mailto:gcc-help (a] gcc.gnu.org 6288 20. mailto:gcc (a] gcc.gnu.org 6289 21. https://gcc.gnu.org/lists.html 6290 22. https://www.fsf.org/ 6291 23. https://gcc.gnu.org/about.html 6292 ====================================================================== 6293 http://gcc.gnu.org/gcc-6/changes.html 6294 GCC 6 Release Series 6295 Changes, New Features, and Fixes 6296 6297 This page is a brief summary of some of the huge number of improvements 6298 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 6299 the [2]full GCC documentation. 6300 6301 Caveats 6302 6303 * The default mode for C++ is now -std=gnu++14 instead of 6304 -std=gnu++98. 6305 * Support for a number of older systems and recently unmaintained or 6306 untested target ports of GCC has been declared obsolete in GCC 6. 6307 Unless there is activity to revive them, the next release of GCC 6308 will have their sources permanently removed. 6309 The following ports for individual systems on particular 6310 architectures have been obsoleted: 6311 + SH5 / SH64 (sh64-*-*) as announced [3]here. 6312 * The AVR port requires binutils version 2.26.1 or later for the fix 6313 for [4]PR71151 to work. 6314 * The GCC 6.5 release has an accidental ABI incompatibility for 6315 nested std::pair objects, for more details see [5]PR 87822. The bug 6316 causes a layout change for pairs where the first member is also a 6317 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 6318 is closed so the bug in GCC 6.5 will not be fixed upstream, but 6319 there is a patch in the bug report to allow it to be fixed by 6320 anybody packaging GCC 6.5 or installing it themselves. 6321 6322 General Optimizer Improvements 6323 6324 * UndefinedBehaviorSanitizer gained a new sanitization option, 6325 -fsanitize=bounds-strict, which enables strict checking of array 6326 bounds. In particular, it enables -fsanitize=bounds as well as 6327 instrumentation of flexible array member-like arrays. 6328 * Type-based alias analysis now disambiguates accesses to different 6329 pointers. This improves precision of the alias oracle by about 6330 20-30% on higher-level C++ programs. Programs doing invalid type 6331 punning of pointer types may now need -fno-strict-aliasing to work 6332 correctly. 6333 * Alias analysis now correctly supports the weakref and alias 6334 attributes. This allows accessing both a variable and its alias in 6335 one translation unit which is common with link-time optimization. 6336 * Value range propagation now assumes that the this pointer in C++ 6337 member functions is non-null. This eliminates common null pointer 6338 checks but also breaks some non-conforming code-bases (such as 6339 Qt-5, Chromium, KDevelop). As a temporary work-around 6340 -fno-delete-null-pointer-checks can be used. Wrong code can be 6341 identified by using -fsanitize=undefined. 6342 * Link-time optimization improvements: 6343 + warning and error attributes are now correctly preserved by 6344 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 6345 supported with -flto. 6346 + Type merging was fixed to handle C and Fortran 6347 interoperability rules as defined by the Fortran 2008 language 6348 standard. 6349 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 6350 with char in all cases because it is an array while char is 6351 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 6352 general, this inter-operability cannot be implemented, for 6353 example on targets where the argument passing convention for 6354 arrays differs from scalars. 6355 + More type information is now preserved at link time, reducing 6356 the loss of accuracy of the type-based alias analysis compared 6357 to builds without link-time optimization. 6358 + Invalid type punning on global variables and declarations is 6359 now reported with -Wodr-type-mismatch. 6360 + The size of LTO object files was reduced by about 11% 6361 (measured by compiling Firefox 46.0). 6362 + Link-time parallelization (enabled using -flto=n) was 6363 significantly improved by decreasing the size of streamed data 6364 when partitioning programs. The size of streamed IL while 6365 compiling Firefox 46.0 was reduced by 66%. 6366 + The linker plugin was extended to pass information about the 6367 type of binary produced to the GCC back end. (That can also be 6368 controlled manually by -flinker-output.) This makes it 6369 possible to properly configure the code generator and support 6370 incremental linking. Incremental linking of LTO objects by gcc 6371 -r is now supported for plugin-enabled setups. 6372 There are two ways to perform incremental linking: 6373 1. Linking by ld -r will result in an object file with all 6374 sections from individual object files mechanically 6375 merged. This delays the actual link-time optimization to 6376 the final linking step and thus permits whole program 6377 optimization. Linking the final binary with such object 6378 files is however slower. 6379 2. Linking by gcc -r will lead to link-time optimization and 6380 emit the final binary into the object file. Linking such 6381 an object file is fast but avoids any benefits from whole 6382 program optimization. 6383 GCC 7 will support incremental link-time optimization with gcc 6384 -r. 6385 * Inter-procedural optimization improvements: 6386 + Basic jump threading is now performed before profile 6387 construction and inline analysis, resulting in more realistic 6388 size and time estimates that drive the heuristics of the 6389 inliner and function cloning passes. 6390 + Function cloning now more aggressively eliminates unused 6391 function parameters. 6392 6393 New Languages and Language specific improvements 6394 6395 Compared to GCC 5, the GCC 6 release series includes a much improved 6396 implementation of the [6]OpenACC 2.0a specification. Highlights are: 6397 * In addition to single-threaded host-fallback execution, offloading 6398 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 6399 little-endian GNU/Linux host systems. For nvptx offloading, with 6400 the OpenACC parallel construct, the execution model allows for an 6401 arbitrary number of gangs, up to 32 workers, and 32 vectors. 6402 * Initial support for parallelized execution of OpenACC kernels 6403 constructs: 6404 + Parallelization of a kernels region is switched on by 6405 -fopenacc combined with -O2 or higher. 6406 + Code is offloaded onto multiple gangs, but executes with just 6407 one worker, and a vector length of 1. 6408 + Directives inside a kernels region are not supported. 6409 + Loops with reductions can be parallelized. 6410 + Only kernels regions with one loop nest are parallelized. 6411 + Only the outer-most loop of a loop nest can be parallelized. 6412 + Loop nests containing sibling loops are not parallelized. 6413 Typically, using the OpenACC parallel construct gives much better 6414 performance, compared to the initial support of the OpenACC kernels 6415 construct. 6416 * The device_type clause is not supported. The bind and nohost 6417 clauses are not supported. The host_data directive is not supported 6418 in Fortran. 6419 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 6420 * Usage of OpenACC constructs inside multithreaded contexts (such as 6421 created by OpenMP, or pthread programming) is not supported. 6422 * If a call to the acc_on_device function has a compile-time constant 6423 argument, the function call evaluates to a compile-time constant 6424 value only for C and C++ but not for Fortran. 6425 6426 See the [7]OpenACC and [8]Offloading wiki pages for further 6427 information. 6428 6429 C family 6430 6431 * Version 4.5 of the [9]OpenMP specification is now supported in the 6432 C and C++ compilers. 6433 * The C and C++ compilers now support attributes on enumerators. For 6434 instance, it is now possible to mark enumerators as deprecated: 6435 6436 enum { 6437 newval, 6438 oldval __attribute__ ((deprecated ("too old"))) 6439 }; 6440 6441 * Source locations for the C and C++ compilers are now tracked as 6442 ranges, rather than just points, making it easier to identify the 6443 subexpression of interest within a complicated expression. For 6444 example: 6445 6446 test.cc: In function 'int test(int, int, foo, int, int)': 6447 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 6448 ) 6449 return p + q * r * s + t; 6450 ~~^~~ 6451 6452 In addition, there is now initial support for precise diagnostic 6453 locations within strings: 6454 6455 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 6456 t' argument [-Wformat=] 6457 printf("%*d"); 6458 ^ 6459 6460 * Diagnostics can now contain "fix-it hints", which are displayed in 6461 context underneath the relevant source code. For example: 6462 6463 fixits.c: In function 'bad_deref': 6464 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 6465 return ptr.x; 6466 ^ 6467 -> 6468 6469 * The C and C++ compilers now offer suggestions for misspelled field 6470 names: 6471 6472 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 6473 you mean 'color'? 6474 return ptr->colour; 6475 ^~~~~~ 6476 6477 * New command-line options have been added for the C and C++ 6478 compilers: 6479 + -Wshift-negative-value warns about left shifting a negative 6480 value. 6481 + -Wshift-overflow warns about left shift overflows. This 6482 warning is enabled by default. -Wshift-overflow=2 also warns 6483 about left-shifting 1 into the sign bit. 6484 + -Wtautological-compare warns if a self-comparison always 6485 evaluates to true or false. This warning is enabled by -Wall. 6486 + -Wnull-dereference warns if the compiler detects paths that 6487 trigger erroneous or undefined behavior due to dereferencing a 6488 null pointer. This option is only active when 6489 -fdelete-null-pointer-checks is active, which is enabled by 6490 optimizations in most targets. The precision of the warnings 6491 depends on the optimization options used. 6492 + -Wduplicated-cond warns about duplicated conditions in an 6493 if-else-if chain. 6494 + -Wmisleading-indentation warns about places where the 6495 indentation of the code gives a misleading idea of the block 6496 structure of the code to a human reader. For example, given 6497 [10]CVE-2014-1266: 6498 6499 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 6500 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 6501 g-indentation] 6502 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 6503 ^~ 6504 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 6505 indented as if it is guarded by the 'if' 6506 goto fail; 6507 ^~~~ 6508 6509 This warning is enabled by -Wall. 6510 * The C and C++ compilers now emit saner error messages if 6511 merge-conflict markers are present in a source file. 6512 6513 test.c:3:1: error: version control conflict marker in file 6514 <<<<<<< HEAD 6515 ^~~~~~~ 6516 6517 C 6518 6519 * It is possible to disable warnings when an initialized field of a 6520 structure or a union with side effects is being overridden when 6521 using designated initializers via a new warning option 6522 -Woverride-init-side-effects. 6523 * A new type attribute scalar_storage_order applying to structures 6524 and unions has been introduced. It specifies the storage order (aka 6525 endianness) in memory of scalar fields in structures or unions. 6526 6527 C++ 6528 6529 * The default mode has been changed to -std=gnu++14. 6530 * [11]C++ Concepts are now supported when compiling with -fconcepts. 6531 * -flifetime-dse is more aggressive in dead-store elimination in 6532 situations where a memory store to a location precedes a 6533 constructor to that memory location. 6534 * G++ now supports [12]C++17 fold expressions, u8 character literals, 6535 extended static_assert, and nested namespace definitions. 6536 * G++ now allows constant evaluation for all non-type template 6537 arguments. 6538 * G++ now supports C++ Transactional Memory when compiling with 6539 -fgnu-tm. 6540 6541 Runtime Library (libstdc++) 6542 6543 * Extensions to the C++ Library to support mathematical special 6544 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 6545 * Experimental support for C++17, including the following new 6546 features: 6547 + std::uncaught_exceptions function (this is also available for 6548 -std=gnu++NN modes); 6549 + new member functions try_emplace and insert_or_assign for 6550 unique_key maps; 6551 + non-member functions std::size, std::empty, and std::data for 6552 accessing containers and arrays; 6553 + std::invoke; 6554 + std::shared_mutex; 6555 + std::void_t and std::bool_constant metaprogramming utilities. 6556 Thanks to Ville Voutilainen for contributing many of the C++17 6557 features. 6558 * An experimental implementation of the File System TS. 6559 * Experimental support for most features of the second version of the 6560 Library Fundamentals TS. This includes polymorphic memory resources 6561 and array support in shared_ptr, thanks to Fan You. 6562 * Some assertions checked by Debug Mode can now also be enabled by 6563 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 6564 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 6565 don't affect the library ABI, so can be enabled per-translation 6566 unit. 6567 * Timed mutex types are supported on more targets, including Darwin. 6568 * Improved std::locale support for DragonFly and FreeBSD, thanks to 6569 John Marino and Andreas Tobler. 6570 6571 Fortran 6572 6573 * Fortran 2008 SUBMODULE support. 6574 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 6575 support. 6576 * Improved support for Fortran 2003 deferred-length character 6577 variables. 6578 * Improved support for OpenMP and OpenACC. 6579 * The MATMUL intrinsic is now inlined for straightforward cases if 6580 front-end optimization is active. The maximum size for inlining can 6581 be set to n with the -finline-matmul-limit=n option and turned off 6582 with -finline-matmul-limit=0. 6583 * The -Wconversion-extra option will warn about REAL constants which 6584 have excess precision for their kind. 6585 * The -Winteger-division option has been added, which warns about 6586 divisions of integer constants which are truncated. This option is 6587 included in -Wall by default. 6588 6589 libgccjit 6590 6591 * The driver code is now run in-process within libgccjit, providing a 6592 small speed-up of the compilation process. 6593 * The API has gained entrypoints for 6594 + [13]timing how long was spent in different parts of code, 6595 + [14]creating switch statements, 6596 + [15]allowing unreachable basic blocks in a function, and 6597 + [16]adding arbitrary command-line options to a compilation. 6598 6599 New Targets and Target Specific Improvements 6600 6601 AArch64 6602 6603 * A number of AArch64-specific options have been added. The most 6604 important ones are summarised in this section; for more detailed 6605 information please refer to the documentation. 6606 * The command-line options -march=native, -mcpu=native and 6607 -mtune=native are now available on native AArch64 GNU/Linux 6608 systems. Specifying these options causes GCC to auto-detect the 6609 host CPU and choose the optimal setting for that system. 6610 * -fpic is now supported when generating code for the small code 6611 model (-mcmodel=small). The size of the global offset table (GOT) 6612 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 6613 ILP32 SysV ABI. 6614 * The AArch64 port now supports target attributes and pragmas. Please 6615 refer to the [17]documentation for details of available attributes 6616 and pragmas as well as usage instructions. 6617 * Link-time optimization across translation units with different 6618 target-specific options is now supported. 6619 * The option -mtls-size= is now supported. It can be used to specify 6620 the bit size of TLS offsets, allowing GCC to generate better TLS 6621 instruction sequences. 6622 * The option -fno-plt is now fully functional. 6623 * The ARMv8.1-A architecture and the Large System Extensions are now 6624 supported. They can be used by specifying the -march=armv8.1-a 6625 option. Additionally, the +lse option extension can be used in a 6626 similar fashion to other option extensions. The Large System 6627 Extensions introduce new instructions that are used in the 6628 implementation of atomic operations. 6629 * The ACLE half-precision floating-point type __fp16 is now supported 6630 in the C and C++ languages. 6631 * The ARM Cortex-A35 processor is now supported via the 6632 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 6633 equivalent target attributes and pragmas. 6634 * The Qualcomm QDF24xx processor is now supported via the 6635 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 6636 target attributes and pragmas. 6637 * Code generation for the ARM Cortex-A57 processor is improved. Among 6638 general code generation improvements, a better algorithm is added 6639 for allocating registers to floating-point multiply-accumulate 6640 instructions offering increased performance when compiling with 6641 -mcpu=cortex-a57 or -mtune=cortex-a57. 6642 * Code generation for the ARM Cortex-A53 processor is improved. A 6643 more accurate instruction scheduling model for the processor is now 6644 used, and a number of compiler tuning parameters have been set to 6645 offer increased performance when compiling with -mcpu=cortex-a53 or 6646 -mtune=cortex-a53. 6647 * Code generation for the Samsung Exynos M1 processor is improved. A 6648 more accurate instruction scheduling model for the processor is now 6649 used, and a number of compiler tuning parameters have been set to 6650 offer increased performance when compiling with -mcpu=exynos-m1 or 6651 -mtune=exynos-m1. 6652 * Improvements in the generation of conditional branches and literal 6653 pools allow the compiler to compile functions of a large size. 6654 Constant pools are now placed into separate rodata sections. The 6655 new option -mpc-relative-literal-loads generates per-function 6656 literal pools, limiting the maximum size of functions to 1MiB. 6657 * Several correctness issues generating Advanced SIMD instructions 6658 for big-endian targets have been fixed resulting in improved code 6659 generation for ACLE intrinsics with -mbig-endian. 6660 6661 ARM 6662 6663 * Support for revisions of the ARM architecture prior to ARMv4t has 6664 been deprecated and will be removed in a future GCC release. The 6665 -mcpu and -mtune values that are deprecated are: arm2, arm250, 6666 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 6667 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 6668 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 6669 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 6670 arm7tdmi is still supported. The values of -march that are 6671 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 6672 * The ARM port now supports target attributes and pragmas. Please 6673 refer to the [18]documentation for details of available attributes 6674 and pragmas as well as usage instructions. 6675 * Support has been added for the following processors (GCC 6676 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 6677 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 6678 identifiers can be used as arguments to the -mcpu or -mtune 6679 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 6680 6681 Heterogeneous Systems Architecture 6682 6683 * GCC can now generate HSAIL (Heterogeneous System Architecture 6684 Intermediate Language) for simple OpenMP device constructs if 6685 configured with --enable-offload-targets=hsa. A new libgomp plugin 6686 then runs the HSA GPU kernels implementing these constructs on HSA 6687 capable GPUs via a standard HSA run time. 6688 If the HSA compilation back end determines it cannot output HSAIL 6689 for a particular input, it gives a warning by default. These 6690 warnings can be suppressed with -Wno-hsa. To give a few examples, 6691 the HSA back end does not implement compilation of code using 6692 function pointers, automatic allocation of variable sized arrays, 6693 functions with variadic arguments as well as a number of other less 6694 common programming constructs. 6695 When compilation for HSA is enabled, the compiler attempts to 6696 compile composite OpenMP constructs 6697 6698 #pragma omp target teams distribute parallel for 6699 6700 into parallel HSA GPU kernels. 6701 6702 IA-32/x86-64 6703 6704 * GCC now supports the Intel CPU named Skylake with AVX-512 6705 extensions through -march=skylake-avx512. The switch enables the 6706 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 6707 AVX-512DQ. 6708 * Support for new AMD instructions monitorx and mwaitx has been 6709 added. This includes new intrinsic and built-in support. It is 6710 enabled through option -mmwaitx. The instructions monitorx and 6711 mwaitx implement the same functionality as the old monitor and 6712 mwait instructions. In addition, mwaitx adds a configurable timer. 6713 The timer value is received as third argument and stored in 6714 register %ebx. 6715 * x86-64 targets now allow stack realignment from a word-aligned 6716 stack pointer using the command-line option -mstackrealign or 6717 __attribute__ ((force_align_arg_pointer)). This allows functions 6718 compiled with a vector-aligned stack to be invoked from objects 6719 that keep only word-alignment. 6720 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 6721 can be used to access data via the %fs and %gs segments without 6722 having to resort to inline assembly. Please refer to the 6723 [19]documentation for usage instructions. 6724 * Support for AMD Zen (family 17h) processors is now available 6725 through the -march=znver1 and -mtune=znver1 options. 6726 6727 MeP 6728 6729 * Support for the MeP (mep-elf) architecture has been deprecated and 6730 will be removed in a future GCC release. 6731 6732 MSP430 6733 6734 * The MSP430 compiler now has the ability to automatically distribute 6735 code and data between low memory (addresses below 64K) and high 6736 memory. This only applies to parts that actually have both memory 6737 regions and only if the linker script for the part has been 6738 specifically set up to support this feature. 6739 A new attribute of either can be applied to both functions and 6740 data, and this tells the compiler to place the object into low 6741 memory if there is room and into high memory otherwise. Two other 6742 new attributes - lower and upper - can be used to explicitly state 6743 that an object should be placed in the specified memory region. If 6744 there is not enough left in that region the compilation will fail. 6745 Two new command-line options - -mcode-region=[lower|upper|either] 6746 and -mdata-region=[lower|upper|either] - can be used to tell the 6747 compiler what to do with objects that do not have one of these new 6748 attributes. 6749 6750 PowerPC / PowerPC64 / RS6000 6751 6752 * PowerPC64 now supports IEEE 128-bit floating-point using the 6753 __float128 data type. In GCC 6, this is not enabled by default, but 6754 you can enable it with -mfloat128. The IEEE 128-bit floating-point 6755 support requires the use of the VSX instruction set. IEEE 128-bit 6756 floating-point values are passed and returned as a single vector 6757 value. The software emulator for IEEE 128-bit floating-point 6758 support is only built on PowerPC GNU/Linux systems where the 6759 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 6760 and later), you will be able to use the -mfloat128-hardware option 6761 to use the ISA 3.0 instructions that support IEEE 128-bit 6762 floating-point. An additional type (__ibm128) has been added to 6763 refer to the IBM extended double type that normally implements long 6764 double. This will allow for a future transition to implementing 6765 long double with IEEE 128-bit floating-point. 6766 * Basic support has been added for POWER9 hardware that will use the 6767 recently published OpenPOWER ISA 3.0 instructions. The following 6768 new switches are available: 6769 + -mcpu=power9: Implement all of the ISA 3.0 instructions 6770 supported by the compiler. 6771 + -mtune=power9: In the future, apply tuning for POWER9 systems. 6772 Currently, POWER8 tunings are used. 6773 + -mmodulo: Generate code using the ISA 3.0 integer instructions 6774 (modulus, count trailing zeros, array index support, integer 6775 multiply/add). 6776 + -mpower9-fusion: Generate code to suitably fuse instruction 6777 sequences for a POWER9 system. 6778 + -mpower9-dform: Generate code to use the new D-form 6779 (register+offset) memory instructions for the vector 6780 registers. 6781 + -mpower9-vector: Generate code using the new ISA 3.0 vector 6782 (VSX or Altivec) instructions. 6783 + -mpower9-minmax: Reserved for future development. 6784 + -mtoc-fusion: Keep TOC entries together to provide more fusion 6785 opportunities. 6786 * New constraints have been added to support IEEE 128-bit 6787 floating-point and ISA 3.0 instructions: 6788 + wb: Altivec register if -mpower9-dform is enabled. 6789 + we: VSX register if -mpower9-vector is enabled for 64-bit code 6790 generation. 6791 + wo: VSX register if -mpower9-vector is enabled. 6792 + wp: Reserved for future use if long double is implemented with 6793 IEEE 128-bit floating-point instead of IBM extended double. 6794 + wq: VSX register if -mfloat128 is enabled. 6795 + wF: Memory operand suitable for POWER9 fusion load/store. 6796 + wG: Memory operand suitable for TOC fusion memory references. 6797 + wL: Integer constant identifying the element number mfvsrld 6798 accesses within a vector. 6799 * Support has been added for __builtin_cpu_is() and 6800 __builtin_cpu_supports(), allowing for very fast access to 6801 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 6802 glibc 2.23 or later. 6803 * All hardware transactional memory builtins now correctly behave as 6804 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 6805 determine whether their "old" compiler treats the builtins as 6806 barriers. 6807 * Split-stack support has been added for gccgo on PowerPC64 for both 6808 big- and little-endian (but not for 32-bit). The gold linker from 6809 at least binutils 2.25.1 must be available in the PATH when 6810 configuring and building gccgo to enable split stack. (The 6811 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 6812 split-stack feature allows a small initial stack size to be 6813 allocated for each goroutine, which increases as needed. 6814 * GCC on PowerPC now supports the standard lround function. 6815 * A new configuration option ---with-advance-toolchain=at was added 6816 for PowerPC 64-bit GNU/Linux systems to use the header files, 6817 library files, and the dynamic linker from a specific Advance 6818 Toolchain release instead of the default versions that are provided 6819 by the GNU/Linux distribution. In general, this option is intended 6820 for the developers of GCC, and it is not intended for general use. 6821 * The "q", "S", "T", and "t" asm-constraints have been removed. 6822 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 6823 6824 S/390, System z, IBM z Systems 6825 6826 * Support for the IBM z13 processor has been added. When using the 6827 -march=z13 option, the compiler will generate code making use of 6828 the new instructions and registers introduced with the vector 6829 extension facility. The -mtune=z13 option enables z13 specific 6830 instruction scheduling without making use of new instructions. 6831 Compiling code with -march=z13 reduces the default alignment of 6832 vector types bigger than 8 bytes to 8. This is an ABI change and 6833 care must be taken when linking modules compiled with different 6834 arch levels which interchange variables containing vector type 6835 values. For newly compiled code the GNU linker will emit a warning. 6836 * The -mzvector option enables a C/C++ language extension. This 6837 extension provides a new keyword vector which can be used to define 6838 vector type variables. (Note: This is not available when enforcing 6839 strict standard compliance e.g. with -std=c99. Either enable GNU 6840 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 6841 Additionally a set of overloaded builtins is provided which is 6842 partially compatible to the PowerPC Altivec builtins. In order to 6843 make use of these builtins the vecintrin.h header file needs to be 6844 included. 6845 * The new command-line options -march=native, and -mtune=native are 6846 now available on native IBM z Systems. Specifying these options 6847 causes GCC to auto-detect the host CPU and choose the optimal 6848 setting for that system. If GCC is unable to detect the host CPU 6849 these options have no effect. 6850 * The IBM z Systems port now supports target attributes and pragmas. 6851 Please refer to the [20]documentation for details of available 6852 attributes and pragmas as well as usage instructions. 6853 * -fsplit-stack is now supported as part of the IBM z Systems port. 6854 This feature requires a recent gold linker to be used. 6855 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 6856 been deprecated and will be removed in a future GCC release. -m31 6857 from now on defaults to -march=z900 if not specified otherwise. 6858 -march=native on a g5/g6 machine will default to -march=z900. 6859 6860 SH 6861 6862 * Support for SH5 / SH64 has been declared obsolete and will be 6863 removed in future releases. 6864 * Support for the FDPIC ABI has been added. It can be enabled using 6865 the new -mfdpic target option and --enable-fdpic configure option. 6866 6867 SPARC 6868 6869 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 6870 change will break binary compatibility with earlier releases for 6871 code it affects, but this should be pretty rare in practice. The 6872 conditions are: a 16-byte structure containing a double or a 8-byte 6873 vector in the second half is passed to a subprogram in slot #15, 6874 for example as 16th parameter if the first 15 ones have at most 8 6875 bytes. The double or vector was wrongly passed in floating-point 6876 register %d32 in lieu of on the stack as per the SPARC calling 6877 conventions. 6878 6879 Operating Systems 6880 6881 AIX 6882 6883 * DWARF debugging support for AIX 7.1 has been enabled as an optional 6884 debugging format. A more recent Technology Level (TL) and GCC built 6885 with that level are required for full exploitation of DWARF 6886 debugging capabilities. 6887 6888 Linux 6889 6890 * Support for the [21]musl C library was added for the AArch64, ARM, 6891 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 6892 x86_64 targets. It can be selected using the new -mmusl option in 6893 case musl is not the default libc. GCC defaults to musl libc if it 6894 is built with a target triplet matching the *-linux-musl* pattern. 6895 6896 RTEMS 6897 6898 * The RTEMS thread model implementation changed. Mutexes now use 6899 self-contained objects defined in newlib <sys/lock.h> instead of 6900 Classic API semaphores. The keys for thread specific data and the 6901 once function are directly defined via <pthread.h>. Self-contained 6902 condition variables are provided via newlib <sys/lock.h>. The RTEMS 6903 thread model also supports C++11 threads. 6904 * OpenMP support now uses self-contained objects provided by newlib 6905 <sys/lock.h> and offers a significantly better performance compared 6906 to the POSIX configuration of libgomp. It is possible to configure 6907 thread pools for each scheduler instance via the environment 6908 variable GOMP_RTEMS_THREAD_POOLS. 6909 6910 Solaris 6911 6912 * Solaris 12 is now fully supported. Minimal support had already been 6913 present in GCC 5.3. 6914 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 6915 crtn.o), which GCC now prefers over its own ones. 6916 * Position independent executables (PIE) are now supported on Solaris 6917 12. 6918 * Constructor priority is now supported on Solaris 12 with the system 6919 linker. 6920 * libvtv has been ported to Solaris 11 and up. 6921 6922 Windows 6923 6924 * The option -mstackrealign is now automatically activated in 32-bit 6925 mode whenever the use of SSE instructions is requested. 6926 6927 Other significant improvements 6928 6929 * The gcc and g++ driver programs will now provide suggestions for 6930 misspelled command-line options. 6931 6932 $ gcc -static-libfortran test.f95 6933 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 6934 '-static-libgfortran'? 6935 6936 * The --enable-default-pie configure option enables generation of PIE 6937 by default. 6938 6939 GCC 6.2 6940 6941 This is the [22]list of problem reports (PRs) from GCC's bug tracking 6942 system that are known to be fixed in the 6.2 release. This list might 6943 not be complete (that is, it is possible that some PRs that have been 6944 fixed are not listed here). 6945 6946 Target Specific Changes 6947 6948 SPARC 6949 6950 * Support for --with-cpu-32 and --with-cpu-64 configure options has 6951 been added on bi-architecture platforms. 6952 * Support for the SPARC M7 (Niagara 7) processor has been added. 6953 * Support for the VIS 4.0 instruction set has been added. 6954 6955 GCC 6.3 6956 6957 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6958 system that are known to be fixed in the 6.3 release. This list might 6959 not be complete (that is, it is possible that some PRs that have been 6960 fixed are not listed here). 6961 6962 Target Specific Changes 6963 6964 IA-32/x86-64 6965 6966 * Support for the [24]deprecated pcommit instruction has been 6967 removed. 6968 6969 GCC 6.4 6970 6971 This is the [25]list of problem reports (PRs) from GCC's bug tracking 6972 system that are known to be fixed in the 6.4 release. This list might 6973 not be complete (that is, it is possible that some PRs that have been 6974 fixed are not listed here). 6975 6976 Operating Systems 6977 6978 RTEMS 6979 6980 * The ABI changes on ARM so that no short enums are used by default. 6981 6982 GCC 6.5 6983 6984 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6985 system that are known to be fixed in the 6.5 release. This list might 6986 not be complete (that is, it is possible that some PRs that have been 6987 fixed are not listed here). 6988 6989 6990 For questions related to the use of GCC, please consult these web 6991 pages and the [27]GCC manuals. If that fails, the 6992 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6993 web pages and the development of GCC are welcome on our developer 6994 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 6995 archives. 6996 6997 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 6998 distribution of this entire article is permitted in any medium, 6999 provided this notice is preserved. 7000 7001 These pages are [32]maintained by the GCC team. Last modified 7002 2023-02-20. 7003 7004 References 7005 7006 1. https://gcc.gnu.org/gcc-6/porting_to.html 7007 2. https://gcc.gnu.org/onlinedocs/index.html#current 7008 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 7009 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 7010 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 7011 6. https://www.openacc.org/ 7012 7. https://gcc.gnu.org/wiki/OpenACC 7013 8. https://gcc.gnu.org/wiki/Offloading 7014 9. https://www.openmp.org/specifications/ 7015 10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 7016 11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 7017 12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 7018 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 7019 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 7020 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 7021 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 7022 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 7023 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 7024 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 7025 20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 7026 21. http://www.musl-libc.org/ 7027 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 7028 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 7029 24. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 7030 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 7031 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 7032 27. https://gcc.gnu.org/onlinedocs/ 7033 28. mailto:gcc-help (a] gcc.gnu.org 7034 29. mailto:gcc (a] gcc.gnu.org 7035 30. https://gcc.gnu.org/lists.html 7036 31. https://www.fsf.org/ 7037 32. https://gcc.gnu.org/about.html 7038 ====================================================================== 7039 http://gcc.gnu.org/gcc-5/index.html 7040 GCC 5 Release Series 7041 7042 (This release series is no longer supported.) 7043 7044 October 10, 2017 7045 7046 The [1]GNU project and the GCC developers are pleased to announce the 7047 release of GCC 5.5. 7048 7049 This release is a bug-fix release, containing fixes for regressions in 7050 GCC 5.4 relative to previous releases of GCC. 7051 7052 Release History 7053 7054 GCC 5.5 7055 October 10, 2017 ([2]changes, [3]documentation) 7056 7057 GCC 5.4 7058 June 3, 2016 ([4]changes, [5]documentation) 7059 7060 GCC 5.3 7061 December 4, 2015 ([6]changes, [7]documentation) 7062 7063 GCC 5.2 7064 July 16, 2015 ([8]changes, [9]documentation) 7065 7066 GCC 5.1 7067 April 22, 2015 ([10]changes, [11]documentation) 7068 7069 References and Acknowledgements 7070 7071 GCC used to stand for the GNU C Compiler, but since the compiler 7072 supports several other languages aside from C, it now stands for the 7073 GNU Compiler Collection. 7074 7075 A list of [12]successful builds is updated as new information becomes 7076 available. 7077 7078 The GCC developers would like to thank the numerous people that have 7079 contributed new features, improvements, bug fixes, and other changes as 7080 well as test results to GCC. This [13]amazing group of volunteers is 7081 what makes GCC successful. 7082 7083 For additional information about GCC please refer to the [14]GCC 7084 project web site or contact the [15]GCC development mailing list. 7085 7086 To obtain GCC please use [16]our mirror sites or [17]our version 7087 control system. 7088 7089 7090 For questions related to the use of GCC, please consult these web 7091 pages and the [18]GCC manuals. If that fails, the 7092 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7093 web pages and the development of GCC are welcome on our developer 7094 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 7095 archives. 7096 7097 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 7098 distribution of this entire article is permitted in any medium, 7099 provided this notice is preserved. 7100 7101 These pages are [23]maintained by the GCC team. Last modified 7102 2024-05-30. 7103 7104 References 7105 7106 1. http://www.gnu.org/ 7107 2. https://gcc.gnu.org/gcc-5/changes.html 7108 3. https://gcc.gnu.org/onlinedocs/5.5.0/ 7109 4. https://gcc.gnu.org/gcc-5/changes.html 7110 5. https://gcc.gnu.org/onlinedocs/5.4.0/ 7111 6. https://gcc.gnu.org/gcc-5/changes.html 7112 7. https://gcc.gnu.org/onlinedocs/5.3.0/ 7113 8. https://gcc.gnu.org/gcc-5/changes.html 7114 9. https://gcc.gnu.org/onlinedocs/5.2.0/ 7115 10. https://gcc.gnu.org/gcc-5/changes.html 7116 11. https://gcc.gnu.org/onlinedocs/5.1.0/ 7117 12. https://gcc.gnu.org/gcc-5/buildstat.html 7118 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7119 14. https://gcc.gnu.org/index.html 7120 15. mailto:gcc (a] gcc.gnu.org 7121 16. https://gcc.gnu.org/mirrors.html 7122 17. https://gcc.gnu.org/git.html 7123 18. https://gcc.gnu.org/onlinedocs/ 7124 19. mailto:gcc-help (a] gcc.gnu.org 7125 20. mailto:gcc (a] gcc.gnu.org 7126 21. https://gcc.gnu.org/lists.html 7127 22. https://www.fsf.org/ 7128 23. https://gcc.gnu.org/about.html 7129 ====================================================================== 7130 http://gcc.gnu.org/gcc-5/changes.html 7131 GCC 5 Release Series 7132 Changes, New Features, and Fixes 7133 7134 Caveats 7135 7136 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 7137 * The C++ runtime library (libstdc++) uses a new ABI by default (see 7138 [1]below). 7139 * The Graphite framework for loop optimizations no longer requires 7140 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 7141 The installation manual contains more information about 7142 requirements to build GCC. 7143 * The non-standard C++0x type traits has_trivial_default_constructor, 7144 has_trivial_copy_constructor and has_trivial_copy_assign have been 7145 deprecated and will be removed in a future version. The standard 7146 C++11 traits is_trivially_default_constructible, 7147 is_trivially_copy_constructible and is_trivially_copy_assignable 7148 should be used instead. 7149 * On AVR, support has been added for the devices 7150 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 7151 * The AVR port uses a new scheme to describe supported devices: For 7152 each supported device the compiler provides a device-specific 7153 [2]spec file. If the compiler is used together with AVR-LibC, this 7154 requires at least GCC 5.2 and a version of AVR-LibC which 7155 implements [3]feature #44574. 7156 7157 General Optimizer Improvements 7158 7159 * Inter-procedural optimization improvements: 7160 + An Identical Code Folding (ICF) pass (controlled via 7161 -fipa-icf) has been added. Compared to the identical code 7162 folding performed by the Gold linker this pass does not 7163 require function sections. It also performs merging before 7164 inlining, so inter-procedural optimizations are aware of the 7165 code re-use. On the other hand not all unifications performed 7166 by a linker are doable by GCC which must honor aliasing 7167 information. During link-time optimization of Firefox, this 7168 pass unifies about 31000 functions, that is 14% overall. 7169 + The devirtualization pass was significantly improved by adding 7170 better support for speculative devirtualization and dynamic 7171 type detection. About 50% of virtual calls in Firefox are now 7172 speculatively devirtualized during link-time optimization. 7173 + A new comdat localization pass allows the linker to eliminate 7174 more dead code in presence of C++ inline functions. 7175 + Virtual tables are now optimized. Local aliases are used to 7176 reduce dynamic linking time of C++ virtual tables on ELF 7177 targets and data alignment has been reduced to limit data 7178 segment bloat. 7179 + A new -fno-semantic-interposition option can be used to 7180 improve code quality of shared libraries where interposition 7181 of exported symbols is not allowed. 7182 + Write-only variables are now detected and optimized out. 7183 + With profile feedback the function inliner can now bypass 7184 --param inline-insns-auto and --param inline-insns-single 7185 limits for hot calls. 7186 + The IPA reference pass was significantly sped up making it 7187 feasible to enable -fipa-reference with -fprofile-generate. 7188 This also solves a bottleneck seen when building Chromium with 7189 link-time optimization. 7190 + The symbol table and call-graph API was reworked to C++ and 7191 simplified. 7192 + The interprocedural propagation of constants now also 7193 propagates alignments of pointer parameters. This for example 7194 means that the vectorizer often does not need to generate loop 7195 prologues and epilogues to make up for potential 7196 misalignments. 7197 * Link-time optimization improvements: 7198 + One Definition Rule based merging of C++ types has been 7199 implemented. Type merging enables better devirtualization and 7200 alias analysis. Streaming extra information needed to merge 7201 types adds about 2-6% of memory size and object size increase. 7202 This can be controlled by -flto-odr-type-merging. 7203 + Command-line optimization and target options are now streamed 7204 on a per-function basis and honored by the link-time 7205 optimizer. This change makes link-time optimization a more 7206 transparent replacement of per-file optimizations. It is now 7207 possible to build projects that require different optimization 7208 settings for different translation units (such as -ffast-math, 7209 -mavx, or -finline). Contrary to earlier GCC releases, the 7210 optimization and target options passed on the link command 7211 line are ignored. 7212 Note that this applies only to those command-line options that 7213 can be passed to optimize and target attributes. Command-line 7214 options affecting global code generation (such as -fpic), 7215 warnings (such as -Wodr), optimizations affecting the way 7216 static variables are optimized (such as -fcommon), debug 7217 output (such as -g), and --param parameters can be applied 7218 only to the whole link-time optimization unit. In these cases, 7219 it is recommended to consistently use the same options at both 7220 compile time and link time. 7221 + GCC bootstrap now uses slim LTO object files. 7222 + Memory usage and link times were improved. Tree merging was 7223 sped up, memory usage of GIMPLE declarations and types was 7224 reduced, and, support for on-demand streaming of variable 7225 constructors was added. 7226 * Feedback directed optimization improvements: 7227 + A new auto-FDO mode uses profiles collected by low overhead 7228 profiling tools (perf) instead of more expensive program 7229 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 7230 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 7231 traditional feedback directed optimization. 7232 + Profile precision was improved in presence of C++ inline and 7233 extern inline functions. 7234 + The new gcov-tool utility allows manipulating profiles. 7235 + Profiles are now more tolerant to source file changes (this 7236 can be controlled by --param profile-func-internal-id). 7237 * Register allocation improvements: 7238 + A new local register allocator (LRA) sub-pass, controlled by 7239 -flra-remat, implements control-flow sensitive global register 7240 rematerialization. Instead of spilling and restoring a 7241 register value, it is recalculated if it is profitable. The 7242 sub-pass improved SPEC2000 generated code by 1% and 0.5% 7243 correspondingly on ARM and x86-64. 7244 + Reuse of the PIC hard register, instead of using a fixed 7245 register, was implemented on x86/x86-64 targets. This improves 7246 generated PIC code performance as more hard registers can be 7247 used. Shared libraries can significantly benefit from this 7248 optimization. Currently it is switched on only for x86/x86-64 7249 targets. As RA infrastructure is already implemented for PIC 7250 register reuse, other targets might follow this in the future. 7251 + A simple form of inter-procedural RA was implemented. When it 7252 is known that a called function does not use caller-saved 7253 registers, save/restore code is not generated around the call 7254 for such registers. This optimization can be controlled by 7255 -fipa-ra 7256 + LRA is now much more effective at generating spills of general 7257 registers into vector registers instead of memory on 7258 architectures (e.g., modern Intel processors) where this is 7259 profitable. 7260 * UndefinedBehaviorSanitizer gained a few new sanitization options: 7261 + -fsanitize=float-divide-by-zero: detect floating-point 7262 division by zero; 7263 + -fsanitize=float-cast-overflow: check that the result of 7264 floating-point type to integer conversions do not overflow; 7265 + -fsanitize=bounds: enable instrumentation of array bounds and 7266 detect out-of-bounds accesses; 7267 + -fsanitize=alignment: enable alignment checking, detect 7268 various misaligned objects; 7269 + -fsanitize=object-size: enable object size checking, detect 7270 various out-of-bounds accesses. 7271 + -fsanitize=vptr: enable checking of C++ member function calls, 7272 member accesses and some conversions between pointers to base 7273 and derived classes, detect if the referenced object does not 7274 have the correct dynamic type. 7275 * Pointer Bounds Checker, a bounds violation detector, has been added 7276 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 7277 instrumented with run-time checks of used pointers against their 7278 bounds to detect pointer bounds violations (overflows). The Pointer 7279 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 7280 new ISA extension Intel MPX support. See the Pointer Bounds Checker 7281 [4]Wiki page for more details. 7282 7283 New Languages and Language specific improvements 7284 7285 * [5]OpenMP 4.0 specification offloading features are now supported 7286 by the C, C++, and Fortran compilers. Generic changes: 7287 + Infrastructure (suitable for any vendor). 7288 + Testsuite which covers offloading from the [6]OpenMP 4.0 7289 Examples document. 7290 Specific for upcoming Intel Xeon Phi products: 7291 + Run-time library. 7292 + Card emulator. 7293 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 7294 specification. OpenACC is intended for programming accelerator 7295 devices such as GPUs. See [7]the OpenACC wiki page for more 7296 information. 7297 7298 C family 7299 7300 * The default setting of the -fdiagnostics-color= command-line option 7301 is now [8]configurable when building GCC using configuration option 7302 --with-diagnostics-color=. The possible values are: never, always, 7303 auto and auto-if-env. The new default auto uses color only when the 7304 standard error is a terminal. The default in GCC 4.9 was 7305 auto-if-env, which is equivalent to auto if there is a non-empty 7306 GCC_COLORS environment variable, and never otherwise. As in GCC 7307 4.9, an empty GCC_COLORS variable in the environment will always 7308 disable colors, no matter what the default is or what command-line 7309 options are used. 7310 * A new command-line option -Wswitch-bool has been added for the C 7311 and C++ compilers, which warns whenever a switch statement has an 7312 index of boolean type. 7313 * A new command-line option -Wlogical-not-parentheses has been added 7314 for the C and C++ compilers, which warns about "logical not" used 7315 on the left hand side operand of a comparison. 7316 * A new command-line option -Wsizeof-array-argument has been added 7317 for the C and C++ compilers, which warns when the sizeof operator 7318 is applied to a parameter that has been declared as an array in a 7319 function definition. 7320 * A new command-line option -Wbool-compare has been added for the C 7321 and C++ compilers, which warns about boolean expressions compared 7322 with an integer value different from true/false. 7323 * Full support for Cilk Plus has been added to the GCC compiler. Cilk 7324 Plus is an extension to the C and C++ languages to support data and 7325 task parallelism. 7326 * A new attribute no_reorder prevents reordering of selected symbols 7327 against other such symbols or inline assembler. This enables to 7328 link-time optimize the Linux kernel without having to resort to 7329 -fno-toplevel-reorder that disables several optimizations. 7330 * New preprocessor constructs, __has_include and __has_include_next, 7331 to test the availability of headers have been added. 7332 This demonstrates a way to include the header <optional> only if it 7333 is available: 7334 7335 #ifdef __has_include 7336 # if __has_include(<optional>) 7337 # include <optional> 7338 # define have_optional 1 7339 # elif __has_include(<experimental/optional>) 7340 # include <experimental/optional> 7341 # define have_optional 1 7342 # define experimental_optional 7343 # else 7344 # define have_optional 0 7345 # endif 7346 #endif 7347 7348 The header search paths for __has_include and __has_include_next 7349 are equivalent to those of the standard directive #include and the 7350 extension #include_next respectively. 7351 * A new built-in function-like macro to determine the existence of an 7352 attribute, __has_attribute, has been added. The equivalent built-in 7353 macro __has_cpp_attribute was added to C++ to support 7354 [9]Feature-testing recommendations for C++. The macro 7355 __has_attribute is added to all C-like languages as an extension: 7356 7357 int 7358 #ifdef __has_attribute 7359 # if __has_attribute(__noinline__) 7360 __attribute__((__noinline__)) 7361 # endif 7362 #endif 7363 foo(int x); 7364 7365 If an attribute exists, a nonzero constant integer is returned. For 7366 standardized C++ attributes a date is returned, otherwise the 7367 constant returned is 1. Both __has_attribute and 7368 __has_cpp_attribute will add underscores to an attribute name if 7369 necessary to resolve the name. For C++11 and onwards the attribute 7370 may be scoped. 7371 * A new set of built-in functions for arithmetics with overflow 7372 checking has been added: __builtin_add_overflow, 7373 __builtin_sub_overflow and __builtin_mul_overflow and for 7374 compatibility with clang also other variants. These builtins have 7375 two integral arguments (which don't need to have the same type), 7376 the arguments are extended to infinite precision signed type, +, - 7377 or * is performed on those, and the result is stored in an integer 7378 variable pointed to by the last argument. If the stored value is 7379 equal to the infinite precision result, the built-in functions 7380 return false, otherwise true. The type of the integer variable that 7381 will hold the result can be different from the types of the first 7382 two arguments. The following snippet demonstrates how this can be 7383 used in computing the size for the calloc function: 7384 7385 void * 7386 calloc (size_t x, size_t y) 7387 { 7388 size_t sz; 7389 if (__builtin_mul_overflow (x, y, &sz)) 7390 return NULL; 7391 void *ret = malloc (sz); 7392 if (ret) memset (res, 0, sz); 7393 return ret; 7394 } 7395 7396 On e.g. i?86 or x86-64 the above will result in a mul instruction 7397 followed by a jump on overflow. 7398 * The option -fextended-identifiers is now enabled by default for 7399 C++, and for C99 and later C versions. Various bugs in the 7400 implementation of extended identifiers have been fixed. 7401 7402 C 7403 7404 * The default mode has been changed to -std=gnu11. 7405 * A new command-line option -Wc90-c99-compat has been added to warn 7406 about features not present in ISO C90, but present in ISO C99. 7407 * A new command-line option -Wc99-c11-compat has been added to warn 7408 about features not present in ISO C99, but present in ISO C11. 7409 * It is possible to disable warnings about conversions between 7410 pointers that have incompatible types via a new warning option 7411 -Wno-incompatible-pointer-types; warnings about implicit 7412 incompatible integer to pointer and pointer to integer conversions 7413 via a new warning option -Wno-int-conversion; and warnings about 7414 qualifiers on pointers being discarded via a new warning option 7415 -Wno-discarded-qualifiers. 7416 * To allow proper use of const qualifiers with multidimensional 7417 arrays, GCC will not warn about incompatible pointer types anymore 7418 for conversions between pointers to arrays with and without const 7419 qualifier (except when using -pedantic). Instead, a new warning is 7420 emitted only if the const qualifier is lost. This can be controlled 7421 with a new warning option -Wno-discarded-array-qualifiers. 7422 * The C front end now generates more precise caret diagnostics. 7423 * The -pg command-line option now only affects the current file in an 7424 LTO build. 7425 7426 C++ 7427 7428 * G++ now supports [10]C++14 variable templates. 7429 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 7430 * Excessive template instantiation depth is now a fatal error. This 7431 prevents excessive diagnostics that usually do not help to identify 7432 the problem. 7433 * G++ and libstdc++ now implement the feature-testing macros from 7434 [11]Feature-testing recommendations for C++. 7435 * G++ now allows typename in a template template parameter. 7436 7437 template<template<typename> typename X> struct D; // OK 7438 7439 * G++ now supports [12]C++14 aggregates with non-static data member 7440 initializers. 7441 7442 struct A { int i, j = i; }; 7443 A a = { 42 }; // a.j is also 42 7444 7445 * G++ now supports [13]C++14 extended constexpr. 7446 7447 constexpr int f (int i) 7448 { 7449 int j = 0; 7450 for (; i > 0; --i) 7451 ++j; 7452 return j; 7453 } 7454 7455 constexpr int i = f(42); // i is 42 7456 7457 * G++ now supports the [14]C++14 sized deallocation functions. 7458 7459 void operator delete (void *, std::size_t) noexcept; 7460 void operator delete[] (void *, std::size_t) noexcept; 7461 7462 * A new One Definition Rule violation warning (controlled by -Wodr) 7463 detects mismatches in type definitions and virtual table contents 7464 during link-time optimization. 7465 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 7466 developers to annotate programs with final specifiers (or anonymous 7467 namespaces) to improve code generation. These warnings can be used 7468 at compile time, but they are more useful in combination with 7469 link-time optimization. 7470 * G++ no longer supports [15]N3639 variable length arrays, as they 7471 were removed from the C++14 working paper prior to ratification. 7472 GNU VLAs are still supported, so VLA support is now the same in 7473 C++14 mode as in C++98 and C++11 modes. 7474 * G++ now allows passing a non-trivially-copyable class via C 7475 varargs, which is conditionally-supported with 7476 implementation-defined semantics in the standard. This uses the 7477 same calling convention as a normal value parameter. 7478 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 7479 various mangling bugs are fixed, but G++ will still emit aliases 7480 with the old, wrong mangling where feasible. -Wabi=2 will warn 7481 about differences between ABI version 2 and the current setting. 7482 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 7483 to be unaffected, but -Wabi=8 will warn about a non-static data 7484 member with type std::nullptr_t which changes position due to this 7485 change. 7486 7487 Runtime Library (libstdc++) 7488 7489 * A [16]Dual ABI is provided by the library. A new ABI is enabled by 7490 default. The old ABI is still supported and can be used by defining 7491 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 7492 standard library headers. 7493 * A new implementation of std::string is enabled by default, using 7494 the small string optimization instead of copy-on-write reference 7495 counting. 7496 * A new implementation of std::list is enabled by default, with an 7497 O(1) size() function; 7498 * [17]Full support for C++11, including the following new features: 7499 + std::deque and std::vector<bool> meet the allocator-aware 7500 container requirements; 7501 + movable and swappable iostream classes; 7502 + support for std::align and std::aligned_union; 7503 + type traits std::is_trivially_copyable, 7504 std::is_trivially_constructible, std::is_trivially_assignable 7505 etc.; 7506 + I/O manipulators std::put_time, std::get_time, std::hexfloat 7507 and std::defaultfloat; 7508 + generic locale-aware std::isblank; 7509 + locale facets for Unicode conversion; 7510 + atomic operations for std::shared_ptr; 7511 + std::notify_all_at_thread_exit() and functions for making 7512 futures ready at thread exit. 7513 * Support for the C++11 hexfloat manipulator changes how the num_put 7514 facet formats floating point types when 7515 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 7516 This change affects all language modes, even though the C++98 7517 standard gave no special meaning to that combination of flags. To 7518 prevent the use of hexadecimal notation for floating point types 7519 use str.unsetf(std::ios_base::floatfield) to clear the relevant 7520 bits in str.flags(). 7521 * [18]Full experimental support for C++14, including the following 7522 new features: 7523 + std::is_final type trait; 7524 + heterogeneous comparison lookup in associative containers. 7525 + global functions cbegin, cend, rbegin, rend, crbegin, and 7526 crend for range access to containers, arrays and initializer 7527 lists. 7528 * [19]Improved experimental support for the Library Fundamentals TS, 7529 including: 7530 + class std::experimental::any; 7531 + function template std::experimental::apply; 7532 + function template std::experimental::sample; 7533 + function template std::experimental::search and related 7534 searcher types; 7535 + variable templates for type traits; 7536 + function template std::experimental::not_fn. 7537 * New random number distributions logistic_distribution and 7538 uniform_on_sphere_distribution as extensions. 7539 * [20]GDB Xmethods for containers and std::unique_ptr. 7540 7541 Fortran 7542 7543 * Compatibility notice: 7544 + The version of the module files (.mod) has been incremented. 7545 + For free-form source files [21]-Werror=line-truncation is now 7546 enabled by default. Note that comments exceeding the line 7547 length are not diagnosed. (For fixed-form source code, the 7548 same warning is available but turned off by default, such that 7549 excess characters are ignored. -ffree-line-length-n and 7550 -ffixed-line-length-n can be used to modify the default line 7551 lengths of 132 and 72 columns, respectively.) 7552 + The -Wtabs option is now more sensible: with -Wtabs the 7553 compiler warns if it encounters tabs and with -Wno-tabs this 7554 warning is turned off. Before, -Wno-tabs warned and -Wtabs 7555 disabled the warning. As before, this warning is also enabled 7556 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 7557 options of -std=. 7558 * Incomplete support for colorizing diagnostics emitted by gfortran 7559 has been added. The option [22]-fdiagnostics-color controls when 7560 color is used in diagnostics. The default value of this option can 7561 be [23]configured when building GCC. The GCC_COLORS environment 7562 variable can be used to customize the colors or disable coloring 7563 completely. Sample diagnostics output: 7564 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 7565 test.f90:6:1: 7566 7567 0 continue 7568 1 7569 Error: Zero is not a valid statement label at (1) 7570 test.f90:9:6: 7571 7572 USE foo 7573 1 7574 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 7575 7576 * The -Wuse-without-only option has been added to warn when a USE 7577 statement has no ONLY qualifier and thus implicitly imports all 7578 public entities of the used module. 7579 * Formatted READ and WRITE statements now work correctly in 7580 locale-aware programs. For more information and potential caveats, 7581 see [24]Section 5.3 Thread-safety of the runtime library in the 7582 manual. 7583 * [25]Fortran 2003: 7584 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 7585 IEEE_ARITHMETIC) are now supported. 7586 * [26]Fortran 2008: 7587 + [27]Coarrays: Full experimental support of Fortran 2008's 7588 coarrays with -fcoarray=lib except for allocatable/pointer 7589 components of derived-type coarrays. GCC currently only ships 7590 with a single-image library (libcaf_single), but multi-image 7591 support based on MPI and GASNet is provided by the libraries 7592 of the [28]OpenCoarrays project. 7593 * TS18508 Additional Parallel Features in Fortran: 7594 + Support for the collective intrinsic subroutines CO_MAX, 7595 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 7596 including -fcoarray=lib support. 7597 + Support for the new atomic intrinsics has been added, 7598 including -fcoarray=lib support. 7599 * Fortran 2015: 7600 + Support for IMPLICIT NONE (external, type). 7601 + ERROR STOP is now permitted in pure procedures. 7602 7603 Go 7604 7605 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 7606 * Building GCC 5 with Go enabled will install two new programs: 7607 [29]go and [30]gofmt. 7608 7609 libgccjit 7610 7611 New in GCC 5 is the ability to build GCC as a shared library for 7612 embedding in other processes (such as interpreters), suitable for 7613 Just-In-Time compilation to machine code. 7614 7615 The shared library has a [31]C API and a [32]C++ wrapper API providing 7616 some "syntactic sugar". There are also bindings available from 3rd 7617 parties for [33]Python and for [34]D. 7618 7619 For example, this library can be used by interpreters for [35]compiling 7620 functions from bytecode to machine code. 7621 7622 The library can also be used for ahead-of-time compilation, enabling 7623 GCC to be plugged into a pre-existing front end. An example of using 7624 this to build a compiler for an esoteric language we'll refer to as 7625 "brainf" can be seen [36]here. 7626 7627 libgccjit is licensed under the GPLv3 (or at your option, any later 7628 version) 7629 7630 It should be regarded as experimental at this time. 7631 7632 New Targets and Target Specific Improvements 7633 7634 Reporting stack usage 7635 7636 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 7637 -fstack-usage option. 7638 7639 AArch64 7640 7641 * Code generation for the ARM Cortex-A57 processor has been improved. 7642 A more accurate instruction scheduling model for the processor is 7643 now used, and a number of compiler tuning parameters have been set 7644 to offer increased performance when compiling with -mcpu=cortex-a57 7645 or -mtune=cortex-a57. 7646 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 7647 and can be enabled by giving the -mfix-cortex-a53-835769 option. 7648 Alternatively it can be enabled by default by configuring GCC with 7649 the --enable-fix-cortex-a53-835769 option. 7650 * The optional cryptographic extensions to the ARMv8-A architecture 7651 are no longer enabled by default when specifying the 7652 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 7653 options. To enable these extensions add +crypto to the value of 7654 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 7655 * Support has been added for the following processors (GCC 7656 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 7657 initial support for its big.LITTLE combination with the ARM 7658 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 7659 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 7660 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7661 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 7662 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 7663 support for the Cortex-A72. 7664 * The transitional options -mlra and -mno-lra have been removed. The 7665 AArch64 back end now uses the local register allocator (LRA) only. 7666 7667 ARM 7668 7669 * Thumb-1 assembly code is now generated in unified syntax. The new 7670 option -masm-syntax-unified specifies whether inline assembly code 7671 is using unified syntax. By default the option is off which means 7672 non-unified syntax is used. However this is subject to change in 7673 future releases. Eventually the non-unified syntax will be 7674 deprecated. 7675 * It is now a configure-time error to use the --with-cpu configure 7676 option with either of --with-tune or --with-arch. 7677 * Code generation for the ARM Cortex-A57 processor has been improved. 7678 A more accurate instruction scheduling model for the processor is 7679 now used, and a number of compiler tuning parameters have been set 7680 to offer increased performance when compiling with -mcpu=cortex-a57 7681 or -mtune=cortex-a57. 7682 * Support has been added for the following processors (GCC 7683 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 7684 initial support for its big.LITTLE combination with the ARM 7685 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 7686 initial support for its big.LITTLE combination with the ARM 7687 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 7688 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 7689 The GCC identifiers can be used as arguments to the -mcpu or -mtune 7690 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 7691 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 7692 support for the Cortex-A72. 7693 * The deprecated option -mwords-little-endian has been removed. 7694 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 7695 -mtpcs-leaf-frame which are only applicable to the old ABI have 7696 been deprecated. 7697 * The transitional options -mlra and -mno-lra have been removed. The 7698 ARM back end now uses the local register allocator (LRA) only. 7699 7700 AVR 7701 7702 * The compiler no more supports individual devices like ATmega8. 7703 Specifying, say, -mmcu=atmega8 triggers the usage of the 7704 device-specific [37]spec file specs-atmega8 which is part of the 7705 installation and describes options for the sub-processes like 7706 compiler proper, assembler and linker. You can add support for a 7707 new device -mmcu=mydevice as follows: 7708 1. In an empty directory /someplace, create a new directory 7709 device-specs. 7710 2. Copy a device spec file from the installed device-specs 7711 folder, follow the comments in that file and then save it as 7712 /someplace/device-specs/specs-mydevice. 7713 3. Add -B /someplace -mmcu=mydevice to the compiler's 7714 command-line options. Notice that /someplace must specify an 7715 absolute path and that mydevice must not start with "avr". 7716 4. Provided you have a device-specific library libmydevice.a 7717 available, you can put it at /someplace, dito for a 7718 device-specific startup file crtmydevice.o. 7719 The contents of the device spec files depend on the compiler's 7720 configuration, in particular on --with-avrlibc=no and whether or 7721 not it is configured for RTEMS. 7722 * A new command-line option -nodevicelib has been added. It prevents 7723 the compiler from linking against AVR-LibC's device-specific 7724 library libdevice.a. 7725 * The following three command-line options have been added: 7726 7727 -mrmw 7728 Set if the device supports the read-modify-write 7729 instructions LAC, LAS, LAT and XCH. 7730 7731 -mn-flash=size 7732 Specify the flash size of the device in units of 64 KiB, 7733 rounded up to the next integer as needed. This option 7734 affects the availability of the [38]AVR address-spaces. 7735 7736 -mskip-bug 7737 Set if the device is affected by the respective silicon 7738 bug. 7739 7740 In general, you don't need to set these options by hand. The new 7741 device-specific spec file will set them as needed. 7742 7743 IA-32/x86-64 7744 7745 * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's 7746 CPU codenamed Skylake Server was added to GCC. That includes inline 7747 assembly support, new intrinsics, and basic autovectorization. 7748 These new AVX-512 extensions are available via the following GCC 7749 switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512 7750 Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword 7751 instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma 7752 and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi. 7753 * New ISA extensions support Intel MPX was added to GCC. This new 7754 extension is available via the -mmpx compiler switch. Intel MPX is 7755 a set of processor features which, with compiler, run-time library 7756 and OS support, brings increased robustness to software by run-time 7757 checking pointer references against their bounds. In GCC Intel MPX 7758 is supported by Pointer Bounds Checker and libmpx run-time 7759 libraries. 7760 * The new -mrecord-mcount option for -pg generates a Linux kernel 7761 style table of pointers to mcount or __fentry__ calls at the 7762 beginning of functions. The new -mnop-mcount option in addition 7763 also generates nops in place of the __fentry__ or mcount call, so 7764 that a call per function can be later patched in. This can be used 7765 for low overhead tracing or hot code patching. 7766 * The new -malign-data option controls how GCC aligns variables. 7767 -malign-data=compat uses increased alignment compatible with GCC 7768 4.8 and earlier, -malign-data=abi uses alignment as specified by 7769 the psABI, and -malign-data=cacheline uses increased alignment to 7770 match the cache line size. -malign-data=compat is the default. 7771 * The new -mskip-rax-setup option skips setting up the RAX register 7772 when SSE is disabled and there are no variable arguments passed in 7773 vector registers. This can be used to optimize the Linux kernel. 7774 7775 MIPS 7776 7777 * MIPS Releases 3 and 5 are now directly supported. Use the 7778 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 7779 to enable code-generation for these processors. 7780 * The Imagination P5600 processor is now supported using the 7781 -march=p5600 command-line option. 7782 * The Cavium Octeon3 processor is now supported using the 7783 -march=octeon3 command-line option. 7784 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 7785 command-line options. 7786 * The o32 ABI has been modified and extended. The o32 64-bit 7787 floating-point register support is now obsolete and has been 7788 removed. It has been replaced by three ABI extensions FPXX, FP64A, 7789 and FP64. The meaning of the -mfp64 command-line option has 7790 changed. It is now used to enable the FP64A and FP64 ABI 7791 extensions. 7792 + The FPXX extension requires that code generated to access 7793 double-precision values use even-numbered registers. Code that 7794 adheres to this extension is link-compatible with all other 7795 o32 double-precision ABI variants and will execute correctly 7796 in all hardware FPU modes. The command-line options -mabi=32 7797 -mfpxx can be used to enable this extension. MIPS II is the 7798 minimum processor required. 7799 + The o32 FP64A extension requires that floating-point registers 7800 be 64-bit and odd-numbered single-precision registers are not 7801 allowed. Code that adheres to the o32 FP64A variant is 7802 link-compatible with all other o32 double-precision ABI 7803 variants. The command-line options -mabi=32 -mfp64 7804 -mno-odd-spreg can be used to enable this extension. MIPS32R2 7805 is the minimum processor required. 7806 + The o32 FP64 extension also requires that floating-point 7807 registers be 64-bit, but permits the use of single-precision 7808 registers. Code that adheres to the o32 FP64 variant is 7809 link-compatible with o32 FPXX and o32 FP64A variants only, 7810 i.e. it is not compatible with the original o32 7811 double-precision ABI. The command-line options -mabi=32 -mfp64 7812 -modd-spreg can be used to enable this extension. MIPS32R2 is 7813 the minimum processor required. 7814 The new ABI variants can be enabled by default using the configure 7815 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 7816 It is strongly recommended that all vendors begin to set o32 FPXX 7817 as the default ABI. This will be required to run the generated code 7818 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 7819 MIPSR6 cores. 7820 * GCC will now pass all floating-point options to the assembler if 7821 GNU binutils 2.25 is used. As a result, any inline assembly code 7822 that uses hard-float instructions should be amended to include a 7823 .set directive to override the global assembler options when 7824 compiling for soft-float targets. 7825 7826 NDS32 7827 7828 * The variadic function ABI implementation is now compatible with 7829 past Andes toolchains where the caller uses registers to pass 7830 arguments and the callee is in charge of pushing them on stack. 7831 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 7832 been removed since they are not yet available in the nds32 port of 7833 GNU binutils. 7834 * A new option -mcmodel=[small|medium|large] supports varied code 7835 models on code generation. The -mgp-direct option became 7836 meaningless and can be discarded. 7837 7838 RX 7839 7840 * A new command line option -mno-allow-string-insns can be used to 7841 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 7842 SWHILE and RMPA instructions. An erratum released by Renesas shows 7843 that it is unsafe to use these instructions on addresses within the 7844 I/O space of the processor. The new option can be used when the 7845 programmer is concerned that the I/O space might be accessed. The 7846 default is still to enable these instructions. 7847 7848 SH 7849 7850 * The compiler will now pass the appropriate --isa= option to the 7851 assembler. 7852 * The default handling for the GBR has been changed from call 7853 clobbered to call preserved. The old behavior can be reinstated by 7854 specifying the option -fcall-used-gbr. 7855 * Support for the SH4A fpchg instruction has been added which will be 7856 utilized when switching between single and double precision FPU 7857 modes. 7858 * The compiler no longer uses the __fpscr_values array for switching 7859 between single and double FPU precision modes on non-SH4A targets. 7860 Instead mode switching will now be performed by storing, modifying 7861 and reloading the FPSCR, so that other FPSCR bits are preserved 7862 across mode switches. The __fpscr_values array that is defined in 7863 libgcc is still present for backwards compatibility, but it will 7864 not be referenced by compiler generated code anymore. 7865 * New builtin functions __builtin_sh_get_fpscr and 7866 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 7867 function will mask the specified bits in such a way that the SZ, PR 7868 and FR mode bits will be preserved, while changing the other bits. 7869 These new functions do not reference the __fpscr_values array. The 7870 old functions __set_fpscr and __get_fpscr in libgcc which access 7871 the __fpscr_values array are still present for backwards 7872 compatibility, but their usage is highly discouraged. 7873 * Some improvements to code generated for __atomic built-in 7874 functions. 7875 * When compiling for SH2E the compiler will no longer force the usage 7876 of delay slots for conditional branch instructions bt and bf. The 7877 old behavior can be reinstated (e.g. to work around a hardware bug 7878 in the original SH7055) by specifying the new option 7879 -mcbranch-force-delay-slot. 7880 7881 Operating Systems 7882 7883 AIX 7884 7885 * GCC now supports stabs debugging continuation lines to allow long 7886 stabs debug information without overflow that generates AIX linker 7887 errors. 7888 7889 DragonFly BSD 7890 7891 * GCC now supports the DragonFly BSD operating system. 7892 7893 FreeBSD 7894 7895 * GCC now supports the FreeBSD operating system for the arm port 7896 through the arm*-*-freebsd* target triplets. 7897 7898 VxWorks MILS 7899 7900 * GCC now supports the MILS (Multiple Independent Levels of Security) 7901 variant of WindRiver's VxWorks operating system for PowerPC 7902 targets. 7903 7904 Other significant improvements 7905 7906 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 7907 to set the compiler to use. 7908 7909 * When the new command-line option -freport-bug is used, GCC 7910 automatically generates a developer-friendly reproducer whenever an 7911 internal compiler error is encountered. 7912 7913 GCC 5.2 7914 7915 This is the [39]list of problem reports (PRs) from GCC's bug tracking 7916 system that are known to be fixed in the 5.2 release. This list might 7917 not be complete (that is, it is possible that some PRs that have been 7918 fixed are not listed here). 7919 7920 Target Specific Changes 7921 7922 IA-32/x86-64 7923 7924 * Support for new AMD instructions monitorx and mwaitx has been 7925 added. This includes new intrinsic and built-in support. It is 7926 enabled through option -mmwaitx. The instructions monitorx and 7927 mwaitx implement the same functionality as the old monitor and 7928 mwait instructions. In addition, mwaitx adds a configurable timer. 7929 The timer value is received as third argument and stored in 7930 register %ebx. 7931 7932 S/390, System z, IBM z Systems 7933 7934 * Support for the IBM z13 processor has been added. When using the 7935 -march=z13 option, the compiler will generate code making use of 7936 the new instructions and registers introduced with the vector 7937 extension facility. The -mtune=z13 option enables z13 specific 7938 instruction scheduling without making use of new instructions. 7939 Compiling code with -march=z13 reduces the default alignment of 7940 vector types bigger than 8 bytes to 8. This is an ABI change and 7941 care must be taken when linking modules compiled with different 7942 arch levels which interchange variables containing vector type 7943 values. For newly compiled code the GNU linker will emit a warning. 7944 * The -mzvector option enables a C/C++ language extension. This 7945 extension provides a new keyword vector which can be used to define 7946 vector type variables. (Note: This is not available when enforcing 7947 strict standard compliance e.g. with -std=c99. Either enable GNU 7948 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 7949 Additionally a set of overloaded builtins is provided which is 7950 partially compatible to the PowerPC Altivec builtins. In order to 7951 make use of these builtins the vecintrin.h header file needs to be 7952 included. 7953 7954 GCC 5.3 7955 7956 This is the [40]list of problem reports (PRs) from GCC's bug tracking 7957 system that are known to be fixed in the 5.3 release. This list might 7958 not be complete (that is, it is possible that some PRs that have been 7959 fixed are not listed here). 7960 7961 Target Specific Changes 7962 7963 IA-32/x86-64 7964 7965 * GCC now supports the Intel CPU named Skylake with AVX-512 7966 extensions through -march=skylake-avx512. The switch enables the 7967 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 7968 AVX-512DQ. 7969 7970 S/390, System z, IBM z Systems 7971 7972 * With this version of GCC IBM z Systems support has been added to 7973 the GO runtime environment. GCC 5.3 has proven to be able to 7974 compile larger GO applications on IBM z Systems. 7975 7976 GCC 5.4 7977 7978 This is the [41]list of problem reports (PRs) from GCC's bug tracking 7979 system that are known to be fixed in the 5.4 release. This list might 7980 not be complete (that is, it is possible that some PRs that have been 7981 fixed are not listed here). 7982 7983 GCC 5.5 7984 7985 This is the [42]list of problem reports (PRs) from GCC's bug tracking 7986 system that are known to be fixed in the 5.5 release. This list might 7987 not be complete (that is, it is possible that some PRs that have been 7988 fixed are not listed here). 7989 7990 Target Specific Changes 7991 7992 IA-32/x86-64 7993 7994 * Support for the [43]deprecated pcommit instruction has been 7995 removed. 7996 7997 7998 For questions related to the use of GCC, please consult these web 7999 pages and the [44]GCC manuals. If that fails, the 8000 [45]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8001 web pages and the development of GCC are welcome on our developer 8002 list at [46]gcc (a] gcc.gnu.org. All of [47]our lists have public 8003 archives. 8004 8005 Copyright (C) [48]Free Software Foundation, Inc. Verbatim copying and 8006 distribution of this entire article is permitted in any medium, 8007 provided this notice is preserved. 8008 8009 These pages are [49]maintained by the GCC team. Last modified 8010 2024-06-08. 8011 8012 References 8013 8014 1. https://gcc.gnu.org/gcc-5/changes.html#libstdcxx 8015 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 8016 3. https://savannah.nongnu.org/bugs/?44574 8017 4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler 8018 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 8019 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 8020 7. https://gcc.gnu.org/wiki/OpenACC 8021 8. https://gcc.gnu.org/install/configure.html 8022 9. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 8023 10. https://gcc.gnu.org/projects/cxx1y.html 8024 11. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 8025 12. https://gcc.gnu.org/projects/cxx1y.html 8026 13. https://gcc.gnu.org/projects/cxx1y.html 8027 14. https://gcc.gnu.org/projects/cxx1y.html 8028 15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 8029 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 8030 17. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 8031 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 8032 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 8033 20. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python 8034 21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 8035 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 8036 23. https://gcc.gnu.org/install/configure.html 8037 24. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 8038 25. https://gcc.gnu.org/wiki/Fortran2003Status 8039 26. https://gcc.gnu.org/wiki/Fortran2008Status 8040 27. https://gcc.gnu.org/wiki/Coarray 8041 28. http://www.opencoarrays.org/ 8042 29. https://pkg.go.dev/cmd/go 8043 30. https://pkg.go.dev/cmd/gofmt 8044 31. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 8045 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 8046 33. https://github.com/davidmalcolm/pygccjit 8047 34. https://github.com/ibuclaw/gccjitd 8048 35. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 8049 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 8050 37. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 8051 38. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 8052 39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 8053 40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 8054 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 8055 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 8056 43. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html 8057 44. https://gcc.gnu.org/onlinedocs/ 8058 45. mailto:gcc-help (a] gcc.gnu.org 8059 46. mailto:gcc (a] gcc.gnu.org 8060 47. https://gcc.gnu.org/lists.html 8061 48. https://www.fsf.org/ 8062 49. https://gcc.gnu.org/about.html 8063 ====================================================================== 8064 http://gcc.gnu.org/gcc-4.9/index.html 8065 GCC 4.9 Release Series 8066 8067 (This release series is no longer supported.) 8068 8069 Aug 3, 2016 8070 8071 The [1]GNU project and the GCC developers are pleased to announce the 8072 release of GCC 4.9.4. 8073 8074 This release is a bug-fix release, containing fixes for regressions in 8075 GCC 4.9.3 relative to previous releases of GCC. 8076 8077 Release History 8078 8079 GCC 4.9.4 8080 Aug 3, 2016 ([2]changes, [3]documentation) 8081 8082 GCC 4.9.3 8083 June 26, 2015 ([4]changes, [5]documentation) 8084 8085 GCC 4.9.2 8086 October 30, 2014 ([6]changes, [7]documentation) 8087 8088 GCC 4.9.1 8089 July 16, 2014 ([8]changes, [9]documentation) 8090 8091 GCC 4.9.0 8092 April 22, 2014 ([10]changes, [11]documentation) 8093 8094 References and Acknowledgements 8095 8096 GCC used to stand for the GNU C Compiler, but since the compiler 8097 supports several other languages aside from C, it now stands for the 8098 GNU Compiler Collection. 8099 8100 A list of [12]successful builds is updated as new information becomes 8101 available. 8102 8103 The GCC developers would like to thank the numerous people that have 8104 contributed new features, improvements, bug fixes, and other changes as 8105 well as test results to GCC. This [13]amazing group of volunteers is 8106 what makes GCC successful. 8107 8108 For additional information about GCC please refer to the [14]GCC 8109 project web site or contact the [15]GCC development mailing list. 8110 8111 To obtain GCC please use [16]our mirror sites or [17]our version 8112 control system. 8113 8114 8115 For questions related to the use of GCC, please consult these web 8116 pages and the [18]GCC manuals. If that fails, the 8117 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8118 web pages and the development of GCC are welcome on our developer 8119 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 8120 archives. 8121 8122 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8123 distribution of this entire article is permitted in any medium, 8124 provided this notice is preserved. 8125 8126 These pages are [23]maintained by the GCC team. Last modified 8127 2024-05-30. 8128 8129 References 8130 8131 1. http://www.gnu.org/ 8132 2. https://gcc.gnu.org/gcc-4.9/changes.html 8133 3. https://gcc.gnu.org/onlinedocs/4.9.4/ 8134 4. https://gcc.gnu.org/gcc-4.9/changes.html 8135 5. https://gcc.gnu.org/onlinedocs/4.9.3/ 8136 6. https://gcc.gnu.org/gcc-4.9/changes.html 8137 7. https://gcc.gnu.org/onlinedocs/4.9.2/ 8138 8. https://gcc.gnu.org/gcc-4.9/changes.html 8139 9. https://gcc.gnu.org/onlinedocs/4.9.1/ 8140 10. https://gcc.gnu.org/gcc-4.9/changes.html 8141 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 8142 12. https://gcc.gnu.org/gcc-4.9/buildstat.html 8143 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8144 14. https://gcc.gnu.org/index.html 8145 15. mailto:gcc (a] gcc.gnu.org 8146 16. https://gcc.gnu.org/mirrors.html 8147 17. https://gcc.gnu.org/git.html 8148 18. https://gcc.gnu.org/onlinedocs/ 8149 19. mailto:gcc-help (a] gcc.gnu.org 8150 20. mailto:gcc (a] gcc.gnu.org 8151 21. https://gcc.gnu.org/lists.html 8152 22. https://www.fsf.org/ 8153 23. https://gcc.gnu.org/about.html 8154 ====================================================================== 8155 http://gcc.gnu.org/gcc-4.9/changes.html 8156 GCC 4.9 Release Series 8157 Changes, New Features, and Fixes 8158 8159 Caveats 8160 8161 * The mudflap run time checker has been removed. The mudflap options 8162 remain, but do nothing. 8163 * Support for a number of older systems and recently unmaintained or 8164 untested target ports of GCC has been declared obsolete in GCC 4.9. 8165 Unless there is activity to revive them, the next release of GCC 8166 will have their sources permanently removed. 8167 The following ports for individual systems on particular 8168 architectures have been obsoleted: 8169 + Solaris 9 (*-*-solaris2.9). Details can be found in the 8170 [1]announcement. 8171 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 8172 float64x1_t exported by arm_neon.h are defined to be the same as 8173 their base types. This results in incorrect application of 8174 parameter passing rules to arguments of types int64x1_t and 8175 uint64x1_t, with respect to the AAPCS64 ABI specification. In 8176 addition, names of C++ functions with parameters of these types 8177 (including float64x1_t) are not mangled correctly. The current 8178 typedef declarations also unintentionally allow implicit casting 8179 between singleton vector types and their base types. These issues 8180 will be resolved in a near future release. See [2]PR60825 for more 8181 information. 8182 8183 More information on porting to GCC 4.9 from previous versions of GCC 8184 can be found in the [3]porting guide for this release. 8185 8186 General Optimizer Improvements 8187 8188 * AddressSanitizer, a fast memory error detector, is now available on 8189 ARM. 8190 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 8191 detector, has been added and can be enabled via 8192 -fsanitize=undefined. Various computations will be instrumented to 8193 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 8194 currently available for the C and C++ languages. 8195 * Link-time optimization (LTO) improvements: 8196 + Type merging was rewritten. The new implementation is 8197 significantly faster and uses less memory. 8198 + Better partitioning algorithm resulting in less streaming 8199 during link time. 8200 + Early removal of virtual methods reduces the size of object 8201 files and improves link-time memory usage and compile time. 8202 + Function bodies are now loaded on-demand and released early 8203 improving overall memory usage at link time. 8204 + C++ hidden keyed methods can now be optimized out. 8205 + When using a linker plugin, compiling with the -flto option 8206 now generates slim object files (.o) which only contain 8207 intermediate language representation for LTO. Use 8208 -ffat-lto-objects to create files which contain additionally 8209 the object code. To generate static libraries suitable for LTO 8210 processing, use gcc-ar and gcc-ranlib; to list symbols from a 8211 slim object file use gcc-nm. (This requires that ar, ranlib 8212 and nm have been compiled with plugin support.) 8213 Memory usage building Firefox with debug enabled was reduced from 8214 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 8215 * Inter-procedural optimization improvements: 8216 + New type inheritance analysis module improving 8217 devirtualization. Devirtualization now takes into account 8218 anonymous name-spaces and the C++11 final keyword. 8219 + New speculative devirtualization pass (controlled by 8220 -fdevirtualize-speculatively. 8221 + Calls that were speculatively made direct are turned back to 8222 indirect where direct call is not cheaper. 8223 + Local aliases are introduced for symbols that are known to be 8224 semantically equivalent across shared libraries improving 8225 dynamic linking times. 8226 * Feedback directed optimization improvements: 8227 + Profiling of programs using C++ inline functions is now more 8228 reliable. 8229 + New time profiling determines typical order in which functions 8230 are executed. 8231 + A new function reordering pass (controlled by 8232 -freorder-functions) significantly reduces startup time of 8233 large applications. Until binutils support is completed, it is 8234 effective only with link-time optimization. 8235 + Feedback driven indirect call removal and devirtualization now 8236 handle cross-module calls when link-time optimization is 8237 enabled. 8238 8239 New Languages and Language specific improvements 8240 8241 * Version 4.0 of the [4]OpenMP specification is now supported in the 8242 C and C++ compilers and starting with the 4.9.1 release also in the 8243 Fortran compiler. The new -fopenmp-simd option can be used to 8244 enable OpenMP's SIMD directives while ignoring other OpenMP 8245 directives. The new [5]-fsimd-cost-model= option permits to tune 8246 the vectorization cost model for loops annotated with OpenMP and 8247 Cilk Plus simd directives. -Wopenmp-simd warns when the current 8248 cost model overrides simd directives set by the user. 8249 * The -Wdate-time option has been added for the C, C++ and Fortran 8250 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 8251 macros are used. Those macros might prevent bit-wise-identical 8252 reproducible compilations. 8253 8254 Ada 8255 8256 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 8257 8258 C family 8259 8260 * Support for colorizing diagnostics emitted by GCC has been added. 8261 The [6]-fdiagnostics-color=auto will enable it when outputting to 8262 terminals, -fdiagnostics-color=always unconditionally. The 8263 GCC_COLORS environment variable can be used to customize the colors 8264 or disable coloring. If GCC_COLORS variable is present in the 8265 environment, the default is -fdiagnostics-color=auto, otherwise 8266 -fdiagnostics-color=never. 8267 Sample diagnostics output: 8268 $ g++ -fdiagnostics-color=always -S -Wall test.C 8269 test.C: In function int foo(): 8270 test.C:1:14: warning: no return statement in function returning non-void [-W 8271 return-type] 8272 int foo () { } 8273 ^ 8274 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 8275 -ftemplate-depth= to increase the maximum) instantiating struct X<100> 8276 template <int N> struct X { static const int value = X<N-1>::value; }; temp 8277 late struct X<1000>; 8278 ^ 8279 test.C:2:46: recursively required from const int X<999>::value 8280 test.C:2:46: required from const int X<1000>::value 8281 test.C:2:88: required from here 8282 8283 test.C:2:46: error: incomplete type X<100> used in nested name specifier 8284 8285 * With the new [7]#pragma GCC ivdep, the user can assert that there 8286 are no loop-carried dependencies which would prevent concurrent 8287 execution of consecutive iterations using SIMD (single instruction 8288 multiple data) instructions. 8289 * Support for Cilk Plus has been added and can be enabled with the 8290 -fcilkplus option. Cilk Plus is an extension to the C and C++ 8291 languages to support data and task parallelism. The present 8292 implementation follows ABI version 1.2; all features but _Cilk_for 8293 have been implemented. 8294 8295 C 8296 8297 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 8298 <stdatomic.h> header) are now supported. 8299 * ISO C11 generic selections (_Generic keyword) are now supported. 8300 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 8301 __thread) is now supported. 8302 * ISO C11 support is now at a similar level of completeness to ISO 8303 C99 support: substantially complete modulo bugs, extended 8304 identifiers (supported except for corner cases when 8305 -fextended-identifiers is used), floating-point issues (mainly but 8306 not entirely relating to optional C99 features from Annexes F and 8307 G) and the optional Annexes K (Bounds-checking interfaces) and L 8308 (Analyzability). 8309 * A new C extension __auto_type provides a subset of the 8310 functionality of C++11 auto in GNU C. 8311 8312 C++ 8313 8314 * The G++ implementation of [8]C++1y return type deduction for normal 8315 functions has been updated to conform to [9]N3638, the proposal 8316 accepted into the working paper. Most notably, it adds 8317 decltype(auto) for getting decltype semantics rather than the 8318 template argument deduction semantics of plain auto: 8319 8320 int& f(); 8321 auto i1 = f(); // int 8322 decltype(auto) i2 = f(); // int& 8323 8324 * G++ supports [10]C++1y lambda capture initializers: 8325 8326 [x = 42]{ ... }; 8327 8328 Actually, they have been accepted since GCC 4.5, but now the 8329 compiler doesn't warn about them with -std=c++1y, and supports 8330 parenthesized and brace-enclosed initializers as well. 8331 * G++ supports [11]C++1y variable length arrays. G++ has supported 8332 GNU/C99-style VLAs for a long time, but now additionally supports 8333 initializers and lambda capture by reference. In C++1y mode G++ 8334 will complain about VLA uses that are not permitted by the draft 8335 standard, such as forming a pointer to VLA type or applying sizeof 8336 to a VLA variable. Note that it now appears that VLAs will not be 8337 part of C++14, but will be part of a separate document and then 8338 perhaps C++17. 8339 8340 void f(int n) { 8341 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 8342 [&a]{ for (int i : a) { cout << i << endl; } }(); 8343 &a; // error, taking address of VLA 8344 } 8345 8346 * G++ supports the [12]C++1y [[deprecated]] attribute modulo bugs in 8347 the underlying [[gnu::deprecated]] attribute. Classes and functions 8348 can be marked deprecated and a diagnostic message added: 8349 8350 class A; 8351 int bar(int n); 8352 #if __cplusplus > 201103 8353 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 8354 [[deprecated("bar is unsafe; use foo() instead")]] 8355 int bar(int n); 8356 8357 int foo(int n); 8358 class B; 8359 #endif 8360 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 8361 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 8362 o() instead 8363 8364 * G++ supports [13]C++1y digit separators. Long numeric literals can 8365 be subdivided with a single quote ' to enhance readability: 8366 8367 int i = 1048576; 8368 int j = 1'048'576; 8369 int k = 0x10'0000; 8370 int m = 0'004'000'000; 8371 int n = 0b0001'0000'0000'0000'0000'0000; 8372 8373 double x = 1.602'176'565e-19; 8374 double y = 1.602'176'565e-1'9; 8375 8376 * G++ supports [14]C++1y generic (polymorphic) lambdas. 8377 8378 // a functional object that will increment any type 8379 auto incr = [](auto x) { return x++; }; 8380 8381 * As a GNU extension, G++ supports explicit template parameter syntax 8382 for generic lambdas. This can be combined in the expected way with 8383 the standard auto syntax. 8384 8385 // a functional object that will add two like-type objects 8386 auto add = [] <typename T> (T a, T b) { return a + b; }; 8387 8388 * G++ supports unconstrained generic functions as specified by 4.1.2 8389 and 5.1.1 of [15]N3889: Concepts Lite Specification. Briefly, auto 8390 may be used as a type-specifier in a parameter declaration of any 8391 function declarator in order to introduce an implicit function 8392 template parameter, akin to generic lambdas. 8393 8394 // the following two function declarations are equivalent 8395 auto incr(auto x) { return x++; } 8396 template <typename T> 8397 auto incr(T x) { return x++; } 8398 8399 Runtime Library (libstdc++) 8400 8401 * [16]Improved support for C++11, including: 8402 + support for <regex>; 8403 + The associative containers in <map> and <set> and the 8404 unordered associative containers in <unordered_map> and 8405 <unordered_set> meet the allocator-aware container 8406 requirements; 8407 * [17]Improved experimental support for the upcoming ISO C++ 8408 standard, C++14, including: 8409 + fixing constexpr member functions without const; 8410 + implementation of the std::exchange() utility function; 8411 + addressing tuples by type; 8412 + implemention of std::make_unique; 8413 + implemention of std::shared_lock; 8414 + making std::result_of SFINAE-friendly; 8415 + adding operator() to std::integral_constant; 8416 + adding user-defined literals for standard library types 8417 std::basic_string, std::chrono::duration, and std::complex; 8418 + adding two range overloads to non-modifying sequence oprations 8419 std::equal and std::mismatch; 8420 + adding IO manipulators for quoted strings; 8421 + adding constexpr members to <utility>, <complex>, <chrono>, 8422 and some containers; 8423 + adding compile-time std::integer_sequence; 8424 + adding cleaner transformation traits; 8425 + making <functional>s operator functors easier to use and more 8426 generic; 8427 * An implementation of std::experimental::optional. 8428 * An implementation of std::experimental::string_view. 8429 * The non-standard function std::copy_exception has been deprecated 8430 and will be removed in a future version. std::make_exception_ptr 8431 should be used instead. 8432 8433 Fortran 8434 8435 * Compatibility notice: 8436 + Module files: The version of the module files (.mod) has been 8437 incremented; additionally, module files are now compressed. 8438 Fortran MODULEs compiled by earlier GCC versions have to be 8439 recompiled, when they are USEd by files compiled with GCC 4.9. 8440 GCC 4.9 is not able to read .mod files of earlier GCC 8441 versions; attempting to do so gives an error message. Note: 8442 The ABI of the produced assembler data itself has not changed: 8443 object files and libraries are fully compatible with older 8444 versions (except as stated below). 8445 + ABI changes: 8446 o The [18]argument passing ABI has changed for scalar dummy 8447 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 8448 which have both the VALUE and the OPTIONAL attributes. 8449 o To support finalization the virtual table associated with 8450 polymorphic variables has changed. Code containing CLASS 8451 should be recompiled, including all files which define 8452 derived types involved in the type definition used by 8453 polymorphic variables. (Note: Due to the incremented 8454 module version, trying to mix old code with new code will 8455 usually give an error message.) 8456 + GNU Fortran no longer deallocates allocatable variables or 8457 allocatable components of variables declared in the main 8458 program. Since Fortran 2008, the standard explicitly states 8459 that variables declared in the Fortran main program 8460 automatically have the SAVE attribute. 8461 + When opening files, the close-on-exec flag is set if the 8462 system supports such a feature. This is generally considered 8463 good practice these days, but if there is a need to pass file 8464 descriptors to child processes the parent process must now 8465 remember to clear the close-on-exec flag by calling fcntl(), 8466 e.g. via ISO_C_BINDING, before executing the child process. 8467 * The deprecated command-line option -fno-whole-file has been 8468 removed. (-fwhole-file is the default since GCC 4.6.) 8469 -fwhole-file/-fno-whole-file continue to be accepted but do not 8470 influence the code generation. 8471 * The compiler no longer unconditionally warns about DO loops with 8472 zero iterations. This warning is now controlled by the -Wzerotrip 8473 option, which is implied by -Wall. 8474 * The new NO_ARG_CHECK attribute of the [19]!GCC$ directive can be 8475 used to disable the type-kind-rank (TKR) argument check for a dummy 8476 argument. The feature is similar to ISO/IEC TS 29133:2012's 8477 TYPE(*), except that it additionally also disables the rank check. 8478 Variables with NO_ARG_CHECK have to be dummy arguments and may only 8479 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 8480 to another NO_ARG_CHECK dummy argument; also the other constraints 8481 of TYPE(*) apply. The dummy arguments should be declared as scalar 8482 or assumed-size variable of type type(*) (recommended) or of type 8483 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 8484 the data without further type or shape information is passed, 8485 similar to C's void*. Note that also TS 29113's 8486 type(*),dimension(..) accepts arguments of any type and rank; 8487 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 8488 descriptor which contains the array shape and stride of the 8489 argument. 8490 * [20]Fortran 2003: 8491 + Finalization is now supported. It is currently only done for a 8492 subset of those situations in which it should occur. 8493 + Experimental support for scalar character components with 8494 deferred length (i.e. allocatable string length) in derived 8495 types has been added. (Deferred-length character variables are 8496 supported since GCC 4.6.) 8497 * [21]Fortran 2008: 8498 + When STOP or ERROR STOP are used to terminate the execution 8499 and any exception (but inexact) is signaling, a warning is 8500 printed to ERROR_UNIT, indicating which exceptions are 8501 signaling. The [22]-ffpe-summary= command-line option can be 8502 used to fine-tune for which exceptions the warning should be 8503 shown. 8504 + Rounding on input (READ) is now handled on systems where 8505 strtod honours the rounding mode. (For output, rounding is 8506 supported since GCC 4.5.) Note that for input, the compatible 8507 rounding mode is handled as nearest (i.e., rounding to an even 8508 least significant [cf. IEC 60559:1989] for a tie, while 8509 compatible rounds away from zero in that case). 8510 8511 Go 8512 8513 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 8514 8515 New Targets and Target Specific Improvements 8516 8517 AArch64 8518 8519 * The ARMv8-A crypto and CRC instructions are now supported through 8520 intrinsics. These are enabled when the architecture supports these 8521 and are available through the -march=armv8-a+crc and 8522 -march=armv8-a+crypto options. 8523 * Initial support for ILP32 has now been added to the compiler. This 8524 is now available through the command-line option -mabi=ilp32. 8525 Support for ILP32 is considered experimental as the ABI 8526 specification is still beta. 8527 * Coverage of more of the ISA including the SIMD extensions has been 8528 added. The Advanced SIMD intrinsics have also been improved. 8529 * The new local register allocator (LRA) is now on by default for the 8530 AArch64 back end. 8531 * The REE (Redundant extension elimination) pass has now been enabled 8532 by default for the AArch64 back end. 8533 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 8534 * Initial big.LITTLE tuning support for the combination of Cortex-A57 8535 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 8536 option. 8537 * A number of structural changes have been made to both the ARM and 8538 AArch64 back ends to facilitate improved code-generation. 8539 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 8540 has been added and can be enabled by giving the 8541 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 8542 default by configuring GCC with the --enable-fix-cortex-a53-835769 8543 option. 8544 8545 ARC 8546 8547 * A port for Synopsys Designware ARC has been contributed by Embecosm 8548 and Synopsys Inc. 8549 8550 ARM 8551 8552 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 8553 disabled by default. This was found to generate better code in only 8554 a small number of cases. It can be turned back on with the 8555 -mneon-for-64bits option. 8556 * Further support for the ARMv8-A architecture, notably implementing 8557 the restriction around IT blocks in the Thumb32 instruction set has 8558 been added. The -mrestrict-it option can be used with 8559 -march=armv7-a or the -march=armv7ve options to make code 8560 generation fully compatible with the deprecated instructions in 8561 ARMv8-A. 8562 * Support has now been added for the ARMv7ve variant of the 8563 architecture. This can be used by the -march=armv7ve option. 8564 * The ARMv8-A crypto and CRC instructions are now supported through 8565 intrinsics and are available through the -march=armv8-a+crc and 8566 mfpu=crypto-neon-fp-armv8 options. 8567 * LRA is now on by default for the ARM target. This can be turned off 8568 using the -mno-lra option. This option is a purely transitionary 8569 command-line option and will be removed in a future release. We are 8570 interested in any bug reports regarding functional and performance 8571 regressions with LRA. 8572 * A new option -mslow-flash-data to improve performance of programs 8573 fetching data on slow flash memory has now been introduced for the 8574 ARMv7-M profile cores. 8575 * A new option -mpic-data-is-text-relative for targets that allows 8576 data segments to be relative to text segments has been added. This 8577 is on by default for all targets except VxWorks RTP. 8578 * A number of infrastructural changes have been made to both the ARM 8579 and AArch64 back ends to facilitate improved code-generation. 8580 * GCC now supports Cortex-A12 and the Cortex-R7 through the 8581 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 8582 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 8583 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 8584 * Initial big.LITTLE tuning support for the combination of Cortex-A57 8585 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 8586 option. Similar support was added for the combination of Cortex-A15 8587 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 8588 * Further performance optimizations for the Cortex-A15 and the 8589 Cortex-M4 have been added. 8590 * A number of code generation improvements for Thumb2 to reduce code 8591 size when compiling for the M-profile processors. 8592 8593 AVR 8594 8595 * A new command-line option -mfract-convert-truncate has been added. 8596 It allows compiler to use truncation instead of rounding towards 8597 zero for fractional fixed-point types. 8598 8599 IA-32/x86-64 8600 8601 * -mfpmath=sse is now implied by -ffast-math on all targets where 8602 SSE2 is supported. 8603 * Intel AVX-512 support was added to GCC. That includes inline 8604 assembly support, new registers and extending existing ones, new 8605 intrinsics (covered by corresponding testsuite), and basic 8606 autovectorization. AVX-512 instructions are available via the 8607 following GCC switches: AVX-512 foundation instructions: -mavx512f, 8608 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 8609 reciprocal instructions: -mavx512er, AVX-512 conflict detection 8610 instructions: -mavx512cd. 8611 * It is now possible to call x86 intrinsics from select functions in 8612 a file that are tagged with the corresponding target attribute 8613 without having to compile the entire file with the -mxxx option. 8614 This improves the usability of x86 intrinsics and is particularly 8615 useful when doing [23]Function Multiversioning. 8616 * GCC now supports the new Intel microarchitecture named Silvermont 8617 through -march=silvermont. 8618 * GCC now supports the new Intel microarchitecture named Broadwell 8619 through -march=broadwell. 8620 * Optimizing for other Intel microarchitectures have been renamed to 8621 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 8622 * -march=generic has been retuned for better support of Intel core 8623 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 8624 Pentium-M, and Pentium4 based CPUs is no longer considered 8625 important for generic. 8626 * -mtune=intel can now be used to generate code running well on the 8627 most current Intel processors, which are Haswell and Silvermont for 8628 GCC 4.9. 8629 * Support to encode 32-bit assembly instructions in 16-bit format is 8630 now available through the -m16 command-line option. 8631 * Better inlining of memcpy and memset that is aware of value ranges 8632 and produces shorter alignment prologues. 8633 * -mno-accumulate-outgoing-args is now honored when unwind 8634 information is output. Argument accumulation is also now turned off 8635 for portions of programs optimized for size. 8636 * Support for new AMD family 15h processors (Excavator core) is now 8637 available through the -march=bdver4 and -mtune=bdver4 options. 8638 8639 MSP430 8640 8641 * A new command-line option -mcpu= has been added to the MSP430 back 8642 end. This option is used to specify the ISA to be used. Accepted 8643 values are msp430 (the default), msp430x and msp430xv2. The ISA is 8644 no longer deduced from the -mmcu= option as there are far too many 8645 different MCU names. The -mmcu= option is still supported, and this 8646 is still used to select linker scripts and generate a C 8647 preprocessor symbol that will be recognised by the msp430.h header 8648 file. 8649 8650 NDS32 8651 8652 * A new nds32 port supports the 32-bit architecture from Andes 8653 Technology Corporation. 8654 * The port provides initial support for the V2, V3, V3m instruction 8655 set architectures. 8656 8657 Nios II 8658 8659 * A port for the Altera Nios II has been contributed by Mentor 8660 Graphics. 8661 8662 PowerPC / PowerPC64 / RS6000 8663 8664 * GCC now supports Power ISA 2.07, which includes support for 8665 Hardware Transactional Memory (HTM), Quadword atomics and several 8666 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 8667 integer and decimal integer operations. 8668 * Support for the POWER8 processor is now available through the 8669 -mcpu=power8 and -mtune=power8 options. 8670 * The libitm library has been modified to add a HTM fastpath that 8671 automatically uses POWER's HTM hardware instructions when it is 8672 executing on a HTM enabled processor. 8673 * Support for the new powerpc64le-linux platform has been added. It 8674 defaults to generating code that conforms to the ELFV2 ABI. 8675 8676 S/390, System z 8677 8678 * Support for the Transactional Execution Facility included with the 8679 IBM zEnterprise zEC12 processor has been added. A set of GCC style 8680 builtins as well as XLC style builtins are provided. The builtins 8681 are enabled by default when using the -march=zEC12 option but can 8682 explicitly be disabled with -mno-htm. Using the GCC builtins also 8683 libitm supports hardware transactions on S/390. 8684 * The hotpatch features allows to prepare functions for hotpatching. 8685 A certain amount of bytes is reserved before the function entry 8686 label plus a NOP is inserted at its very beginning to implement a 8687 backward jump when applying a patch. The feature can either be 8688 enabled per compilation unit via the command-line option -mhotpatch 8689 or per function using the hotpatch attribute. 8690 * The shrink wrap optimization is now supported on S/390 and enabled 8691 by default. 8692 * A major rework of the routines to determine which registers need to 8693 be saved and restored in function prologue/epilogue now allow to 8694 use floating point registers as save slots. This will happen for 8695 certain leaf function with -march=z10 or higher. 8696 * The LRA rtl pass replaces reload by default on S/390. 8697 8698 RX 8699 8700 * The port now allows to specify the RX100, RX200, and RX600 8701 processors with the command-line options -mcpu=rx100, -mcpu=rx200 8702 and -mcpu=rx600. 8703 8704 SH 8705 8706 * Minor improvements to code generated for integer arithmetic and 8707 code that involves the T bit. 8708 * Added support for the SH2A clips and clipu instructions. The 8709 compiler will now try to utilize them for min/max expressions such 8710 as max (-128, min (127, x)). 8711 * Added support for the cmp/str instruction through built-in 8712 functions such as __builtin_strlen. When not optimizing for size, 8713 the compiler will now expand calls to e.g. strlen as an inlined 8714 sequences which utilize the cmp/str instruction. 8715 * Improved code generated around volatile memory loads and stores. 8716 * The option -mcbranchdi has been deprecated. Specifying it will 8717 result in a warning and will not influence code generation. 8718 * The option -mcmpeqdi has been deprecated. Specifying it will result 8719 in a warning and will not influence code generation. 8720 8721 GCC 4.9.1 8722 8723 This is the [24]list of problem reports (PRs) from GCC's bug tracking 8724 system that are known to be fixed in the 4.9.1 release. This list might 8725 not be complete (that is, it is possible that some PRs that have been 8726 fixed are not listed here). 8727 8728 Version 4.0 of the OpenMP specification is supported even in Fortran, 8729 not just C and C++. 8730 8731 GCC 4.9.2 8732 8733 This is the [25]list of problem reports (PRs) from GCC's bug tracking 8734 system that are known to be fixed in the 4.9.2 release. This list might 8735 not be complete (that is, it is possible that some PRs that have been 8736 fixed are not listed here). 8737 8738 GCC 4.9.3 8739 8740 This is the [26]list of problem reports (PRs) from GCC's bug tracking 8741 system that are known to be fixed in the 4.9.3 release. This list might 8742 not be complete (that is, it is possible that some PRs that have been 8743 fixed are not listed here). 8744 8745 GCC 4.9.4 8746 8747 This is the [27]list of problem reports (PRs) from GCC's bug tracking 8748 system that are known to be fixed in the 4.9.4 release. This list might 8749 not be complete (that is, it is possible that some PRs that have been 8750 fixed are not listed here). 8751 8752 8753 For questions related to the use of GCC, please consult these web 8754 pages and the [28]GCC manuals. If that fails, the 8755 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8756 web pages and the development of GCC are welcome on our developer 8757 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 8758 archives. 8759 8760 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 8761 distribution of this entire article is permitted in any medium, 8762 provided this notice is preserved. 8763 8764 These pages are [33]maintained by the GCC team. Last modified 8765 2023-09-02. 8766 8767 References 8768 8769 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 8770 2. https://gcc.gnu.org/PR60825 8771 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 8772 4. https://www.openmp.org/specifications/ 8773 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 8774 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 8775 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 8776 8. https://gcc.gnu.org/projects/cxx1y.html 8777 9. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 8778 10. https://gcc.gnu.org/projects/cxx1y.html 8779 11. https://gcc.gnu.org/projects/cxx1y.html 8780 12. https://gcc.gnu.org/projects/cxx1y.html 8781 13. https://gcc.gnu.org/projects/cxx1y.html 8782 14. https://gcc.gnu.org/projects/cxx1y.html 8783 15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 8784 16. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 8785 17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 8786 18. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 8787 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 8788 20. https://gcc.gnu.org/wiki/Fortran2003Status 8789 21. https://gcc.gnu.org/wiki/Fortran2008Status 8790 22. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 8791 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 8792 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 8793 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 8794 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 8795 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 8796 28. https://gcc.gnu.org/onlinedocs/ 8797 29. mailto:gcc-help (a] gcc.gnu.org 8798 30. mailto:gcc (a] gcc.gnu.org 8799 31. https://gcc.gnu.org/lists.html 8800 32. https://www.fsf.org/ 8801 33. https://gcc.gnu.org/about.html 8802 ====================================================================== 8803 http://gcc.gnu.org/gcc-4.8/index.html 8804 GCC 4.8 Release Series 8805 8806 (This release series is no longer supported.) 8807 8808 June 23, 2015 8809 8810 The [1]GNU project and the GCC developers are pleased to announce the 8811 release of GCC 4.8.5. 8812 8813 This release is a bug-fix release, containing fixes for regressions in 8814 GCC 4.8.4 relative to previous releases of GCC. 8815 8816 Release History 8817 8818 GCC 4.8.5 8819 June 23, 2015 ([2]changes, [3]documentation) 8820 8821 GCC 4.8.4 8822 December 19, 2014 ([4]changes, [5]documentation) 8823 8824 GCC 4.8.3 8825 May 22, 2014 ([6]changes, [7]documentation) 8826 8827 GCC 4.8.2 8828 October 16, 2013 ([8]changes, [9]documentation) 8829 8830 GCC 4.8.1 8831 May 31, 2013 ([10]changes, [11]documentation) 8832 8833 GCC 4.8.0 8834 March 22, 2013 ([12]changes, [13]documentation) 8835 8836 References and Acknowledgements 8837 8838 GCC used to stand for the GNU C Compiler, but since the compiler 8839 supports several other languages aside from C, it now stands for the 8840 GNU Compiler Collection. 8841 8842 A list of [14]successful builds is updated as new information becomes 8843 available. 8844 8845 The GCC developers would like to thank the numerous people that have 8846 contributed new features, improvements, bug fixes, and other changes as 8847 well as test results to GCC. This [15]amazing group of volunteers is 8848 what makes GCC successful. 8849 8850 For additional information about GCC please refer to the [16]GCC 8851 project web site or contact the [17]GCC development mailing list. 8852 8853 To obtain GCC please use [18]our mirror sites or [19]our version 8854 control system. 8855 8856 8857 For questions related to the use of GCC, please consult these web 8858 pages and the [20]GCC manuals. If that fails, the 8859 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8860 web pages and the development of GCC are welcome on our developer 8861 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 8862 archives. 8863 8864 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 8865 distribution of this entire article is permitted in any medium, 8866 provided this notice is preserved. 8867 8868 These pages are [25]maintained by the GCC team. Last modified 8869 2022-10-26. 8870 8871 References 8872 8873 1. http://www.gnu.org/ 8874 2. https://gcc.gnu.org/gcc-4.8/changes.html 8875 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 8876 4. https://gcc.gnu.org/gcc-4.8/changes.html 8877 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 8878 6. https://gcc.gnu.org/gcc-4.8/changes.html 8879 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 8880 8. https://gcc.gnu.org/gcc-4.8/changes.html 8881 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 8882 10. https://gcc.gnu.org/gcc-4.8/changes.html 8883 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 8884 12. https://gcc.gnu.org/gcc-4.8/changes.html 8885 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 8886 14. https://gcc.gnu.org/gcc-4.8/buildstat.html 8887 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8888 16. https://gcc.gnu.org/index.html 8889 17. mailto:gcc (a] gcc.gnu.org 8890 18. https://gcc.gnu.org/mirrors.html 8891 19. https://gcc.gnu.org/git.html 8892 20. https://gcc.gnu.org/onlinedocs/ 8893 21. mailto:gcc-help (a] gcc.gnu.org 8894 22. mailto:gcc (a] gcc.gnu.org 8895 23. https://gcc.gnu.org/lists.html 8896 24. https://www.fsf.org/ 8897 25. https://gcc.gnu.org/about.html 8898 ====================================================================== 8899 http://gcc.gnu.org/gcc-4.8/changes.html 8900 GCC 4.8 Release Series 8901 Changes, New Features, and Fixes 8902 8903 Caveats 8904 8905 GCC now uses C++ as its implementation language. This means that to 8906 build GCC from sources, you will need a C++ compiler that understands 8907 C++ 2003. For more details on the rationale and specific changes, 8908 please refer to the [1]C++ conversion page. 8909 8910 To enable the Graphite framework for loop optimizations you now need 8911 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 8912 the [2]GCC infrastructure directory. The installation manual contains 8913 more information about requirements to build GCC. 8914 8915 GCC now uses a more aggressive analysis to derive an upper bound for 8916 the number of iterations of loops using constraints imposed by language 8917 standards. This may cause non-conforming programs to no longer work as 8918 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 8919 option, -fno-aggressive-loop-optimizations, was added to disable this 8920 aggressive analysis. In some loops that have known constant number of 8921 iterations, but undefined behavior is known to occur in the loop before 8922 reaching or during the last iteration, GCC will warn about the 8923 undefined behavior in the loop instead of deriving lower upper bound of 8924 the number of iterations for the loop. The warning can be disabled with 8925 -Wno-aggressive-loop-optimizations. 8926 8927 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 8928 for the layout of vectors that could lead to wrong code being 8929 generated. Vectors larger than 8 bytes in size are now by default 8930 aligned to an 8-byte boundary. This is an ABI change: code that makes 8931 explicit use of vector types may be incompatible with binary objects 8932 built with older versions of GCC. Auto-vectorized code is not affected 8933 by this change. 8934 8935 On AVR, support has been removed for the command-line option 8936 -mshort-calls deprecated in GCC 4.7. 8937 8938 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 8939 is turned on per default for all non-RTEMS configurations. This option 8940 arranges for a better integration of [3]AVR Libc with avr-gcc. For 8941 technical details, see [4]PR54461. To turn off the option in non-RTEMS 8942 configurations, use --with-avrlibc=no. If the compiler is configured 8943 for RTEMS, the option is always turned off. 8944 8945 More information on porting to GCC 4.8 from previous versions of GCC 8946 can be found in the [5]porting guide for this release. 8947 8948 General Optimizer Improvements (and Changes) 8949 8950 * DWARF4 is now the default when generating DWARF debug information. 8951 When -g is used on a platform that uses DWARF debugging 8952 information, GCC will now default to -gdwarf-4 8953 -fno-debug-types-section. 8954 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 8955 consumers support DWARF4 by default. Before GCC 4.8 the default 8956 version used was DWARF2. To make GCC 4.8 generate an older DWARF 8957 version use -g together with -gdwarf-2 or -gdwarf-3. The default 8958 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 8959 * A new general optimization level, -Og, has been introduced. It 8960 addresses the need for fast compilation and a superior debugging 8961 experience while providing a reasonable level of run-time 8962 performance. Overall experience for development should be better 8963 than the default optimization level -O0. 8964 * A new option -ftree-partial-pre was added to control the partial 8965 redundancy elimination (PRE) optimization. This option is enabled 8966 by default at the -O3 optimization level, and it makes PRE more 8967 aggressive. 8968 * The option -fconserve-space has been removed; it was no longer 8969 useful on most targets since GCC supports putting variables into 8970 BSS without making them common. 8971 * The struct reorg and matrix reorg optimizations (command-line 8972 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 8973 removed. They did not always work correctly, nor did they work with 8974 link-time optimization (LTO), hence were only applicable to 8975 programs consisting of a single translation unit. 8976 * Several scalability bottle-necks have been removed from GCC's 8977 optimization passes. Compilation of extremely large functions, e.g. 8978 due to the use of the flatten attribute in the "Eigen" C++ linear 8979 algebra templates library, is significantly faster than previous 8980 releases of GCC. 8981 * Link-time optimization (LTO) improvements: 8982 + LTO partitioning has been rewritten for better reliability and 8983 maintanibility. Several important bugs leading to link 8984 failures have been fixed. 8985 * Interprocedural optimization improvements: 8986 + A new symbol table has been implemented. It builds on existing 8987 callgraph and varpool modules and provide a new API. Unusual 8988 symbol visibilities and aliases are handled more consistently 8989 leading to, for example, more aggressive unreachable code 8990 removal with LTO. 8991 + The inline heuristic can now bypass limits on the size of of 8992 inlined functions when the inlining is particularly 8993 profitable. This happens, for example, when loop bounds or 8994 array strides get propagated. 8995 + Values passed through aggregates (either by value or 8996 reference) are now propagated at the inter-procedural level 8997 leading to better inlining decisions (for example in the case 8998 of Fortran array descriptors) and devirtualization. 8999 * [6]AddressSanitizer , a fast memory error detector, has been added 9000 and can be enabled via -fsanitize=address. Memory access 9001 instructions will be instrumented to detect heap-, stack-, and 9002 global-buffer overflow as well as use-after-free bugs. To get nicer 9003 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 9004 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 9005 x86-64 Darwin. 9006 * [7]ThreadSanitizer has been added and can be enabled via 9007 -fsanitize=thread. Instructions will be instrumented to detect data 9008 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 9009 * A new local register allocator (LRA) has been implemented, which 9010 replaces the 26 year old reload pass and improves generated code 9011 quality. For now it is active on the IA-32 and x86-64 targets. 9012 * Support for transactional memory has been implemented on the 9013 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 9014 Alpha. 9015 9016 New Languages and Language specific improvements 9017 9018 C family 9019 9020 * Each diagnostic emitted now includes the original source line and a 9021 caret '^' indicating the column. The option 9022 -fno-diagnostics-show-caret suppresses this information. 9023 * The option -ftrack-macro-expansion=2 is now enabled by default. 9024 This allows the compiler to display the macro expansion stack in 9025 diagnostics. Combined with the caret information, an example 9026 diagnostic showing these two features is: 9027 9028 t.c:1:94: error: invalid operands to binary < (have struct mystruct and float 9029 ) 9030 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 9031 _b = (B); __a < __b ? __b : __a; }) 9032 9033 ^ 9034 t.c:7:7: note: in expansion of macro 'MYMAX' 9035 X = MYMAX(P, F); 9036 ^ 9037 9038 * A new -Wsizeof-pointer-memaccess warning has been added (also 9039 enabled by -Wall) to warn about suspicious length parameters to 9040 certain string and memory built-in functions if the argument uses 9041 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 9042 (ptr)); if ptr is not an array, but a pointer, and suggests a 9043 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 9044 * The new option -Wpedantic is an alias for -pedantic, which is now 9045 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 9046 -Wno-error=pedantic work in the same way as for any other -W 9047 option. One caveat is that -Werror=pedantic is not equivalent to 9048 -pedantic-errors, since the latter makes into errors some warnings 9049 that are not controlled by -Wpedantic, and the former only affects 9050 diagnostics that are disabled when using -Wno-pedantic. 9051 * The option -Wshadow no longer warns if a declaration shadows a 9052 function declaration, unless the former declares a function or 9053 pointer to function, because this is [8]a common and valid case in 9054 real-world code. 9055 9056 C++ 9057 9058 * G++ now implements the [9]C++11 thread_local keyword; this differs 9059 from the GNU __thread keyword primarily in that it allows dynamic 9060 initialization and destruction semantics. Unfortunately, this 9061 support requires a run-time penalty for references to 9062 non-function-local thread_local variables defined in a different 9063 translation unit even if they don't need dynamic initialization, so 9064 users may want to continue to use __thread for TLS variables with 9065 static initialization semantics. 9066 If the programmer can be sure that no use of the variable in a 9067 non-defining TU needs to trigger dynamic initialization (either 9068 because the variable is statically initialized, or a use of the 9069 variable in the defining TU will be executed before any uses in 9070 another TU), they can avoid this overhead with the 9071 -fno-extern-tls-init option. 9072 OpenMP threadprivate variables now also support dynamic 9073 initialization and destruction by the same mechanism. 9074 * G++ now implements the [10]C++11 attribute syntax, e.g. 9075 9076 [[noreturn]] void f(); 9077 9078 and also the alignment specifier, e.g. 9079 9080 alignas(double) int i; 9081 9082 * G++ now implements [11]C++11 inheriting constructors, e.g. 9083 9084 struct A { A(int); }; 9085 struct B: A { using A::A; }; // defines B::B(int) 9086 B b(42); // OK 9087 9088 * As of GCC 4.8.1, G++ implements the change to decltype semantics 9089 from [12]N3276. 9090 9091 struct A f(); 9092 decltype(f()) g(); // OK, return type of f() is not required to be complete. 9093 9094 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 9095 9096 struct A { int f() &; }; 9097 int i = A().f(); // error, f() requires an lvalue object 9098 9099 * G++ now supports a -std=c++1y option for experimentation with 9100 features proposed for the next revision of the standard, expected 9101 around 2014. Currently the only difference from -std=c++11 is 9102 support for return type deduction in normal functions, as proposed 9103 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 9104 [15]here. 9105 * The G++ namespace association extension, __attribute ((strong)), 9106 has been deprecated. Inline namespaces should be used instead. 9107 * G++ now supports a -fext-numeric-literal option to control whether 9108 GNU numeric literal suffixes are accepted as extensions or 9109 processed as C++11 user-defined numeric literal suffixes. The flag 9110 is on (use suffixes for GNU literals) by default for -std=gnu++*, 9111 and -std=c++98. The flag is off (use suffixes for user-defined 9112 literals) by default for -std=c++11 and later. 9113 9114 Runtime Library (libstdc++) 9115 9116 * [16]Improved experimental support for the new ISO C++ standard, 9117 C++11, including: 9118 + forward_list meets the allocator-aware container requirements; 9119 + this_thread::sleep_for(), this_thread::sleep_until() and 9120 this_thread::yield() are defined without requiring the 9121 configure option --enable-libstdcxx-time; 9122 * Improvements to <random>: 9123 + SSE optimized normal_distribution. 9124 + Use of hardware RNG instruction for random_device on new x86 9125 processors (requires the assembler to support the 9126 instruction.) 9127 and <ext/random>: 9128 + New random number engine simd_fast_mersenne_twister_engine 9129 with an optimized SSE implementation. 9130 + New random number distributions beta_distribution, 9131 normal_mv_distribution, rice_distribution, 9132 nakagami_distribution, pareto_distribution, k_distribution, 9133 arcsine_distribution, hoyt_distribution. 9134 * Added --disable-libstdcxx-verbose configure option to disable 9135 diagnostic messages issued when a process terminates abnormally. 9136 This may be useful for embedded systems to reduce the size of 9137 executables that link statically to the library. 9138 9139 Fortran 9140 9141 * Compatibility notice: 9142 + Module files: The version of module files (.mod) has been 9143 incremented. Fortran MODULEs compiled by earlier GCC versions 9144 have to be recompiled, when they are USEd by files compiled 9145 with GCC 4.8. GCC 4.8 is not able to read .mod files created 9146 by earlier versions; attempting to do so gives an error 9147 message. 9148 Note: The ABI of the produced assembler data itself has not 9149 changed; object files and libraries are fully compatible with 9150 older versions except as noted below. 9151 + ABI: Some internal names (used in the assembler/object file) 9152 have changed for symbols declared in the specification part of 9153 a module. If an affected module or a file using it via use 9154 association is recompiled, the module and all files which 9155 directly use such symbols have to be recompiled as well. This 9156 change only affects the following kind of module symbols: 9157 o Procedure pointers. Note: C-interoperable function 9158 pointers (type(c_funptr)) are not affected nor are 9159 procedure-pointer components. 9160 o Deferred-length character strings. 9161 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 9162 backtrace at an arbitrary place in user code; program execution 9163 continues normally afterwards. 9164 * The [18]-Wc-binding-type warning option has been added (disabled by 9165 default). It warns if the a variable might not be C interoperable; 9166 in particular, if the variable has been declared using an intrinsic 9167 type with default kind instead of using a kind parameter defined 9168 for C interoperability in the intrinsic ISO_C_Binding module. 9169 Before, this warning was always printed. The -Wc-binding-type 9170 option is enabled by -Wall. 9171 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 9172 options have been added, which diagnose when code is inserted for 9173 automatic (re)allocation of a variable during assignment. This 9174 option can be used to decide whether it is safe to use 9175 [20]-fno-realloc-lhs. Additionally, it can be used to find 9176 automatic (re)allocation in hot loops. (For arrays, replacing 9177 "var=" by "var(:)=" disables the automatic reallocation.) 9178 * The [21]-Wcompare-reals command-line option has been added. When 9179 this is set, warnings are issued when comparing REAL or COMPLEX 9180 types for equality and inequality; consider replacing a == b by 9181 abs(ab) < eps with a suitable eps. -Wcompare-reals is enabled by 9182 -Wextra. 9183 * The [22]-Wtarget-lifetime command-line option has been added 9184 (enabled with -Wall), which warns if the pointer in a pointer 9185 assignment might outlive its target. 9186 * Reading floating point numbers which use "q" for the exponential 9187 (such as 4.0q0) is now supported as vendor extension for better 9188 compatibility with old data files. It is strongly recommended to 9189 use for I/O the equivalent but standard conforming "e" (such as 9190 4.0e0). 9191 (For Fortran source code, consider replacing the "q" in 9192 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 9193 suitable qp). Note that in Fortran source code replacing "q" by 9194 a simple "e" is not equivalent.) 9195 * The GFORTRAN_TMPDIR environment variable for specifying a 9196 non-default directory for files opened with STATUS="SCRATCH", is 9197 not used anymore. Instead gfortran checks the POSIX/GNU standard 9198 TMPDIR environment variable. If TMPDIR is not defined, gfortran 9199 falls back to other methods to determine the directory for 9200 temporary files as documented in the [23]user manual. 9201 * [24]Fortran 2003: 9202 + Support for unlimited polymorphic variables (CLASS(*)) has 9203 been added. Nonconstant character lengths are not yet 9204 supported. 9205 * [25]TS 29113: 9206 + Assumed types (TYPE(*)) are now supported. 9207 + Experimental support for assumed-rank arrays (dimension(..)) 9208 has been added. Note that currently gfortran's own array 9209 descriptor is used, which is different from the one defined in 9210 TS29113, see [26]gfortran's header file or use the [27]Chasm 9211 Language Interoperability Tools. 9212 9213 Go 9214 9215 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 9216 release. 9217 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 9218 release. The library support is not quite complete. 9219 * Go has been tested on GNU/Linux and Solaris platforms for various 9220 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 9221 work on other platforms as well. 9222 9223 New Targets and Target Specific Improvements 9224 9225 AArch64 9226 9227 * A new port has been added to support AArch64, the new 64-bit 9228 architecture from ARM. Note that this is a separate port from the 9229 existing 32-bit ARM port. 9230 * The port provides initial support for the Cortex-A53 and the 9231 Cortex-A57 processors with the command line options 9232 -mcpu=cortex-a53 and -mcpu=cortex-a57. 9233 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 9234 has been added and can be enabled by giving the 9235 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 9236 default by configuring GCC with the --enable-fix-cortex-a53-835769 9237 option. 9238 9239 ARM 9240 9241 * Initial support has been added for the AArch32 extensions defined 9242 in the ARMv8 architecture. 9243 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 9244 * A new option, -mcpu=marvell-pj4, has been added to generate code 9245 for the Marvell PJ4 processor. 9246 * The compiler can now automatically generate the VFMA, VFMS, REVSH 9247 and REV16 instructions. 9248 * A new vectorizer cost model for Advanced SIMD configurations to 9249 improve the auto-vectorization strategies used. 9250 * The scheduler now takes into account the number of live registers 9251 to reduce the amount of spilling that can occur. This should 9252 improve code performance in large functions. The limit can be 9253 removed by using the option -fno-sched-pressure. 9254 * Improvements have been made to the Marvell iWMMX code generation 9255 and support for the iWMMX2 SIMD unit has been added. The option 9256 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 9257 * A number of code generation improvements for Thumb2 to reduce code 9258 size when compiling for the M-profile processors. 9259 * The RTEMS (arm-rtems) port has been updated to use the EABI. 9260 * Code generation support for the old FPA and Maverick floating-point 9261 architectures has been removed. Ports that previously relied on 9262 these features have also been removed. This includes the targets: 9263 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 9264 + arm*-*-elf (use arm*-*-eabi) 9265 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 9266 + arm*-*-ecos-elf (no alternative) 9267 + arm*-*-freebsd (no alternative) 9268 + arm*-wince-pe* (no alternative). 9269 9270 AVR 9271 9272 * Support for the "Embedded C" fixed-point has been added. For 9273 details, see the [28]GCC wiki and the [29]user manual. The support 9274 is not complete. 9275 * A new print modifier %r for register operands in inline assembler 9276 is supported. It will print the raw register number without the 9277 register prefix 'r': 9278 /* Return the most significant byte of 'val', a 64-bit value. */ 9279 9280 unsigned char msb (long long val) 9281 { 9282 unsigned char c; 9283 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 9284 return c; 9285 } 9286 The inline assembler in this example will generate code like 9287 mov r24, 8+7 9288 provided c is allocated to R24 and val is allocated to R8R15. This 9289 works because the GNU assembler accepts plain register numbers 9290 without register prefix. 9291 * Static initializers with 3-byte symbols are supported now: 9292 extern const __memx char foo; 9293 const __memx void *pfoo = &foo; 9294 This requires at least Binutils 2.23. 9295 9296 IA-32/x86-64 9297 9298 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 9299 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 9300 stack alignment, this is ABI incompatible and intended to be used 9301 in controlled environments where stack space is an important 9302 limitation. This option will lead to wrong code when functions 9303 compiled with 16 byte stack alignment (such as functions from a 9304 standard library) are called with misaligned stack. In this case, 9305 SSE instructions may lead to misaligned memory access traps. In 9306 addition, variable arguments will be handled incorrectly for 16 9307 byte aligned objects (including x87 long double and __int128), 9308 leading to wrong results. You must build all modules with 9309 -mpreferred-stack-boundary=3, including any libraries. This 9310 includes the system libraries and startup modules. 9311 * Support for the new Intel processor codename Broadwell with RDSEED, 9312 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 9313 -mrdseed command-line options. 9314 * Support for the Intel RTM and HLE intrinsics, built-in functions 9315 and code generation is available via -mrtm and -mhle. 9316 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 9317 Intrinsics and built-in functions are available via -mfxsr, -mxsave 9318 and -mxsaveopt respectively. 9319 * New -maddress-mode=[short|long] options for x32. 9320 -maddress-mode=short overrides default 64-bit addresses to 32-bit 9321 by emitting the 0x67 address-size override prefix. This is the 9322 default address mode for x32. 9323 * New built-in functions to detect run-time CPU type and ISA: 9324 + A built-in function __builtin_cpu_is has been added to detect 9325 if the run-time CPU is of a particular type. It returns a 9326 positive integer on a match and zero otherwise. It accepts one 9327 string literal argument, the CPU name. For example, 9328 __builtin_cpu_is("westmere") returns a positive integer if the 9329 run-time CPU is an Intel Core i7 Westmere processor. Please 9330 refer to the [30]user manual for the list of valid CPU names 9331 recognized. 9332 + A built-in function __builtin_cpu_supports has been added to 9333 detect if the run-time CPU supports a particular ISA feature. 9334 It returns a positive integer on a match and zero otherwise. 9335 It accepts one string literal argument, the ISA feature. For 9336 example, __builtin_cpu_supports("ssse3") returns a positive 9337 integer if the run-time CPU supports SSSE3 instructions. 9338 Please refer to the [31]user manual for the list of valid ISA 9339 names recognized. 9340 Caveat: If these built-in functions are called before any static 9341 constructors are invoked, like during IFUNC initialization, then 9342 the CPU detection initialization must be explicitly run using this 9343 newly provided built-in function, __builtin_cpu_init. The 9344 initialization needs to be done only once. For example, this is how 9345 the invocation would look like inside an IFUNC initializer: 9346 static void (*some_ifunc_resolver(void))(void) 9347 { 9348 __builtin_cpu_init(); 9349 if (__builtin_cpu_is("amdfam10h") ... 9350 if (__builtin_cpu_supports("popcnt") ... 9351 } 9352 9353 * Function Multiversioning Support with G++: 9354 It is now possible to create multiple function versions each 9355 targeting a specific processor and/or ISA. Function versions have 9356 the same signature but different target attributes. For example, 9357 here is a program with function versions: 9358 __attribute__ ((target ("default"))) 9359 int foo(void) 9360 { 9361 return 1; 9362 } 9363 9364 __attribute__ ((target ("sse4.2"))) 9365 int foo(void) 9366 { 9367 return 2; 9368 } 9369 9370 int main (void) 9371 { 9372 int (*p) = &foo; 9373 assert ((*p)() == foo()); 9374 return 0; 9375 } 9376 9377 Please refer to this [32]wiki for more information. 9378 * The x86 back end has been improved to allow option -fschedule-insns 9379 to work reliably. This option can be used to schedule instructions 9380 better and leads to improved performace in certain cases. 9381 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 9382 from the Mingw-w64 trunk. 9383 * Support for new AMD family 15h processors (Steamroller core) is now 9384 available through the -march=bdver3 and -mtune=bdver3 options. 9385 * Support for new AMD family 16h processors (Jaguar core) is now 9386 available through the -march=btver2 and -mtune=btver2 options. 9387 9388 FRV 9389 9390 * This target now supports the -fstack-usage command-line option. 9391 9392 MIPS 9393 9394 * GCC can now generate code specifically for the R4700, Broadcom XLP 9395 and MIPS 34kn processors. The associated -march options are 9396 -march=r4700, -march=xlp and -march=34kn respectively. 9397 * GCC now generates better DSP code for MIPS 74k cores thanks to 9398 further scheduling optimizations. 9399 * The MIPS port now supports the -fstack-check option. 9400 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 9401 * Previous versions of GCC would silently accept -fpic and -fPIC for 9402 -mno-abicalls targets like mips*-elf. This combination was not 9403 intended or supported, and did not generate position-independent 9404 code. GCC 4.8 now reports an error when this combination is used. 9405 9406 PowerPC / PowerPC64 / RS6000 9407 9408 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 9409 restore or update the VRSAVE register by default. The respective 9410 operating systems manage the VRSAVE register directly. 9411 * Large TOC support has been added for AIX through the command line 9412 option -mcmodel=large. 9413 * Native Thread-Local Storage support has been added for AIX. 9414 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 9415 when targetting processors that support those hardware features on 9416 AIX 6.1 and above. 9417 9418 RX 9419 9420 * This target will now issue a warning message whenever multiple fast 9421 interrupt handlers are found in the same compilation unit. This 9422 feature can be turned off by the new 9423 -mno-warn-multiple-fast-interrupts command-line option. 9424 9425 S/390, System z 9426 9427 * Support for the IBM zEnterprise zEC12 processor has been added. 9428 When using the -march=zEC12 option, the compiler will generate code 9429 making use of the following new instructions: 9430 + load and trap instructions 9431 + 2 new compare and trap instructions 9432 + rotate and insert selected bits - without CC clobber 9433 The -mtune=zEC12 option enables zEC12 specific instruction 9434 scheduling without making use of new instructions. 9435 * Register pressure sensitive instruction scheduling is enabled by 9436 default. 9437 * The ifunc function attribute is enabled by default. 9438 * memcpy and memcmp invokations on big memory chunks or with run time 9439 lengths are not generated inline anymore when tuning for z10 or 9440 higher. The purpose is to make use of the IFUNC optimized versions 9441 in Glibc. 9442 9443 SH 9444 9445 * The default alignment settings have been reduced to be less 9446 aggressive. This results in more compact code for optimization 9447 levels other than -Os. 9448 * Improved support for the __atomic built-in functions: 9449 + A new option -matomic-model=model selects the model for the 9450 generated atomic sequences. The following models are 9451 supported: 9452 9453 soft-gusa 9454 Software gUSA sequences (SH3* and SH4* only). On 9455 SH4A targets this will now also partially utilize 9456 the movco.l and movli.l instructions. This is the 9457 default when the target is sh3*-*-linux* or 9458 sh4*-*-linux*. 9459 9460 hard-llcs 9461 Hardware movco.l / movli.l sequences (SH4A only). 9462 9463 soft-tcb 9464 Software thread control block sequences. 9465 9466 soft-imask 9467 Software interrupt flipping sequences (privileged 9468 mode only). This is the default when the target is 9469 sh1*-*-linux* or sh2*-*-linux*. 9470 9471 none 9472 Generates function calls to the respective __atomic 9473 built-in functions. This is the default for SH64 9474 targets or when the target is not sh*-*-linux*. 9475 9476 + The option -msoft-atomic has been deprecated. It is now an 9477 alias for -matomic-model=soft-gusa. 9478 + A new option -mtas makes the compiler generate the tas.b 9479 instruction for the __atomic_test_and_set built-in function 9480 regardless of the selected atomic model. 9481 + The __sync functions in libgcc now reflect the selected atomic 9482 model when building the toolchain. 9483 * Added support for the mov.b and mov.w instructions with 9484 displacement addressing. 9485 * Added support for the SH2A instructions movu.b and movu.w. 9486 * Various improvements to code generated for integer arithmetic. 9487 * Improvements to conditional branches and code that involves the T 9488 bit. A new option -mzdcbranch tells the compiler to favor 9489 zero-displacement branches. This is enabled by default for SH4* 9490 targets. 9491 * The pref instruction will now be emitted by the __builtin_prefetch 9492 built-in function for SH3* targets. 9493 * The fmac instruction will now be emitted by the fmaf standard 9494 function and the __builtin_fmaf built-in function. 9495 * The -mfused-madd option has been deprecated in favor of the 9496 machine-independent -ffp-contract option. Notice that the fmac 9497 instruction will now be generated by default for expressions like a 9498 * b + c. This is due to the compiler default setting 9499 -ffp-contract=fast. 9500 * Added new options -mfsrra and -mfsca to allow the compiler using 9501 the fsrra and fsca instructions on targets other than SH4A (where 9502 they are already enabled by default). 9503 * Added support for the __builtin_bswap32 built-in function. It is 9504 now expanded as a sequence of swap.b and swap.w instructions 9505 instead of a library function call. 9506 * The behavior of the -mieee option has been fixed and the negative 9507 form -mno-ieee has been added to control the IEEE conformance of 9508 floating point comparisons. By default -mieee is now enabled and 9509 the option -ffinite-math-only implicitly sets -mno-ieee. 9510 * Added support for the built-in functions __builtin_thread_pointer 9511 and __builtin_set_thread_pointer. This assumes that GBR is used to 9512 hold the thread pointer of the current thread. Memory loads and 9513 stores relative to the address returned by __builtin_thread_pointer 9514 will now also utilize GBR based displacement address modes. 9515 * The -mdiv= option for targets other than SHmedia has been fixed and 9516 documented. 9517 9518 SPARC 9519 9520 * Added optimized instruction scheduling for Niagara4. 9521 9522 TILE-Gx 9523 9524 * Added support for the -mcmodel=MODEL command-line option. The 9525 models supported are small and large. 9526 9527 V850 9528 9529 * This target now supports the E3V5 architecture via the use of the 9530 new -mv850e3v5 command-line option. It also has experimental 9531 support for the e3v5 LOOP instruction which can be enabled via the 9532 new -mloop command-line option. 9533 9534 XStormy16 9535 9536 * This target now supports the -fstack-usage command-line option. 9537 9538 Operating Systems 9539 9540 OpenBSD 9541 9542 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 9543 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 9544 9545 Windows (Cygwin) 9546 9547 * Executables are now linked against shared libgcc by default. The 9548 previous default was to link statically, which can still be done by 9549 explicitly specifying -static or static-libgcc on the command line. 9550 However it is strongly advised against, as it will cause problems 9551 for any application that makes use of DLLs compiled by GCC. It 9552 should be alright for a monolithic stand-alone application that 9553 only links against the Windows DLLs, but offers little or no 9554 benefit. 9555 9556 GCC 4.8.1 9557 9558 This is the [33]list of problem reports (PRs) from GCC's bug tracking 9559 system that are known to be fixed in the 4.8.1 release. This list might 9560 not be complete (that is, it is possible that some PRs that have been 9561 fixed are not listed here). 9562 9563 The C++11 <chrono> std::chrono::system_clock and 9564 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 9565 both are now separate (never typedefs of each other), both use 9566 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 9567 std::chrono::steady_clock is now finally monotonic, and both classes 9568 are mangled differently than in the previous GCC releases. 9569 std::chrono::system_clock::now() with std::chrono::microseconds resp. 9570 std::chrono::seconds resolution is still exported for backwards 9571 compatibility with default configured libstdc++. Note that libstdc++ 9572 configured with --enable-libstdcxx-time= used to be ABI incompatible 9573 with default configured libstdc++ for those two classes and no ABI 9574 compatibility can be offered for those configurations, so any C++11 9575 code that uses those classes and has been compiled and linked against 9576 libstdc++ configured with the non-default --enable-libstdcxx-time= 9577 configuration option needs to be recompiled. 9578 9579 GCC 4.8.2 9580 9581 This is the [34]list of problem reports (PRs) from GCC's bug tracking 9582 system that are known to be fixed in the 4.8.2 release. This list might 9583 not be complete (that is, it is possible that some PRs that have been 9584 fixed are not listed here). 9585 9586 GCC 4.8.3 9587 9588 This is the [35]list of problem reports (PRs) from GCC's bug tracking 9589 system that are known to be fixed in the 4.8.3 release. This list might 9590 not be complete (that is, it is possible that some PRs that have been 9591 fixed are not listed here). 9592 9593 Support for the new powerpc64le-linux platform has been added. It 9594 defaults to generating code that conforms to the ELFV2 ABI. 9595 9596 GCC 4.8.4 9597 9598 This is the [36]list of problem reports (PRs) from GCC's bug tracking 9599 system that are known to be fixed in the 4.8.4 release. This list might 9600 not be complete (that is, it is possible that some PRs that have been 9601 fixed are not listed here). 9602 9603 GCC 4.8.5 9604 9605 This is the [37]list of problem reports (PRs) from GCC's bug tracking 9606 system that are known to be fixed in the 4.8.5 release. This list might 9607 not be complete (that is, it is possible that some PRs that have been 9608 fixed are not listed here). 9609 9610 9611 For questions related to the use of GCC, please consult these web 9612 pages and the [38]GCC manuals. If that fails, the 9613 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9614 web pages and the development of GCC are welcome on our developer 9615 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 9616 archives. 9617 9618 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 9619 distribution of this entire article is permitted in any medium, 9620 provided this notice is preserved. 9621 9622 These pages are [43]maintained by the GCC team. Last modified 9623 2022-11-05. 9624 9625 References 9626 9627 1. https://gcc.gnu.org/wiki/cxx-conversion 9628 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 9629 3. http://www.nongnu.org/avr-libc/ 9630 4. https://gcc.gnu.org/PR54461 9631 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 9632 6. https://github.com/google/sanitizers 9633 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 9634 8. https://lkml.org/lkml/2006/11/28/239 9635 9. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9636 10. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9637 11. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9638 12. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 9639 13. https://gcc.gnu.org/gcc-4.8/cxx0x_status.html 9640 14. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 9641 15. https://gcc.gnu.org/projects/cxx1y.html 9642 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 9643 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 9644 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9645 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9646 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 9647 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9648 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 9649 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 9650 24. https://gcc.gnu.org/wiki/Fortran2003Status 9651 25. https://gcc.gnu.org/wiki/TS29113Status 9652 26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h 9653 27. https://chasm-interop.sourceforge.net/ 9654 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 9655 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 9656 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 9657 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 9658 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 9659 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 9660 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 9661 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 9662 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 9663 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 9664 38. https://gcc.gnu.org/onlinedocs/ 9665 39. mailto:gcc-help (a] gcc.gnu.org 9666 40. mailto:gcc (a] gcc.gnu.org 9667 41. https://gcc.gnu.org/lists.html 9668 42. https://www.fsf.org/ 9669 43. https://gcc.gnu.org/about.html 9670 ====================================================================== 9671 http://gcc.gnu.org/gcc-4.7/index.html 9672 GCC 4.7 Release Series 9673 9674 (This release series is no longer supported.) 9675 9676 June 12, 2014 9677 9678 The [1]GNU project and the GCC developers are pleased to announce the 9679 release of GCC 4.7.4. 9680 9681 This release is a bug-fix release, containing fixes for regressions in 9682 GCC 4.7.3 relative to previous releases of GCC. 9683 9684 Release History 9685 9686 GCC 4.7.4 9687 June 12, 2014 ([2]changes, [3]documentation) 9688 9689 GCC 4.7.3 9690 April 11, 2013 ([4]changes, [5]documentation) 9691 9692 GCC 4.7.2 9693 September 20, 2012 ([6]changes, [7]documentation) 9694 9695 GCC 4.7.1 9696 June 14, 2012 ([8]changes, [9]documentation) 9697 9698 GCC 4.7.0 9699 March 22, 2012 ([10]changes, [11]documentation) 9700 9701 References and Acknowledgements 9702 9703 GCC used to stand for the GNU C Compiler, but since the compiler 9704 supports several other languages aside from C, it now stands for the 9705 GNU Compiler Collection. 9706 9707 A list of [12]successful builds is updated as new information becomes 9708 available. 9709 9710 The GCC developers would like to thank the numerous people that have 9711 contributed new features, improvements, bug fixes, and other changes as 9712 well as test results to GCC. This [13]amazing group of volunteers is 9713 what makes GCC successful. 9714 9715 For additional information about GCC please refer to the [14]GCC 9716 project web site or contact the [15]GCC development mailing list. 9717 9718 To obtain GCC please use [16]our mirror sites or [17]our version 9719 control system. 9720 9721 9722 For questions related to the use of GCC, please consult these web 9723 pages and the [18]GCC manuals. If that fails, the 9724 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9725 web pages and the development of GCC are welcome on our developer 9726 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 9727 archives. 9728 9729 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9730 distribution of this entire article is permitted in any medium, 9731 provided this notice is preserved. 9732 9733 These pages are [23]maintained by the GCC team. Last modified 9734 2022-10-26. 9735 9736 References 9737 9738 1. http://www.gnu.org/ 9739 2. https://gcc.gnu.org/gcc-4.7/changes.html 9740 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 9741 4. https://gcc.gnu.org/gcc-4.7/changes.html 9742 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 9743 6. https://gcc.gnu.org/gcc-4.7/changes.html 9744 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 9745 8. https://gcc.gnu.org/gcc-4.7/changes.html 9746 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 9747 10. https://gcc.gnu.org/gcc-4.7/changes.html 9748 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 9749 12. https://gcc.gnu.org/gcc-4.7/buildstat.html 9750 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9751 14. https://gcc.gnu.org/index.html 9752 15. mailto:gcc (a] gcc.gnu.org 9753 16. https://gcc.gnu.org/mirrors.html 9754 17. https://gcc.gnu.org/git.html 9755 18. https://gcc.gnu.org/onlinedocs/ 9756 19. mailto:gcc-help (a] gcc.gnu.org 9757 20. mailto:gcc (a] gcc.gnu.org 9758 21. https://gcc.gnu.org/lists.html 9759 22. https://www.fsf.org/ 9760 23. https://gcc.gnu.org/about.html 9761 ====================================================================== 9762 http://gcc.gnu.org/gcc-4.7/changes.html 9763 GCC 4.7 Release Series 9764 Changes, New Features, and Fixes 9765 9766 Caveats 9767 9768 * The -fconserve-space flag has been deprecated. The flag had no 9769 effect for most targets: only targets without a global .bss section 9770 and without support for switchable sections. Furthermore, the flag 9771 only had an effect for G++, where it could result in wrong 9772 semantics (please refer to the GCC manual for further details). The 9773 flag will be removed in GCC 4.8 9774 * Support for a number of older systems and recently unmaintained or 9775 untested target ports of GCC has been declared obsolete in GCC 4.7. 9776 Unless there is activity to revive them, the next release of GCC 9777 will have their sources permanently removed. 9778 All GCC ports for the following processor architectures have been 9779 declared obsolete: 9780 + picoChip (picochip-*) 9781 The following ports for individual systems on particular 9782 architectures have been obsoleted: 9783 + IRIX 6.5 (mips-sgi-irix6.5) 9784 + MIPS OpenBSD (mips*-*-openbsd*) 9785 + Solaris 8 (*-*-solaris2.8). Details can be found in the 9786 [1]announcement. 9787 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 9788 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 9789 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 9790 default, which for some sources generates code that accesses memory 9791 on unaligned addresses. This requires the kernel of those systems 9792 to enable such accesses (controlled by CP15 register c1, refer to 9793 ARM documentation). Alternatively, or for compatibility with 9794 kernels where unaligned accesses are not supported, all code has to 9795 be compiled with -mno-unaligned-access. Upstream Linux kernel 9796 releases have automatically and unconditionally supported unaligned 9797 accesses as emitted by GCC due to this option being active since 9798 version 2.6.28. 9799 * Support on ARM for the legacy floating-point accelerator (FPA) and 9800 the mixed-endian floating-point format that it used has been 9801 obsoleted. The ports that still use this format have been obsoleted 9802 as well. Many legacy ARM ports already provide an alternative that 9803 uses the VFP floating-point format. The obsolete ports will be 9804 deleted in the next release. 9805 The obsolete ports with alternatives are: 9806 + arm*-*-rtems (use arm*-*-rtemseabi) 9807 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 9808 + arm*-*-elf (use arm*-*-eabi) 9809 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 9810 Note, however, that these alternatives are not binary compatible 9811 with their legacy counterparts (although some can support running 9812 legacy applications). 9813 The obsolete ports that currently lack a modern alternative are: 9814 + arm*-*-ecos-elf 9815 + arm*-*-freebsd 9816 + arm*-wince-pe* 9817 New ports that support more recent versions of the architecture are 9818 welcome. 9819 * Support for the Maverick co-processor on ARM has been obsoleted. 9820 Code to support it will be deleted in the next release. 9821 * Support has been removed for Unix International threads on Solaris 9822 2, so the --enable-threads=solaris configure option and the 9823 -threads compiler option don't work any longer. 9824 * Support has been removed for the Solaris BSD Compatibility Package, 9825 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 9826 from Solaris 11, and was only intended as a migration aid from 9827 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 9828 recognized any longer. 9829 * The AVR port's libgcc has been improved and its multilib structure 9830 has been enhanced. As a result, all objects contributing to an 9831 application must either be compiled with GCC versions up to 4.6.x 9832 or with GCC versions 4.7.1 or later. If the compiler is used with 9833 AVR Libc, you need a version that supports the new layout, i.e. 9834 implements [2]#35407. 9835 * The AVR port's -mshort-calls command-line option has been 9836 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 9837 for a replacement. 9838 * The AVR port only references startup code that clears .bss and the 9839 common section resp. initializes the .data and .rodata section 9840 provided respective sections (or subsections thereof) are not 9841 empty, see [3]PR18145. Applications that put all static storage 9842 objects into non-standard sections and / or define all static 9843 storage objects in assembler modules, must reference __do_clear_bss 9844 resp. __do_copy_data by hand or undefine the symbol(s) by means of 9845 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 9846 * The ARM port's -mwords-little-endian option has been deprecated. It 9847 will be removed in a future release. 9848 * Support has been removed for the NetWare x86 configuration 9849 obsoleted in GCC 4.6. 9850 * It is no longer possible to use the "l" constraint in MIPS16 asm 9851 statements. 9852 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 9853 library which affected the ABI in C++11 mode: a data member was 9854 added to std::list changing its size and altering the definitions 9855 of some member functions, and std::pair's move constructor was 9856 non-trivial which altered the calling convention for functions with 9857 std::pair arguments or return types. The ABI incompatibilities have 9858 been fixed for GCC version 4.7.2 but as a result C++11 code 9859 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 9860 code compiled with different GCC versions and with C++98/C++03 code 9861 compiled with any version. 9862 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 9863 rules for the layout of vectors that could lead to wrong code being 9864 generated. Vectors larger than 8 bytes in size are now by default 9865 aligned to an 8-byte boundary. This is an ABI change: code that 9866 makes explicit use of vector types may be incompatible with binary 9867 objects built with older versions of GCC. Auto-vectorized code is 9868 not affected by this change. (This change affects GCC versions 9869 4.7.2 and later.) 9870 * More information on porting to GCC 4.7 from previous versions of 9871 GCC can be found in the [4]porting guide for this release. 9872 9873 General Optimizer Improvements 9874 9875 * Support for a new parameter --param case-values-threshold=n was 9876 added to allow users to control the cutoff between doing switch 9877 statements as a series of if statements and using a jump table. 9878 * Link-time optimization (LTO) improvements: 9879 + Improved scalability and reduced memory usage. Link time 9880 optimization of Firefox now requires 3GB of RAM on a 64-bit 9881 system, while over 8GB was needed previously. Linking time has 9882 been improved, too. The serial stage of linking Firefox has 9883 been sped up by about a factor of 10. 9884 + Reduced size of object files and temporary storage used during 9885 linking. 9886 + Streaming performance (both outbound and inbound) has been 9887 improved. 9888 + ld -r is now supported with LTO. 9889 + Several bug fixes, especially in symbol table handling and 9890 merging. 9891 * Interprocedural optimization improvements: 9892 + Heuristics now take into account that after inlining code will 9893 be optimized out because of known values (or properties) of 9894 function parameters. For example: 9895 void foo(int a) 9896 { 9897 if (a > 10) 9898 ... huge code ... 9899 } 9900 void bar (void) 9901 { 9902 foo (0); 9903 } 9904 9905 The call of foo will be inlined into bar even when optimizing 9906 for code size. Constructs based on __builtin_constant_p are 9907 now understood by the inliner and code size estimates are 9908 evaluated a lot more realistically. 9909 + The representation of C++ virtual thunks and aliases (both 9910 implicit and defined via the alias attribute) has been 9911 re-engineered. Aliases no longer pose optimization barriers 9912 and calls to an alias can be inlined and otherwise optimized. 9913 + The inter-procedural constant propagation pass has been 9914 rewritten. It now performs generic function specialization. 9915 For example when compiling the following: 9916 void foo(bool flag) 9917 { 9918 if (flag) 9919 ... do something ... 9920 else 9921 ... do something else ... 9922 } 9923 void bar (void) 9924 { 9925 foo (false); 9926 foo (true); 9927 foo (false); 9928 foo (true); 9929 foo (false); 9930 foo (true); 9931 } 9932 9933 GCC will now produce two copies of foo. One with flag being 9934 true, while other with flag being false. This leads to 9935 performance improvements previously possible only by inlining 9936 all calls. Cloning causes a lot less code size growth. 9937 * A string length optimization pass has been added. It attempts to 9938 track string lengths and optimize various standard C string 9939 functions like strlen, strchr, strcpy, strcat, stpcpy and their 9940 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 9941 enabled by default at -O2 or above, unless optimizing for size, and 9942 can be disabled by the -fno-optimize-strlen option. The pass can 9943 e.g. optimize 9944 char *bar (const char *a) 9945 { 9946 size_t l = strlen (a) + 2; 9947 char *p = malloc (l); if (p == NULL) return p; 9948 strcpy (p, a); strcat (p, "/"); return p; 9949 } 9950 9951 into: 9952 char *bar (const char *a) 9953 { 9954 size_t tmp = strlen (a); 9955 char *p = malloc (tmp + 2); if (p == NULL) return p; 9956 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 9957 } 9958 9959 or for hosted compilations where stpcpy is available in the runtime 9960 and headers provide its prototype, e.g. 9961 void foo (char *a, const char *b, const char *c, const char *d) 9962 { 9963 strcpy (a, b); strcat (a, c); strcat (a, d); 9964 } 9965 9966 can be optimized into: 9967 void foo (char *a, const char *b, const char *c, const char *d) 9968 { 9969 strcpy (stpcpy (stpcpy (a, b), c), d); 9970 } 9971 9972 New Languages and Language specific improvements 9973 9974 * Version 3.1 of the OpenMP specification is now supported for the C, 9975 C++, and Fortran compilers. 9976 9977 Ada 9978 9979 * The command-line option -feliminate-unused-debug-types has been 9980 re-enabled by default, as it is for the other languages, leading to 9981 a reduction in debug info size of 12.5% and more for relevant 9982 cases, as well as to a small compilation speedup. 9983 9984 C family 9985 9986 * A new built-in, __builtin_assume_aligned, has been added, through 9987 which the compiler can be hinted about pointer alignment and can 9988 use it to improve generated code. 9989 * A new warning option -Wunused-local-typedefs was added for C, C++, 9990 Objective-C and Objective-C++. This warning diagnoses typedefs 9991 locally defined in a function, and otherwise not used. 9992 * A new experimental command-line option -ftrack-macro-expansion was 9993 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 9994 the compiler to emit diagnostic about the current macro expansion 9995 stack when a compilation error occurs in a macro expansion. 9996 * Experimental support for transactional memory has been added. It 9997 includes support in the compiler, as well as a supporting runtime 9998 library called libitm. To compile code with transactional memory 9999 constructs, use the -fgnu-tm option. 10000 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 10001 and 32-bit/64-bit x86 platforms. 10002 For more details on transactional memory see [5]the GCC WiKi. 10003 * Support for atomic operations specifying the C++11/C11 memory model 10004 has been added. These new __atomic routines replace the existing 10005 __sync built-in routines. 10006 Atomic support is also available for memory blocks. Lock-free 10007 instructions will be used if a memory block is the same size and 10008 alignment as a supported integer type. Atomic operations which do 10009 not have lock-free support are left as function calls. A set of 10010 library functions is available on the GCC atomic wiki in the 10011 "External Atomics Library" section. 10012 For more details on the memory models and features, see the 10013 [6]atomic wiki. 10014 * When a binary operation is performed on vector types and one of the 10015 operands is a uniform vector, it is possible to replace the vector 10016 with the generating element. For example: 10017 typedef int v4si __attribute__ ((vector_size (16))); 10018 v4si res, a = {1,2,3,4}; 10019 int x; 10020 10021 res = 2 + a; /* means {2,2,2,2} + a */ 10022 res = a - x; /* means a - {x,x,x,x} */ 10023 10024 C 10025 10026 * There is support for some more features from the C11 revision of 10027 the ISO C standard. GCC now accepts the options -std=c11 and 10028 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 10029 + Unicode strings (previously supported only with options such 10030 as -std=gnu11, now supported with -std=c11), and the 10031 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 10032 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 10033 + Alignment support (_Alignas, _Alignof, max_align_t, 10034 <stdalign.h>). 10035 + A built-in function __builtin_complex is provided to support C 10036 library implementation of the CMPLX family of macros. 10037 10038 C++ 10039 10040 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 10041 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 10042 -Wc++0x-compat, respectively. 10043 * G++ now implements [7]C++11 extended friend syntax: 10044 10045 template<class W> 10046 class Q 10047 { 10048 static const int I = 2; 10049 public: 10050 friend W; 10051 }; 10052 10053 struct B 10054 { 10055 int ar[Q<B>::I]; 10056 }; 10057 10058 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 10059 override control. 10060 10061 struct B { 10062 virtual void f() const final; 10063 virtual void f(int); 10064 }; 10065 10066 struct D : B { 10067 void f() const; // error: D::f attempts to override final B::f 10068 void f(long) override; // error: doesn't override anything 10069 void f(int) override; // ok 10070 }; 10071 10072 struct E final { }; 10073 struct F: E { }; // error: deriving from final class 10074 10075 * G++ now implements [9]C++11 non-static data member initializers. 10076 10077 struct A { 10078 int i = 42; 10079 } a; // initializes a.i to 42 10080 10081 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 10082 user-defined literals. 10083 10084 // Not actually a good approximation. :) 10085 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 10086 long double pi = 180.0_degrees; 10087 10088 * G++ now implements [11]C++11 alias-declarations. 10089 10090 template <class T> using Ptr = T*; 10091 Ptr<int> ip; // decltype(ip) is int* 10092 10093 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 10094 [12]C++11 delegating constructors. 10095 10096 struct A { 10097 A(int); 10098 A(): A(42) { } // delegate to the A(int) constructor 10099 }; 10100 10101 * G++ now fully implements C++11 atomic classes rather than just 10102 integer derived classes. 10103 10104 class POD { 10105 int a; 10106 int b; 10107 }; 10108 std::atomic<POD> my_atomic_POD; 10109 10110 * G++ now sets the predefined macro __cplusplus to the correct value, 10111 199711L for C++98/03, and 201103L for C++11. 10112 * G++ now correctly implements the two-phase lookup rules such that 10113 an unqualified name used in a template must have an appropriate 10114 declaration found either in scope at the point of definition of the 10115 template or by argument-dependent lookup at the point of 10116 instantiation. As a result, code that relies on a second 10117 unqualified lookup at the point of instantiation to find functions 10118 declared after the template or in dependent bases will be rejected. 10119 The compiler will suggest ways to fix affected code, and using the 10120 -fpermissive compiler flag will allow the code to compile with a 10121 warning. 10122 10123 template <class T> 10124 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 10125 void g(int) { } // fix by moving this declaration before the declaration of f 10126 10127 template <class T> 10128 struct A: T { 10129 // error, B::g(B) not found by argument-dependent lookup 10130 void f() { g(T()); } // fix by using this->g or A::g 10131 }; 10132 10133 struct B { void g(B); }; 10134 10135 int main() 10136 { 10137 f<int>(); 10138 A<B>().f(); 10139 } 10140 10141 * G++ now properly re-uses stack space allocated for temporary 10142 objects when their lifetime ends, which can significantly lower 10143 stack consumption for some C++ functions. As a result of this, some 10144 code with undefined behavior will now break: 10145 10146 const int &f(const int &i) { return i; } 10147 .... 10148 const int &x = f(1); 10149 const int &y = f(2); 10150 10151 Here, x refers to the temporary allocated to hold the 1 argument, 10152 which only lives until the end of the initialization; it 10153 immediately becomes a dangling reference. So the next statement 10154 re-uses the stack slot to hold the 2 argument, and users of x get 10155 that value instead. 10156 Note that this should not cause any change of behavior for 10157 temporaries of types with non-trivial destructors, as they are 10158 already destroyed at end of full-expression; the change is that now 10159 the storage is released as well. 10160 * A new command-line option -Wdelete-non-virtual-dtor has been added 10161 to warn when delete is used to destroy an instance of a class which 10162 has virtual functions and non-virtual destructor. It is unsafe to 10163 delete an instance of a derived class through a pointer to a base 10164 class if the base class does not have a virtual destructor. This 10165 warning is enabled by -Wall. 10166 * A new command-line option -Wzero-as-null-pointer-constant has been 10167 added to warn when a literal '0' is used as null pointer constant. 10168 It can be useful to facilitate the conversion to nullptr in C++11. 10169 * As per C++98, access-declarations are now deprecated by G++. 10170 Using-declarations are to be used instead. Furthermore, some 10171 efforts have been made to improve the support of class scope 10172 using-declarations. In particular, using-declarations referring to 10173 a dependent type now work as expected ([13]bug c++/14258). 10174 * The ELF symbol visibility of a template instantiation is now 10175 properly constrained by the visibility of its template arguments 10176 ([14]bug c++/35688). 10177 10178 Runtime Library (libstdc++) 10179 10180 * [15]Improved experimental support for the new ISO C++ standard, 10181 C++11, including: 10182 + using noexcept in most of the library; 10183 + implementations of pointer_traits, allocator_traits and 10184 scoped_allocator_adaptor; 10185 + uses-allocator construction for tuple; 10186 + vector meets the allocator-aware container requirements; 10187 + replacing monotonic_clock with steady_clock; 10188 + enabling the thread support library on most POSIX targets; 10189 + many small improvements to conform to the FDIS. 10190 * Added --enable-clocale=newlib configure option. 10191 * Debug Mode iterators for unordered associative containers. 10192 * Avoid polluting the global namespace and do not include <unistd.h>. 10193 10194 Fortran 10195 10196 * The compile flag [16]-fstack-arrays has been added, which causes 10197 all local arrays to be put on stack memory. For some programs this 10198 will improve the performance significantly. If your program uses 10199 very large local arrays, it is possible that you will have to 10200 extend your runtime limits for stack memory. 10201 * The [17]-Ofast flag now also implies [18]-fno-protect-parens and 10202 [19]-fstack-arrays. 10203 * Front-end optimizations can now be selected by the 10204 [20]-ffrontend-optimize option and deselected by the 10205 -fno-frontend-optimize option. 10206 * When front-end optimization removes a function call, 10207 [21]-Wfunction-elimination warns about that. 10208 * When performing front-end-optimization, the 10209 [22]-faggressive-function-elimination option allows the removal of 10210 duplicate function calls even for impure functions. 10211 * The flag [23]-Wreal-q-constant has been added, which warns if 10212 floating-point literals have been specified using q (such as 10213 1.0q0); the q marker is now supported as a vendor extension to 10214 denote quad precision (REAL(16) or, if not available, REAL(10)). 10215 Consider using a kind parameter (such as in 1.0_qp) instead, which 10216 can be obtained via [24]SELECTED_REAL_KIND. 10217 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 10218 Fortran now always prints error messages to standard error. If you 10219 wish to redirect standard error, please consult the manual for your 10220 OS, shell, batch environment etc. as appropriate. 10221 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 10222 variable have been removed. When encountering a serious error, 10223 gfortran will now always abort the program. Whether a core dump is 10224 generated depends on the user environment settings; see the ulimit 10225 -c setting for POSIX shells, limit coredumpsize for C shells, and 10226 the [25]WER user-mode dumps settings on Windows. 10227 * The [26]-fbacktrace option is now enabled by default. When 10228 encountering a fatal error, gfortran will attempt to print a 10229 backtrace to standard error before aborting. It can be disabled 10230 with -fno-backtrace. Note: On POSIX targets with the addr2line 10231 utility from GNU binutils, GNU Fortran can print a backtrace with 10232 function name, file name, line number information in addition to 10233 the addresses; otherwise only the addresses are printed. 10234 * [27]Fortran 2003: 10235 + Generic interface names which have the same name as derived 10236 types are now supported, which allows to write constructor 10237 functions. Note that Fortran does not support static 10238 constructor functions; only default initialization or an 10239 explicit structure-constructor initialization are available. 10240 + [28]Polymorphic (class) arrays are now supported. 10241 * [29]Fortran 2008: 10242 + Support for the DO CONCURRENT construct has been added, which 10243 allows the user to specify that individual loop iterations 10244 have no interdependencies. 10245 + [30]Coarrays: Full single-image support except for polymorphic 10246 coarrays. Additionally, preliminary support for multiple 10247 images via an MPI-based [31]coarray communication library has 10248 been added. Note: The library version is not yet usable as 10249 remote coarray access is not yet possible. 10250 * [32]TS 29113: 10251 + New flag [33]-std=f2008ts permits programs that are expected 10252 to conform to the Fortran 2008 standard and the draft 10253 Technical Specification (TS) 29113 on Further Interoperability 10254 of Fortran with C. 10255 + The OPTIONAL attribute is now allowed for dummy arguments of 10256 BIND(C) procedures. 10257 + The RANK intrinsic has been added. 10258 + The implementation of the ASYNCHRONOUS attribute in GCC is 10259 compatible with the candidate draft of TS 29113 (since GCC 10260 4.6). 10261 10262 Go 10263 10264 * GCC 4.7 implements the [34]Go 1 language standard. The library 10265 support in 4.7.0 is not quite complete, due to release timing. 10266 Release 4.7.1 includes complete support for Go 1. The Go library is 10267 from the Go 1.0.1 release. 10268 * Go has been tested on GNU/Linux and Solaris platforms. It may work 10269 on other platforms as well. 10270 10271 New Targets and Target Specific Improvements 10272 10273 ARM 10274 10275 * GCC now supports the Cortex-A7 processor implementing the v7-a 10276 version of the architecture using the option -mcpu=cortex-a7. 10277 * The default vector size in auto-vectorization for NEON is now 128 10278 bits. If vectorization fails thusly, the vectorizer tries again 10279 with 64-bit vectors. 10280 * A new option -mvectorize-with-neon-double was added to allow users 10281 to change the vector size to 64 bits. 10282 10283 AVR 10284 10285 * GCC now supports the XMEGA architecture. This requires GNU binutils 10286 2.22 or later. 10287 * Support for the [35]named address spaces __flash, __flash1, , 10288 __flash5 and __memx has been added. These address spaces locate 10289 read-only data in flash memory and allow reading from flash memory 10290 by means of ordinary C code, i.e. without the need of (inline) 10291 assembler code: 10292 10293 const __flash int values[] = { 42, 31 }; 10294 10295 int add_values (const __flash int *p, int i) 10296 { 10297 return values[i] + *p; 10298 } 10299 10300 * Support has been added for the AVR-specific configure option 10301 --with-avrlibc=yes in order to arrange for better integration of 10302 [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 10303 and newer and will only take effect in non-RTEMS configurations. If 10304 avr-gcc is configured for RTEMS, the option will be ignored which 10305 is the same as specifying --with-avrlibc=no. See [37]PR54461 for 10306 more technical details. 10307 * Support for AVR-specific [38]built-in functions has been added. 10308 * Support has been added for the signed and unsigned 24-bit scalar 10309 integer types __int24 and __uint24. 10310 * New command-line options -maccumulate-args, -mbranch-cost=cost and 10311 -mstrict-X were added to allow better fine-tuning of code 10312 optimization. 10313 * The command-line option -fdata-sections now also takes affect on 10314 the section names of variables with the progmem attribute. 10315 * A new inline assembler print modifier %i to print a RAM address as 10316 I/O address has been added: 10317 10318 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 10319 10320 void set_portb (uint8_t value) 10321 { 10322 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 10323 } 10324 10325 The offset between an I/O address and the RAM address for that I/O 10326 location is device-specific. This offset is taken into account when 10327 printing a RAM address with the %i modifier so that the address is 10328 suitable to be used as operand in an I/O command. The address must 10329 be a constant integer known at compile time. 10330 * The inline assembler constraint "R" to represent integers in the 10331 range 6 5 has been removed without replacement. 10332 * Many optimizations to: 10333 + 64-bit integer arithmetic 10334 + Widening multiplication 10335 + Integer division by a constant 10336 + Avoid constant reloading in multi-byte instructions. 10337 + Micro-optimizations for special instruction sequences. 10338 + Generic built-in functions like __builtin_ffs*, 10339 __builtin_clz*, etc. 10340 + If-else decision trees generated by switch instructions 10341 + Merging of data located in flash memory 10342 + New libgcc variants for devices with 8-bit wide stack pointer 10343 + 10344 * Better documentation: 10345 + Handling of EIND and indirect jumps on devices with more than 10346 128 KiB of program memory. 10347 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 10348 registers. 10349 + Function attributes OS_main and OS_task. 10350 + AVR-specific built-in macros. 10351 10352 C6X 10353 10354 * Support has been added for the Texas Instruments C6X family of 10355 processors. 10356 10357 CR16 10358 10359 * Support has been added for National Semiconductor's CR16 10360 architecture. 10361 10362 Epiphany 10363 10364 * Support has been added for Adapteva's Epiphany architecture. 10365 10366 IA-32/x86-64 10367 10368 * Support for Intel AVX2 intrinsics, built-in functions and code 10369 generation is available via -mavx2. 10370 * Support for Intel BMI2 intrinsics, built-in functions and code 10371 generation is available via -mbmi2. 10372 * Implementation and automatic generation of __builtin_clz* using the 10373 lzcnt instruction is available via -mlzcnt. 10374 * Support for Intel FMA3 intrinsics and code generation is available 10375 via -mfma. 10376 * A new -mfsgsbase command-line option is available that makes GCC 10377 generate new segment register read/write instructions through 10378 dedicated built-ins. 10379 * Support for the new Intel rdrnd instruction is available via 10380 -mrdrnd. 10381 * Two additional AVX vector conversion instructions are available via 10382 -mf16c. 10383 * Support for new Intel processor codename IvyBridge with RDRND, 10384 FSGSBASE and F16C is available through -march=core-avx-i. 10385 * Support for the new Intel processor codename Haswell with AVX2, 10386 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 10387 * Support for new AMD family 15h processors (Piledriver core) is now 10388 available through -march=bdver2 and -mtune=bdver2 options. 10389 * Support for [39]the x32 psABI is now available through the -mx32 10390 option. 10391 * Windows mingw targets are using the -mms-bitfields option by 10392 default. 10393 * Windows x86 targets are using the __thiscall calling convention for 10394 C++ class-member functions. 10395 * Support for the configure option --with-threads=posix for Windows 10396 mingw targets. 10397 10398 MIPS 10399 10400 * GCC now supports thread-local storage (TLS) for MIPS16. This 10401 requires GNU binutils 2.22 or later. 10402 * GCC can now generate code specifically for the Cavium Octeon+ and 10403 Octeon2 processors. The associated command-line options are 10404 -march=octeon+ and -march=octeon2 respectively. Both options 10405 require GNU binutils 2.22 or later. 10406 * GCC can now work around certain 24k errata, under the control of 10407 the command-line option -mfix-24k. These workarounds require GNU 10408 binutils 2.20 or later. 10409 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 10410 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 10411 toolchain that generates 32-bit code by default. Use the 10412 configure-time option --enable-targets=all to select these extra 10413 multilibs. 10414 * Passing -fno-delayed-branch now also stops the assembler from 10415 automatically filling delay slots. 10416 10417 PowerPC/PowerPC64 10418 10419 * Vectors of type vector long long or vector long are passed and 10420 returned using the same method as other vectors with the VSX 10421 instruction set. Previously GCC did not adhere to the ABI for 10422 128-bit vectors with 64-bit integer base types (PR 48857). This 10423 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 10424 * A new option -mno-pointers-to-nested-functions was added to allow 10425 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 10426 that the compiler should not load up the chain register (r11) 10427 before calling a function through a pointer. If you use this 10428 option, you cannot call nested functions through a pointer, or call 10429 other languages that might use the static chain. 10430 * A new option msave-toc-indirect was added to allow AIX 10431 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 10432 save the TOC in the prologue for indirect calls or generate the 10433 save inline. This can speed up some programs that call through a 10434 function pointer a lot, but it can slow down other functions that 10435 only call through a function pointer in exceptional cases. 10436 * The PowerPC port will now enable machine-specific built-in 10437 functions when the user switches the target machine using the 10438 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 10439 sequences. In addition, the target macros are updated. However, due 10440 to the way the -save-temps switch is implemented, you won't see the 10441 effect of these additional macros being defined in preprocessor 10442 output. 10443 10444 SH 10445 10446 * A new option -msoft-atomic has been added. When it is specified, 10447 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 10448 the new __atomic routines. 10449 * Since it is neither supported by GAS nor officially documented, 10450 code generation for little endian SH2A has been disabled. 10451 Specifying -ml with -m2a* will now result in a compiler error. 10452 * The defunct -mbranch-cost option has been fixed. 10453 * Some improvements to the generated code of: 10454 + Utilization of the tst #imm,R0 instruction. 10455 + Dynamic shift instructions on SH2A. 10456 + Integer absolute value calculations. 10457 * The -mdiv= option for targets other than SHmedia has been fixed and 10458 documented. 10459 10460 SPARC 10461 10462 * The option -mflat has been reinstated. When it is specified, the 10463 compiler will generate code for a single register window model. 10464 This is essentially a new implementation and the corresponding 10465 debugger support has been added to GDB 7.4. 10466 * Support for the options -mtune=native and -mcpu=native has been 10467 added on selected native platforms (GNU/Linux and Solaris). 10468 * Support for the SPARC T3 (Niagara 3) processor has been added. 10469 * VIS: 10470 + An intrinsics header visintrin.h has been added. 10471 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 10472 compare instructions have been added. 10473 + The little-endian version of alignaddr is now supported. 10474 + When possible, VIS builtins are marked const, which should 10475 increase the compiler's ability to optimize VIS operations. 10476 + The compiler now properly tracks the %gsr register and how it 10477 behaves as an input for various VIS instructions. 10478 + Akin to fzero, the compiler can now generate fone instructions 10479 in order to set all of the bits of a floating-point register 10480 to 1. 10481 + The documentation for the VIS intrinsics in the GCC manual has 10482 been brought up to date and many inaccuracies were fixed. 10483 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 10484 non-condition-code setting edge instructions have been added. 10485 Their availability is controlled by the new -mvis2 and 10486 -mno-vis2 options. They are enabled by default on 10487 UltraSPARC-III and later CPUs. 10488 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 10489 has been added. These instructions are enabled by default on SPARC 10490 T3 (Niagara 3) and later CPUs. 10491 10492 TILE-Gx/TILEPro 10493 10494 * Support has been added for the Tilera TILE-Gx and TILEPro families 10495 of processors. 10496 10497 Other significant improvements 10498 10499 * A new option (-grecord-gcc-switches) was added that appends 10500 compiler command-line options that might affect code generation to 10501 the DW_AT_producer attribute string in the DWARF debugging 10502 information. 10503 * GCC now supports various new GNU extensions to the DWARF debugging 10504 information format, like [40]entry value and [41]call site 10505 information, [42]typed DWARF stack or [43]a more compact macro 10506 representation. Support for these extensions has been added to GDB 10507 7.4. They can be disabled through the -gstrict-dwarf command-line 10508 option. 10509 10510 GCC 4.7.1 10511 10512 This is the [44]list of problem reports (PRs) from GCC's bug tracking 10513 system that are known to be fixed in the 4.7.1 release. This list might 10514 not be complete (that is, it is possible that some PRs that have been 10515 fixed are not listed here). 10516 10517 The Go front end in the 4.7.1 release fully supports the [45]Go 1 10518 language standard. 10519 10520 GCC 4.7.2 10521 10522 This is the [46]list of problem reports (PRs) from GCC's bug tracking 10523 system that are known to be fixed in the 4.7.2 release. This list might 10524 not be complete (that is, it is possible that some PRs that have been 10525 fixed are not listed here). 10526 10527 GCC 4.7.3 10528 10529 This is the [47]list of problem reports (PRs) from GCC's bug tracking 10530 system that are known to be fixed in the 4.7.3 release. This list might 10531 not be complete (that is, it is possible that some PRs that have been 10532 fixed are not listed here). 10533 10534 GCC 4.7.4 10535 10536 This is the [48]list of problem reports (PRs) from GCC's bug tracking 10537 system that are known to be fixed in the 4.7.4 release. This list might 10538 not be complete (that is, it is possible that some PRs that have been 10539 fixed are not listed here). 10540 10541 10542 For questions related to the use of GCC, please consult these web 10543 pages and the [49]GCC manuals. If that fails, the 10544 [50]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10545 web pages and the development of GCC are welcome on our developer 10546 list at [51]gcc (a] gcc.gnu.org. All of [52]our lists have public 10547 archives. 10548 10549 Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and 10550 distribution of this entire article is permitted in any medium, 10551 provided this notice is preserved. 10552 10553 These pages are [54]maintained by the GCC team. Last modified 10554 2023-03-29. 10555 10556 References 10557 10558 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 10559 2. http://savannah.nongnu.org/bugs/?35407 10560 3. https://gcc.gnu.org/PR18145 10561 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 10562 5. https://gcc.gnu.org/wiki/TransactionalMemory 10563 6. https://gcc.gnu.org/wiki/Atomic/GCCMM 10564 7. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10565 8. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10566 9. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10567 10. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10568 11. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10569 12. https://gcc.gnu.org/gcc-4.7/cxx0x_status.html 10570 13. https://gcc.gnu.org/PR14258 10571 14. https://gcc.gnu.org/PR35688 10572 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 10573 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 10574 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 10575 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 10576 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 10577 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 10578 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 10579 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 10580 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 10581 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 10582 25. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 10583 26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 10584 27. https://gcc.gnu.org/wiki/Fortran2003Status 10585 28. https://gcc.gnu.org/wiki/OOP 10586 29. https://gcc.gnu.org/wiki/Fortran2008Status 10587 30. https://gcc.gnu.org/wiki/Coarray 10588 31. https://gcc.gnu.org/wiki/CoarrayLib 10589 32. https://gcc.gnu.org/wiki/TS29113Status 10590 33. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 10591 34. https://go.dev/doc/go1 10592 35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 10593 36. http://www.nongnu.org/avr-libc/ 10594 37. https://gcc.gnu.org/PR54461 10595 38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html 10596 39. https://sites.google.com/site/x32abi/ 10597 40. https://dwarfstd.org/issues/100909.1.html 10598 41. https://dwarfstd.org/issues/100909.2.html 10599 42. https://dwarfstd.org/issues/140425.1.html 10600 43. https://dwarfstd.org/issues/110722.1.html 10601 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 10602 45. https://go.dev/doc/go1 10603 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 10604 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 10605 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 10606 49. https://gcc.gnu.org/onlinedocs/ 10607 50. mailto:gcc-help (a] gcc.gnu.org 10608 51. mailto:gcc (a] gcc.gnu.org 10609 52. https://gcc.gnu.org/lists.html 10610 53. https://www.fsf.org/ 10611 54. https://gcc.gnu.org/about.html 10612 ====================================================================== 10613 http://gcc.gnu.org/gcc-4.6/index.html 10614 GCC 4.6 Release Series 10615 10616 (This release series is no longer supported.) 10617 10618 April 12, 2013 10619 10620 The [1]GNU project and the GCC developers are pleased to announce the 10621 release of GCC 4.6.4. 10622 10623 This release is a bug-fix release, containing fixes for regressions in 10624 GCC 4.6.3 relative to previous releases of GCC. 10625 10626 Release History 10627 10628 GCC 4.6.4 10629 April 12, 2013 ([2]changes, [3]documentation) 10630 10631 GCC 4.6.3 10632 March 1, 2012 ([4]changes, [5]documentation) 10633 10634 GCC 4.6.2 10635 October 26, 2011 ([6]changes, [7]documentation) 10636 10637 GCC 4.6.1 10638 June 27, 2011 ([8]changes, [9]documentation) 10639 10640 GCC 4.6.0 10641 March 25, 2011 ([10]changes, [11]documentation) 10642 10643 References and Acknowledgements 10644 10645 GCC used to stand for the GNU C Compiler, but since the compiler 10646 supports several other languages aside from C, it now stands for the 10647 GNU Compiler Collection. 10648 10649 A list of [12]successful builds is updated as new information becomes 10650 available. 10651 10652 The GCC developers would like to thank the numerous people that have 10653 contributed new features, improvements, bug fixes, and other changes as 10654 well as test results to GCC. This [13]amazing group of volunteers is 10655 what makes GCC successful. 10656 10657 For additional information about GCC please refer to the [14]GCC 10658 project web site or contact the [15]GCC development mailing list. 10659 10660 To obtain GCC please use [16]our mirror sites or [17]our version 10661 control system. 10662 10663 10664 For questions related to the use of GCC, please consult these web 10665 pages and the [18]GCC manuals. If that fails, the 10666 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10667 web pages and the development of GCC are welcome on our developer 10668 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 10669 archives. 10670 10671 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 10672 distribution of this entire article is permitted in any medium, 10673 provided this notice is preserved. 10674 10675 These pages are [23]maintained by the GCC team. Last modified 10676 2022-10-26. 10677 10678 References 10679 10680 1. http://www.gnu.org/ 10681 2. https://gcc.gnu.org/gcc-4.6/changes.html 10682 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 10683 4. https://gcc.gnu.org/gcc-4.6/changes.html 10684 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 10685 6. https://gcc.gnu.org/gcc-4.6/changes.html 10686 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 10687 8. https://gcc.gnu.org/gcc-4.6/changes.html 10688 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 10689 10. https://gcc.gnu.org/gcc-4.6/changes.html 10690 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 10691 12. https://gcc.gnu.org/gcc-4.6/buildstat.html 10692 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10693 14. https://gcc.gnu.org/index.html 10694 15. mailto:gcc (a] gcc.gnu.org 10695 16. https://gcc.gnu.org/mirrors.html 10696 17. https://gcc.gnu.org/git.html 10697 18. https://gcc.gnu.org/onlinedocs/ 10698 19. mailto:gcc-help (a] gcc.gnu.org 10699 20. mailto:gcc (a] gcc.gnu.org 10700 21. https://gcc.gnu.org/lists.html 10701 22. https://www.fsf.org/ 10702 23. https://gcc.gnu.org/about.html 10703 ====================================================================== 10704 http://gcc.gnu.org/gcc-4.6/changes.html 10705 GCC 4.6 Release Series 10706 Changes, New Features, and Fixes 10707 10708 Caveats 10709 10710 * The options -b <machine> and -V <version> have been removed because 10711 they were unreliable. Instead, users should directly run 10712 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 10713 run a different version of gcc. 10714 * GCC now has stricter checks for invalid command-line options. In 10715 particular, when gcc was called to link object files rather than 10716 compile source code, it would previously accept and ignore all 10717 options starting with --, including linker options such as 10718 --as-needed and --export-dynamic, although such options would 10719 result in errors if any source code was compiled. Such options, if 10720 unknown to the compiler, are now rejected in all cases; if the 10721 intent was to pass them to the linker, options such as 10722 -Wl,--as-needed should be used. 10723 * Versions of the GNU C library up to and including 2.11.1 included 10724 an [1]incorrect implementation of the cproj function. GCC optimizes 10725 its builtin cproj according to the behavior specified and allowed 10726 by the ISO C99 standard. If you want to avoid discrepancies between 10727 the C library and GCC's builtin transformations when using cproj in 10728 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 10729 and actually rely on the incorrect behavior of cproj, then you can 10730 disable GCC's transformations using -fno-builtin-cproj. 10731 * The C-only intermodule optimization framework (IMA, enabled by 10732 -combine) has been removed in favor of the new generic link-time 10733 optimization framework (LTO) introduced in [2]GCC 4.5.0. 10734 * GCC now ships with the LGPL-licensed libquadmath library, which 10735 provides quad-precision mathematical functions for targets with a 10736 __float128 datatype. __float128 is available for targets on 32-bit 10737 x86, x86-64 and Itanium architectures. The libquadmath library is 10738 automatically built on such targets when building the Fortran 10739 compiler. 10740 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 10741 warnings were added for C, C++, Objective-C and Objective-C++. 10742 These warnings diagnose variables respective parameters which are 10743 only set in the code and never otherwise used. Usually such 10744 variables are useless and often even the value assigned to them is 10745 computed needlessly, sometimes expensively. The 10746 -Wunused-but-set-variable warning is enabled by default by -Wall 10747 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 10748 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 10749 rules for the layout of vectors that could lead to wrong code being 10750 generated. Vectors larger than 8 bytes in size are now by default 10751 aligned to an 8-byte boundary. This is an ABI change: code that 10752 makes explicit use of vector types may be incompatible with binary 10753 objects built with older versions of GCC. Auto-vectorized code is 10754 not affected by this change. (This change affects GCC versions 10755 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 10756 * On AVR, variables with the progmem attribute to locate data in 10757 flash memory must be qualified as const. 10758 * Support for a number of older systems and recently unmaintained or 10759 untested target ports of GCC has been declared obsolete in GCC 4.6. 10760 Unless there is activity to revive them, the next release of GCC 10761 will have their sources permanently removed. 10762 All GCC ports for the following processor architectures have been 10763 declared obsolete: 10764 + Argonaut ARC (arc-*) 10765 + National Semiconductor CRX (crx-*) 10766 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 10767 m68hc12-*-*, m6812-*-*) 10768 + Sunplus S+core (score-*) 10769 The following ports for individual systems on particular 10770 architectures have been obsoleted: 10771 + Interix (i[34567]86-*-interix3*) 10772 + NetWare x86 (i[3456x]86-*-netware*) 10773 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 10774 + MCore PE (mcore-*-pe*) 10775 + SH SymbianOS (sh*-*-symbianelf*) 10776 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 10777 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 10778 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 10779 vax-*-netbsd*, but not *-*-netbsdelf*) 10780 The i[34567]86-*-pe alias for Cygwin targets has also been 10781 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 10782 Certain configure options to control the set of libraries built 10783 with GCC on some targets have been obsoleted. On ARM targets, the 10784 options --disable-fpu, --disable-26bit, --disable-underscore, 10785 --disable-interwork, --disable-biendian and --disable-nofmult have 10786 been obsoleted. On MIPS targets, the options 10787 --disable-single-float, --disable-biendian and --disable-softfloat 10788 have been obsoleted. 10789 * Support has been removed for all the [3]configurations obsoleted in 10790 GCC 4.5. 10791 * More information on porting to GCC 4.6 from previous versions of 10792 GCC can be found in the [4]porting guide for this release. 10793 10794 General Optimizer Improvements 10795 10796 * A new general optimization level, -Ofast, has been introduced. It 10797 combines the existing optimization level -O3 with options that can 10798 affect standards compliance but result in better optimized code. 10799 For example, -Ofast enables -ffast-math. 10800 * Link-time optimization improvements: 10801 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 10802 stabilized to the point of being usable. It has become the 10803 default mode when using the LTO optimization model. Link time 10804 optimization can now split itself into multiple parallel 10805 compilations. Parallelism is controlled with -flto=n (where n 10806 specifies the number of compilations to execute in parallel). 10807 GCC can also cooperate with a GNU make job server by 10808 specifying the -flto=jobserver option and adding + to the 10809 beginning of the Makefile rule executing the linker. 10810 Classical LTO mode can be enforced by -flto-partition=none. 10811 This may result in small code quality improvements. 10812 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 10813 and other large applications can be built with LTO enabled. 10814 + The linker plugin support improvements 10815 o Linker plugin is now enabled by default when the linker 10816 is detected to have plugin support. This is the case for 10817 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 10818 the Gold linker on ELF targets. Plugin support of the 10819 Apple linker on Darwin is not compatible with GCC. The 10820 linker plugin can also be controlled by the 10821 -fuse-linker-plugin command-line option. 10822 o Resolution information from the linker plugin is used to 10823 drive whole program assumptions. Use of the linker plugin 10824 results in more aggressive optimization on binaries and 10825 on shared libraries that use the hidden visibility 10826 attribute. Consequently the use of -fwhole-program is not 10827 necessary in addition to LTO. 10828 + Hidden symbols used from non-LTO objects now have to be 10829 explicitly annotated with externally_visible when the linker 10830 plugin is not used. 10831 + C++ inline functions and virtual tables are now privatized 10832 more aggressively, leading to better inter-procedural 10833 optimization and faster dynamic linking. 10834 + Memory usage and intermediate language streaming performance 10835 have been improved. 10836 + Static constructors and destructors from individual units are 10837 inlined into a single function. This can significantly improve 10838 startup times of large C++ applications where static 10839 constructors are very common. For example, static constructors 10840 are used when including the iostream header. 10841 + Support for the Ada language has been added. 10842 * Interprocedural optimization improvements 10843 + The interprocedural framework was re-tuned for link time 10844 optimization. Several scalability issues were resolved. 10845 + Improved auto-detection of const and pure functions. Newly, 10846 noreturn functions are auto-detected. 10847 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 10848 available that informs users when adding attributes to headers 10849 might improve code generation. 10850 + A number of inlining heuristic improvements. In particular: 10851 o Partial inlining is now supported and enabled by default 10852 at -O2 and greater. The feature can be controlled via 10853 -fpartial-inlining. 10854 Partial inlining splits functions with short hot path to 10855 return. This allows more aggressive inlining of the hot 10856 path leading to better performance and often to code size 10857 reductions (because cold parts of functions are not 10858 duplicated). 10859 o Scalability for large compilation units was improved 10860 significantly. 10861 o Inlining of callbacks is now more aggressive. 10862 o Virtual methods are considered for inlining when the 10863 caller is inlined and devirtualization is then possible. 10864 o Inlining when optimizing for size (either in cold regions 10865 of a program or when compiling with -Os) was improved to 10866 better handle C++ programs with larger abstraction 10867 penalty, leading to smaller and faster code. 10868 + The IPA reference optimization pass detecting global variables 10869 used or modified by functions was strengthened and sped up. 10870 + Functions whose address was taken are now optimized out when 10871 all references to them are dead. 10872 + A new inter-procedural static profile estimation pass detects 10873 functions that are executed once or unlikely to be executed. 10874 Unlikely executed functions are optimized for size. Functions 10875 executed once are optimized for size except for the inner 10876 loops. 10877 + On most targets with named section support, functions used 10878 only at startup (static constructors and main), functions used 10879 only at exit and functions detected to be cold are placed into 10880 separate text segment subsections. This extends the 10881 -freorder-functions feature and is controlled by the same 10882 switch. The goal is to improve the startup time of large C++ 10883 programs. 10884 Proper function placement requires linker support. GNU ld 10885 2.21.51 on ELF targets was updated to place those functions 10886 together within the text section leading to better code 10887 locality and faster startup times of large C++ programs. The 10888 feature is also supported in the Apple linker. Support in the 10889 gold linker is planned. 10890 * A new switch -fstack-usage has been added. It makes the compiler 10891 output stack usage information for the program, on a per-function 10892 basis, in an auxiliary file. 10893 * A new switch -fcombine-stack-adjustments has been added. It can be 10894 used to enable or disable the compiler's stack-slot combining pass 10895 which before was enabled automatically at -O1 and above, but could 10896 not be controlled on its own. 10897 * A new switch -fstrict-volatile-bitfields has been added. Using it 10898 indicates that accesses to volatile bitfields should use a single 10899 access of the width of the field's type. This option can be useful 10900 for precisely defining and accessing memory-mapped peripheral 10901 registers from C or C++. 10902 10903 Compile time and memory usage improvements 10904 10905 * Datastructures used by the dataflow framework in GCC were 10906 reorganized for better memory usage and more cache locality. 10907 Compile time is improved especially on units with large functions 10908 (possibly resulting from a lot of inlining) not fitting into the 10909 processor cache. The compile time of the GCC C compiler binary with 10910 link-time optimization went down by over 10% (benchmarked on x86-64 10911 target). 10912 10913 New Languages and Language specific improvements 10914 10915 Ada 10916 10917 * Stack checking has been improved on selected architectures (Alpha, 10918 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 10919 overflows in all cases on these architectures. 10920 * Initial support for Ada 2012 has been added. 10921 10922 C family 10923 10924 * A new warning, enabled by -Wdouble-promotion, has been added that 10925 warns about cases where a value of type float is implicitly 10926 promoted to double. This is especially helpful for CPUs that handle 10927 the former in hardware, but emulate the latter in software. 10928 * A new function attribute leaf was introduced. This attribute allows 10929 better inter-procedural optimization across calls to functions that 10930 return to the current unit only via returning or exception 10931 handling. This is the case for most library functions that have no 10932 callbacks. 10933 * Support for a new data type __int128 for targets having wide enough 10934 machine-mode support. 10935 * The new function attribute callee_pop_aggregate allows to specify 10936 if the caller or callee is responsible for popping the aggregate 10937 return pointer value from the stack. 10938 * Support for selectively enabling and disabling warnings via #pragma 10939 GCC diagnostic has been added. For instance: 10940 #pragma GCC diagnostic error "-Wuninitialized" 10941 foo(a); /* error is given for this one */ 10942 #pragma GCC diagnostic push 10943 #pragma GCC diagnostic ignored "-Wuninitialized" 10944 foo(b); /* no diagnostic for this one */ 10945 #pragma GCC diagnostic pop 10946 foo(c); /* error is given for this one */ 10947 #pragma GCC diagnostic pop 10948 foo(d); /* depends on command-line options */ 10949 10950 * The -fmax-errors=N option is now supported. Using this option 10951 causes the compiler to exit after N errors have been issued. 10952 10953 C 10954 10955 * There is now experimental support for some features from the 10956 upcoming C1X revision of the ISO C standard. This support may be 10957 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 10958 Note that this support is experimental and may change incompatibly 10959 in future releases for consistency with changes to the C1X standard 10960 draft. The following features are newly supported as described in 10961 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 10962 meeting); some other features were already supported with no 10963 compiler changes being needed, or have some support but not in full 10964 accord with N1539 (as amended). 10965 + Static assertions (_Static_assert keyword) 10966 + Typedef redefinition 10967 + New macros in <float.h> 10968 + Anonymous structures and unions 10969 * The new -fplan9-extensions option directs the compiler to support 10970 some extensions for anonymous struct fields which are implemented 10971 by the Plan 9 compiler. A pointer to a struct may be automatically 10972 converted to a pointer to an anonymous field when calling a 10973 function, in order to make the types match. An anonymous struct 10974 field whose type is a typedef name may be referred to using the 10975 typedef name. 10976 10977 C++ 10978 10979 * Improved [7]experimental support for the upcoming C++0x ISO C++ 10980 standard, including support for constexpr (thanks to Gabriel Dos 10981 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 10982 noexcept, unrestricted unions, range-based for loops (thanks to 10983 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 10984 Rodrigo), implicitly deleted functions and implicit move 10985 constructors. 10986 * When an extern declaration within a function does not match a 10987 declaration in the enclosing context, G++ now properly declares the 10988 name within the namespace of the function rather than the namespace 10989 which was open just before the function definition ([8]c++/43145). 10990 * GCC now warns by default when casting integers to larger pointer 10991 types. These warnings can be disabled with the option 10992 -Wno-int-to-pointer-cast, which is now also available in C++. 10993 * G++ no longer optimizes using the assumption that a value of 10994 enumeration type will fall within the range specified by the 10995 standard, since that assumption is easily violated with a 10996 conversion from integer type ([9]c++/43680). The old behavior can 10997 be restored with -fstrict-enums. 10998 * The new -fnothrow-opt flag changes the semantics of a throw() 10999 exception specification to match the proposed semantics of the 11000 noexcept specification: just call terminate if an exception tries 11001 to propagate out of a function with such an exception 11002 specification. This dramatically reduces or eliminates the code 11003 size overhead from adding the exception specification. 11004 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 11005 a function that the compiler can tell doesn't throw if it would 11006 change the value of a noexcept expression. 11007 * The -Wshadow option now warns if a local variable or type 11008 declaration shadows another type in C++. Note that the compiler 11009 will not warn if a local variable shadows a struct/class/enum, but 11010 will warn if it shadows an explicit typedef. 11011 * When an identifier is not found in the current scope, G++ now 11012 offers suggestions about which identifier might have been intended. 11013 * G++ now issues clearer diagnostics for missing semicolons after 11014 class, struct, and union definitions. 11015 * G++ now issues clearer diagnostics for missing semicolons after 11016 class member declarations. 11017 * G++ now issues clearer diagnostics when a colon is used in a place 11018 where a double-colon was intended. 11019 * G++ no longer accepts mutable on reference members ([10]c++/33558). 11020 Use -fpermissive to allow the old, non-conforming behaviour. 11021 * A few mangling fixes have been made, to attribute const/volatile on 11022 function pointer types, decltype of a plain decl, and use of a 11023 function parameter in the declaration of another parameter. By 11024 default the compiler still uses the old mangling, but emits aliases 11025 with the new mangling on targets that support strong aliases. Users 11026 can switch over entirely to the new mangling with -fabi-version=5 11027 or -fabi-version=0. -Wabi will now warn about code that uses the 11028 old mangling. 11029 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 11030 type to be default initialized unless the type has a user-declared 11031 default constructor. In 4.6.2 G++ implements the proposed 11032 resolution of [11]DR 253, so default initialization is allowed if 11033 it initializes all subobjects. Code that fails to compile can be 11034 fixed by providing an initializer e.g. 11035 struct A { A(); }; 11036 struct B : A { int i; }; 11037 const B b = B(); 11038 Use -fpermissive to allow the old, non-conforming behaviour. 11039 11040 Runtime Library (libstdc++) 11041 11042 * [12]Improved experimental support for the upcoming ISO C++ 11043 standard, C++0x, including using constexpr and nullptr. 11044 * Performance improvements to the [13]Debug Mode, thanks to Franois 11045 Dumont. 11046 * Atomic operations used for reference-counting are annotated so that 11047 they can be understood by race detectors such as Helgrind, see 11048 [14]Data Race Hunting. 11049 * Most libstdc++ standard headers have been changed to no longer 11050 include the cstddef header as an implementation detail. Code that 11051 relied on that header being included as side-effect of including 11052 other standard headers will need to include cstddef explicitly. 11053 11054 Fortran 11055 11056 * On systems supporting the libquadmath library, GNU Fortran now also 11057 supports a quad-precision, kind=16 floating-point data type 11058 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 11059 hardware, calculations might be one to two orders of magnitude 11060 slower than with the 4, 8 or 10 bytes floating-point data types. 11061 This change does not affect systems which support REAL(16) in 11062 hardware nor those which do not support libquadmath. 11063 * Much improved compile time for large array constructors. 11064 * In order to reduce execution time and memory consumption, use of 11065 temporary arrays in assignment expressions is avoided for many 11066 cases. The compiler now reverses loops in order to avoid generating 11067 a temporary array where possible. 11068 * Improved diagnostics, especially with -fwhole-file. 11069 * The -fwhole-file flag is now enabled by default. This improves code 11070 generation and diagnostics. It can be disabled using the deprecated 11071 -fno-whole-file flag. 11072 * Support the generation of Makefile dependencies via the [15]-M... 11073 flags of GCC; you may need to specify the -cpp option in addition. 11074 The dependencies take modules, Fortran's include, and CPP's 11075 #include into account. Note: Using -M for the module path is no 11076 longer supported, use -J instead. 11077 * The flag -Wconversion has been modified to only issue warnings 11078 where a conversion leads to information loss. This drastically 11079 reduces the number of warnings; -Wconversion is thus now enabled 11080 with -Wall. The flag -Wconversion-extra has been added and also 11081 warns about other conversions; -Wconversion-extra typically issues 11082 a huge number of warnings, most of which can be ignored. 11083 * A new command-line option -Wunused-dummy-argument warns about 11084 unused dummy arguments and is included in -Wall. Before, 11085 -Wunused-variable also warned about unused dummy arguments. 11086 * Fortran 2003 support has been extended: 11087 + Improved support for polymorphism between libraries and 11088 programs and for complicated inheritance patterns (cf. 11089 [16]object-oriented programming). 11090 + Experimental support of the ASSOCIATE construct. 11091 + In pointer assignments it is now possible to specify the lower 11092 bounds of the pointer and, for a rank-1 or a simply contiguous 11093 data-target, to remap the bounds. 11094 + Automatic (re)allocation: In intrinsic assignments to 11095 allocatable variables the left-hand side will be automatically 11096 allocated (if unallocated) or reallocated (if the shape or 11097 type parameter is different). To avoid the small performance 11098 penalty, you can use a(:) = ... instead of a = ... for arrays 11099 and character strings or disable the feature using -std=f95 11100 or -fno-realloc-lhs. 11101 + Deferred type parameter: For scalar allocatable and pointer 11102 variables the character length can be deferred. 11103 + Namelist variables with allocatable and pointer attribute and 11104 nonconstant length type parameter are supported. 11105 * Fortran 2008 support has been extended: 11106 + Experimental [17]coarray support (for one image only, i.e. 11107 num_images() == 1); use the [18]-fcoarray=single flag to 11108 enable it. 11109 + The STOP and the new ERROR STOP statements now support all 11110 constant expressions. 11111 + Support for the CONTIGUOUS attribute. 11112 + Support for ALLOCATE with MOLD. 11113 + Support for the STORAGE_SIZE intrinsic inquiry function. 11114 + Support of the NORM2 and PARITY intrinsic functions. 11115 + The following bit intrinsics were added: POPCNT and POPPAR for 11116 counting the number of 1 bits and returning the parity; BGE, 11117 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 11118 for combined left and right shifts, MASKL and MASKR for simple 11119 left and right justified masks, MERGE_BITS for a bitwise merge 11120 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 11121 and the transformational bit intrinsics IALL, IANY and 11122 IPARITY. 11123 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 11124 + Support for the IMPURE attribute for procedures, which allows 11125 for ELEMENTAL procedures without the restrictions of PURE. 11126 + Null pointers (including NULL()) and not allocated variables 11127 can be used as actual argument to optional non-pointer, 11128 non-allocatable dummy arguments, denoting an absent argument. 11129 + Non-pointer variables with TARGET attribute can be used as 11130 actual argument to POINTER dummies with INTENT(IN) 11131 + Pointers including procedure pointers and those in a derived 11132 type (pointer components) can now be initialized by a target 11133 instead of only by NULL. 11134 + The EXIT statement (with construct-name) can now be used to 11135 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 11136 SELECT CASE and SELECT TYPE constructs. 11137 + Internal procedures can now be used as actual argument. 11138 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 11139 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 11140 have been added; these arrays contain the supported kind 11141 values for the respective types. 11142 + The module procedures C_SIZEOF of the intrinsic module 11143 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 11144 ISO_FORTRAN_ENV have been implemented. 11145 + Minor changes: obsolescence diagnostics for ENTRY was added 11146 for -std=f2008; a line may start with a semicolon; for 11147 internal and module procedures END can be used instead of END 11148 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 11149 a RADIX argument; intrinsic types are supported for 11150 TYPE(intrinsic-type-spec); multiple type-bound procedures can 11151 be declared in a single PROCEDURE statement; implied-shape 11152 arrays are supported for named constants (PARAMETER). The 11153 transformational, three argument versions of BESSEL_JN and 11154 BESSEL_YN were added the elemental, two-argument version had 11155 been added in GCC 4.4; note that the transformational 11156 functions use a recurrence algorithm. 11157 11158 Go 11159 11160 Support for the Go programming language has been added to GCC. It is 11161 not enabled by default when you build GCC; use the --enable-languages 11162 configure option to build it. The driver program for compiling Go code 11163 is gccgo. 11164 11165 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 11166 is in progress. It may or may not work on other platforms. 11167 11168 Objective-C and Objective-C++ 11169 11170 * The -fobjc-exceptions flag is now required to enable Objective-C 11171 exception and synchronization syntax (introduced by the keywords 11172 @try, @catch, @finally and @synchronized). 11173 * A number of Objective-C 2.0 features and extensions are now 11174 supported by GCC. These features are enabled by default; you can 11175 disable them by using the new -fobjc-std=objc1 command-line option. 11176 * The Objective-C 2.0 dot-syntax is now supported. It is an 11177 alternative syntax for using getters and setters; object.count is 11178 automatically converted into [object count] or [object setCount: 11179 ...] depending on context; for example if (object.count > 0) is 11180 automatically compiled into the equivalent of if ([object count] > 11181 0) while object.count = 0; is automatically compiled into the 11182 equivalent ot [object setCount: 0];. The dot-syntax can be used 11183 with instance and class objects and with any setters or getters, no 11184 matter if they are part of a declared property or not. 11185 * Objective-C 2.0 declared properties are now supported. They are 11186 declared using the new @property keyword, and are most commonly 11187 used in conjunction with the new Objective-C 2.0 dot-syntax. The 11188 nonatomic, readonly, readwrite, assign, retain, copy, setter and 11189 getter attributes are all supported. Marking declared properties 11190 with __attribute__ ((deprecated)) is supported too. 11191 * The Objective-C 2.0 @synthesize and @dynamic keywords are 11192 supported. @synthesize causes the compiler to automatically 11193 synthesize a declared property, while @dynamic is used to disable 11194 all warnings for a declared property for which no implementation is 11195 provided at compile time. Synthesizing declared properties requires 11196 runtime support in most useful cases; to be able to use it with the 11197 GNU runtime, appropriate helper functions have been added to the 11198 GNU Objective-C runtime ABI, and are implemented by the GNU 11199 Objective-C runtime library shipped with GCC. 11200 * The Objective-C 2.0 fast enumeration syntax is supported in 11201 Objective-C. This is currently not yet available in Objective-C++. 11202 Fast enumeration requires support in the runtime, and such support 11203 has been added to the GNU Objective-C runtime library (shipped with 11204 GCC). 11205 * The Objective-C 2.0 @optional keyword is supported. It allows you 11206 to mark methods or properties in a protocol as optional as opposed 11207 to required. 11208 * The Objective-C 2.0 @package keyword is supported. It has currently 11209 the same effect as the @public keyword. 11210 * Objective-C 2.0 method attributes are supported. Currently the 11211 supported attributes are deprecated, sentinel, noreturn and format. 11212 * Objective-C 2.0 method argument attributes are supported. The most 11213 widely used attribute is unused, to mark an argument as unused in 11214 the implementation. 11215 * Objective-C 2.0 class and protocol attributes are supported. 11216 Currently the only supported attribute is deprecated. 11217 * Objective-C 2.0 class extensions are supported. A class extension 11218 has the same syntax as a category declaration with no category 11219 name, and the methods and properties declared in it are added 11220 directly to the main class. It is mostly used as an alternative to 11221 a category to add methods to a class without advertising them in 11222 the public headers, with the advantage that for class extensions 11223 the compiler checks that all the privately declared methods are 11224 actually implemented. 11225 * As a result of these enhancements, GCC can now be used to build 11226 Objective-C and Objective-C++ software that uses Foundation and 11227 other important system frameworks with the NeXT runtime on Darwin 9 11228 and Darwin 10 (OSX 10.5 and 10.6). 11229 * Many bugs in the compiler have been fixed in this release; in 11230 particular, LTO can now be used when compiling Objective-C and 11231 Objective-C++ and the parser is much more robust in dealing with 11232 invalid code. 11233 11234 Runtime Library (libobjc) 11235 11236 * The GNU Objective-C runtime library now defines the macro 11237 __GNU_LIBOBJC__ (with a value that is increased at every release 11238 where there is any change to the API) in objc/objc.h, making it 11239 easy to determine if the GNU Objective-C runtime library is being 11240 used, and if so, which version. Previous versions of the GNU 11241 Objective-C runtime library (and other Objective-C runtime 11242 libraries such as the Apple one) do not define this macro. 11243 * A new Objective-C 2.0 API, almost identical to the one implemented 11244 by the Apple Objective-C runtime, has been implemented in the GNU 11245 Objective-C runtime library. The new API hides the internals of 11246 most runtime structures but provides a more extensive set of 11247 functions to operate on them. It is much easier, for example, to 11248 create or modify classes at runtime. The new API also makes it 11249 easier to port software from Apple to GNU as almost no changes 11250 should be required. The old API is still supported for backwards 11251 compatibility; including the old objc/objc-api.h header file 11252 automatically selects the old API, while including the new 11253 objc/runtime.h header file automatically selects the new API. 11254 Support for the old API is being phased out and upgrading the 11255 software to use the new API is strongly recommended. To check for 11256 the availability of the new API, the __GNU_LIBOBJC__ macro can be 11257 used as older versions of the GNU Objective-C runtime library, 11258 which do not support the new API, do not define such a macro. 11259 * Runtime support for @synchronized has been added. 11260 * Runtime support for Objective-C 2.0 synthesized property accessors 11261 has been added. 11262 * Runtime support for Objective-C 2.0 fast enumeration has been 11263 added. 11264 11265 New Targets and Target Specific Improvements 11266 11267 ARM 11268 11269 * GCC now supports the Cortex-M4 processor implementing the v7-em 11270 version of the architecture using the option -mcpu=cortex-m4. 11271 * Scheduling descriptions for the Cortex-M4, the Neon and the 11272 floating point units of the Cortex-A9 and a pipeline description 11273 for the Cortex-A5 have been added. 11274 * Synchronization primitives such as __sync_fetch_and_add and friends 11275 are now inlined for supported architectures rather than calling 11276 into a kernel helper function. 11277 * SSA loop prefetching is enabled by default for the Cortex-A9 at 11278 -O3. 11279 * Several improvements were committed to improve code generation for 11280 the ARM architecture including a rewritten implementation for load 11281 and store multiples. 11282 * Several enhancements were committed to improve SIMD code generation 11283 for NEON by adding support for widening instructions, misaligned 11284 loads and stores, vector conditionals and support for 64-bit 11285 arithmetic. 11286 * Support was added for the Faraday cores fa526, fa606te, fa626te, 11287 fmp626te, fmp626 and fa726te and can be used with the respective 11288 names as parameters to the -mcpu= option. 11289 * Basic support was added for Cortex-A15 and is available through 11290 -mcpu=cortex-a15. 11291 * GCC for AAPCS configurations now more closely adheres to the AAPCS 11292 specification by enabling -fstrict-volatile-bitfields by default. 11293 11294 IA-32/x86-64 11295 11296 * The new -fsplit-stack option permits programs to use a 11297 discontiguous stack. This is useful for threaded programs, in that 11298 it is no longer necessary to specify the maximum stack size when 11299 creating a thread. This feature is currently only implemented for 11300 32-bit and 64-bit x86 GNU/Linux targets. 11301 * Support for emitting profiler counter calls before function 11302 prologues. This is enabled via a new command-line option -mfentry. 11303 * Optimization for the Intel Core 2 processors is now available 11304 through the -march=core2 and -mtune=core2 options. 11305 * Support for Intel Core i3/i5/i7 processors is now available through 11306 the -march=corei7 and -mtune=corei7 options. 11307 * Support for Intel Core i3/i5/i7 processors with AVX is now 11308 available through the -march=corei7-avx and -mtune=corei7-avx 11309 options. 11310 * Support for AMD Bobcat (family 14) processors is now available 11311 through the -march=btver1 and -mtune=btver1 options. 11312 * Support for AMD Bulldozer (family 15) processors is now available 11313 through the -march=bdver1 and -mtune=bdver1 options. 11314 * The default setting (when not optimizing for size) for 32-bit 11315 GNU/Linux and Darwin x86 targets has been changed to 11316 -fomit-frame-pointer. The default can be reverted to 11317 -fno-omit-frame-pointer by configuring GCC with the 11318 --enable-frame-pointer configure option. 11319 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 11320 __float128 on 32-bit and 64-bit x86 targets. 11321 * AVX floating-point arithmetic can now be enabled by default at 11322 configure time with the new --with-fpmath=avx option. 11323 * The SSA loop prefetching pass is enabled when using -O3 when 11324 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 11325 than K6). 11326 * Support for TBM (Trailing Bit Manipulation) built-in functions and 11327 code generation is available via -mtbm. 11328 * Support for AMD's BMI (Bit Manipulation) built-in functions and 11329 code generation is available via -mbmi. 11330 11331 MicroBlaze 11332 11333 * Support has been added for the Xilinx MicroBlaze softcore processor 11334 (microblaze-elf) embedded target. This configurable processor is 11335 supported on several Xilinx Spartan and Virtex FPGAs. 11336 11337 MIPS 11338 11339 * GCC now supports the Loongson 3A processor. Its canonical -march= 11340 and -mtune= name is loongson3a. 11341 11342 MN10300 / AM33 11343 11344 * The inline assembly register constraint "A" has been renamed "c". 11345 This constraint is used to select a floating-point register that 11346 can be used as the destination of a multiply-accumulate 11347 instruction. 11348 * New inline assembly register constraints "A" and "D" have been 11349 added. These constraint letters resolve to all general registers 11350 when compiling for AM33, and resolve to address registers only or 11351 data registers only when compiling for MN10300. 11352 * The MDR register is represented in the compiler. One can access the 11353 register via the "z" constraint in inline assembly. It can be 11354 marked as clobbered or used as a local register variable via the 11355 "mdr" name. The compiler uses the RETF instruction if the function 11356 does not modify the MDR register, so it is important that inline 11357 assembly properly annotate any usage of the register. 11358 11359 PowerPC/PowerPC64 11360 11361 * GCC now supports the Applied Micro Titan processor with 11362 -mcpu=titan. 11363 * The -mrecip option has been added, which indicates whether the 11364 reciprocal and reciprocal square root instructions should be used. 11365 * The -mveclibabi=mass option can be used to enable the compiler to 11366 autovectorize mathematical functions using the Mathematical 11367 Acceleration Subsystem library. 11368 * The -msingle-pic-base option has been added, which instructs the 11369 compiler to avoid loading the PIC base register in function 11370 prologues. The PIC base register must be initialized by the runtime 11371 system. 11372 * The -mblock-move-inline-limit option has been added, which enables 11373 the user to control the maximum size of inlined memcpy calls and 11374 similar. 11375 * PowerPC64 GNU/Linux support for applications requiring a large TOC 11376 section has been improved. A new command-line option, 11377 -mcmodel=MODEL, controls this feature; valid values for MODEL are 11378 small, medium, or large. 11379 * The Altivec builtin functions vec_ld and vec_st have been modified 11380 to generate the Altivec memory instructions LVX and STVX, even if 11381 the -mvsx option is used. In the initial GCC 4.5 release, these 11382 builtin functions were changed to generate VSX memory reference 11383 instructions instead of Altivec memory instructions, but there are 11384 differences between the two instructions. If the VSX instruction 11385 set is available, you can now use the new builtin functions 11386 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 11387 instructions. 11388 * The GCC compiler on AIX now defaults to a process layout with a 11389 larger data space allowing larger programs to be compiled. 11390 * The GCC long double type on AIX 6.1 and above has reverted to 11391 64-bit double precision, matching the AIX XL compiler default, 11392 because of missing C99 symbols required by the GCC runtime. 11393 * The default processor scheduling model and tuning for PowerPC64 11394 GNU/Linux and for AIX 6.1 and above now is POWER7. 11395 * Starting with GCC 4.6.1, vectors of type vector long long or vector 11396 long are passed and returned in the same method as other vectors 11397 with the VSX instruction set. Previously the GCC compiler did not 11398 adhere to the ABI for 128-bit vectors with 64-bit integer base 11399 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 11400 11401 S/390, zSeries and System z9/z10, IBM zEnterprise z196 11402 11403 * Support for the zEnterprise z196 processor has been added. When 11404 using the -march=z196 option, the compiler will generate code 11405 making use of the following instruction facilities: 11406 + Conditional load/store 11407 + Distinct-operands 11408 + Floating-point-extension 11409 + Interlocked-access 11410 + Population-count 11411 The -mtune=z196 option avoids the compare and branch instructions 11412 as well as the load address instruction with an index register as 11413 much as possible and performs instruction scheduling appropriate 11414 for the new out-of-order pipeline architecture. 11415 * When using the -m31 -mzarch options the generated code still 11416 conforms to the 32-bit ABI but uses the general purpose registers 11417 as 64-bit registers internally. This requires a Linux kernel saving 11418 the whole 64-bit registers when doing a context switch. Kernels 11419 providing that feature indicate that by the 'highgprs' string in 11420 /proc/cpuinfo. 11421 * The SSA loop prefetching pass is enabled when using -O3. 11422 11423 SPARC 11424 11425 * GCC now supports the LEON series of SPARC V8 processors. The code 11426 generated by the compiler can either be tuned to it by means of the 11427 --with-tune=leon configure option and -mtune=leon compilation 11428 option, or the compiler can be built for the sparc-leon-{elf,linux} 11429 and sparc-leon3-{elf,linux} targets directly. 11430 * GCC has stopped sign/zero-extending parameter registers in the 11431 callee for functions taking parameters with sub-word size in 32-bit 11432 mode, since this is redundant with the specification of the ABI. 11433 GCC has never done so in 64-bit mode since this is also redundant. 11434 * The command-line option -mfix-at697f has been added to enable the 11435 documented workaround for the single erratum of the Atmel AT697F 11436 processor. 11437 11438 Operating Systems 11439 11440 Android 11441 11442 * GCC now supports the Bionic C library and provides a convenient way 11443 of building native libraries and applications for the Android 11444 platform. Refer to the documentation of the -mandroid and -mbionic 11445 options for details on building native code. At the moment, Android 11446 support is enabled only for ARM. 11447 11448 Darwin/Mac OS X 11449 11450 * General 11451 + Initial support for CFString types has been added. 11452 This allows GCC to build projects including the system Core 11453 Foundation frameworks. The GCC Objective-C family supports 11454 CFString "toll-free bridged" as per the Mac OS X system tools. 11455 CFString is also recognized in the context of format 11456 attributes and arguments (see the documentation for format 11457 attributes for limitations). At present, 8-bit character types 11458 are supported. 11459 + Object file size reduction. 11460 The Darwin zeroed memory allocators have been re-written to 11461 make more use of .zerofill sections. For non-debug code, this 11462 can reduce object file size significantly. 11463 + Objective-C family 64-bit support (NeXT ABI 2). 11464 Initial support has been added to support 64-bit Objective-C 11465 code using the Darwin/OS X native (NeXT) runtime. ABI version 11466 2 will be selected automatically when 64-bit code is built. 11467 + Objective-C family 32-bit ABI 1. 11468 For 32-bit code ABI 1 is also now also allowed. At present it 11469 must be selected manually using -fobjc-abi-version=1 where 11470 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 11471 * x86 Architecture 11472 + The -mdynamic-no-pic option has been enabled. 11473 Code supporting -mdynamic-no-pic optimization has been added 11474 and is applicable to -m32 builds. The compiler bootstrap uses 11475 the option where appropriate. 11476 + The default value for -mtune= has been changed. 11477 Since Darwin systems are primarily Xeon, Core-2 or similar the 11478 default tuning has been changed to -mtune=core2. 11479 + Enable 128-bit long double (__float128) support on Darwin. 11480 * PPC Architecture 11481 + Darwin64 ABI. 11482 Several significant bugs have been fixed, such that GCC now 11483 produces code compatible with the Darwin64 PowerPC ABI. 11484 + libffi and boehm-gc. 11485 The Darwin ports of the libffi and boehm-gc libraries have 11486 been upgraded to include a Darwin64 implementation. This means 11487 that powerpc*-*-darwin9 platforms may now, for example, build 11488 Java applications with -m64 enabled. 11489 + Plug-in support has been enabled. 11490 + The -fsection-anchors option is now available although, 11491 presently, not heavily tested. 11492 11493 Solaris 2 11494 11495 New Features 11496 11497 * Support symbol versioning with the Sun linker. 11498 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 11499 * Support thread-local storage (TLS) with the Sun assembler on 11500 Solaris 2/x86. 11501 * Support TLS on Solaris 8/9 if prerequisites are met. 11502 * Support COMDAT group with the GNU assembler and recent Sun linker. 11503 * Support the Sun assembler visibility syntax. 11504 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 11505 -march=pentiumpro (Solaris 8/9). 11506 * Don't use SSE on Solaris 8/9 x86 by default. 11507 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 11508 11509 ABI Change 11510 11511 * Change the ABI for returning 8-byte vectors like __m64 in MMX 11512 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 11513 compilers. This is an incompatible change. If you use such types, 11514 you must either recompile all your code with the new compiler or 11515 use the new -mvect8-ret-in-mem option to remain compatible with 11516 previous versions of GCC and Sun Studio. 11517 11518 Windows x86/x86_64 11519 11520 * Initial support for decimal floating point. 11521 * Support for the __thiscall calling-convention. 11522 * Support for hot-patchable function prologues via the 11523 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 11524 * Improvements of stack-probing and stack-allocation mechanisms. 11525 * Support of push/pop-macro pragma as preprocessor command. 11526 With #pragma push_macro("macro-name") the current definition of 11527 macro-name is saved and can be restored with #pragma 11528 pop_macro("macro-name") to its saved definition. 11529 * Enable 128-bit long double (__float128) support on MinGW and 11530 Cygwin. 11531 11532 Other significant improvements 11533 11534 Installation changes 11535 11536 * An install-strip make target is provided that installs stripped 11537 executables, and may install libraries with unneeded or debugging 11538 sections stripped. 11539 * On Power7 systems, there is a potential problem if you build the 11540 GCC compiler with a host compiler using options that enable the VSX 11541 instruction set generation. If the host compiler has been patched 11542 so that the vec_ld and vec_st builtin functions generate Altivec 11543 memory instructions instead of VSX memory instructions, then you 11544 should be able to build the compiler with VSX instruction 11545 generation. 11546 11547 Changes for GCC Developers 11548 11549 Note: these changes concern developers that develop GCC itself or 11550 software that integrates with GCC, such as plugins, and not the general 11551 GCC users. 11552 * The gengtype utility, which previously was internal to the GCC 11553 build process, has been enchanced to provide GC root information 11554 for plugins as necessary. 11555 * The old GC allocation interface of ggc_alloc and friends was 11556 replaced with a type-safe alternative. 11557 11558 GCC 4.6.1 11559 11560 This is the [19]list of problem reports (PRs) from GCC's bug tracking 11561 system that are known to be fixed in the 4.6.1 release. This list might 11562 not be complete (that is, it is possible that some PRs that have been 11563 fixed are not listed here). 11564 11565 GCC 4.6.2 11566 11567 This is the [20]list of problem reports (PRs) from GCC's bug tracking 11568 system that are known to be fixed in the 4.6.2 release. This list might 11569 not be complete (that is, it is possible that some PRs that have been 11570 fixed are not listed here). 11571 11572 GCC 4.6.3 11573 11574 This is the [21]list of problem reports (PRs) from GCC's bug tracking 11575 system that are known to be fixed in the 4.6.3 release. This list might 11576 not be complete (that is, it is possible that some PRs that have been 11577 fixed are not listed here). 11578 11579 GCC 4.6.4 11580 11581 This is the [22]list of problem reports (PRs) from GCC's bug tracking 11582 system that are known to be fixed in the 4.6.4 release. This list might 11583 not be complete (that is, it is possible that some PRs that have been 11584 fixed are not listed here). 11585 11586 11587 For questions related to the use of GCC, please consult these web 11588 pages and the [23]GCC manuals. If that fails, the 11589 [24]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11590 web pages and the development of GCC are welcome on our developer 11591 list at [25]gcc (a] gcc.gnu.org. All of [26]our lists have public 11592 archives. 11593 11594 Copyright (C) [27]Free Software Foundation, Inc. Verbatim copying and 11595 distribution of this entire article is permitted in any medium, 11596 provided this notice is preserved. 11597 11598 These pages are [28]maintained by the GCC team. Last modified 11599 2024-05-17. 11600 11601 References 11602 11603 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 11604 2. https://gcc.gnu.org/gcc-4.5/changes.html 11605 3. https://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 11606 4. https://gcc.gnu.org/gcc-4.6/porting_to.html 11607 5. https://gcc.gnu.org/projects/lto/whopr.pdf 11608 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 11609 7. https://gcc.gnu.org/gcc-4.6/cxx0x_status.html 11610 8. https://gcc.gnu.org/PR43145 11611 9. https://gcc.gnu.org/PR43680 11612 10. https://gcc.gnu.org/PR33558 11613 11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 11614 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 11615 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 11616 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 11617 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 11618 16. https://gcc.gnu.org/wiki/OOP 11619 17. https://gcc.gnu.org/wiki/Coarray 11620 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 11621 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 11622 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 11623 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 11624 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 11625 23. https://gcc.gnu.org/onlinedocs/ 11626 24. mailto:gcc-help (a] gcc.gnu.org 11627 25. mailto:gcc (a] gcc.gnu.org 11628 26. https://gcc.gnu.org/lists.html 11629 27. https://www.fsf.org/ 11630 28. https://gcc.gnu.org/about.html 11631 ====================================================================== 11632 http://gcc.gnu.org/gcc-4.5/index.html 11633 GCC 4.5 Release Series 11634 11635 (This release series is no longer supported.) 11636 11637 Jul 2, 2012 11638 11639 The [1]GNU project and the GCC developers are pleased to announce the 11640 release of GCC 4.5.4. 11641 11642 This release is a bug-fix release, containing fixes for regressions in 11643 GCC 4.5.3 relative to previous releases of GCC. 11644 11645 Release History 11646 11647 GCC 4.5.4 11648 Jul 2, 2012 ([2]changes) 11649 11650 GCC 4.5.3 11651 Apr 28, 2011 ([3]changes) 11652 11653 GCC 4.5.2 11654 Dec 16, 2010 ([4]changes) 11655 11656 GCC 4.5.1 11657 Jul 31, 2010 ([5]changes) 11658 11659 GCC 4.5.0 11660 April 14, 2010 ([6]changes) 11661 11662 References and Acknowledgements 11663 11664 GCC used to stand for the GNU C Compiler, but since the compiler 11665 supports several other languages aside from C, it now stands for the 11666 GNU Compiler Collection. 11667 11668 A list of [7]successful builds is updated as new information becomes 11669 available. 11670 11671 The GCC developers would like to thank the numerous people that have 11672 contributed new features, improvements, bug fixes, and other changes as 11673 well as test results to GCC. This [8]amazing group of volunteers is 11674 what makes GCC successful. 11675 11676 For additional information about GCC please refer to the [9]GCC project 11677 web site or contact the [10]GCC development mailing list. 11678 11679 To obtain GCC please use [11]our mirror sites or [12]our version 11680 control system. 11681 11682 11683 For questions related to the use of GCC, please consult these web 11684 pages and the [13]GCC manuals. If that fails, the 11685 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11686 web pages and the development of GCC are welcome on our developer 11687 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 11688 archives. 11689 11690 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 11691 distribution of this entire article is permitted in any medium, 11692 provided this notice is preserved. 11693 11694 These pages are [18]maintained by the GCC team. Last modified 11695 2022-10-26. 11696 11697 References 11698 11699 1. http://www.gnu.org/ 11700 2. https://gcc.gnu.org/gcc-4.5/changes.html 11701 3. https://gcc.gnu.org/gcc-4.5/changes.html 11702 4. https://gcc.gnu.org/gcc-4.5/changes.html 11703 5. https://gcc.gnu.org/gcc-4.5/changes.html 11704 6. https://gcc.gnu.org/gcc-4.5/changes.html 11705 7. https://gcc.gnu.org/gcc-4.5/buildstat.html 11706 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11707 9. https://gcc.gnu.org/index.html 11708 10. mailto:gcc (a] gcc.gnu.org 11709 11. https://gcc.gnu.org/mirrors.html 11710 12. https://gcc.gnu.org/git.html 11711 13. https://gcc.gnu.org/onlinedocs/ 11712 14. mailto:gcc-help (a] gcc.gnu.org 11713 15. mailto:gcc (a] gcc.gnu.org 11714 16. https://gcc.gnu.org/lists.html 11715 17. https://www.fsf.org/ 11716 18. https://gcc.gnu.org/about.html 11717 ====================================================================== 11718 http://gcc.gnu.org/gcc-4.5/changes.html 11719 GCC 4.5 Release Series 11720 Changes, New Features, and Fixes 11721 11722 Caveats 11723 11724 * GCC now requires the [1]MPC library in order to build. See the 11725 [2]prerequisites page for version requirements. 11726 * Support for a number of older systems and recently unmaintained or 11727 untested target ports of GCC has been declared obsolete in GCC 4.5. 11728 Unless there is activity to revive them, the next release of GCC 11729 will have their sources permanently removed. 11730 The following ports for individual systems on particular 11731 architectures have been obsoleted: 11732 + IRIX releases before 6.5 (mips-sgi-irix5*, 11733 mips-sgi-irix6.[0-4]) 11734 + Solaris 7 (*-*-solaris2.7) 11735 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 11736 alpha-dec-osf5.0*) 11737 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 11738 can be found in the [3]announcement. 11739 Support for the classic POWER architecture implemented in the 11740 original RIOS and RIOS2 processors of the old IBM RS/6000 product 11741 line has been obsoleted in the rs6000 port. This does not affect 11742 the new generation Power and PowerPC architectures. 11743 * Support has been removed for all the [4]configurations obsoleted in 11744 GCC 4.4. 11745 * Support has been removed for the protoize and unprotoize utilities, 11746 obsoleted in GCC 4.4. 11747 * Support has been removed for tuning for Itanium1 (Merced) variants. 11748 Note that code tuned for Itanium2 should also run correctly on 11749 Itanium1. 11750 * GCC now generates unwind info also for epilogues. DWARF debuginfo 11751 generated by GCC now uses more features of DWARF3 than before, and 11752 also some DWARF4 features. GDB older than 7.0 is not able to handle 11753 either of these, so to debug GCC 4.5 generated binaries or 11754 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 11755 features with the -gdwarf-3 -gstrict-dwarf options, or use 11756 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 11757 epilogue unwind info is emitted unconditionally whenever unwind 11758 info is emitted. 11759 * On x86 targets, code containing floating-point calculations may run 11760 significantly more slowly when compiled with GCC 4.5 in strict C99 11761 conformance mode than they did with earlier GCC versions. This is 11762 due to stricter standard conformance of the compiler and can be 11763 avoided by using the option -fexcess-precision=fast; also see 11764 [5]below. 11765 * The function attribute noinline no longer prevents GCC from cloning 11766 the function. A new attribute noclone has been introduced for this 11767 purpose. Cloning a function means that it is duplicated and the new 11768 copy is specialized for certain contexts (for example when a 11769 parameter is a known constant). 11770 11771 General Optimizer Improvements 11772 11773 * The -save-temps now takes an optional argument. The -save-temps and 11774 -save-temps=cwd switches write the temporary files in the current 11775 working directory based on the original source file. The 11776 -save-temps=obj switch will write files into the directory 11777 specified with the -o option, and the intermediate filenames are 11778 based on the output file. This will allow the user to get the 11779 compiler intermediate files when doing parallel builds without two 11780 builds of the same filename located in different directories from 11781 interfering with each other. 11782 * Debugging dumps are now created in the same directory as the object 11783 file rather than in the current working directory. This allows the 11784 user to get debugging dumps when doing parallel builds without two 11785 builds of the same filename interfering with each other. 11786 * GCC has been integrated with the MPC library. This allows GCC to 11787 evaluate complex arithmetic at compile time [6]more accurately. It 11788 also allows GCC to evaluate calls to complex built-in math 11789 functions having constant arguments and replace them at compile 11790 time with their mathematically equivalent results. In doing so, GCC 11791 can generate correct results regardless of the math library 11792 implementation or floating point precision of the host platform. 11793 This also allows GCC to generate identical results regardless of 11794 whether one compiles in native or cross-compile configurations to a 11795 particular target. The following built-in functions take advantage 11796 of this new capability: cacos, cacosh, casin, casinh, catan, 11797 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 11798 and ctanh. The float and long double variants of these functions 11799 (e.g. csinf and csinl) are also handled. 11800 * A new link-time optimizer has been added ([7]-flto). When this 11801 option is used, GCC generates a bytecode representation of each 11802 input file and writes it to specially-named sections in each object 11803 file. When the object files are linked together, all the function 11804 bodies are read from these named sections and instantiated as if 11805 they had been part of the same translation unit. This enables 11806 interprocedural optimizations to work across different files (and 11807 even different languages), potentially improving the performance of 11808 the generated code. To use the link-timer optimizer, -flto needs to 11809 be specified at compile time and during the final link. If the 11810 program does not require any symbols to be exported, it is possible 11811 to combine -flto and the experimental [8]-fwhopr with 11812 [9]-fwhole-program to allow the interprocedural optimizers to use 11813 more aggressive assumptions. 11814 * The automatic parallelization pass was enhanced to support 11815 parallelization of outer loops. 11816 * Automatic parallelization can be enabled as part of Graphite. In 11817 addition to -ftree-parallelize-loops=, specify 11818 -floop-parallelize-all to enable the Graphite-based optimization. 11819 * The infrastructure for optimizing based on [10]restrict qualified 11820 pointers has been rewritten and should result in code generation 11821 improvements. Optimizations based on restrict qualified pointers 11822 are now also available when using -fno-strict-aliasing. 11823 * There is a new optimization pass that attempts to change prototype 11824 of functions to avoid unused parameters, pass only relevant parts 11825 of structures and turn arguments passed by reference to arguments 11826 passed by value when possible. It is enabled by -O2 and above as 11827 well as -Os and can be manually invoked using the new command-line 11828 switch -fipa-sra. 11829 * GCC now optimize exception handling code. In particular cleanup 11830 regions that are proved to not have any effect are optimized out. 11831 11832 New Languages and Language specific improvements 11833 11834 All languages 11835 11836 * The -fshow-column option is now on by default. This means error 11837 messages now have a column associated with them. 11838 11839 Ada 11840 11841 * Compilation of programs heavily using discriminated record types 11842 with variant parts has been sped up and generates more compact 11843 code. 11844 * Stack checking now works reasonably well on most plaforms. In some 11845 specific cases, stack overflows may still fail to be detected, but 11846 a compile-time warning will be issued for these cases. 11847 11848 C family 11849 11850 * If a header named in a #include directive is not found, the 11851 compiler exits immediately. This avoids a cascade of errors arising 11852 from declarations expected to be found in that header being 11853 missing. 11854 * A new built-in function __builtin_unreachable() has been added that 11855 tells the compiler that control will never reach that point. It may 11856 be used after asm statements that terminate by transferring control 11857 elsewhere, and in other places that are known to be unreachable. 11858 * The -Wlogical-op option now warns for logical expressions such as 11859 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 11860 mistakes. This option is disabled by default. 11861 * An asm goto feature has been added to allow asm statements that 11862 jump to C labels. 11863 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 11864 * The deprecated attribute now takes an optional string argument, for 11865 example, __attribute__((deprecated("text string"))), that will be 11866 printed together with the deprecation warning. 11867 11868 C 11869 11870 * The -Wenum-compare option, which warns when comparing values of 11871 different enum types, now works for C. It formerly only worked for 11872 C++. This warning is enabled by -Wall. It may be avoided by using a 11873 type cast. 11874 * The -Wcast-qual option now warns about casts which are unsafe in 11875 that they permit const-correctness to be violated without further 11876 warnings. Specifically, it warns about cases where a qualifier is 11877 added when all the lower types are not const. For example, it warns 11878 about a cast from char ** to const char **. 11879 * The -Wc++-compat option is significantly improved. It issues new 11880 warnings for: 11881 + Using C++ reserved operator names as identifiers. 11882 + Conversions to enum types without explicit casts. 11883 + Using va_arg with an enum type. 11884 + Using different enum types in the two branches of ?:. 11885 + Using ++ or -- on a variable of enum type. 11886 + Using the same name as both a struct, union or enum tag and a 11887 typedef, unless the typedef refers to the tagged type itself. 11888 + Using a struct, union, or enum which is defined within another 11889 struct or union. 11890 + A struct field defined using a typedef if there is a field in 11891 the struct, or an enclosing struct, whose name is the typedef 11892 name. 11893 + Duplicate definitions at file scope. 11894 + Uninitialized const variables. 11895 + A global variable with an anonymous struct, union, or enum 11896 type. 11897 + Using a string constant to initialize a char array whose size 11898 is the length of the string. 11899 * The new -Wjump-misses-init option warns about cases where a goto or 11900 switch skips the initialization of a variable. This sort of branch 11901 is an error in C++ but not in C. This warning is enabled by 11902 -Wc++-compat. 11903 * GCC now ensures that a C99-conforming <stdint.h> is present on most 11904 targets, and uses information about the types in this header to 11905 implement the Fortran bindings to those types. GCC does not ensure 11906 the presence of such a header, and does not implement the Fortran 11907 bindings, on the following targets: NetBSD, VxWorks, VMS, 11908 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 11909 * GCC now implements C90- and C99-conforming rules for constant 11910 expressions. This may cause warnings or errors for some code using 11911 expressions that can be folded to a constant but are not constant 11912 expressions as defined by ISO C. 11913 * All known target-independent C90 and C90 Amendment 1 conformance 11914 bugs, and all known target-independent C99 conformance bugs not 11915 related to floating point or extended identifiers, have been fixed. 11916 * The C decimal floating point support now includes support for the 11917 FLOAT_CONST_DECIMAL64 pragma. 11918 * The named address space feature from ISO/IEC TR 18037 is now 11919 supported. This is currently only implemented for the SPU 11920 processor. 11921 11922 C++ 11923 11924 * Improved [11]experimental support for the upcoming C++0x ISO C++ 11925 standard, including support for raw strings, lambda expressions and 11926 explicit type conversion operators. 11927 * When printing the name of a class template specialization, G++ will 11928 now omit any template arguments which come from default template 11929 arguments. This behavior (and the pretty-printing of function 11930 template specializations as template signature and arguments) can 11931 be disabled with the -fno-pretty-templates option. 11932 * Access control is now applied to typedef names used in a template, 11933 which may cause G++ to reject some ill-formed code that was 11934 accepted by earlier releases. The -fno-access-control option can be 11935 used as a temporary workaround until the code is corrected. 11936 * Compilation time for code that uses templates should now scale 11937 linearly with the number of instantiations rather than 11938 quadratically, as template instantiations are now looked up using 11939 hash tables. 11940 * Declarations of functions that look like builtin declarations of 11941 library functions are only considered to be redeclarations if they 11942 are declared with extern "C". This may cause problems with code 11943 that omits extern "C" on hand-written declarations of C library 11944 functions such as abort or memcpy. Such code is ill-formed, but was 11945 accepted by earlier releases. 11946 * Diagnostics that used to complain about passing non-POD types to 11947 ... or jumping past the declaration of a non-POD variable now check 11948 for triviality rather than PODness, as per C++0x. 11949 * In C++0x mode local and anonymous classes are now allowed as 11950 template arguments, and in declarations of variables and functions 11951 with linkage, so long as any such declaration that is used is also 11952 defined ([12]DR 757). 11953 * Labels may now have attributes, as has been permitted for a while 11954 in C. This is only permitted when the label definition and the 11955 attribute specifier is followed by a semicoloni.e., the label 11956 applies to an empty statement. The only useful attribute for a 11957 label is unused. 11958 * G++ now implements [13]DR 176. Previously G++ did not support using 11959 the injected-class-name of a template base class as a type name, 11960 and lookup of the name found the declaration of the template in the 11961 enclosing scope. Now lookup of the name finds the 11962 injected-class-name, which can be used either as a type or as a 11963 template, depending on whether or not the name is followed by a 11964 template argument list. As a result of this change, some code that 11965 was previously accepted may be ill-formed because 11966 1. The injected-class-name is not accessible because it's from a 11967 private base, or 11968 2. The injected-class-name cannot be used as an argument for a 11969 template template parameter. 11970 In either of these cases, the code can be fixed by adding a 11971 nested-name-specifier to explicitly name the template. The first 11972 can be worked around with -fno-access-control; the second is only 11973 rejected with -pedantic. 11974 * A new standard mangling for SIMD vector types has been added, to 11975 avoid name clashes on systems with vectors of varying length. By 11976 default the compiler still uses the old mangling, but emits aliases 11977 with the new mangling on targets that support strong aliases. Users 11978 can switch over entirely to the new mangling with -fabi-version=4 11979 or -fabi-version=0. -Wabi will now warn about code that uses the 11980 old mangling. 11981 * The command-line option -ftemplate-depth-N is now written as 11982 -ftemplate-depth=N and the old form is deprecated. 11983 * Conversions between NULL and non-pointer types are now warned by 11984 default. The new option -Wno-conversion-null disables these 11985 warnings. Previously these warnings were only available when using 11986 -Wconversion explicitly. 11987 11988 Runtime Library (libstdc++) 11989 11990 * Improved experimental support for the upcoming ISO C++ standard, 11991 C++0x, including: 11992 + Support for <future>, <functional>, and <random>. 11993 + Existing facilities now exploit explicit operators and the 11994 newly implemented core C++0x features. 11995 + The header <cstdatomic> has been renamed to <atomic>. 11996 * An experimental [14]profile mode has been added. This is an 11997 implementation of many C++ standard library constructs with an 11998 additional analysis layer that gives performance improvement advice 11999 based on recognition of suboptimal usage patterns. For example, 12000 #include <vector> 12001 int main() 12002 { 12003 std::vector<int> v; 12004 for (int k = 0; k < 1024; ++k) 12005 v.insert(v.begin(), k); 12006 } 12007 12008 When instrumented via the profile mode, can return suggestions 12009 about the initial size and choice of the container used as follows: 12010 vector-to-list: improvement = 5: call stack = 0x804842c ... 12011 : advice = change std::vector to std::list 12012 vector-size: improvement = 3: call stack = 0x804842c ... 12013 : advice = change initial container size from 0 to 1024 12014 12015 These constructs can be substituted for the normal libstdc++ 12016 constructs on a piecemeal basis, or all existing components can be 12017 transformed via the -D_GLIBCXX_PROFILE macro. 12018 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 12019 24733) has been added. This support is in header file 12020 <decimal/decimal>, uses namespace std::decimal, and includes 12021 classes decimal32, decimal64, and decimal128. 12022 * Sources have been audited for application of function attributes 12023 nothrow, const, pure, and noreturn. 12024 * Python pretty-printers have been added for many standard library 12025 components that simplify the internal representation and present a 12026 more intuitive view of components when used with 12027 appropriately-advanced versions of GDB. For more information, 12028 please consult the more [16]detailed description. 12029 * The default behavior for comparing typeinfo names has changed, so 12030 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 12031 * The new -static-libstdc++ option directs g++ to link the C++ 12032 library statically, even if the default would normally be to link 12033 it dynamically. 12034 12035 Fortran 12036 12037 * The COMMON default padding has been changed instead of adding the 12038 padding before a variable it is now added afterwards, which 12039 increases the compatibility with other vendors and helps to obtain 12040 the correct output in some cases. Cf. also the -falign-commons 12041 option ([17]added in 4.4). 12042 * The -finit-real= option now also supports the value snan for 12043 signaling not-a-number; to be effective, one additionally needs to 12044 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 12045 optimizations can turn a signaling NaN into a quiet one. 12046 * The new option -fcheck= has been added with the options bounds, 12047 array-temps, do, pointer, and recursive. The bounds and array-temps 12048 options are equivalent to -fbounds-check and 12049 -fcheck-array-temporaries. The do option checks for invalid 12050 modification of loop iteration variables, and the recursive option 12051 tests for recursive calls to subroutines/functions which are not 12052 marked as recursive. With pointer pointer association checks in 12053 calls are performed; however, neither undefined pointers nor 12054 pointers in expressions are handled. Using -fcheck=all enables all 12055 these run-time checks. 12056 * The run-time checking -fcheck=bounds now warns about invalid string 12057 lengths of character dummy arguments. Additionally, more 12058 compile-time checks have been added. 12059 * The new option [18]-fno-protect-parens has been added; if set, the 12060 compiler may reorder REAL and COMPLEX expressions without regard to 12061 parentheses. 12062 * GNU Fortran no longer links against libgfortranbegin. As before, 12063 MAIN__ (assembler symbol name) is the actual Fortran main program, 12064 which is invoked by the main function. However, main is now 12065 generated and put in the same object file as MAIN__. For the time 12066 being, libgfortranbegin still exists for backward compatibility. 12067 For details see the new [19]Mixed-Language Programming chapter in 12068 the manual. 12069 * The I/O library was restructured for performance and cleaner code. 12070 * Array assignments and WHERE are now run in parallel when OpenMP's 12071 WORKSHARE is used. 12072 * The experimental option -fwhole-file was added. The option allows 12073 whole-file checking of procedure arguments and allows for better 12074 optimizations. It can also be used with -fwhole-program, which is 12075 now also supported in gfortran. 12076 * More Fortran 2003 and Fortran 2008 mathematical functions can now 12077 be used as initialization expressions. 12078 * Some extended attributes such as STDCALL are now supported via the 12079 [20]GCC$ compiler directive. 12080 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 12081 intrinsic behaves now as if zero were always positive. 12082 * For legacy compatibiliy: On Cygwin and MinGW, the special files 12083 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 12084 supported. 12085 * Fortran 2003 support has been extended: 12086 + Procedure-pointer function results and procedure-pointer 12087 components (including PASS), 12088 + allocatable scalars (experimental), 12089 + DEFERRED type-bound procedures, 12090 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 12091 have been implemented. 12092 + The ALLOCATE statement supports type-specs and the SOURCE= 12093 argument. 12094 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 12095 type-bound procedure (i.e. as type-bound operators). 12096 + Rounding (ROUND=, RZ, ...) for output is now supported. 12097 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 12098 intrinsic module ISO_C_BINDING are now supported, except for 12099 the targets listed above as ones where GCC does not have 12100 <stdint.h> type information. 12101 + Extensible derived types with type-bound procedure or 12102 procedure pointer with PASS attribute now have to use CLASS in 12103 line with the Fortran 2003 standard; the workaround to use 12104 TYPE is no longer supported. 12105 + [21]Experimental, incomplete support for polymorphism, 12106 including CLASS, SELECT TYPE and dynamic dispatch of 12107 type-bound procedure calls. Some features do not work yet such 12108 as unlimited polymorphism (CLASS(*)). 12109 * Fortran 2008 support has been extended: 12110 + The OPEN statement now supports the NEWUNIT= option, which 12111 returns a unique file unit, thus preventing inadvertent use of 12112 the same unit in different parts of the program. 12113 + Support for unlimited format items has been added. 12114 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 12115 the intrinsic module ISO_FORTRAN_ENV are now supported. 12116 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 12117 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 12118 and ATANH have been added (for real and complex arguments) and 12119 ATAN(Y,X) is now an alias for ATAN2(Y,X). 12120 + The BLOCK construct has been implemented. 12121 12122 New Targets and Target Specific Improvements 12123 12124 AIX 12125 12126 * Full cross-toolchain support now available with GNU Binutils 12127 12128 ARM 12129 12130 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 12131 * GCC now supports the ARM v7E-M architecture. 12132 * GCC now supports VFPv4-based FPUs and FPUs with 12133 single-precision-only VFP. 12134 * GCC has many improvements to optimization for other ARM processors, 12135 including scheduling support for the integer pipeline on Cortex-A9. 12136 * GCC now supports the IEEE 754-2008 half-precision floating-point 12137 type, and a variant ARM-specific half-precision type. This type is 12138 specified using __fp16, with the layout determined by 12139 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 12140 VFPv4 half-precision instructions will be used. 12141 * GCC now supports the variant of AAPCS that uses VFP registers for 12142 parameter passing and return values. 12143 12144 AVR 12145 12146 * The -mno-tablejump option has been removed because it has the same 12147 effect as the -fno-jump-tables option. 12148 * Added support for these new AVR devices: 12149 + ATmega8U2 12150 + ATmega16U2 12151 + ATmega32U2 12152 12153 IA-32/x86-64 12154 12155 * GCC now will set the default for -march= based on the configure 12156 target. 12157 * GCC now supports handling floating-point excess precision arising 12158 from use of the x87 floating-point unit in a way that conforms to 12159 ISO C99. This is enabled with -fexcess-precision=standard and with 12160 standards conformance options such as -std=c99, and may be disabled 12161 using -fexcess-precision=fast. 12162 * Support for the Intel Atom processor is now available through the 12163 -march=atom and -mtune=atom options. 12164 * A new -mcrc32 option is now available to enable crc32 intrinsics. 12165 * A new -mmovbe option is now available to enable GCC to use the 12166 movbe instruction to implement __builtin_bswap32 and 12167 __builtin_bswap64. 12168 * SSE math now can be enabled by default at configure time with the 12169 new --with-fpmath=sse option. 12170 * There is a new intrinsic header file, <x86intrin.h>. It should be 12171 included before using any IA-32/x86-64 intrinsics. 12172 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 12173 Orochi processors are now available with the -mxop, -mfma4, and 12174 -mlwp options. 12175 * The -mabm option enables GCC to use the popcnt and lzcnt 12176 instructions on AMD processors. 12177 * The -mpopcnt option enables GCC to use the popcnt instructions on 12178 both AMD and Intel processors. 12179 12180 M68K/ColdFire 12181 12182 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 12183 and 5441x devices. 12184 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 12185 processors. 12186 12187 MeP 12188 12189 Support has been added for the Toshiba Media embedded Processor (MeP, 12190 or mep-elf) embedded target. 12191 12192 MIPS 12193 12194 * GCC now supports MIPS 1004K processors. 12195 * GCC can now be configured with options --with-arch-32, 12196 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 12197 default optimization separately for 32-bit and 64-bit modes. 12198 * MIPS targets now support an alternative _mcount interface, in which 12199 register $12 points to the function's save slot for register $31. 12200 This interface is selected by the -mcount-ra-address option; see 12201 the documentation for more details. 12202 * GNU/Linux targets can now generate read-only .eh_frame sections. 12203 This optimization requires GNU binutils 2.20 or above, and is only 12204 available if GCC is configured with a suitable version of binutils. 12205 * GNU/Linux targets can now attach special relocations to indirect 12206 calls, so that the linker can turn them into direct jumps or 12207 branches. This optimization requires GNU binutils 2.20 or later, 12208 and is automatically selected if GCC is configured with an 12209 appropriate version of binutils. It can be explicitly enabled or 12210 disabled using the -mrelax-pic-calls command-line option. 12211 * GCC now generates more heavily-optimized atomic operations on 12212 Octeon processors. 12213 * MIPS targets now support the -fstack-protector option. 12214 * GCC now supports an -msynci option, which specifies that synci is 12215 enough to flush the instruction cache, without help from the 12216 operating system. GCC uses this information to optimize 12217 automatically-generated cache flush operations, such as those used 12218 for nested functions in C. There is also a --with-synci 12219 configure-time option, which makes -msynci the default. 12220 * GCC supports four new function attributes for interrupt handlers: 12221 interrupt, use_shadow_register_set, keep_interrupts_masked and 12222 use_debug_exception_return. See the documentation for more details 12223 about these attributes. 12224 12225 RS/6000 (POWER/PowerPC) 12226 12227 * GCC now supports the Power ISA 2.06, which includes the VSX 12228 instructions that add vector 64-bit floating point support, new 12229 population count instructions, and conversions between floating 12230 point and unsigned types. 12231 * Support for the power7 processor is now available through the 12232 -mcpu=power7 and -mtune=power7. 12233 * GCC will now vectorize loops that contain simple math functions 12234 like copysign when generating code for altivec or VSX targets. 12235 * Support for the A2 processor is now available through the -mcpu=a2 12236 and -mtune=a2 options. 12237 * Support for the 476 processor is now available through the 12238 -mcpu={476,476fp} and -mtune={476,476fp} options. 12239 * Support for the e500mc64 processor is now available through the 12240 -mcpu=e500mc64 and -mtune=e500mc64 options. 12241 * GCC can now be configured with options --with-cpu-32, 12242 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 12243 default optimization separately for 32-bit and 64-bit modes. 12244 * Starting with GCC 4.5.4, vectors of type vector long long or vector 12245 long are passed and returned in the same method as other vectors 12246 with the VSX instruction set. Previously the GCC compiler did not 12247 adhere to the ABI for 128-bit vectors with 64-bit integer base 12248 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 12249 12250 RX 12251 12252 Support has been added for the Renesas RX Processor (rx-elf) target. 12253 12254 Operating Systems 12255 12256 Windows (Cygwin and MinGW) 12257 12258 * GCC now installs all the major language runtime libraries as DLLs 12259 when configured with the --enable-shared option. 12260 * GCC now makes use of the new support for aligned common variables 12261 in versions of binutils >= 2.20 to fix bugs in the support for SSE 12262 data types. 12263 * Improvements to the libffi support library increase the reliability 12264 of code generated by GCJ on all Windows platforms. Libgcj is 12265 enabled by default for the first time. 12266 * Libtool improvements simplify installation by placing the generated 12267 DLLs in the correct binaries directory. 12268 * Numerous other minor bugfixes and improvements, and substantial 12269 enhancements to the Fortran language support library. 12270 12271 > 12272 12273 Other significant improvements 12274 12275 Plugins 12276 12277 * It is now possible to extend the compiler without having to modify 12278 its source code. A new option -fplugin=file.so tells GCC to load 12279 the shared object file.so and execute it as part of the compiler. 12280 The internal documentation describes the details on how plugins can 12281 interact with the compiler. 12282 12283 Installation changes 12284 12285 * The move to newer autotools changed default installation 12286 directories and switches to control them: The --with-datarootdir, 12287 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 12288 used any more. Instead, you can now use --datarootdir, --docdir, 12289 --htmldir, and --pdfdir. The default installation directories have 12290 changed as follows according to the GNU Coding Standards: 12291 12292 datarootdir read-only architecture-independent data root [PREFIX/share] 12293 localedir locale-specific message catalogs [DATAROOTDIR/locale] 12294 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 12295 htmldir html documentation [DOCDIR] 12296 dvidir dvi documentation [DOCDIR] 12297 pdfdir pdf documentation [DOCDIR] 12298 psdir ps documentation [DOCDIR] 12299 The following variables have new default values: 12300 12301 datadir read-only architecture-independent data [DATAROOTDIR] 12302 infodir info documentation [DATAROOTDIR/info] 12303 mandir man documentation [DATAROOTDIR/man] 12304 12305 GCC 4.5.1 12306 12307 This is the [22]list of problem reports (PRs) from GCC's bug tracking 12308 system that are known to be fixed in the 4.5.1 release. This list might 12309 not be complete (that is, it is possible that some PRs that have been 12310 fixed are not listed here). 12311 12312 All languages 12313 12314 * GCC's new link-time optimizer ([23]-flto) now also works on a few 12315 non-ELF targets: 12316 + Cygwin (*-cygwin*) 12317 + MinGW (*-mingw*) 12318 + Darwin on x86-64 (x86_64-apple-darwin*) 12319 LTO is not enabled by default for these targets. To enable LTO, you 12320 should configure with the --enable-lto option. 12321 12322 GCC 4.5.2 12323 12324 This is the [24]list of problem reports (PRs) from GCC's bug tracking 12325 system that are known to be fixed in the 4.5.2 release. This list might 12326 not be complete (that is, it is possible that some PRs that have been 12327 fixed are not listed here). 12328 12329 GCC 4.5.3 12330 12331 This is the [25]list of problem reports (PRs) from GCC's bug tracking 12332 system that are known to be fixed in the 4.5.3 release. This list might 12333 not be complete (that is, it is possible that some PRs that have been 12334 fixed are not listed here). 12335 12336 On the PowerPC compiler, the Altivec builtin functions vec_ld and 12337 vec_st have been modified to generate the Altivec memory instructions 12338 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 12339 release, these builtin functions were changed to generate VSX memory 12340 reference instructions instead of Altivec memory instructions, but 12341 there are differences between the two instructions. If the VSX 12342 instruction set is available, you can now use the new builtin functions 12343 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 12344 instructions. 12345 12346 GCC 4.5.4 12347 12348 This is the [26]list of problem reports (PRs) from GCC's bug tracking 12349 system that are known to be fixed in the 4.5.4 release. This list might 12350 not be complete (that is, it is possible that some PRs that have been 12351 fixed are not listed here). 12352 12353 12354 For questions related to the use of GCC, please consult these web 12355 pages and the [27]GCC manuals. If that fails, the 12356 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12357 web pages and the development of GCC are welcome on our developer 12358 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 12359 archives. 12360 12361 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 12362 distribution of this entire article is permitted in any medium, 12363 provided this notice is preserved. 12364 12365 These pages are [32]maintained by the GCC team. Last modified 12366 2023-07-30. 12367 12368 References 12369 12370 1. https://www.multiprecision.org/ 12371 2. https://gcc.gnu.org/install/prerequisites.html 12372 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 12373 4. https://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 12374 5. https://gcc.gnu.org/gcc-4.5/changes.html#x86 12375 6. https://gcc.gnu.org/PR30789 12376 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 12377 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 12378 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 12379 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 12380 11. https://gcc.gnu.org/gcc-4.5/cxx0x_status.html 12381 12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 12382 13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 12383 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 12384 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 12385 16. https://sourceware.org/gdb/wiki/STLSupport 12386 17. https://gcc.gnu.org/gcc-4.4/changes.html 12387 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 12388 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 12389 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 12390 21. https://gcc.gnu.org/wiki/OOP 12391 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 12392 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 12393 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 12394 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 12395 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 12396 27. https://gcc.gnu.org/onlinedocs/ 12397 28. mailto:gcc-help (a] gcc.gnu.org 12398 29. mailto:gcc (a] gcc.gnu.org 12399 30. https://gcc.gnu.org/lists.html 12400 31. https://www.fsf.org/ 12401 32. https://gcc.gnu.org/about.html 12402 ====================================================================== 12403 http://gcc.gnu.org/gcc-4.4/index.html 12404 GCC 4.4 Release Series 12405 12406 This release series is no longer maintained. 12407 12408 March 13, 2012 12409 12410 The [1]GNU project and the GCC developers are pleased to announce the 12411 release of GCC 4.4.7. 12412 12413 This release is a bug-fix release, containing fixes for regressions in 12414 GCC 4.4.6 relative to previous releases of GCC. 12415 12416 Release History 12417 12418 GCC 4.4.7 12419 March 13, 2012 ([2]changes) 12420 12421 GCC 4.4.6 12422 April 16, 2011 ([3]changes) 12423 12424 GCC 4.4.5 12425 October 1, 2010 ([4]changes) 12426 12427 GCC 4.4.4 12428 April 29, 2010 ([5]changes) 12429 12430 GCC 4.4.3 12431 January 21, 2010 ([6]changes) 12432 12433 GCC 4.4.2 12434 October 15, 2009 ([7]changes) 12435 12436 GCC 4.4.1 12437 July 22, 2009 ([8]changes) 12438 12439 GCC 4.4.0 12440 April 21, 2009 ([9]changes) 12441 12442 References and Acknowledgements 12443 12444 GCC used to stand for the GNU C Compiler, but since the compiler 12445 supports several other languages aside from C, it now stands for the 12446 GNU Compiler Collection. 12447 12448 A list of [10]successful builds is updated as new information becomes 12449 available. 12450 12451 The GCC developers would like to thank the numerous people that have 12452 contributed new features, improvements, bug fixes, and other changes as 12453 well as test results to GCC. This [11]amazing group of volunteers is 12454 what makes GCC successful. 12455 12456 For additional information about GCC please refer to the [12]GCC 12457 project web site or contact the [13]GCC development mailing list. 12458 12459 To obtain GCC please use [14]our mirror sites or [15]our version 12460 control system. 12461 12462 12463 For questions related to the use of GCC, please consult these web 12464 pages and the [16]GCC manuals. If that fails, the 12465 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12466 web pages and the development of GCC are welcome on our developer 12467 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 12468 archives. 12469 12470 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 12471 distribution of this entire article is permitted in any medium, 12472 provided this notice is preserved. 12473 12474 These pages are [21]maintained by the GCC team. Last modified 12475 2022-10-26. 12476 12477 References 12478 12479 1. http://www.gnu.org/ 12480 2. https://gcc.gnu.org/gcc-4.4/changes.html 12481 3. https://gcc.gnu.org/gcc-4.4/changes.html 12482 4. https://gcc.gnu.org/gcc-4.4/changes.html 12483 5. https://gcc.gnu.org/gcc-4.4/changes.html 12484 6. https://gcc.gnu.org/gcc-4.4/changes.html 12485 7. https://gcc.gnu.org/gcc-4.4/changes.html 12486 8. https://gcc.gnu.org/gcc-4.4/changes.html 12487 9. https://gcc.gnu.org/gcc-4.4/changes.html 12488 10. https://gcc.gnu.org/gcc-4.4/buildstat.html 12489 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12490 12. https://gcc.gnu.org/index.html 12491 13. mailto:gcc (a] gcc.gnu.org 12492 14. https://gcc.gnu.org/mirrors.html 12493 15. https://gcc.gnu.org/git.html 12494 16. https://gcc.gnu.org/onlinedocs/ 12495 17. mailto:gcc-help (a] gcc.gnu.org 12496 18. mailto:gcc (a] gcc.gnu.org 12497 19. https://gcc.gnu.org/lists.html 12498 20. https://www.fsf.org/ 12499 21. https://gcc.gnu.org/about.html 12500 ====================================================================== 12501 http://gcc.gnu.org/gcc-4.4/changes.html 12502 GCC 4.4 Release Series 12503 Changes, New Features, and Fixes 12504 12505 The latest release in the 4.4 release series is [1]GCC 4.4.7. 12506 12507 Caveats 12508 12509 * __builtin_stdarg_start has been completely removed from GCC. 12510 Support for <varargs.h> had been deprecated since GCC 4.0. Use 12511 __builtin_va_start as a replacement. 12512 * Some of the errors issued by the C++ front end that could be 12513 downgraded to warnings in previous releases by using -fpermissive 12514 are now warnings by default. They can be converted into errors by 12515 using -pedantic-errors. 12516 * Use of the cpp assertion extension will now emit a warning when 12517 -Wdeprecated or -pedantic is used. This extension has been 12518 deprecated for many years, but never warned about. 12519 * Packed bit-fields of type char were not properly bit-packed on many 12520 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 12521 causes an ABI change. For example there is no longer a 4-bit 12522 padding between field a and b in this structure: 12523 struct foo 12524 { 12525 char a:4; 12526 char b:8; 12527 } __attribute__ ((packed)); 12528 There is a new warning to help identify fields that are affected: 12529 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 12530 The warning can be disabled with -Wno-packed-bitfield-compat. 12531 * On ARM EABI targets, the C++ mangling of the va_list type has been 12532 changed to conform to the current revision of the EABI. This does 12533 not affect the libstdc++ library included with GCC. 12534 * The SCOUNT and POS bits of the MIPS DSP control register are now 12535 treated as global. Previous versions of GCC treated these fields as 12536 call-clobbered instead. 12537 * The MIPS port no longer recognizes the h asm constraint. It was 12538 necessary to remove this constraint in order to avoid generating 12539 unpredictable code sequences. 12540 One of the main uses of the h constraint was to extract the high 12541 part of a multiplication on 64-bit targets. For example: 12542 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 12543 You can now achieve the same effect using 128-bit types: 12544 typedef unsigned int uint128_t __attribute__((mode(TI))); 12545 result = ((uint128_t) x * y) >> 64; 12546 The second sequence is better in many ways. For example, if x and y 12547 are constants, the compiler can perform the multiplication at 12548 compile time. If x and y are not constants, the compiler can 12549 schedule the runtime multiplication better than it can schedule an 12550 asm statement. 12551 * Support for a number of older systems and recently unmaintained or 12552 untested target ports of GCC has been declared obsolete in GCC 4.4. 12553 Unless there is activity to revive them, the next release of GCC 12554 will have their sources permanently removed. 12555 The following ports for individual systems on particular 12556 architectures have been obsoleted: 12557 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 12558 m68k-*-aout*) 12559 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 12560 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 12561 sh-*-*). This does not affect other more specific targets 12562 using the COFF object format on those architectures, or the 12563 more specific H8300 and SH targets (h8300-*-rtems*, 12564 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 12565 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 12566 + 2BSD on PDP-11 (pdp11-*-bsd) 12567 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 12568 powerpc-ibm-aix4.[12]*) 12569 + Tuning support for Itanium1 (Merced) variants. Note that code 12570 tuned for Itanium2 should also run correctly on Itanium1. 12571 * The protoize and unprotoize utilities have been obsoleted and will 12572 be removed in GCC 4.5. These utilities have not been installed by 12573 default since GCC 3.0. 12574 * Support has been removed for all the [2]configurations obsoleted in 12575 GCC 4.3. 12576 * Unknown -Wno-* options are now silently ignored by GCC if no other 12577 diagnostics are issued. If other diagnostics are issued, then GCC 12578 warns about the unknown options. 12579 * More information on porting to GCC 4.4 from previous versions of 12580 GCC can be found in the [3]porting guide for this release. 12581 12582 General Optimizer Improvements 12583 12584 * A new command-line switch -findirect-inlining has been added. When 12585 turned on it allows the inliner to also inline indirect calls that 12586 are discovered to have known targets at compile time thanks to 12587 previous inlining. 12588 * A new command-line switch -ftree-switch-conversion has been added. 12589 This new pass turns simple initializations of scalar variables in 12590 switch statements into initializations from a static array, given 12591 that all the values are known at compile time and the ratio between 12592 the new array size and the original switch branches does not exceed 12593 the parameter --param switch-conversion-max-branch-ratio (default 12594 is eight). 12595 * A new command-line switch -ftree-builtin-call-dce has been added. 12596 This optimization eliminates unnecessary calls to certain builtin 12597 functions when the return value is not used, in cases where the 12598 calls can not be eliminated entirely because the function may set 12599 errno. This optimization is on by default at -O2 and above. 12600 * A new command-line switch -fconserve-stack directs the compiler to 12601 minimize stack usage even if it makes the generated code slower. 12602 This affects inlining decisions. 12603 * When the assembler supports it, the compiler will now emit unwind 12604 information using assembler .cfi directives. This makes it possible 12605 to use such directives in inline assembler code. The new option 12606 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 12607 directives. 12608 * The [4]Graphite branch has been merged. This merge has brought in a 12609 new framework for loop optimizations based on a polyhedral 12610 intermediate representation. These optimizations apply to all the 12611 languages supported by GCC. The following new code transformations 12612 are available in GCC 4.4: 12613 + -floop-interchange performs loop interchange transformations 12614 on loops. Interchanging two nested loops switches the inner 12615 and outer loops. For example, given a loop like: 12616 DO J = 1, M 12617 DO I = 1, N 12618 A(J, I) = A(J, I) * C 12619 ENDDO 12620 ENDDO 12621 12622 loop interchange will transform the loop as if the user had 12623 written: 12624 DO I = 1, N 12625 DO J = 1, M 12626 A(J, I) = A(J, I) * C 12627 ENDDO 12628 ENDDO 12629 12630 which can be beneficial when N is larger than the caches, 12631 because in Fortran, the elements of an array are stored in 12632 memory contiguously by column, and the original loop iterates 12633 over rows, potentially creating at each access a cache miss. 12634 + -floop-strip-mine performs loop strip mining transformations 12635 on loops. Strip mining splits a loop into two nested loops. 12636 The outer loop has strides equal to the strip size and the 12637 inner loop has strides of the original loop within a strip. 12638 For example, given a loop like: 12639 DO I = 1, N 12640 A(I) = A(I) + C 12641 ENDDO 12642 12643 loop strip mining will transform the loop as if the user had 12644 written: 12645 DO II = 1, N, 4 12646 DO I = II, min (II + 3, N) 12647 A(I) = A(I) + C 12648 ENDDO 12649 ENDDO 12650 12651 + -floop-block performs loop blocking transformations on loops. 12652 Blocking strip mines each loop in the loop nest such that the 12653 memory accesses of the element loops fit inside caches. For 12654 example, given a loop like: 12655 DO I = 1, N 12656 DO J = 1, M 12657 A(J, I) = B(I) + C(J) 12658 ENDDO 12659 ENDDO 12660 12661 loop blocking will transform the loop as if the user had 12662 written: 12663 DO II = 1, N, 64 12664 DO JJ = 1, M, 64 12665 DO I = II, min (II + 63, N) 12666 DO J = JJ, min (JJ + 63, M) 12667 A(J, I) = B(I) + C(J) 12668 ENDDO 12669 ENDDO 12670 ENDDO 12671 ENDDO 12672 12673 which can be beneficial when M is larger than the caches, 12674 because the innermost loop will iterate over a smaller amount 12675 of data that can be kept in the caches. 12676 * A new register allocator has replaced the old one. It is called 12677 integrated register allocator (IRA) because coalescing, register 12678 live range splitting, and hard register preferencing are done 12679 on-the-fly during coloring. It also has better integration with the 12680 reload pass. IRA is a regional register allocator which uses modern 12681 Chaitin-Briggs coloring instead of Chow's priority coloring used in 12682 the old register allocator. More info about IRA internals and 12683 options can be found in the GCC manuals. 12684 * A new instruction scheduler and software pipeliner, based on the 12685 selective scheduling approach, has been added. The new pass 12686 performs instruction unification, register renaming, substitution 12687 through register copies, and speculation during scheduling. The 12688 software pipeliner is able to pipeline non-countable loops. The new 12689 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 12690 it is available for the Intel Itanium platform working by default 12691 as the second scheduling pass (after register allocation) at the 12692 -O3 optimization level. 12693 * When using -fprofile-generate with a multi-threaded program, the 12694 profile counts may be slightly wrong due to race conditions. The 12695 new -fprofile-correction option directs the compiler to apply 12696 heuristics to smooth out the inconsistencies. By default the 12697 compiler will give an error message when it finds an inconsistent 12698 profile. 12699 * The new -fprofile-dir=PATH option permits setting the directory 12700 where profile data files are stored when using -fprofile-generate 12701 and friends, and the directory used when reading profile data files 12702 using -fprofile-use and friends. 12703 12704 New warning options 12705 12706 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 12707 warning if any stack frame is larger than NUMBER bytes. This may be 12708 used to help ensure that code fits within a limited amount of stack 12709 space. 12710 * The command-line option -Wlarger-than-N is now written as 12711 -Wlarger-than=N and the old form is deprecated. 12712 * The new -Wno-mudflap option disables warnings about constructs 12713 which can not be instrumented when using -fmudflap. 12714 12715 New Languages and Language specific improvements 12716 12717 * Version 3.0 of the OpenMP specification is now supported for the C, 12718 C++, and Fortran compilers. 12719 * New character data types, per [5]TR 19769: New character types in 12720 C, are now supported for the C compiler in -std=gnu99 mode, as 12721 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 12722 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 12723 12724 C family 12725 12726 * A new optimize attribute was added to allow programmers to change 12727 the optimization level and particular optimization options for an 12728 individual function. You can also change the optimization options 12729 via the GCC optimize pragma for functions defined after the pragma. 12730 The GCC push_options pragma and the GCC pop_options pragma allow 12731 you temporarily save and restore the options used. The GCC 12732 reset_options pragma restores the options to what was specified on 12733 the command line. 12734 * Uninitialized warnings do not require enabling optimization 12735 anymore, that is, -Wuninitialized can be used together with -O0. 12736 Nonetheless, the warnings given by -Wuninitialized will probably be 12737 more accurate if optimization is enabled. 12738 * -Wparentheses now warns about expressions such as (!x | y) and (!x 12739 & y). Using explicit parentheses, such as in ((!x) | y), silences 12740 this warning. 12741 * -Wsequence-point now warns within if, while,do while and for 12742 conditions, and within for begin/end expressions. 12743 * A new option -dU is available to dump definitions of preprocessor 12744 macros that are tested or expanded. 12745 12746 C++ 12747 12748 * [6]Improved experimental support for the upcoming ISO C++ standard, 12749 C++0x. Including support for auto, inline namespaces, generalized 12750 initializer lists, defaulted and deleted functions, new character 12751 types, and scoped enums. 12752 * Those errors that may be downgraded to warnings to build legacy 12753 code now mention -fpermissive when -fdiagnostics-show-option is 12754 enabled. 12755 * -Wconversion now warns if the result of a static_cast to enumeral 12756 type is unspecified because the value is outside the range of the 12757 enumeral type. 12758 * -Wuninitialized now warns if a non-static reference or non-static 12759 const member appears in a class without constructors. 12760 * G++ now properly implements value-initialization, so objects with 12761 an initializer of () and an implicitly defined default constructor 12762 will be zero-initialized before the default constructor is called. 12763 12764 Runtime Library (libstdc++) 12765 12766 * Improved experimental support for the upcoming ISO C++ standard, 12767 C++0x, including: 12768 + Support for <chrono>, <condition_variable>, <cstdatomic>, 12769 <forward_list>, <initializer_list>, <mutex>, <ratio>, 12770 <system_error>, and <thread>. 12771 + unique_ptr, <algorithm> additions, exception propagation, and 12772 support for the new character types in <string> and <limits>. 12773 + Existing facilities now exploit initializer lists, defaulted 12774 and deleted functions, and the newly implemented core C++0x 12775 features. 12776 + Some standard containers are more efficient together with 12777 stateful allocators, i.e., no allocator is constructed on the 12778 fly at element construction time. 12779 * Experimental support for non-standard pointer types in containers. 12780 * The long standing libstdc++/30928 has been fixed for targets 12781 running glibc 2.10 or later. 12782 * As usual, many small and larger bug fixes, in particular quite a 12783 few corner cases in <locale>. 12784 12785 Fortran 12786 12787 * GNU Fortran now employs libcpp directly instead of using cc1 as an 12788 external preprocessor. The [7]-cpp option was added to allow manual 12789 invocation of the preprocessor without relying on filename 12790 extensions. 12791 * The [8]-Warray-temporaries option warns about array temporaries 12792 generated by the compiler, as an aid to optimization. 12793 * The [9]-fcheck-array-temporaries option has been added, printing a 12794 notification at run time, when an array temporary had to be created 12795 for an function argument. Contrary to -Warray-temporaries the 12796 warning is only printed if the array is noncontiguous. 12797 * Improved generation of DWARF debugging symbols 12798 * If using an intrinsic not part of the selected standard (via -std= 12799 and -fall-intrinsics) gfortran will now treat it as if this 12800 procedure were declared EXTERNAL and try to link to a user-supplied 12801 procedure. -Wintrinsics-std will warn whenever this happens. The 12802 now-useless option -Wnonstd-intrinsic was removed. 12803 * The flag -falign-commons has been added to control the alignment of 12804 variables in COMMON blocks, which is enabled by default in line 12805 with previous GCC version. Using -fno-align-commons one can force 12806 commons to be contiguous in memory as required by the Fortran 12807 standard, however, this slows down the memory access. The option 12808 -Walign-commons, which is enabled by default, warns when padding 12809 bytes were added for alignment. The proper solution is to sort the 12810 common objects by decreasing storage size, which avoids the 12811 alignment problems. 12812 * Fortran 2003 support has been extended: 12813 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 12814 now supported (except internal reads from/writes to wide 12815 strings). [10]-fbackslash now supports also \unnnn and 12816 \Unnnnnnnn to enter Unicode characters. 12817 + Asynchronous I/O (implemented as synchronous I/O) and the 12818 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 12819 are now supported in I/O statements. 12820 + Support for Fortran 2003 structure constructors and for array 12821 constructor with typespec has been added. 12822 + Procedure Pointers (but not yet as component in derived types 12823 and as function results) are now supported. 12824 + Abstract types, type extension, and type-bound procedures 12825 (both PROCEDURE and GENERIC but not as operators). Note: As 12826 CLASS/polymorphyic types are not implemented, type-bound 12827 procedures with PASS accept as non-standard extension TYPE 12828 arguments. 12829 * Fortran 2008 support has been added: 12830 + The -std=f2008 option and support for the file extensions 12831 .f2008 and .F2008 has been added. 12832 + The g0 format descriptor is now supported. 12833 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 12834 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 12835 are now available (some of them existed as GNU extension 12836 before). Note: The hyperbolic functions are not yet supporting 12837 complex arguments and the three- argument version of BESSEL_*N 12838 is not available. 12839 + The bit intrinsics LEADZ and TRAILZ have been added. 12840 12841 Java (GCJ) 12842 12843 Ada 12844 12845 * The Ada runtime now supports multilibs on many platforms including 12846 x86_64, SPARC and PowerPC. Their build is enabled by default. 12847 12848 New Targets and Target Specific Improvements 12849 12850 ARM 12851 12852 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 12853 Cortex-R4F processors and has many other improvements to 12854 optimization for ARM processors. 12855 * GCC now supports the VFPv3 variant with 16 double-precision 12856 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 12857 renamed to -mfpu=vfpv3. 12858 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 12859 erratum on Cortex-M3 processors. 12860 * GCC now supports the __sync_* atomic operations for ARM EABI 12861 GNU/Linux. 12862 * The section anchors optimization is now enabled by default when 12863 optimizing for ARM. 12864 * GCC now uses a new EABI-compatible profiling interface for EABI 12865 targets. This requires a function __gnu_mcount_nc, which is 12866 provided by GNU libc versions 2.8 and later. 12867 12868 AVR 12869 12870 * The -mno-tablejump option has been deprecated because it has the 12871 same effect as the -fno-jump-tables option. 12872 * Added support for these new AVR devices: 12873 + ATA6289 12874 + ATtiny13A 12875 + ATtiny87 12876 + ATtiny167 12877 + ATtiny327 12878 + ATmega8C1 12879 + ATmega16C1 12880 + ATmega32C1 12881 + ATmega8M1 12882 + ATmega16M1 12883 + ATmega32M1 12884 + ATmega32U4 12885 + ATmega16HVB 12886 + ATmega4HVD 12887 + ATmega8HVD 12888 + ATmega64C1 12889 + ATmega64M1 12890 + ATmega16U4 12891 + ATmega32U6 12892 + ATmega128RFA1 12893 + AT90PWM81 12894 + AT90SCR100 12895 + M3000F 12896 + M3000S 12897 + M3001B 12898 12899 IA-32/x86-64 12900 12901 * Support for Intel AES built-in functions and code generation is 12902 available via -maes. 12903 * Support for Intel PCLMUL built-in function and code generation is 12904 available via -mpclmul. 12905 * Support for Intel AVX built-in functions and code generation is 12906 available via -mavx. 12907 * Automatically align the stack for local variables with alignment 12908 requirement. 12909 * GCC can now utilize the SVML library for vectorizing calls to a set 12910 of C99 functions if -mveclibabi=svml is specified and you link to 12911 an SVML ABI compatible library. 12912 * On x86-64, the ABI has been changed in the following cases to 12913 conform to the x86-64 ABI: 12914 + Passing/returning structures with flexible array member: 12915 struct foo 12916 { 12917 int i; 12918 int flex[]; 12919 }; 12920 + Passing/returning structures with complex float member: 12921 struct foo 12922 { 12923 int i; 12924 __complex__ float f; 12925 }; 12926 + Passing/returning unions with long double member: 12927 union foo 12928 { 12929 int x; 12930 long double ld; 12931 }; 12932 Code built with previous versions of GCC that uses any of these is 12933 not compatible with code built with GCC 4.4.0 or later. 12934 * A new target attribute was added to allow programmers to change the 12935 target options like -msse2 or -march=k8 for an individual function. 12936 You can also change the target options via the GCC target pragma 12937 for functions defined after the pragma. 12938 * GCC can now be configured with options --with-arch-32, 12939 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 12940 --with-tune-64 to control the default optimization separately for 12941 32-bit and 64-bit modes. 12942 12943 IA-32/IA64 12944 12945 * Support for __float128 (TFmode) IEEE quad type and corresponding 12946 TCmode IEEE complex quad type is available via the soft-fp library 12947 on IA-32/IA64 targets. This includes basic arithmetic operations 12948 (addition, subtraction, negation, multiplication and division) on 12949 __float128 real and TCmode complex values, the full set of IEEE 12950 comparisons between __float128 values, conversions to and from 12951 float, double and long double floating point types, as well as 12952 conversions to and from signed or unsigned integer, signed or 12953 unsigned long integer and signed or unsigned quad (TImode, IA64 12954 only) integer types. Additionally, all operations generate the full 12955 set of IEEE exceptions and support the full set of IEEE rounding 12956 modes. 12957 12958 M68K/ColdFire 12959 12960 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 12961 processors. (Scheduling support for ColdFire V2 processors was 12962 added in GCC 4.3.) 12963 * GCC now supports the -mxgot option to support programs requiring 12964 many GOT entries on ColdFire. 12965 * The m68k-*-linux-gnu target now builds multilibs by default. 12966 12967 MIPS 12968 12969 * MIPS Technologies have extended the original MIPS SVR4 ABI to 12970 include support for procedure linkage tables (PLTs) and copy 12971 relocations. These extensions allow GNU/Linux executables to use a 12972 significantly more efficient code model than the one defined by the 12973 original ABI. 12974 GCC support for this code model is available via a new command-line 12975 option, -mplt. There is also a new configure-time option, 12976 --with-mips-plt, to make -mplt the default. 12977 The new code model requires support from the assembler, the linker, 12978 and the runtime C library. This support is available in binutils 12979 2.19 and GLIBC 2.9. 12980 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 12981 and 32-bit GNU/Linux shared libraries. This feature requires GNU 12982 binutils 2.19 or above. 12983 * Support for RMI's XLR processor is now available through the 12984 -march=xlr and -mtune=xlr options. 12985 * 64-bit targets can now perform 128-bit multiplications inline, 12986 instead of relying on a libgcc function. 12987 * Native GNU/Linux toolchains now support -march=native and 12988 -mtune=native, which select the host processor. 12989 * GCC now supports the R10K, R12K, R14K and R16K processors. The 12990 canonical -march= and -mtune= names for these processors are 12991 r10000, r12000, r14000 and r16000 respectively. 12992 * GCC can now work around the side effects of speculative execution 12993 on R10K processors. Please see the documentation of the 12994 -mr10k-cache-barrier option for details. 12995 * Support for the MIPS64 Release 2 instruction set has been added. 12996 The option -march=mips64r2 enables generation of these 12997 instructions. 12998 * GCC now supports Cavium Networks' Octeon processor. This support is 12999 available through the -march=octeon and -mtune=octeon options. 13000 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 13001 canonical -march= and -mtune= names for these processors are 13002 loongson2e and loongson2f. 13003 13004 picochip 13005 13006 Picochip is a 16-bit processor. A typical picoChip contains over 250 13007 small cores, each with small amounts of memory. There are three 13008 processor variants (STAN, MEM and CTRL) with different instruction sets 13009 and memory configurations and they can be chosen using the -mae option. 13010 13011 This port is intended to be a "C" only port. 13012 13013 Power Architecture and PowerPC 13014 13015 * GCC now supports the e300c2, e300c3 and e500mc processors. 13016 * GCC now supports Xilinx processors with a single-precision FPU. 13017 * Decimal floating point is now supported for e500 processors. 13018 13019 S/390, zSeries and System z9/z10 13020 13021 * Support for the IBM System z10 EC/BC processor has been added. When 13022 using the -march=z10 option, the compiler will generate code making 13023 use of instructions provided by the General-Instruction-Extension 13024 Facility and the Execute-Extension Facility. 13025 13026 VxWorks 13027 13028 * GCC now supports the thread-local storage mechanism used on 13029 VxWorks. 13030 13031 Xtensa 13032 13033 * GCC now supports thread-local storage (TLS) for Xtensa processor 13034 configurations that include the Thread Pointer option. TLS also 13035 requires support from the assembler and linker; this support is 13036 provided in the GNU binutils beginning with version 2.19. 13037 13038 Documentation improvements 13039 13040 Other significant improvements 13041 13042 GCC 4.4.1 13043 13044 This is the [11]list of problem reports (PRs) from GCC's bug tracking 13045 system that are known to be fixed in the 4.4.1 release. This list might 13046 not be complete (that is, it is possible that some PRs that have been 13047 fixed are not listed here). 13048 13049 GCC 4.4.2 13050 13051 This is the [12]list of problem reports (PRs) from GCC's bug tracking 13052 system that are known to be fixed in the 4.4.2 release. This list might 13053 not be complete (that is, it is possible that some PRs that have been 13054 fixed are not listed here). 13055 13056 GCC 4.4.3 13057 13058 This is the [13]list of problem reports (PRs) from GCC's bug tracking 13059 system that are known to be fixed in the 4.4.3 release. This list might 13060 not be complete (that is, it is possible that some PRs that have been 13061 fixed are not listed here). 13062 13063 GCC 4.4.4 13064 13065 This is the [14]list of problem reports (PRs) from GCC's bug tracking 13066 system that are known to be fixed in the 4.4.4 release. This list might 13067 not be complete (that is, it is possible that some PRs that have been 13068 fixed are not listed here). 13069 13070 GCC 4.4.5 13071 13072 This is the [15]list of problem reports (PRs) from GCC's bug tracking 13073 system that are known to be fixed in the 4.4.5 release. This list might 13074 not be complete (that is, it is possible that some PRs that have been 13075 fixed are not listed here). 13076 13077 GCC 4.4.6 13078 13079 This is the [16]list of problem reports (PRs) from GCC's bug tracking 13080 system that are known to be fixed in the 4.4.6 release. This list might 13081 not be complete (that is, it is possible that some PRs that have been 13082 fixed are not listed here). 13083 13084 GCC 4.4.7 13085 13086 This is the [17]list of problem reports (PRs) from GCC's bug tracking 13087 system that are known to be fixed in the 4.4.7 release. This list might 13088 not be complete (that is, it is possible that some PRs that have been 13089 fixed are not listed here). 13090 13091 13092 For questions related to the use of GCC, please consult these web 13093 pages and the [18]GCC manuals. If that fails, the 13094 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13095 web pages and the development of GCC are welcome on our developer 13096 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 13097 archives. 13098 13099 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 13100 distribution of this entire article is permitted in any medium, 13101 provided this notice is preserved. 13102 13103 These pages are [23]maintained by the GCC team. Last modified 13104 2022-11-01. 13105 13106 References 13107 13108 1. https://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 13109 2. https://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 13110 3. https://gcc.gnu.org/gcc-4.4/porting_to.html 13111 4. https://gcc.gnu.org/wiki/Graphite 13112 5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 13113 6. https://gcc.gnu.org/gcc-4.4/cxx0x_status.html 13114 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 13115 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 13116 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 13117 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 13118 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 13119 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 13120 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 13121 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 13122 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 13123 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 13124 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 13125 18. https://gcc.gnu.org/onlinedocs/ 13126 19. mailto:gcc-help (a] gcc.gnu.org 13127 20. mailto:gcc (a] gcc.gnu.org 13128 21. https://gcc.gnu.org/lists.html 13129 22. https://www.fsf.org/ 13130 23. https://gcc.gnu.org/about.html 13131 ====================================================================== 13132 http://gcc.gnu.org/gcc-4.3/index.html 13133 GCC 4.3 Release Series 13134 13135 (This release series is no longer supported.) 13136 13137 Jun 27, 2011 13138 13139 The [1]GNU project and the GCC developers are pleased to announce the 13140 release of GCC 4.3.6. 13141 13142 This release is a bug-fix release, containing fixes for regressions in 13143 GCC 4.3.5 relative to previous releases of GCC. 13144 13145 Release History 13146 13147 GCC 4.3.6 13148 Jun 27, 2011 ([2]changes) 13149 13150 GCC 4.3.5 13151 May 22, 2010 ([3]changes) 13152 13153 GCC 4.3.4 13154 August 4, 2009 ([4]changes) 13155 13156 GCC 4.3.3 13157 January 24, 2009 ([5]changes) 13158 13159 GCC 4.3.2 13160 August 27, 2008 ([6]changes) 13161 13162 GCC 4.3.1 13163 June 6, 2008 ([7]changes) 13164 13165 GCC 4.3.0 13166 March 5, 2008 ([8]changes) 13167 13168 References and Acknowledgements 13169 13170 GCC used to stand for the GNU C Compiler, but since the compiler 13171 supports several other languages aside from C, it now stands for the 13172 GNU Compiler Collection. 13173 13174 A list of [9]successful builds is updated as new information becomes 13175 available. 13176 13177 The GCC developers would like to thank the numerous people that have 13178 contributed new features, improvements, bug fixes, and other changes as 13179 well as test results to GCC. This [10]amazing group of volunteers is 13180 what makes GCC successful. 13181 13182 For additional information about GCC please refer to the [11]GCC 13183 project web site or contact the [12]GCC development mailing list. 13184 13185 To obtain GCC please use [13]our mirror sites or [14]our version 13186 control system. 13187 13188 13189 For questions related to the use of GCC, please consult these web 13190 pages and the [15]GCC manuals. If that fails, the 13191 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13192 web pages and the development of GCC are welcome on our developer 13193 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 13194 archives. 13195 13196 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 13197 distribution of this entire article is permitted in any medium, 13198 provided this notice is preserved. 13199 13200 These pages are [20]maintained by the GCC team. Last modified 13201 2022-10-26. 13202 13203 References 13204 13205 1. http://www.gnu.org/ 13206 2. https://gcc.gnu.org/gcc-4.3/changes.html 13207 3. https://gcc.gnu.org/gcc-4.3/changes.html 13208 4. https://gcc.gnu.org/gcc-4.3/changes.html 13209 5. https://gcc.gnu.org/gcc-4.3/changes.html 13210 6. https://gcc.gnu.org/gcc-4.3/changes.html 13211 7. https://gcc.gnu.org/gcc-4.3/changes.html 13212 8. https://gcc.gnu.org/gcc-4.3/changes.html 13213 9. https://gcc.gnu.org/gcc-4.3/buildstat.html 13214 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13215 11. https://gcc.gnu.org/index.html 13216 12. mailto:gcc (a] gcc.gnu.org 13217 13. https://gcc.gnu.org/mirrors.html 13218 14. https://gcc.gnu.org/git.html 13219 15. https://gcc.gnu.org/onlinedocs/ 13220 16. mailto:gcc-help (a] gcc.gnu.org 13221 17. mailto:gcc (a] gcc.gnu.org 13222 18. https://gcc.gnu.org/lists.html 13223 19. https://www.fsf.org/ 13224 20. https://gcc.gnu.org/about.html 13225 ====================================================================== 13226 http://gcc.gnu.org/gcc-4.3/changes.html 13227 GCC 4.3 Release Series 13228 Changes, New Features, and Fixes 13229 13230 The latest release in the 4.3 release series is [1]GCC 4.3.5. 13231 13232 Caveats 13233 13234 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 13235 various front-end languages it supports. See the [4]prerequisites 13236 page for version requirements. 13237 * ColdFire targets now treat long double as having the same format as 13238 double. In earlier versions of GCC, they used the 68881 long double 13239 format instead. 13240 * The m68k-uclinux target now uses the same calling conventions as 13241 m68k-linux-gnu. You can select the original calling conventions by 13242 configuring for m68k-uclinuxoldabi instead. Note that 13243 m68k-uclinuxoldabi also retains the original 80-bit long double on 13244 ColdFire targets. 13245 * The -fforce-mem option has been removed because it has had no 13246 effect in the last few GCC releases. 13247 * The i386 -msvr3-shlib option has been removed since it is no longer 13248 used. 13249 * Fastcall for i386 has been changed not to pass aggregate arguments 13250 in registers, following Microsoft compilers. 13251 * Support for the AOF assembler has been removed from the ARM back 13252 end; this affects only the targets arm-semi-aof and armel-semi-aof, 13253 which are no longer recognized. We removed these targets without a 13254 deprecation period because we discovered that they have been 13255 unusable since GCC 4.0.0. 13256 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 13257 has been removed. This support had been deprecated since GCC 4.0.0. 13258 * Support for a number of older systems and recently unmaintained or 13259 untested target ports of GCC has been declared obsolete in GCC 4.3. 13260 Unless there is activity to revive them, the next release of GCC 13261 will have their sources permanently removed. 13262 All GCC ports for the following processor architectures have been 13263 declared obsolete: 13264 + Morpho MT (mt-*) 13265 The following aliases for processor architectures have been 13266 declared obsolete. Users should use the indicated generic target 13267 names instead, with compile-time options such as -mcpu or 13268 configure-time options such as --with-cpu to control the 13269 configuration more precisely. 13270 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 13271 instead). 13272 + parisc*-*-* (use hppa*-*-* instead). 13273 + m680[012]0-*-* (use m68k-*-* instead). 13274 All GCC ports for the following operating systems have been 13275 declared obsolete: 13276 + BeOS (*-*-beos*) 13277 + kaOS (*-*-kaos*) 13278 + GNU/Linux using the a.out object format (*-*-linux*aout*) 13279 + GNU/Linux using version 1 of the GNU C Library 13280 (*-*-linux*libc1*) 13281 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 13282 *-*-solaris2.[0-6].*) 13283 + Miscellaneous System V (*-*-sysv*) 13284 + WindISS (*-*-windiss*) 13285 Also, those for some individual systems on particular architectures 13286 have been obsoleted: 13287 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 13288 + CRIS with a.out object format (cris-*-aout) 13289 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 13290 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 13291 + PRO on PA-RISC (hppa1.1-*-pro*) 13292 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 13293 i[34567]86-sequent-sysv4*) 13294 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 13295 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 13296 was previously [5]removed in 2001, leaving only the support 13297 for UWIN as a target now being deprecated) 13298 + ChorusOS on PowerPC (powerpc-*-chorusos*) 13299 + All VAX configurations apart from NetBSD and OpenBSD 13300 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 13301 * The [6]-Wconversion option has been modified. Its purpose now is to 13302 warn for implicit conversions that may alter a value. This new 13303 behavior is available for both C and C++. Warnings about 13304 conversions between signed and unsigned integers can be disabled by 13305 using -Wno-sign-conversion. In C++, they are disabled by default 13306 unless -Wsign-conversion is explicitly requested. The old behavior 13307 of -Wconversion, that is, warn for prototypes causing a type 13308 conversion that is different from what would happen to the same 13309 argument in the absence of a prototype, has been moved to a new 13310 option -Wtraditional-conversion, which is only available for C. 13311 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 13312 been removed because they were deprecated for more than 3 GCC major 13313 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 13314 -mtune=pentiumpro as a replacement. 13315 * The -funsafe-math-optimizations option now automatically turns on 13316 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 13317 reassociation and thus may introduce or remove traps. 13318 * The -ftree-vectorize option is now on by default under -O3. In 13319 order to generate code for a SIMD extension, it has to be enabled 13320 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 13321 i?86 and x86_64. 13322 * More information on porting to GCC 4.3 from previous versions of 13323 GCC can be found in the [7]porting guide for this release. 13324 13325 General Optimizer Improvements 13326 13327 * The GCC middle-end has been integrated with the MPFR library. This 13328 allows GCC to evaluate and replace at compile-time calls to 13329 built-in math functions having constant arguments with their 13330 mathematically equivalent results. In making use of MPFR, GCC can 13331 generate correct results regardless of the math library 13332 implementation or floating point precision of the host platform. 13333 This also allows GCC to generate identical results regardless of 13334 whether one compiles in native or cross-compile configurations to a 13335 particular target. The following built-in functions take advantage 13336 of this new capability: acos, acosh, asin, asinh, atan2, atan, 13337 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 13338 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 13339 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 13340 tan, tanh, tgamma, y0, y1 and yn. The float and long double 13341 variants of these functions (e.g. sinf and sinl) are also handled. 13342 The sqrt and cabs functions with constant arguments were already 13343 optimized in prior GCC releases. Now they also use MPFR. 13344 * A new forward propagation pass on RTL was added. The new pass 13345 replaces several slower transformations, resulting in compile-time 13346 improvements as well as better code generation in some cases. 13347 * A new command-line switch -frecord-gcc-switches has been added to 13348 GCC, although it is only enabled for some targets. The switch 13349 causes the command line that was used to invoke the compiler to be 13350 recorded into the object file that is being created. The exact 13351 format of this recording is target and binary file format 13352 dependent, but it usually takes the form of a note section 13353 containing ASCII text. The switch is related to the -fverbose-asm 13354 switch, but that one only records the information in the assembler 13355 output file as comments, so the information never reaches the 13356 object file. 13357 * The inliner heuristic is now aware of stack frame consumption. New 13358 command-line parameters --param large-stack-frame and --param 13359 large-stack-frame-growth can be used to limit stack frame size 13360 growth caused by inlining. 13361 * During feedback directed optimizations, the expected block size the 13362 memcpy, memset and bzero functions operate on is discovered and for 13363 cases of commonly used small sizes, specialized inline code is 13364 generated. 13365 * __builtin_expect no longer requires its argument to be a compile 13366 time constant. 13367 * Interprocedural optimization was reorganized to work on functions 13368 in SSA form. This enables more precise and cheaper dataflow 13369 analysis and makes writing interprocedural optimizations easier. 13370 The following improvements have been implemented on top of this 13371 framework: 13372 + Pre-inline optimization: Selected local optimization passes 13373 are run before the inliner (and other interprocedural passes) 13374 are executed. This significantly improves the accuracy of code 13375 growth estimates used by the inliner and reduces the overall 13376 memory footprint for large compilation units. 13377 + Early inlining (a simple bottom-up inliner pass inlining only 13378 functions whose body is smaller than the expected call 13379 overhead) is now executed with the early optimization passes, 13380 thus inlining already optimized function bodies into an 13381 unoptimized function that is subsequently optimized by early 13382 optimizers. This enables the compiler to quickly eliminate 13383 abstraction penalty in C++ programs. 13384 + Interprocedural constant propagation now operate on SSA form 13385 increasing accuracy of the analysis. 13386 * A new internal representation for GIMPLE statements has been 13387 contributed, resulting in compile-time memory savings. 13388 * The vectorizer was enhanced to support vectorization of outer 13389 loops, intra-iteration parallelism (loop-aware SLP), vectorization 13390 of strided accesses and loops with multiple data-types. Run-time 13391 dependency testing using loop versioning was added. The cost model, 13392 turned on by -fvect-cost-model, was developed. 13393 13394 New Languages and Language specific improvements 13395 13396 * We have added new command-line options 13397 -finstrument-functions-exclude-function-list and 13398 -finstrument-functions-exclude-file-list. They provide more control 13399 over which functions are annotated by the -finstrument-functions 13400 option. 13401 13402 C family 13403 13404 * Implicit conversions between generic vector types are now only 13405 permitted when the two vectors in question have the same number of 13406 elements and compatible element types. (Note that the restriction 13407 involves compatible element types, not implicitly-convertible 13408 element types: thus, a vector type with element type int may not be 13409 implicitly converted to a vector type with element type unsigned 13410 int.) This restriction, which is in line with specifications for 13411 SIMD architectures such as AltiVec, may be relaxed using the flag 13412 -flax-vector-conversions. This flag is intended only as a 13413 compatibility measure and should not be used for new code. 13414 * -Warray-bounds has been added and is now enabled by default for 13415 -Wall . It produces warnings for array subscripts that can be 13416 determined at compile time to be always out of bounds. 13417 -Wno-array-bounds will disable the warning. 13418 * The constructor and destructor function attributes now accept 13419 optional priority arguments which control the order in which the 13420 constructor and destructor functions are run. 13421 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 13422 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 13423 -Wignored-qualifiers have been added for finer control of the 13424 diverse warnings enabled by -Wextra. 13425 * A new function attribute alloc_size has been added to mark up 13426 malloc style functions. For constant sized allocations this can be 13427 used to find out the size of the returned pointer using the 13428 __builtin_object_size() function for buffer overflow checking and 13429 similar. This supplements the already built-in malloc and calloc 13430 constant size handling. 13431 * Integer constants written in binary are now supported as a GCC 13432 extension. They consist of a prefix 0b or 0B, followed by a 13433 sequence of 0 and 1 digits. 13434 * A new predefined macro __COUNTER__ has been added. It expands to 13435 sequential integral values starting from 0. In conjunction with the 13436 ## operator, this provides a convenient means to generate unique 13437 identifiers. 13438 * A new command-line option -fdirectives-only has been added. It 13439 enables a special preprocessing mode which improves the performance 13440 of applications like distcc and ccache. 13441 * Fixed-point data types and operators have been added. They are 13442 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 13443 Currently, only MIPS targets are supported. 13444 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 13445 N1241, is now supported as a GCC extension to C for targets 13446 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 13447 and x86_64-*-linux-gnu. The feature introduces new data types 13448 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 13449 DD, and DL. 13450 13451 C++ 13452 13453 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 13454 * -Wc++0x-compat has been added and is now enabled by default for 13455 -Wall. It produces warnings for constructs whose meaning differs 13456 between ISO C++ 1998 and C++0x. 13457 * The -Wparentheses option now works for C++ as it does for C. It 13458 warns if parentheses are omitted when operators with confusing 13459 precedence are nested. It also warns about ambiguous else 13460 statements. Since -Wparentheses is enabled by -Wall, this may cause 13461 additional warnings with existing C++ code which uses -Wall. These 13462 new warnings may be disabled by using -Wall -Wno-parentheses. 13463 * The -Wmissing-declarations now works for C++ as it does for C. 13464 * The -fvisibility-ms-compat flag was added, to make it easier to 13465 port larger projects using shared libraries from Microsoft's Visual 13466 Studio to ELF and Mach-O systems. 13467 * C++ attribute handling has been overhauled for template arguments 13468 (ie dependent types). In particular, __attribute__((aligned(T))); 13469 works for C++ types. 13470 13471 Runtime Library (libstdc++) 13472 13473 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 13474 * Support for TR1 mathematical special functions and regular 13475 expressions. 13476 * Default what implementations give more elaborate exception strings 13477 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 13478 * Header dependencies have been streamlined, reducing unnecessary 13479 includes and pre-processed bloat. 13480 * Variadic template implementations of items in <tuple> and 13481 <functional>. 13482 * An experimental [11]parallel mode has been added. This is a 13483 parallel implementation of many C++ Standard library algorithms, 13484 like std::accumulate, std::for_each, std::transform, or std::sort, 13485 to give but four examples. These algorithms can be substituted for 13486 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 13487 or all existing algorithms can be transformed via the 13488 -D_GLIBCXX_PARALLEL macro. 13489 * Debug mode versions of classes in <unordered_set> and 13490 <unordered_map>. 13491 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 13492 now <backward/hash_set> and <backward/hash_map>. This code: 13493 #include <ext/hash_set> 13494 __gnu_cxx::hash_set<int> s; 13495 13496 Can be transformed (in order of preference) to: 13497 #include <tr1/unordered_set> 13498 std::tr1::unordered_set<int> s; 13499 13500 or 13501 #include <backward/hash_set> 13502 __gnu_cxx::hash_set<int> s; 13503 13504 Similar transformations apply to __gnu_cxx::hash_map, 13505 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 13506 __gnu_cxx::hash_multiset. 13507 13508 Fortran 13509 13510 * Due to the fact that the GMP and MPFR libraries are required for 13511 all languages, Fortran is no longer special in this regard and is 13512 available by default. 13513 * The [12]-fexternal-blas option has been added, which generates 13514 calls to BLAS routines for intrinsic matrix operations such as 13515 matmul rather than using the built-in algorithms. 13516 * Support to give a backtrace (compiler flag -fbacktrace or 13517 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 13518 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 13519 run-time error occured. 13520 * GNU Fortran now defines __GFORTRAN__ when it runs the C 13521 preprocessor (CPP). 13522 * The [13]-finit-local-zero, -finit-real, -finit-integer, 13523 -finit-character, and -finit-logical options have been added, which 13524 can be used to initialize local variables. 13525 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 13526 which calculate the Gamma function and its logarithm. Use EXTERNAL 13527 gamma if you want to use your own gamma function. 13528 * GNU Fortran now regards the backslash character as literal (as 13529 required by the Fortran 2003 standard); using [16]-fbackslash GNU 13530 Fortran interprets backslashes as C-style escape characters. 13531 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 13532 literal constants has been changed. Before they were always 13533 interpreted as integer; now they are bit-wise transferred as 13534 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 13535 2003 standard, and for real and complex variables in DATA 13536 statements or when directly assigned to real and complex variables. 13537 Everywhere else and especially in expressions they are still 13538 regarded as integer constants. 13539 * Fortran 2003 support has been extended: 13540 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 13541 + Pointer intent 13542 + Intrinsic module ISO_ENV_FORTRAN 13543 + Interoperability with C (ISO C Bindings) 13544 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 13545 attribute) 13546 + Fortran 2003 BOZ 13547 13548 Java (GCJ) 13549 13550 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 13551 This enables the use of all 1.5 language features, and fixes most 13552 existing front end bugs. 13553 * libgcj now supports all 1.5 language features which require runtime 13554 support: foreach, enum, annotations, generics, and auto-boxing. 13555 * We've made many changes to the tools shipped with gcj. 13556 + The old jv-scan tool has been removed. This tool never really 13557 worked properly. There is no replacement. 13558 + gcjh has been rewritten. Some of its more obscure options no 13559 longer work, but are still recognized in an attempt at 13560 compatibility. gjavah is a new program with similar 13561 functionality but different command-line options. 13562 + grmic and grmiregistry have been rewritten. grmid has been 13563 added. 13564 + gjar replaces the old fastjar. 13565 + gjarsigner (used for signing jars), gkeytool (used for key 13566 management), gorbd (for CORBA), gserialver (computes 13567 serialization UIDs), and gtnameserv (also for CORBA) are now 13568 installed. 13569 * The ability to dump the contents of the java run time heap to a 13570 file for off-line analysis has been added. The heap dumps may be 13571 analyzed with the new gc-analyze tool. They may be generated on 13572 out-of-memory conditions or on demand and are controlled by the new 13573 run time class gnu.gcj.util.GCInfo. 13574 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 13575 provide correct, updated, timezone information. This means that 13576 packagers no longer have to update libgcj when a time zone change 13577 is published. 13578 13579 New Targets and Target Specific Improvements 13580 13581 IA-32/x86-64 13582 13583 * Tuning for Intel Core 2 processors is available via -mtune=core2 13584 and -march=core2. 13585 * Tuning for AMD Geode processors is available via -mtune=geode and 13586 -march=geode. 13587 * Code generation of block move (memcpy) and block set (memset) was 13588 rewritten. GCC can now pick the best algorithm (loop, unrolled 13589 loop, instruction with rep prefix or a library call) based on the 13590 size of the block being copied and the CPU being optimized for. A 13591 new option -minline-stringops-dynamically has been added. With this 13592 option string operations of unknown size are expanded such that 13593 small blocks are copied by in-line code, while for large blocks a 13594 library call is used. This results in faster code than 13595 -minline-all-stringops when the library implementation is capable 13596 of using cache hierarchy hints. The heuristic choosing the 13597 particular algorithm can be overwritten via -mstringop-strategy. 13598 Newly also memset of values different from 0 is inlined. 13599 * GCC no longer places the cld instruction before string operations. 13600 Both i386 and x86-64 ABI documents mandate the direction flag to be 13601 clear at the entry of a function. It is now invalid to set the flag 13602 in asm statement without reseting it afterward. 13603 * Support for SSSE3 built-in functions and code generation are 13604 available via -mssse3. 13605 * Support for SSE4.1 built-in functions and code generation are 13606 available via -msse4.1. 13607 * Support for SSE4.2 built-in functions and code generation are 13608 available via -msse4.2. 13609 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 13610 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 13611 allow explicit control of x87 floating point precision. 13612 * Support for __float128 (TFmode) IEEE quad type and corresponding 13613 TCmode IEEE complex quad type is available via the soft-fp library 13614 on x86_64 targets. This includes basic arithmetic operations 13615 (addition, subtraction, negation, multiplication and division) on 13616 __float128 real and TCmode complex values, the full set of IEEE 13617 comparisons between __float128 values, conversions to and from 13618 float, double and long double floating point types, as well as 13619 conversions to and from signed or unsigned integer, signed or 13620 unsigned long integer and signed or unsigned quad (TImode) integer 13621 types. Additionally, all operations generate the full set of IEEE 13622 exceptions and support the full set of IEEE rounding modes. 13623 * GCC can now utilize the ACML library for vectorizing calls to a set 13624 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 13625 link to an ACML ABI compatible library. 13626 13627 ARM 13628 13629 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 13630 has been added. 13631 13632 CRIS 13633 13634 New features 13635 13636 * Compiler and Library support for the CRIS v32 architecture, as 13637 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 13638 added. 13639 13640 Configuration changes 13641 13642 * The cris-*-elf target now includes support for CRIS v32, including 13643 libraries, through the -march=v32 option. 13644 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 13645 * A new crisv32-*-linux* target defaults to generate code for CRIS 13646 v32. 13647 * The cris-*-aout target has been obsoleted. 13648 13649 Improved support for built-in functions 13650 13651 * GCC can now use the lz and swapwbr instructions to implement the 13652 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 13653 * __builtin_bswap32 is now implemented using the swapwb instruction, 13654 when available. 13655 13656 m68k and ColdFire 13657 13658 New features 13659 13660 * Support for several new ColdFire processors has been added. You can 13661 generate code for them using the new -mcpu option. 13662 * All targets now support ColdFire processors. 13663 * m68k-uclinux targets have improved support for C++ constructors and 13664 destructors, and for shared libraries. 13665 * It is now possible to set breakpoints on the first or last line of 13666 a function, even if there are no statements on that line. 13667 13668 Optimizations 13669 13670 * Support for sibling calls has been added. 13671 * More use is now made of the ColdFire mov3q instruction. 13672 * __builtin_clz is now implemented using the ff1 ColdFire 13673 instruction, when available. 13674 * GCC now honors the -m68010 option. 68010 code now uses clr rather 13675 than move to zero volatile memory. 13676 * 68020 targets and above can now use symbol(index.size*scale) 13677 addresses for indexed array accesses. Earlier compilers would 13678 always load the symbol into a base register first. 13679 13680 Configuration changes 13681 13682 * All m68k and ColdFire targets now allow the default processor to be 13683 set at configure time using --with-cpu. 13684 * A --with-arch configuration option has been added. This option 13685 allows you to restrict a target to ColdFire or non-ColdFire 13686 processors. 13687 13688 Preprocessor macros 13689 13690 * An __mcfv*__ macro is now defined for all ColdFire targets. 13691 (Earlier versions of GCC only defined __mcfv4e__.) 13692 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 13693 * All targets now define __mc68010 and __mc68010__ when generating 13694 68010 code. 13695 13696 Command-line changes 13697 13698 * New command-line options -march, -mcpu, -mtune and -mhard-float 13699 have been added. These options apply to both m68k and ColdFire 13700 targets. 13701 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 13702 versions of -mshort, etc. 13703 * -fforce-addr has been removed. It is now ignored by the compiler. 13704 13705 Other improvements 13706 13707 * ColdFire targets now try to maintain a 4-byte-aligned stack where 13708 possible. 13709 * m68k-uclinux targets now try to avoid situations that lead to the 13710 load-time error: BINFMT_FLAT: reloc outside program. 13711 13712 MIPS 13713 13714 Changes to existing configurations 13715 13716 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 13717 and n64. Every GNU/Linux configuration now builds these libraries 13718 by default. 13719 * GNU/Linux configurations now generate -mno-shared code unless 13720 overridden by -fpic, -fPIC, -fpie or -fPIE. 13721 * mipsisa32*-linux-gnu configurations now generate hard-float code by 13722 default, just like other mipsisa32* and mips*-linux-gnu 13723 configurations. You can build a soft-float version of any 13724 mips*-linux-gnu configuration by passing --with-float=soft to 13725 configure. 13726 * mips-wrs-vxworks now supports run-time processes (RTPs). 13727 13728 Changes to existing command-line options 13729 13730 * The -march and -mtune options no longer accept 24k as a processor 13731 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 13732 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 13733 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 13734 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 13735 24kx, 24kex and 34kx. 13736 13737 New configurations 13738 13739 GCC now supports the following configurations: 13740 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 13741 default. Earlier releases also recognized this configuration, but 13742 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 13743 you can customize any mips*-linux-gnu* configuration to a 13744 particular ISA or processor by passing an appropriate --with-arch 13745 option to configure. 13746 * mipsisa*-sde-elf*, which provides compatibility with MIPS 13747 Technologies' SDE toolchains. The configuration uses the SDE 13748 libraries by default, but you can use it like other newlib-based 13749 ELF configurations by passing --with-newlib to configure. It is the 13750 only configuration besides mips64vr*-elf* to build MIPS16 as well 13751 as non-MIPS16 libraries. 13752 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 13753 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 13754 and 64-bit forms of the EABI. 13755 13756 New processors and application-specific extensions 13757 13758 * Support for the SmartMIPS ASE is available through the new 13759 -msmartmips option. 13760 * Support for revision 2 of the DSP ASE is available through the new 13761 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 13762 indicates the revision of the ASE in use. 13763 * Support for the 4KS and 74K families of processors is available 13764 through the -march and -mtune options. 13765 13766 Improved support for built-in functions 13767 13768 * GCC can now use load-linked, store-conditional and sync 13769 instructions to implement atomic built-in functions such as 13770 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 13771 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 13772 * GCC can now use the clz and dclz instructions to implement the 13773 __builtin_ctz and __builtin_ffs families of functions. 13774 * There is a new __builtin___clear_cache function for flushing the 13775 instruction cache. GCC expands this function inline on MIPS32 13776 revision 2 targets, otherwise it calls the function specified by 13777 -mcache-flush-func. 13778 13779 MIPS16 improvements 13780 13781 * GCC can now compile objects that contain a mixture of MIPS16 and 13782 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 13783 for specifying which mode a function should use. 13784 * A new option called -minterlink-mips16 makes non-MIPS16 code 13785 link-compatible with MIPS16 code. 13786 * After many bug fixes, the long-standing MIPS16 -mhard-float support 13787 should now work fairly reliably. 13788 * GCC can now use the MIPS16e save and restore instructions. 13789 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 13790 with -G0 -fsection-anchors is often smaller than code compiled with 13791 -G8. However, please note that you must usually compile all objects 13792 in your application with the same -G option; see the documentation 13793 of -G for details. 13794 * A new option called-mcode-readable specifies which instructions are 13795 allowed to load from the code segment. -mcode-readable=yes is the 13796 default and says that any instruction may load from the code 13797 segment. The other alternatives are -mcode-readable=pcrel, which 13798 says that only PC-relative MIPS16 instructions may load from the 13799 code segment, and -mcode-readable=no, which says that no 13800 instruction may do so. Please see the documentation for more 13801 details, including example uses. 13802 13803 Small-data improvements 13804 13805 There are three new options for controlling small data: 13806 * -mno-extern-sdata, which disables small-data accesses for 13807 externally-defined variables. Code compiled with -Gn 13808 -mno-extern-sdata will be link-compatible with any -G setting 13809 between -G0 and -Gn inclusive. 13810 * -mno-local-sdata, which disables the use of small-data sections for 13811 data that is not externally visible. This option can be a useful 13812 way of reducing small-data usage in less performance-critical parts 13813 of an application. 13814 * -mno-gpopt, which disables the use of the $gp register while still 13815 honoring the -G limit when placing externally-visible data. This 13816 option implies -mno-extern-sdata and -mno-local-sdata and it can be 13817 useful in situations where $gp does not necessarily hold the 13818 expected value. 13819 13820 Miscellaneous improvements 13821 13822 * There is a new option called -mbranch-cost for tweaking the 13823 perceived cost of branches. 13824 * If GCC is configured to use a version of GAS that supports the 13825 .gnu_attribute directive, it will use that directive to record 13826 certain properties of the output code. .gnu_attribute is new to GAS 13827 2.18. 13828 * There are two new function attributes, near and far, for overriding 13829 the command-line setting of -mlong-calls on a function-by-function 13830 basis. 13831 * -mfp64, which previously required a 64-bit target, now works with 13832 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 13833 mipsisa*-sde-elf* configurations provide suitable library support. 13834 * GCC now recognizes the -mdmx and -mmt options and passes them down 13835 to the assembler. It does nothing else with the options at present. 13836 13837 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 13838 (BEA) 13839 13840 * Support has been added for this new architecture. 13841 13842 RS6000 (POWER/PowerPC) 13843 13844 * Support for the PowerPC 750CL paired-single instructions has been 13845 added with a new powerpc-*-linux*paired* target configuration. It 13846 is enabled by an associated -mpaired option and can be accessed 13847 using new built-in functions. 13848 * Support for auto-detecting architecture and system configuration to 13849 auto-select processor optimization tuning. 13850 * Support for VMX on AIX 5.3 has been added. 13851 * Support for AIX Version 6.1 has been added. 13852 13853 S/390, zSeries and System z9 13854 13855 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 13856 added. When using the -march=z9-ec option, the compiler will 13857 generate code making use of instructions provided by the decimal 13858 floating point facility and the floating point conversion facility 13859 (pfpo). Besides the instructions used to implement decimal floating 13860 point operations these facilities also contain instructions to move 13861 between general purpose and floating point registers and to modify 13862 and copy the sign-bit of floating point values. 13863 * When the -march=z9-ec option is used the new 13864 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 13865 decimal floating point hardware instructions will be used or not. 13866 If none of them is given the hardware support is enabled by 13867 default. 13868 * The -mstack-guard option can now be omitted when using stack 13869 checking via -mstack-size in order to let GCC choose a sensible 13870 stack guard value according to the frame size of each function. 13871 * Various changes to improve performance of generated code have been 13872 implemented, including: 13873 + The condition code set by an add logical with carry 13874 instruction is now available for overflow checks like: a + b + 13875 carry < b. 13876 + The test data class instruction is now used to implement 13877 sign-bit and infinity checks of binary and decimal floating 13878 point numbers. 13879 13880 SPARC 13881 13882 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 13883 added. 13884 13885 Xtensa 13886 13887 * Stack unwinding for exception handling now uses by default a 13888 specialized version of DWARF unwinding. This is not 13889 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 13890 Xtensa with previous versions of GCC. 13891 * For Xtensa processors that include the Conditional Store option, 13892 the built-in functions for atomic memory access are now implemented 13893 using S32C1I instructions. 13894 * If the Xtensa NSA option is available, GCC will use it to implement 13895 the __builtin_ctz and __builtin_clz functions. 13896 13897 Documentation improvements 13898 13899 * Existing libstdc++ documentation has been edited and restructured 13900 into a single DocBook XML manual. The results can be viewed online 13901 [18]here. 13902 13903 Other significant improvements 13904 13905 * The compiler's --help command-line option has been extended so that 13906 it now takes an optional set of arguments. These arguments restrict 13907 the information displayed to specific classes of command-line 13908 options, and possibly only a subset of those options. It is also 13909 now possible to replace the descriptive text associated with each 13910 displayed option with an indication of its current value, or for 13911 binary options, whether it has been enabled or disabled. 13912 Here are some examples. The following will display all the options 13913 controlling warning messages: 13914 --help=warnings 13915 13916 Whereas this will display all the undocumented, target specific 13917 options: 13918 --help=target,undocumented 13919 13920 This sequence of commands will display the binary optimizations 13921 that are enabled by -O3: 13922 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 13923 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 13924 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 13925 13926 * The configure options --with-pkgversion and --with-bugurl have been 13927 added. These allow distributors of GCC to include a 13928 distributor-specific string in manuals and --version output and to 13929 specify the URL for reporting bugs in their versions of GCC. 13930 13931 GCC 4.3.1 13932 13933 This is the [19]list of problem reports (PRs) from GCC's bug tracking 13934 system that are known to be fixed in the 4.3.1 release. This list might 13935 not be complete (that is, it is possible that some PRs that have been 13936 fixed are not listed here). 13937 13938 Target Specific Changes 13939 13940 IA-32/x86-64 13941 13942 ABI changes 13943 13944 * Starting with GCC 4.3.1, decimal floating point variables are 13945 aligned to their natural boundaries when they are passed on the 13946 stack for i386. 13947 13948 Command-line changes 13949 13950 * Starting with GCC 4.3.1, the -mcld option has been added to 13951 automatically generate a cld instruction in the prologue of 13952 functions that use string instructions. This option is used for 13953 backward compatibility on some operating systems and can be enabled 13954 by default for 32-bit x86 targets by configuring GCC with the 13955 --enable-cld configure option. 13956 13957 GCC 4.3.2 13958 13959 This is the [20]list of problem reports (PRs) from GCC's bug tracking 13960 system that are known to be fixed in the 4.3.2 release. This list might 13961 not be complete (that is, it is possible that some PRs that have been 13962 fixed are not listed here). 13963 13964 GCC 4.3.3 13965 13966 This is the [21]list of problem reports (PRs) from GCC's bug tracking 13967 system that are known to be fixed in the 4.3.3 release. This list might 13968 not be complete (that is, it is possible that some PRs that have been 13969 fixed are not listed here). 13970 13971 GCC 4.3.4 13972 13973 This is the [22]list of problem reports (PRs) from GCC's bug tracking 13974 system that are known to be fixed in the 4.3.4 release. This list might 13975 not be complete (that is, it is possible that some PRs that have been 13976 fixed are not listed here). 13977 13978 GCC 4.3.5 13979 13980 This is the [23]list of problem reports (PRs) from GCC's bug tracking 13981 system that are known to be fixed in the 4.3.5 release. This list might 13982 not be complete (that is, it is possible that some PRs that have been 13983 fixed are not listed here). 13984 13985 GCC 4.3.6 13986 13987 This is the [24]list of problem reports (PRs) from GCC's bug tracking 13988 system that are known to be fixed in the 4.3.6 release. This list might 13989 not be complete (that is, it is possible that some PRs that have been 13990 fixed are not listed here). 13991 13992 13993 For questions related to the use of GCC, please consult these web 13994 pages and the [25]GCC manuals. If that fails, the 13995 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13996 web pages and the development of GCC are welcome on our developer 13997 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 13998 archives. 13999 14000 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 14001 distribution of this entire article is permitted in any medium, 14002 provided this notice is preserved. 14003 14004 These pages are [30]maintained by the GCC team. Last modified 14005 2022-10-26. 14006 14007 References 14008 14009 1. https://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 14010 2. https://gmplib.org/ 14011 3. https://www.mpfr.org/ 14012 4. https://gcc.gnu.org/install/prerequisites.html 14013 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 14014 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 14015 7. https://gcc.gnu.org/gcc-4.3/porting_to.html 14016 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 14017 9. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 14018 10. https://gcc.gnu.org/gcc-4.3/cxx0x_status.html 14019 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 14020 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 14021 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 14022 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 14023 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 14024 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 14025 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 14026 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 14027 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 14028 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 14029 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 14030 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 14031 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 14032 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 14033 25. https://gcc.gnu.org/onlinedocs/ 14034 26. mailto:gcc-help (a] gcc.gnu.org 14035 27. mailto:gcc (a] gcc.gnu.org 14036 28. https://gcc.gnu.org/lists.html 14037 29. https://www.fsf.org/ 14038 30. https://gcc.gnu.org/about.html 14039 ====================================================================== 14040 http://gcc.gnu.org/gcc-4.2/index.html 14041 GCC 4.2 Release Series 14042 14043 (This release series is no longer supported.) 14044 14045 May 19, 2008 14046 14047 The [1]GNU project and the GCC developers are pleased to announce the 14048 release of GCC 4.2.4. 14049 14050 This release is a bug-fix release, containing fixes for regressions in 14051 GCC 4.2.3 relative to previous releases of GCC. 14052 14053 Release History 14054 14055 GCC 4.2.4 14056 May 19, 2008 ([2]changes) 14057 14058 GCC 4.2.3 14059 February 1, 2008 ([3]changes) 14060 14061 GCC 4.2.2 14062 October 7, 2007 ([4]changes) 14063 14064 GCC 4.2.1 14065 July 18, 2007 ([5]changes) 14066 14067 GCC 4.2.0 14068 May 13, 2007 ([6]changes) 14069 14070 References and Acknowledgements 14071 14072 GCC used to stand for the GNU C Compiler, but since the compiler 14073 supports several other languages aside from C, it now stands for the 14074 GNU Compiler Collection. 14075 14076 A list of [7]successful builds is updated as new information becomes 14077 available. 14078 14079 The GCC developers would like to thank the numerous people that have 14080 contributed new features, improvements, bug fixes, and other changes as 14081 well as test results to GCC. This [8]amazing group of volunteers is 14082 what makes GCC successful. 14083 14084 For additional information about GCC please refer to the [9]GCC project 14085 web site or contact the [10]GCC development mailing list. 14086 14087 To obtain GCC please use [11]our mirror sites or [12]our version 14088 control system. 14089 14090 14091 For questions related to the use of GCC, please consult these web 14092 pages and the [13]GCC manuals. If that fails, the 14093 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14094 web pages and the development of GCC are welcome on our developer 14095 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 14096 archives. 14097 14098 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14099 distribution of this entire article is permitted in any medium, 14100 provided this notice is preserved. 14101 14102 These pages are [18]maintained by the GCC team. Last modified 14103 2022-10-26. 14104 14105 References 14106 14107 1. http://www.gnu.org/ 14108 2. https://gcc.gnu.org/gcc-4.2/changes.html 14109 3. https://gcc.gnu.org/gcc-4.2/changes.html 14110 4. https://gcc.gnu.org/gcc-4.2/changes.html 14111 5. https://gcc.gnu.org/gcc-4.2/changes.html 14112 6. https://gcc.gnu.org/gcc-4.2/changes.html 14113 7. https://gcc.gnu.org/gcc-4.2/buildstat.html 14114 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14115 9. https://gcc.gnu.org/index.html 14116 10. mailto:gcc (a] gcc.gnu.org 14117 11. https://gcc.gnu.org/mirrors.html 14118 12. https://gcc.gnu.org/git.html 14119 13. https://gcc.gnu.org/onlinedocs/ 14120 14. mailto:gcc-help (a] gcc.gnu.org 14121 15. mailto:gcc (a] gcc.gnu.org 14122 16. https://gcc.gnu.org/lists.html 14123 17. https://www.fsf.org/ 14124 18. https://gcc.gnu.org/about.html 14125 ====================================================================== 14126 http://gcc.gnu.org/gcc-4.2/changes.html 14127 GCC 4.2 Release Series 14128 Changes, New Features, and Fixes 14129 14130 Caveats 14131 14132 * GCC no longer accepts the -fshared-data option. This option has had 14133 no effect in any GCC 4 release; the targets to which the option 14134 used to apply had been removed before GCC 4.0. 14135 14136 General Optimizer Improvements 14137 14138 * New command-line options specify the possible relationships among 14139 parameters and between parameters and global data. For example, 14140 -fargument-noalias-anything specifies that arguments do not alias 14141 any other storage. 14142 Each language will automatically use whatever option is required by 14143 the language standard. You should not need to use these options 14144 yourself. 14145 14146 New Languages and Language specific improvements 14147 14148 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 14149 * New command-line options -fstrict-overflow and -Wstrict-overflow 14150 have been added. -fstrict-overflow tells the compiler that it may 14151 assume that the program follows the strict signed overflow 14152 semantics permitted for the language: for C and C++ this means that 14153 the compiler may assume that signed overflow does not occur. For 14154 example, a loop like 14155 for (i = 1; i > 0; i *= 2) 14156 14157 is presumably intended to continue looping until i overflows. With 14158 -fstrict-overflow, the compiler may assume that signed overflow 14159 will not occur, and transform this into an infinite loop. 14160 -fstrict-overflow is turned on by default at -O2, and may be 14161 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 14162 be used to warn about cases where the compiler assumes that signed 14163 overflow will not occur. It takes five different levels: 14164 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 14165 -Wstrict-overflow=1 is enabled by -Wall. 14166 * The new command-line option -fno-toplevel-reorder directs GCC to 14167 emit top-level functions, variables, and asm statements in the same 14168 order that they appear in the input file. This is intended to 14169 support existing code which relies on a particular ordering (for 14170 example, code which uses top-level asm statements to switch 14171 sections). For new code, it is generally better to use function and 14172 variable attributes. The -fno-toplevel-reorder option may be used 14173 for most cases which currently use -fno-unit-at-a-time. The 14174 -fno-unit-at-a-time option will be removed in some future version 14175 of GCC. If you know of a case which requires -fno-unit-at-a-time 14176 which is not fixed by -fno-toplevel-reorder, please open a bug 14177 report. 14178 14179 C family 14180 14181 * The pragma redefine_extname will now macro expand its tokens for 14182 compatibility with SunPRO. 14183 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 14184 GCC to handle inline functions as specified in the C99 standard. In 14185 preparation for this, GCC 4.2 will warn about any use of non-static 14186 inline functions in gnu99 or c99 mode. This new warning may be 14187 disabled with the new gnu_inline function attribute or the new 14188 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 14189 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 14190 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 14191 in the current compilation. 14192 * A new command-line option -Waddress has been added to warn about 14193 suspicious uses of memory addresses as, for example, using the 14194 address of a function in a conditional expression, and comparisons 14195 against the memory address of a string literal. This warning is 14196 enabled by -Wall. 14197 14198 C++ 14199 14200 * C++ visibility handling has been overhauled. 14201 Restricted visiblity is propagated from classes to members, from 14202 functions to local statics, and from templates and template 14203 arguments to instantiations, unless the latter has explicitly 14204 declared visibility. 14205 The visibility attribute for a class must come between the 14206 class-key and the name, not after the closing brace. 14207 Attributes are now allowed for enums and elaborated-type-specifiers 14208 that only declare a type. 14209 Members of the anonymous namespace are now local to a particular 14210 translation unit, along with any other declarations which use them, 14211 though they are still treated as having external linkage for 14212 language semantics. 14213 * The (undocumented) extension which permitted templates with default 14214 arguments to be bound to template template parameters with fewer 14215 parameters has been removed. For example: 14216 template <template <typename> class C> 14217 void f(C<double>) {} 14218 14219 template <typename T, typename U = int> 14220 struct S {}; 14221 14222 template void f(S<double>); 14223 14224 is no longer accepted by G++. The reason this code is not accepted 14225 is that S is a template with two parameters; therefore, it cannot 14226 be bound to C which has only one parameter. 14227 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 14228 releases, have been removed. 14229 * The command-line option -fconst-strings, deprecated in previous GCC 14230 releases, has been removed. 14231 * The configure variable enable-__cxa_atexit is now enabled by 14232 default for more targets. Enabling this variable is necessary in 14233 order for static destructors to be executed in the correct order, 14234 but it depends upon the presence of a non-standard C library in the 14235 target library in order to work. The variable is now enabled for 14236 more targets which are known to have suitable C libraries. 14237 * -Wextra will produce warnings for if statements with a semicolon as 14238 the only body, to catch code like: 14239 if (a); 14240 return 1; 14241 return 0; 14242 14243 To suppress the warning in valid cases, use { } instead. 14244 * The C++ front end now also produces strict aliasing warnings when 14245 -fstrict-aliasing -Wstrict-aliasing is in effect. 14246 14247 Runtime Library (libstdc++) 14248 14249 * Added support for TR1 <random>, <complex>, and C compatibility 14250 headers. In addition, a lock-free version of shared_ptr was 14251 contributed as part of Phillip Jordan's Google Summer of Code 14252 project on lock-free containers. 14253 * In association with the Summer of Code work on lock-free 14254 containers, the interface for atomic builtins was adjusted, 14255 creating simpler alternatives for non-threaded code paths. Also, 14256 usage was consolidated and all elements were moved from namespace 14257 std to namespace__gnu_cxx. Affected interfaces are the functions 14258 __exchange_and_add, __atomic_add, and the objects __mutex, 14259 __recursive_mutex, and __scoped_lock. 14260 * Support for versioning weak symbol names via namespace association 14261 was added. However, as this changes the names of exported symbols, 14262 this is turned off by default in the current ABI. Intrepid users 14263 can enable this feature by using 14264 --enable-symvers=gnu-versioned-namespace during configuration. 14265 * Revised, simplified, and expanded policy-based associative 14266 containers, including data types for tree and trie forms 14267 (basic_tree, tree, trie), lists (list_update), and both 14268 collision-chaining and probing hash-based containers 14269 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 14270 the [3]documentation. 14271 * The implementation of the debug mode was modified, whereby the 14272 debug namespaces were nested inside of namespace std and namespace 14273 __gnu_cxx in order to resolve some long standing corner cases 14274 involving name lookup. Debug functionality from the policy-based 14275 data structures was consolidated and enabled with the single macro, 14276 _GLIBCXX_DEBUG. See PR 26142 for more information. 14277 * Added extensions for type traits: __conditional_type, 14278 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 14279 * Added a typelist implementation for compile-time meta-programming. 14280 Elements for typelist construction and operation can be found 14281 within namespace __gnu_cxx::typelist. 14282 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 14283 exception-safety. 14284 * Enabled library-wide visibility control, allowing -fvisibility to 14285 be used. 14286 * Consolidated all nested namespaces and the conversion of 14287 __gnu_internal implementation-private details to anonymous 14288 namespaces whenever possible. 14289 * Implemented LWG resolutions DR 431 and DR 538. 14290 14291 Fortran 14292 14293 * Support for allocatable components has been added (TR 15581 and 14294 Fortran 2003). 14295 * Support for the Fortran 2003 streaming IO extension has been added. 14296 * The GNU Fortran compiler now uses 4-byte record markers by default 14297 for unformatted files to be compatible with g77 and most other 14298 compilers. The implementation allows for records greater than 2 GB 14299 and is compatible with several other compilers. Older versions of 14300 gfortran used 8-byte record markers by default (on most systems). 14301 In order to change the length of the record markers, e.g. to read 14302 unformatted files created by older gfortran versions, the 14303 [4]-frecord-marker=8 option can be used. 14304 14305 Java (GCJ) 14306 14307 * A new command-line option -static-libgcj has been added for targets 14308 that use a linker compatible with GNU Binutils. As its name 14309 implies, this causes libgcj to be linked statically. In some cases 14310 this causes the resulting executable to start faster and use less 14311 memory than if the shared version of libgcj were used. However 14312 caution should be used as it can also cause essential parts of the 14313 library to be omitted. Some of these issues are discussed in: 14314 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 14315 * fastjar is no longer bundled with GCC. To build libgcj, you will 14316 need either InfoZIP (both zip and unzip) or an external jar 14317 program. In the former case, the GCC build will install a jar shell 14318 script that is based on InfoZIP and provides the same functionality 14319 as fastjar. 14320 14321 New Targets and Target Specific Improvements 14322 14323 IA-32/x86-64 14324 14325 * -mtune=generic can now be used to generate code running well on 14326 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 14327 Pentium-M, Intel Pentium 4 and Intel Core 2. 14328 * -mtune=native and -march=native will produce code optimized for the 14329 host architecture as detected using the cpuid instruction. 14330 * Added a new command-line option -fstackrealign and and 14331 __attribute__ ((force_align_arg_pointer)) to realign the stack at 14332 runtime. This allows functions compiled with a vector-aligned stack 14333 to be invoked from legacy objects that keep only word-alignment. 14334 14335 SPARC 14336 14337 * The default CPU setting has been changed from V7 to V9 in 32-bit 14338 mode on Solaris 7 and above. This is already the case in 64-bit 14339 mode. It can be overridden by specifying --with-cpu at configure 14340 time. 14341 * Back-end support of built-in functions for atomic memory access has 14342 been implemented. 14343 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 14344 added. 14345 14346 M32C 14347 14348 * Various bug fixes have made some functions (notably, functions 14349 returning structures) incompatible with previous releases. 14350 Recompiling all libraries is recommended. Note that code quality 14351 has considerably improved since 4.1, making a recompile even more 14352 beneficial. 14353 14354 MIPS 14355 14356 * Added support for the Broadcom SB-1A core. 14357 14358 IA-64 14359 14360 * Added support for IA-64 data and control speculation. By default 14361 speculation is enabled only during second scheduler pass. A number 14362 of machine flags was introduced to control the usage of speculation 14363 for both scheduler passes. 14364 14365 HPPA 14366 14367 * Added Java language support (libffi and libjava) for 32-bit HP-UX 14368 11 target. 14369 14370 Obsolete Systems 14371 14372 Documentation improvements 14373 14374 PDF Documentation 14375 14376 * A make pdf target has been added to the top-level makefile, 14377 enabling automated production of PDF documentation files. 14378 (Front-ends external to GCC should modify their Make-lang.in file 14379 to add a lang.pdf: target.) 14380 14381 Other significant improvements 14382 14383 Build system improvements 14384 14385 * All the components of the compiler are now bootstrapped by default. 14386 This improves the resilience to bugs in the system compiler or 14387 binary compatibility problems, as well as providing better testing 14388 of GCC 4.2 itself. In addition, if you build the compiler from a 14389 combined tree, the assembler, linker, etc. will also be 14390 bootstrapped (i.e. built with themselves). 14391 You can disable this behavior, and go back to the pre-GCC 4.2 set 14392 up, by configuring GCC with --disable-bootstrap. 14393 * The rules that configure follows to find target tools resemble more 14394 closely the locations that the built compiler will search. In 14395 addition, you can use the new configure option --with-target-tools 14396 to specify where to find the target tools used during the build, 14397 without affecting what the built compiler will use. 14398 This can be especially useful when building packages of GCC. For 14399 example, you may want to build GCC with GNU as or ld, even if the 14400 resulting compiler to work with the native assembler and linker. To 14401 do so, you can use --with-target-tools to point to the native 14402 tools. 14403 14404 Incompatible changes to the build system 14405 14406 * Front-ends external to GCC should modify their Make-lang.in file to 14407 replace double-colon rules (e.g. dvi::) with normal rules (like 14408 lang.dvi:). Front-end makefile hooks do not use double-colon rules 14409 anymore. 14410 * Up to GCC 4.1, a popular way to specify the target tools used 14411 during the build was to create directories named gas, binutils, 14412 etc. in the build tree, and create links to the tools from there. 14413 This does not work any more when the compiler is bootstrapped. The 14414 new configure option --with-target-tools provides a better way to 14415 achieve the same effect, and works for all native and cross 14416 settings. 14417 14418 14419 For questions related to the use of GCC, please consult these web 14420 pages and the [6]GCC manuals. If that fails, the 14421 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14422 web pages and the development of GCC are welcome on our developer 14423 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 14424 14425 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 14426 distribution of this entire article is permitted in any medium, 14427 provided this notice is preserved. 14428 14429 These pages are [11]maintained by the GCC team. Last modified 14430 2022-10-26. 14431 14432 References 14433 14434 1. https://gcc.gnu.org/projects/gomp/ 14435 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 14436 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 14437 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 14438 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 14439 6. https://gcc.gnu.org/onlinedocs/ 14440 7. mailto:gcc-help (a] gcc.gnu.org 14441 8. mailto:gcc (a] gcc.gnu.org 14442 9. https://gcc.gnu.org/lists.html 14443 10. https://www.fsf.org/ 14444 11. https://gcc.gnu.org/about.html 14445 ====================================================================== 14446 http://gcc.gnu.org/gcc-4.1/index.html 14447 GCC 4.1 Release Series 14448 14449 (This release series is no longer supported.) 14450 14451 February 13, 2007 14452 14453 The [1]GNU project and the GCC developers are pleased to announce the 14454 release of GCC 4.1.2. 14455 14456 This release is a bug-fix release, containing fixes for regressions in 14457 GCC 4.1.1 relative to previous releases of GCC. 14458 14459 Release History 14460 14461 GCC 4.1.2 14462 February 13, 2007 ([2]changes) 14463 14464 GCC 4.1.1 14465 May 24, 2006 ([3]changes) 14466 14467 GCC 4.1.0 14468 February 28, 2006 ([4]changes) 14469 14470 References and Acknowledgements 14471 14472 GCC used to stand for the GNU C Compiler, but since the compiler 14473 supports several other languages aside from C, it now stands for the 14474 GNU Compiler Collection. 14475 14476 A list of [5]successful builds is updated as new information becomes 14477 available. 14478 14479 The GCC developers would like to thank the numerous people that have 14480 contributed new features, improvements, bug fixes, and other changes as 14481 well as test results to GCC. This [6]amazing group of volunteers is 14482 what makes GCC successful. 14483 14484 For additional information about GCC please refer to the [7]GCC project 14485 web site or contact the [8]GCC development mailing list. 14486 14487 To obtain GCC please use [9]our mirror sites or [10]our version control 14488 system. 14489 14490 14491 For questions related to the use of GCC, please consult these web 14492 pages and the [11]GCC manuals. If that fails, the 14493 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14494 web pages and the development of GCC are welcome on our developer 14495 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 14496 archives. 14497 14498 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 14499 distribution of this entire article is permitted in any medium, 14500 provided this notice is preserved. 14501 14502 These pages are [16]maintained by the GCC team. Last modified 14503 2022-10-26. 14504 14505 References 14506 14507 1. http://www.gnu.org/ 14508 2. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 14509 3. https://gcc.gnu.org/gcc-4.1/changes.html 14510 4. https://gcc.gnu.org/gcc-4.1/changes.html 14511 5. https://gcc.gnu.org/gcc-4.1/buildstat.html 14512 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14513 7. https://gcc.gnu.org/index.html 14514 8. mailto:gcc (a] gcc.gnu.org 14515 9. https://gcc.gnu.org/mirrors.html 14516 10. https://gcc.gnu.org/git.html 14517 11. https://gcc.gnu.org/onlinedocs/ 14518 12. mailto:gcc-help (a] gcc.gnu.org 14519 13. mailto:gcc (a] gcc.gnu.org 14520 14. https://gcc.gnu.org/lists.html 14521 15. https://www.fsf.org/ 14522 16. https://gcc.gnu.org/about.html 14523 ====================================================================== 14524 http://gcc.gnu.org/gcc-4.1/changes.html 14525 GCC 4.1 Release Series 14526 Changes, New Features, and Fixes 14527 14528 The latest release in the 4.1 release series is [1]GCC 4.1.2. 14529 14530 Caveats 14531 14532 General Optimizer Improvements 14533 14534 * GCC now has infrastructure for inter-procedural optimizations and 14535 the following inter-procedural optimizations are implemented: 14536 + Profile guided inlining. When doing profile feedback guided 14537 optimization, GCC can now use the profile to make better 14538 informed decisions on whether inlining of a function is 14539 profitable or not. This means that GCC will no longer inline 14540 functions at call sites that are not executed very often, and 14541 that functions at hot call sites are more likely to be 14542 inlined. 14543 A new parameter min-inline-recursive-probability is also now 14544 available to throttle recursive inlining of functions with 14545 small average recursive depths. 14546 + Discovery of pure and const functions, a form of side-effects 14547 analysis. While older GCC releases could also discover such 14548 special functions, the new IPA-based pass runs earlier so that 14549 the results are available to more optimizers. The pass is also 14550 simply more powerful than the old one. 14551 + Analysis of references to static variables and type escape 14552 analysis, also forms of side-effects analysis. The results of 14553 these passes allow the compiler to be less conservative about 14554 call-clobbered variables and references. This results in more 14555 redundant loads being eliminated and in making static 14556 variables candidates for register promotion. 14557 + Improvement of RTL-based alias analysis. The results of type 14558 escape analysis are fed to the RTL type-based alias analyzer, 14559 allowing it to disambiguate more memory references. 14560 + Interprocedural constant propagation and function versioning. 14561 This pass looks for functions that are always called with the 14562 same constant value for one or more of the function arguments, 14563 and propagates those constants into those functions. 14564 + GCC will now eliminate static variables whose usage was 14565 optimized out. 14566 + -fwhole-program --combine can now be used to make all 14567 functions in program static allowing whole program 14568 optimization. As an exception, the main function and all 14569 functions marked with the new externally_visible attribute are 14570 kept global so that programs can link with runtime libraries. 14571 * GCC can now do a form of partial dead code elimination (PDCE) that 14572 allows code motion of expressions to the paths where the result of 14573 the expression is actually needed. This is not always a win, so the 14574 pass has been limited to only consider profitable cases. Here is an 14575 example: 14576 int foo (int *, int *); 14577 int 14578 bar (int d) 14579 { 14580 int a, b, c; 14581 b = d + 1; 14582 c = d + 2; 14583 a = b + c; 14584 if (d) 14585 { 14586 foo (&b, &c); 14587 a = b + c; 14588 } 14589 printf ("%d\n", a); 14590 } 14591 14592 The a = b + c can be sunk to right before the printf. Normal code 14593 sinking will not do this, it will sink the first one above into the 14594 else-branch of the conditional jump, which still gives you two 14595 copies of the code. 14596 * GCC now has a value range propagation pass. This allows the 14597 compiler to eliminate bounds checks and branches. The results of 14598 the pass can also be used to accurately compute branch 14599 probabilities. 14600 * The pass to convert PHI nodes to straight-line code (a form of 14601 if-conversion for GIMPLE) has been improved significantly. The two 14602 most significant improvements are an improved algorithm to 14603 determine the order in which the PHI nodes are considered, and an 14604 improvement that allow the pass to consider if-conversions of basic 14605 blocks with more than two predecessors. 14606 * Alias analysis improvements. GCC can now differentiate between 14607 different fields of structures in Tree-SSA's virtual operands form. 14608 This lets stores/loads from non-overlapping structure fields not 14609 conflict. A new algorithm to compute points-to sets was contributed 14610 that can allows GCC to see now that p->a and p->b, where p is a 14611 pointer to a structure, can never point to the same field. 14612 * Various enhancements to auto-vectorization: 14613 + Incrementally preserve SSA form when vectorizing. 14614 + Incrementally preserve loop-closed form when vectorizing. 14615 + Improvements to peeling for alignment: generate better code 14616 when the misalignment of an access is known at compile time, 14617 or when different accesses are known to have the same 14618 misalignment, even if the misalignment amount itself is 14619 unknown. 14620 + Consider dependence distance in the vectorizer. 14621 + Externalize generic parts of data reference analysis to make 14622 this analysis available to other passes. 14623 + Vectorization of conditional code. 14624 + Reduction support. 14625 * GCC can now partition functions in sections of hot and cold code. 14626 This can significantly improve performance due to better 14627 instruction cache locality. This feature works best together with 14628 profile feedback driven optimization. 14629 * A new pass to avoid saving of unneeded arguments to the stack in 14630 vararg functions if the compiler can prove that they will not be 14631 needed. 14632 * Transition of basic block profiling to tree level implementation 14633 has been completed. The new implementation should be considerably 14634 more reliable (hopefully avoiding profile mismatch errors when 14635 using -fprofile-use or -fbranch-probabilities) and can be used to 14636 drive higher level optimizations, such as inlining. 14637 The -ftree-based-profiling command-line option was removed and 14638 -fprofile-use now implies disabling old RTL level loop optimizer 14639 (-fno-loop-optimize). Speculative prefetching optimization 14640 (originally enabled by -fspeculative-prefetching) was removed. 14641 14642 New Languages and Language specific improvements 14643 14644 C and Objective-C 14645 14646 * The old Bison-based C and Objective-C parser has been replaced by a 14647 new, faster hand-written recursive-descent parser. 14648 14649 Ada 14650 14651 * The build infrastructure for the Ada runtime library and tools has 14652 been changed to be better integrated with the rest of the build 14653 infrastructure of GCC. This should make doing cross builds of Ada a 14654 bit easier. 14655 14656 C++ 14657 14658 * ARM-style name-injection of friend declarations is no longer the 14659 default. For example: 14660 struct S { 14661 friend void f(); 14662 }; 14663 14664 void g() { f(); } 14665 will not be accepted; instead a declaration of f will need to be 14666 present outside of the scope of S. The new -ffriend-injection 14667 option will enable the old behavior. 14668 * The (undocumented) extension which permitted templates with default 14669 arguments to be bound to template template parameters with fewer 14670 parameters has been deprecated, and will be removed in the next 14671 major release of G++. For example: 14672 template <template <typename> class C> 14673 void f(C<double>) {} 14674 14675 template <typename T, typename U = int> 14676 struct S {}; 14677 14678 template void f(S<double>); 14679 14680 makes use of the deprecated extension. The reason this code is not 14681 valid ISO C++ is that S is a template with two parameters; 14682 therefore, it cannot be bound to C which has only one parameter. 14683 14684 Runtime Library (libstdc++) 14685 14686 * Optimization work: 14687 + A new implementation of std::search_n is provided, better 14688 performing in case of random access iterators. 14689 + Added further efficient specializations of istream functions, 14690 i.e., character array and string extractors. 14691 + Other smaller improvements throughout. 14692 * Policy-based associative containers, designed for high-performance, 14693 flexibility and semantic safety are delivered in ext/pb_assoc. 14694 * A versatile string class, __gnu_cxx::__versa_string, providing 14695 facilities conforming to the standard requirements for 14696 basic_string, is delivered in <ext/vstring.h>. In particular: 14697 + Two base classes are provided: the default one avoids 14698 reference counting and is optimized for short strings; the 14699 alternate one, still uses it while improving in a few low 14700 level areas (e.g., alignment). See vstring_fwd.h for some 14701 useful typedefs. 14702 + Various algorithms have been rewritten (e.g., replace), the 14703 code streamlined and simple optimizations added. 14704 + Option 3 of DR 431 is implemented for both available bases, 14705 thus improving the support for stateful allocators. 14706 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 14707 libstdc++/23953) and LWG resolutions put into effect for the first 14708 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 14709 Issue 6.19). The implementation status of TR1 is now tracked in the 14710 docs in tr1.html. 14711 14712 Objective-C++ 14713 14714 * A new language front end for Objective-C++ has been added. This 14715 language allows users to mix the object oriented features of 14716 Objective-C with those of C++. 14717 14718 Java (GCJ) 14719 14720 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 14721 features (plus some 0.20 bug-fixes) 14722 + Networking 14723 o The java.net.HttpURLConnection implementation no longer 14724 buffers the entire response body in memory. This means 14725 that response bodies larger than available memory can now 14726 be handled. 14727 + (N)IO 14728 o NIO FileChannel.map implementation, fast bulk put 14729 implementation for DirectByteBuffer (speeds up this 14730 method 10x). 14731 o FileChannel.lock() and FileChannel.force() implemented. 14732 + XML 14733 o gnu.xml fix for nodes created outside a namespace 14734 context. 14735 o Add support for output indenting and 14736 cdata-section-elements output instruction in 14737 xml.transform. 14738 o xml.xpath corrections for cases where elements/attributes 14739 might have been created in non-namespace-aware mode. 14740 Corrections to handling of XSL variables and minor 14741 conformance updates. 14742 + AWT 14743 o GNU JAWT implementation, the AWT Native Interface, which 14744 allows direct access to native screen resources from 14745 within a Canvas's paint method. GNU Classpath Examples 14746 comes with a Demo, see libjava/classpath/examples/README. 14747 o awt.datatransfer updated to 1.5 with support for 14748 FlavorEvents. The gtk+ awt peers now allow copy/paste of 14749 text, images, URIs/files and serialized objects with 14750 other applications and tracking clipboard change events 14751 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 14752 objects are supported). A GNU Classpath Examples 14753 datatransfer Demo was added to show the new 14754 functionality. 14755 o Split gtk+ awt peers event handling in two threads and 14756 improve gdk lock handling (solves several awt lock ups). 14757 o Speed up awt Image loading. 14758 o Better gtk+ scrollbar peer implementation when using gtk+ 14759 >= 2.6. 14760 o Handle image loading errors correctly for gdkpixbuf and 14761 MediaTracker. 14762 o Better handle GDK lock. Properly prefix gtkpeer native 14763 functions (cp_gtk). 14764 o GdkGraphics2D has been updated to use Cairo 0.5.x or 14765 higher. 14766 o BufferedImage and GtkImage rewrites. All image drawing 14767 operations should now work correctly (flipping requires 14768 gtk+ >= 2.6) 14769 o When gtk+ 2.6 or higher is installed the default log 14770 handler will produce stack traces whenever a WARNING, 14771 CRITICAL or ERROR message is produced. 14772 + Free Swing 14773 o The RepaintManager has been reworked for more efficient 14774 painting, especially for large GUIs. 14775 o The layout manager OverlayLayout has been implemented, 14776 the BoxLayout has been rewritten to make use of the 14777 SizeRequirements utility class and caching for more 14778 efficient layout. 14779 o Improved accessibility support. 14780 o Significant progress has been made in the implementation 14781 of the javax.swing.plaf.metal package, with most UI 14782 delegates in a working state now. Please test this with 14783 your own applications and provide feedback that will help 14784 us to improve this package. 14785 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 14786 extended to highlight various features in our Free Swing 14787 implementation. And it includes a look and feel switcher 14788 for Metal (default), Ocean and GNU themes. 14789 o The javax.swing.plaf.multi package is now implemented. 14790 o Editing and several key actions for JTree and JTable were 14791 implemented. 14792 o Lots of icons and look and feel improvements for Free 14793 Swing basic and metal themes were added. Try running the 14794 GNU Classpath Swing Demo in examples 14795 (gnu.classpath.examples.swing.Demo) with: 14796 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 14797 l or 14798 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 14799 l 14800 o Start of styled text capabilites for java.swing.text. 14801 o DefaultMutableTreeNode pre-order, post-order, depth-first 14802 and breadth-first traversal enumerations implemented. 14803 o JInternalFrame colors and titlebar draw properly. 14804 o JTree is working up to par (icons, selection and keyboard 14805 traversal). 14806 o JMenus were made more compatible in visual and 14807 programmatic behavior. 14808 o JTable changeSelection and multiple selections 14809 implemented. 14810 o JButton and JToggleButton change states work properly 14811 now. 14812 o JFileChooser fixes. 14813 o revalidate() and repaint() fixes which make Free Swing 14814 much more responsive. 14815 o MetalIconFactory implemented. 14816 o Free Swing Top-Level Compatibility. JFrame, JDialog, 14817 JApplet, JInternalFrame, and JWindow are now 1.5 14818 compatible in the sense that you can call add() and 14819 setLayout() directly on them, which will have the same 14820 effect as calling getContentPane().add() and 14821 getContentPane().setLayout(). 14822 o The JTree interface has been completed. JTrees now 14823 recognizes mouse clicks and selections work. 14824 o BoxLayout works properly now. 14825 o Fixed GrayFilter to actually work. 14826 o Metal SplitPane implemented. 14827 o Lots of Free Swing text and editor stuff work now. 14828 + Free RMI and Corba 14829 o Andrew Watson, Vice President and Technical Director of 14830 the Object Management Group, has officially assigned us 14831 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 14832 mark remote classpath-specific system exceptions. 14833 Obtaining the VMCID means that GNU Classpath now is a 14834 recogniseable type of node in a highly interoperable 14835 CORBA world. 14836 o GNU Classpath now includes the first working draft to 14837 support the RMI over IIOP protocol. The current 14838 implementation is capable of remote invocations, 14839 transferring various Serializables and Externalizables 14840 via RMI-IIOP protocol. It can flatten graphs and, at 14841 least for the simple cases, is interoperable with 1.5 14842 JDKs. 14843 o org.omg.PortableInterceptor and related functionality in 14844 other packages is now implemented: 14845 # The sever and client interceptors work as required 14846 since 1.4. 14847 # The IOR interceptor works as needed for 1.5. 14848 o The org.omg.DynamicAny package is completed and passes 14849 the prepared tests. 14850 o The Portable Object Adapter should now support the output 14851 of the recent IDL to java compilers. These compilers now 14852 generate servants and not CORBA objects as before, making 14853 the output depend on the existing POA implementation. 14854 Completing POA means that such code can already be tried 14855 to run on Classpath. Our POA is tested for the following 14856 usager scenarios: 14857 # POA converts servant to the CORBA object. 14858 # Servant provides to the CORBA object. 14859 # POA activates new CORBA object with the given Object 14860 Id (byte array) that is later accessible for the 14861 servant. 14862 # During the first call, the ServantActivator provides 14863 servant for this and all subsequent calls on the 14864 current object. 14865 # During each call, the ServantLocator provides 14866 servant for this call only. 14867 # ServantLocator or ServantActivator forwards call to 14868 another server. 14869 # POA has a single servant, responsible for all 14870 objects. 14871 # POA has a default servant, but some objects are 14872 explicitly connected to they specific servants. 14873 The POA is verified using tests from the former 14874 cost.omg.org. 14875 o The CORBA implementation is now a working prototype that 14876 should support features up to 1.3 inclusive. We invite 14877 groups writing CORBA dependent applications to try 14878 Classpath implementation, reporting any possible bugs. 14879 The CORBA prototype is interoperable with Sun's 14880 implementation v 1.4, transferring object references, 14881 primitive types, narrow and wide strings, arrays, 14882 structures, trees, abstract interfaces and value types 14883 (feature of CORBA 2.3) between these two platforms. 14884 Remote exceptions are transferred and handled correctly. 14885 The stringified object references (IORs) from various 14886 sources are parsed as required. The transient (for 14887 current session) and permanent (till jre restart) 14888 redirections work. Both Little and Big Endian encoded 14889 messages are accepted. The implementation is verified 14890 using tests from the former cost.omg.org. The current 14891 release includes working examples (see the examples 14892 directory), demonstrating the client-server 14893 communication, using either CORBA Request or IDL-based 14894 stub (usually generated by a IDL to java compiler). These 14895 examples also show how to use the Classpath CORBA naming 14896 service. The IDL to java compiler is not yet written, but 14897 as our library must be compatible, it naturally accepts 14898 the output of other idlj implementations. 14899 + Misc 14900 o Updated TimeZone data against Olson tzdata2005l. 14901 o Make zip and jar packages UTF-8 clean. 14902 o "native" code builds and compiles (warning free) on 14903 Darwin and Solaris. 14904 o java.util.logging.FileHandler now rotates files. 14905 o Start of a generic JDWP framework in gnu/classpath/jdwp. 14906 This is unfinished, but feedback (at classpath (a] gnu.org) 14907 from runtime hackers is greatly appreciated. Although 14908 most of the work is currently being done around gcj/gij 14909 we want this framework to be as VM neutral as possible. 14910 Early design is described in: 14911 [2]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 14912 o QT4 AWT peers, enable by giving configure 14913 --enable-qt-peer. Included, but not ready for production 14914 yet. They are explicitly disabled and not supported. But 14915 if you want to help with the development of these new 14916 features we are interested in feedback. You will have to 14917 explicitly enable them to try them out (and they will 14918 most likely contain bugs). 14919 o Documentation fixes all over the place. See 14920 [3]https://developer.classpath.org/doc/ 14921 14922 New Targets and Target Specific Improvements 14923 14924 IA-32/x86-64 14925 14926 * The x86-64 medium model (that allows building applications whose 14927 data segment exceeds 4GB) was redesigned to match latest ABI draft. 14928 New implementation split large datastructures into separate segment 14929 improving performance of accesses to small datastructures and also 14930 allows linking of small model libraries into medium model programs 14931 as long as the libraries are not accessing the large datastructures 14932 directly. Medium model is also supported in position independent 14933 code now. 14934 The ABI change results in partial incompatibility among medium 14935 model objects. Linking medium model libraries (or objects) compiled 14936 with new compiler into medium model program compiled with older 14937 will likely result in exceeding ranges of relocations. 14938 Binutils 2.16.91 or newer are required for compiling medium model 14939 now. 14940 14941 RS6000 (POWER/PowerPC) 14942 14943 * The AltiVec vector primitives in <altivec.h> are now implemented in 14944 a way that puts a smaller burden on the preprocessor, instead 14945 processing the "overloading" in the front ends. This should benefit 14946 compilation speed on AltiVec vector code. 14947 * AltiVec initializers now are generated more efficiently. 14948 * The popcountb instruction available on POWER5 now is generated. 14949 * The floating point round to integer instructions available on 14950 POWER5+ now is generated. 14951 * Floating point divides can be synthesized using the floating point 14952 reciprocal estimate instructions. 14953 * Double precision floating point constants are initialized as single 14954 precision values if they can be represented exactly. 14955 14956 S/390, zSeries and System z9 14957 14958 * Support for the IBM System z9 109 processor has been added. When 14959 using the -march=z9-109 option, the compiler will generate code 14960 making use of instructions provided by the extended immediate 14961 facility. 14962 * Support for 128-bit IEEE floating point has been added. When using 14963 the -mlong-double-128 option, the compiler will map the long double 14964 data type to 128-bit IEEE floating point. Using this option 14965 constitutes an ABI change, and requires glibc support. 14966 * Various changes to improve performance of generated code have been 14967 implemented, including: 14968 + In functions that do not require a literal pool, register %r13 14969 (which is traditionally reserved as literal pool pointer), can 14970 now be freely used for other purposes by the compiler. 14971 + More precise tracking of register use allows the compiler to 14972 generate more efficient function prolog and epilog code in 14973 certain cases. 14974 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 14975 instructions are now used to implement C string functions. 14976 + The MOVE CHARACTER instruction with single byte overlap is now 14977 used to implement the memset function with non-zero fill byte. 14978 + The LOAD ZERO instructions are now used where appropriate. 14979 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 14980 and INSERT IMMEDIATE instructions are now used more frequently 14981 to optimize bitfield operations. 14982 + The BRANCH ON COUNT instruction is now used more frequently. 14983 In particular, the fact that a loop contains a subroutine call 14984 no longer prevents the compiler from using this instruction. 14985 + The compiler is now aware that all shift and rotate 14986 instructions implicitly truncate the shift count to six bits. 14987 * Back-end support for the following generic features has been 14988 implemented: 14989 + The full set of [4]built-in functions for atomic memory 14990 access. 14991 + The -fstack-protector feature. 14992 + The optimization pass avoiding unnecessary stores of incoming 14993 argument registers in functions with variable argument list. 14994 14995 SPARC 14996 14997 * The default code model in 64-bit mode has been changed from 14998 Medium/Anywhere to Medium/Middle on Solaris. 14999 * TLS support is disabled by default on Solaris prior to release 10. 15000 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 15001 and later) by specifying --enable-tls at configure time. 15002 15003 MorphoSys 15004 15005 * Support has been added for this new architecture. 15006 15007 Obsolete Systems 15008 15009 Documentation improvements 15010 15011 Other significant improvements 15012 15013 * GCC can now emit code for protecting applications from 15014 stack-smashing attacks. The protection is realized by buffer 15015 overflow detection and reordering of stack variables to avoid 15016 pointer corruption. 15017 * Some built-in functions have been fortified to protect them against 15018 various buffer overflow (and format string) vulnerabilities. 15019 Compared to the mudflap bounds checking feature, the safe builtins 15020 have far smaller overhead. This means that programs built using 15021 safe builtins should not experience any measurable slowdown. 15022 15023 GCC 4.1.2 15024 15025 This is the [5]list of problem reports (PRs) from GCC's bug tracking 15026 system that are known to be fixed in the 4.1.2 release. This list might 15027 not be complete (that is, it is possible that some PRs that have been 15028 fixed are not listed here). 15029 15030 When generating code for a shared library, GCC now recognizes that 15031 global functions may be replaced when the program runs. Therefore, it 15032 is now more conservative in deducing information from the bodies of 15033 functions. For example, in this example: 15034 void f() {} 15035 void g() { 15036 try { f(); } 15037 catch (...) { 15038 cout << "Exception"; 15039 } 15040 } 15041 15042 G++ would previously have optimized away the catch clause, since it 15043 would have concluded that f cannot throw exceptions. Because users may 15044 replace f with another function in the main body of the program, this 15045 optimization is unsafe, and is no longer performed. If you wish G++ to 15046 continue to optimize as before, you must add a throw() clause to the 15047 declaration of f to make clear that it does not throw exceptions. 15048 15049 15050 For questions related to the use of GCC, please consult these web 15051 pages and the [6]GCC manuals. If that fails, the 15052 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15053 web pages and the development of GCC are welcome on our developer 15054 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 15055 15056 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 15057 distribution of this entire article is permitted in any medium, 15058 provided this notice is preserved. 15059 15060 These pages are [11]maintained by the GCC team. Last modified 15061 2022-10-26. 15062 15063 References 15064 15065 1. https://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 15066 2. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 15067 3. https://developer.classpath.org/doc/ 15068 4. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 15069 5. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 15070 6. https://gcc.gnu.org/onlinedocs/ 15071 7. mailto:gcc-help (a] gcc.gnu.org 15072 8. mailto:gcc (a] gcc.gnu.org 15073 9. https://gcc.gnu.org/lists.html 15074 10. https://www.fsf.org/ 15075 11. https://gcc.gnu.org/about.html 15076 ====================================================================== 15077 http://gcc.gnu.org/gcc-4.0/index.html 15078 GCC 4.0 Release Series 15079 15080 (This release series is no longer supported.) 15081 15082 January 31, 2007 15083 15084 The [1]GNU project and the GCC developers are pleased to announce the 15085 release of GCC 4.0.4. 15086 15087 This release is a bug-fix release, containing fixes for regressions in 15088 GCC 4.0.3 relative to previous releases of GCC. 15089 15090 Release History 15091 15092 GCC 4.0.4 15093 January 31, 2007 ([2]changes) 15094 15095 GCC 4.0.3 15096 March 10, 2006 ([3]changes) 15097 15098 GCC 4.0.2 15099 September 28, 2005 ([4]changes) 15100 15101 GCC 4.0.1 15102 July 7, 2005 ([5]changes) 15103 15104 GCC 4.0.0 15105 April 20, 2005 ([6]changes) 15106 15107 References and Acknowledgements 15108 15109 GCC used to stand for the GNU C Compiler, but since the compiler 15110 supports several other languages aside from C, it now stands for the 15111 GNU Compiler Collection. 15112 15113 A list of [7]successful builds is updated as new information becomes 15114 available. 15115 15116 The GCC developers would like to thank the numerous people that have 15117 contributed new features, improvements, bug fixes, and other changes as 15118 well as test results to GCC. This [8]amazing group of volunteers is 15119 what makes GCC successful. 15120 15121 For additional information about GCC please refer to the [9]GCC project 15122 web site or contact the [10]GCC development mailing list. 15123 15124 To obtain GCC please use [11]our mirror sites, or [12]our version 15125 control system. 15126 15127 15128 For questions related to the use of GCC, please consult these web 15129 pages and the [13]GCC manuals. If that fails, the 15130 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15131 web pages and the development of GCC are welcome on our developer 15132 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 15133 archives. 15134 15135 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 15136 distribution of this entire article is permitted in any medium, 15137 provided this notice is preserved. 15138 15139 These pages are [18]maintained by the GCC team. Last modified 15140 2022-10-26. 15141 15142 References 15143 15144 1. http://www.gnu.org/ 15145 2. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 15146 3. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 15147 4. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 15148 5. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 15149 6. https://gcc.gnu.org/gcc-4.0/changes.html 15150 7. https://gcc.gnu.org/gcc-4.0/buildstat.html 15151 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15152 9. https://gcc.gnu.org/index.html 15153 10. mailto:gcc (a] gcc.gnu.org 15154 11. https://gcc.gnu.org/mirrors.html 15155 12. https://gcc.gnu.org/git.html 15156 13. https://gcc.gnu.org/onlinedocs/ 15157 14. mailto:gcc-help (a] gcc.gnu.org 15158 15. mailto:gcc (a] gcc.gnu.org 15159 16. https://gcc.gnu.org/lists.html 15160 17. https://www.fsf.org/ 15161 18. https://gcc.gnu.org/about.html 15162 ====================================================================== 15163 http://gcc.gnu.org/gcc-4.0/changes.html 15164 GCC 4.0 Release Series 15165 Changes, New Features, and Fixes 15166 15167 The latest release in the 4.0 release series is [1]GCC 4.0.4. 15168 15169 Caveats 15170 15171 * GCC now generates location lists by default when compiling with 15172 debug info and optimization. 15173 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 15174 or later is needed to debug binaries containing location 15175 lists. 15176 + When you are trying to view a value of a variable in a part of 15177 a function where it has no location (for example when the 15178 variable is no longer used and thus its location was used for 15179 something else) GDB will say that it is not available. 15180 You can disable generating location lists by -fno-var-tracking. 15181 * GCC no longer accepts the -fwritable-strings option. Use named 15182 character arrays when you need a writable string. 15183 * The options -freduce-all-givs and -fmove-all-movables have been 15184 discontinued. They were used to circumvent a shortcoming in the 15185 heuristics of the old loop optimization code with respect to common 15186 Fortran constructs. The new (tree) loop optimizer works differently 15187 and doesn't need those work-arounds. 15188 * The graph-coloring register allocator, formerly enabled by the 15189 option -fnew-ra, has been discontinued. 15190 * -I- has been deprecated. -iquote is meant to replace the need for 15191 this option. 15192 * The MIPS -membedded-pic and -mrnames options have been removed. 15193 * All MIPS targets now require the GNU assembler. In particular, IRIX 15194 configurations can no longer use the MIPSpro assemblers, although 15195 they do still support the MIPSpro linkers. 15196 * The SPARC option -mflat has been removed. 15197 * English-language diagnostic messages will now use Unicode quotation 15198 marks in UTF-8 locales. (Non-English messages already used the 15199 quotes appropriate for the language in previous releases.) If your 15200 terminal does not support UTF-8 but you are using a UTF-8 locale 15201 (such locales are the default on many GNU/Linux systems) then you 15202 should set LC_CTYPE=C in the environment to disable that locale. 15203 Programs that parse diagnostics and expect plain ASCII 15204 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 15205 explanation of Unicode quotation marks for more information. 15206 * The specs file is no longer installed on most platforms. Most users 15207 will be totally unaffected. However, if you are accustomed to 15208 editing the specs file yourself, you will now have to use the 15209 -dumpspecs option to generate the specs file, and then edit the 15210 resulting file. 15211 15212 General Optimizer Improvements 15213 15214 * The [3]tree ssa branch has been merged. This merge has brought in a 15215 completely new optimization framework based on a higher level 15216 intermediate representation than the existing RTL representation. 15217 Numerous new code transformations based on the new framework are 15218 available in GCC 4.0, including: 15219 + Scalar replacement of aggregates 15220 + Constant propagation 15221 + Value range propagation 15222 + Partial redundancy elimination 15223 + Load and store motion 15224 + Strength reduction 15225 + Dead store elimination 15226 + Dead and unreachable code elimination 15227 + [4]Autovectorization 15228 + Loop interchange 15229 + Tail recursion by accumulation 15230 Many of these passes outperform their counterparts from previous 15231 GCC releases. 15232 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 15233 scheduling optimization intended for loops that perform heavy 15234 computations. 15235 15236 New Languages and Language specific improvements 15237 15238 C family 15239 15240 * The sentinel attribute has been added to GCC. This function 15241 attribute allows GCC to warn when variadic functions such as execl 15242 are not NULL terminated. See the GCC manual for a complete 15243 description of its behavior. 15244 * Given __attribute__((alias("target"))) it is now an error if target 15245 is not a symbol, defined in the same translation unit. This also 15246 applies to aliases created by #pragma weak alias=target. This is 15247 because it's meaningless to define an alias to an undefined symbol. 15248 On Solaris, the native assembler would have caught this error, but 15249 GNU as does not. 15250 15251 C and Objective-C 15252 15253 * The -Wstrict-aliasing=2 option has been added. This warning catches 15254 all unsafe cases, but it may also give a warning for some cases 15255 that are safe. 15256 * The cast-as-lvalue, conditional-expression-as-lvalue and 15257 compound-expression-as-lvalue extensions, which were deprecated in 15258 3.3.4 and 3.4, have been removed. 15259 * The -fwritable-strings option, which was deprecated in 3.4, has 15260 been removed. 15261 * #pragma pack() semantics have been brought closer to those used by 15262 other compilers. This also applies to C++. 15263 * Taking the address of a variable with register storage is invalid 15264 in C. GCC now issues an error instead of a warning. 15265 * Arrays of incomplete element type are invalid in C. GCC now issues 15266 an error for such arrays. Declarations such as extern struct s x[]; 15267 (where struct s has not been defined) can be moved after the 15268 definition of struct s. Function parameters declared as arrays of 15269 incomplete type can instead be declared as pointers. 15270 15271 C++ 15272 15273 * When compiling without optimizations (-O0), the C++ front end is 15274 much faster than in any previous versions of GCC. Independent 15275 testers have measured speed-ups up to 25% in real-world production 15276 code, compared to the 3.4 family (which was already the fastest 15277 version to date). Upgrading from older versions might show even 15278 bigger improvements. 15279 * ELF visibility attributes can now be applied to a class type, so 15280 that it affects every member function of a class at once, without 15281 having to specify each individually: 15282 class __attribute__ ((visibility("hidden"))) Foo 15283 { 15284 int foo1(); 15285 void foo2(); 15286 }; 15287 The syntax is deliberately similar to the __declspec() system used 15288 by Microsoft Windows based compilers, allowing cross-platform 15289 projects to easily reuse their existing macro system for denoting 15290 exports and imports. By explicitly marking internal classes never 15291 used outside a binary as hidden, one can completely avoid PLT 15292 indirection overheads during their usage by the compiler. You can 15293 find out more about the advantages of this at 15294 [6]https://www.akkadia.org/drepper/dsohowto.pdf 15295 * The -fvisibility-inlines-hidden option has been added which marks 15296 all inlineable functions as having hidden ELF visibility, thus 15297 removing their symbol and typeinfo from the exported symbol table 15298 of the output ELF binary. Using this option can reduce the exported 15299 symbol count of template-heavy code by up to 40% with no code 15300 change at all, thus notably improving link and load times for the 15301 binary as well as a reduction in size of up to 10%. Also, check the 15302 new [7]-fvisibility option. 15303 * The compiler now uses the library interface specified by the [8]C++ 15304 ABI for thread-safe initialization of function-scope static 15305 variables. Most users should leave this alone, but embedded 15306 programmers may want to disable this by specifying 15307 -fno-threadsafe-statics for a small savings in code size. 15308 * Taking the address of an explicit register variable is no longer 15309 supported. Note that C++ allows taking the address of variables 15310 with register storage so this will continue to compile with a 15311 warning. For example, assuming that r0 is a machine register: 15312 register int foo asm ("r0"); 15313 register int bar; 15314 &foo; // error, no longer accepted 15315 &bar; // OK, with a warning 15316 * G++ has an undocumented extension to virtual function covariancy 15317 rules that allowed the overrider to return a type that was 15318 implicitly convertable to the overridden function's return type. 15319 For instance a function returning void * could be overridden by a 15320 function returning T *. This is now deprecated and will be removed 15321 in a future release. 15322 * The G++ minimum and maximum operators (<? and >?) and their 15323 compound forms (<?=) and >?=) have been deprecated and will be 15324 removed in a future version. Code using these operators should be 15325 modified to use std::min and std::max instead. 15326 * Declaration of nested classes of class templates as friends are 15327 supported: 15328 template <typename T> struct A { 15329 class B {}; 15330 }; 15331 class C { 15332 template <typename T> friend class A<T>::B; 15333 }; 15334 This complements the feature member functions of class templates as 15335 friends introduced in GCC 3.4.0. 15336 * When declaring a friend class using an unqualified name, classes 15337 outside the innermost non-class scope are not searched: 15338 class A; 15339 namespace N { 15340 class B { 15341 friend class A; // Refer to N::A which has not been declared yet 15342 // because name outside namespace N are not searched 15343 friend class ::A; // Refer to ::A 15344 }; 15345 } 15346 Hiding the friend name until declaration is still not implemented. 15347 * Friends of classes defined outside their namespace are correctly 15348 handled: 15349 namespace N { 15350 class A; 15351 } 15352 class N::A { 15353 friend class B; // Refer to N::B in GCC 4.0.0 15354 // but ::B in earlier versions of GCC 15355 }; 15356 15357 Runtime Library (libstdc++) 15358 15359 * Optimization work: 15360 + Added efficient specializations of istream functions for char 15361 and wchar_t. 15362 + Further performance tuning of strings, in particular wrt 15363 single-char append and getline. 15364 + iter_swap - and therefore most of the mutating algorithms - 15365 now makes an unqualified call to swap when the value_type of 15366 the two iterators is the same. 15367 * A large subset of the features in Technical Report 1 (TR1 for 15368 short) is experimentally delivered (i.e., no guarantees about the 15369 implementation are provided. In particular it is not promised that 15370 the library will remain link-compatible when code using TR1 is 15371 used): 15372 + General utilities such as reference_wrapper and shared_ptr. 15373 + Function objects, i.e., result_of, mem_fn, bind, function. 15374 + Support for metaprogramming. 15375 + New containers such as tuple, array, unordered_set, 15376 unordered_map, unordered_multiset, unordered_multimap. 15377 * As usual, many bugs have been fixed and LWG resolutions implemented 15378 for the first time (e.g., DR 409). 15379 15380 Java 15381 15382 * In order to prevent naming conflicts with other implementations of 15383 these tools, some GCJ binaries have been renamed: 15384 + rmic is now grmic, 15385 + rmiregistry is now grmiregistry, and 15386 + jar is now fastjar. 15387 In particular, these names were problematic for the jpackage.org 15388 packaging conventions which install symlinks in /usr/bin that point 15389 to the preferred versions of these tools. 15390 * The -findirect-dispatch argument to the compiler now works and 15391 generates code following a new "binary compatibility" ABI. Code 15392 compiled this way follows the binary compatibility rules of the 15393 Java Language Specification. 15394 * libgcj now has support for using GCJ as a JIT, using the 15395 gnu.gcj.jit family of system properties. 15396 * libgcj can now find a shared library corresponding to the bytecode 15397 representation of a class. See the documentation for the new 15398 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 15399 property. 15400 * There have been many improvements to the class library. Here are 15401 some highlights: 15402 + Much more of AWT and Swing exist. 15403 + Many new packages and classes were added, including 15404 java.util.regex, java.net.URI, javax.crypto, 15405 javax.crypto.interfaces, javax.crypto.spec, javax.net, 15406 javax.net.ssl, javax.security.auth, 15407 javax.security.auth.callback, javax.security.auth.login, 15408 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 15409 javax.imageio, javax.imageio.event, javax.imageio.spi, 15410 javax.print, javax.print.attribute, 15411 javax.print.attribute.standard, javax.print.event, and 15412 javax.xml 15413 + Updated SAX and DOM, and imported GNU JAXP 15414 15415 Fortran 15416 15417 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 15418 front end. The new front end supports Fortran 90 and Fortran 95. It 15419 may not yet be as stable as the old Fortran front end. 15420 15421 Ada 15422 15423 * Ada (with tasking and Zero Cost Exceptions) is now available on 15424 many more targets, including but not limited to: alpha-linux, 15425 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 15426 s390x-linux, sparc-linux. 15427 * Some of the new Ada 2005 features are now implemented like 15428 Wide_Wide_Character and Ada.Containers. 15429 * Many bugs have been fixed, tools and documentation improved. 15430 * To compile Ada from the sources, install an older working Ada 15431 compiler and then use --enable-languages=ada at configuration time, 15432 since the Ada front end is not currently activated by default. See 15433 the [10]Installing GCC for details. 15434 15435 New Targets and Target Specific Improvements 15436 15437 H8/300 15438 15439 * The frame layout has changed. In the new layout, the prologue of a 15440 function first saves registers and then allocate space for locals, 15441 resulting in an 1% improvement on code size. 15442 15443 IA-32/x86-64 (AMD64) 15444 15445 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 15446 log1p, log2, logb and tan mathematical builtins (and their float 15447 and long double variants) are now implemented as inline x87 15448 intrinsics when using -ffast-math. 15449 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 15450 (and their float and long double variants) are now implemented as 15451 inline x87 intrinsics when using -ffast-math. 15452 * The x87's fsincos instruction is now used automatically with 15453 -ffast-math when calculating both the sin and cos of the same 15454 argument. 15455 * Instruction selection for multiplication and division by constants 15456 has been improved. 15457 15458 IA-64 15459 15460 * Floating point division, integer division and sqrt are now inlined, 15461 resulting in significant performance improvements on some codes. 15462 15463 MIPS 15464 15465 * Division by zero checks now use conditional traps if the target 15466 processor supports them. This decreases code size by one word per 15467 division operation. The old behavior (branch and break) can be 15468 obtained either at configure time by passing --with-divide=breaks 15469 to configure or at runtime by passing -mdivide-breaks to GCC. 15470 * Support for MIPS64 paired-single instructions has been added. It is 15471 enabled by -mpaired-single and can be accessed using both the 15472 target-independent vector extensions and new MIPS-specific built-in 15473 functions. 15474 * Support for the MIPS-3D ASE has been added. It is enabled by 15475 -mips3d and provides new MIPS-3D-specific built-in functions. 15476 * The -mexplicit-relocs option now supports static n64 code (as is 15477 used, for example, in 64-bit linux kernels). -mexplicit-relocs 15478 should now be feature-complete and is enabled by default when GCC 15479 is configured to use a compatible assembler. 15480 * Support for the NEC VR4130 series has been added. This support 15481 includes the use of VR-specific instructions and a new VR4130 15482 scheduler. Full VR4130 support can be selected with -march=vr4130 15483 while code for any ISA can be tuned for the VR4130 using 15484 -mtune=vr4130. There is also a new -mvr4130-align option that 15485 produces better schedules at the cost of increased code size. 15486 * Support for the Broadcom SB-1 has been extended. There is now an 15487 SB-1 scheduler as well as support for the SB-1-specific 15488 paired-single instructions. Full SB-1 support can be selected with 15489 -march=sb1 while code for any ISA can be optimized for the SB-1 15490 using -mtune=sb1. 15491 * The compiler can now work around errata in R4000, R4400, VR4120 and 15492 VR4130 processors. These workarounds are enabled by -mfix-r4000, 15493 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 15494 and VR4130 workarounds need binutils 2.16 or above. 15495 * IRIX shared libraries are now installed into the standard library 15496 directories: o32 libraries go into lib/, n32 libraries go into 15497 lib32/ and n64 libraries go into lib64/. 15498 * The compiler supports a new -msym32 option. It can be used to 15499 optimize n64 code in which all symbols are known to have 32-bit 15500 values. 15501 15502 S/390 and zSeries 15503 15504 * New command-line options help to generate code intended to run in 15505 an environment where stack space is restricted, e.g. Linux kernel 15506 code: 15507 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 15508 warnings for single functions that require large or dynamic 15509 stack frames. 15510 + -mstack-size and -mstack-guard generate code that checks for 15511 stack overflow at run time. 15512 + -mpacked-stack generates code that reduces the stack frame 15513 size of many functions by reusing unneeded parts of the stack 15514 bias area. 15515 * The -msoft-float option now ensures that generated code never 15516 accesses floating point registers. 15517 * The s390x-ibm-tpf target now fully supports C++, including 15518 exceptions and threads. 15519 * Various changes to improve performance of the generated code have 15520 been implemented, including: 15521 + GCC now uses sibling calls where possible. 15522 + Condition code handling has been optimized, allowing GCC to 15523 omit redundant comparisons in certain cases. 15524 + The cost function guiding many optimizations has been refined 15525 to more accurately represent the z900 and z990 processors. 15526 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 15527 instructions are now used to avoid conditional branches in 15528 certain cases. 15529 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 15530 optimize address arithmetic required to access large stack 15531 frames. 15532 + GCC now makes more efficient use of memory-to-memory type 15533 instructions (MVC, CLC, ...). 15534 + More precise tracking of special register use allows better 15535 instruction scheduling, in particular of the function prologue 15536 and epilogue sequences. 15537 + The Java front end now generates inline code to implement 15538 integer division, instead of calling library routines. 15539 15540 SPARC 15541 15542 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 15543 -mf934 have been removed. They have been replaced with -mcpu=xxx. 15544 * The internal model used to estimate the relative cost of each 15545 instruction has been updated. It is expected to give better results 15546 on recent UltraSPARC processors. 15547 * Code generation for function prologues and epilogues has been 15548 improved, resulting in better scheduling and allowing multiple exit 15549 points in functions. 15550 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 15551 It is enabled by -mvis and provides new built-in functions for VIS 15552 instructions on UltraSPARC processors. 15553 * The option -mapp-regs has been turned on by default on Solaris too. 15554 15555 NetWare 15556 15557 * Novell NetWare (on ix86, no other hardware platform was ever really 15558 supported by this OS) has been re-enabled and the ABI supported by 15559 GCC has been brought into sync with that of MetroWerks CodeWarrior 15560 (the ABI previously supported was that of some Unix systems, which 15561 NetWare never tried to support). 15562 15563 Obsolete Systems 15564 15565 Support for a number of older systems has been declared obsolete in GCC 15566 4.0. Unless there is activity to revive them, the next release of GCC 15567 will have their sources permanently removed. 15568 15569 All GCC ports for the following processor architectures have been 15570 declared obsolete: 15571 * Intel i860 15572 * Ubicom IP2022 15573 * National Semiconductor NS32K (ns32k) 15574 * Texas Instruments TMS320C[34]x 15575 15576 Also, those for some individual systems have been obsoleted: 15577 * SPARC family 15578 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 15579 sparc86x-*-elf) 15580 + OpenBSD 32-bit (sparc-*-openbsd*) 15581 15582 Documentation improvements 15583 15584 Other significant improvements 15585 15586 * Location lists are now generated by default when compiling with 15587 debug info and optimization. Location lists provide more accurate 15588 debug info about locations of variables and they allow debugging 15589 code compiled with -fomit-frame-pointer. 15590 * The -fvisibility option has been added which allows the default ELF 15591 visibility of all symbols to be set per compilation and the new 15592 #pragma GCC visibility preprocessor command allows the setting of 15593 default ELF visibility for a region of code. Using 15594 -fvisibility=hidden especially in combination with the new 15595 -fvisibility-inlines-hidden can yield substantial improvements in 15596 output binary quality including avoiding PLT indirection overheads, 15597 reduction of the exported symbol count by up to 60% (with resultant 15598 improvements to link and load times), better scope for the 15599 optimizer to improve code and up to a 20% reduction in binary size. 15600 Using these options correctly yields a binary with a similar symbol 15601 count to a Windows DLL. 15602 Perhaps more importantly, this new feature finally allows (with 15603 careful planning) complete avoidance of symbol clashes when 15604 manually loading shared objects with RTLD_GLOBAL, thus finally 15605 solving problems many projects such as python were forced to use 15606 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 15607 can find more information about using these options at 15608 [11]https://gcc.gnu.org/wiki/Visibility. 15609 __________________________________________________________________ 15610 15611 GCC 4.0.1 15612 15613 This is the [12]list of problem reports (PRs) from GCC's bug tracking 15614 system that are known to be fixed in the 4.0.1 release. This list might 15615 not be complete (that is, it is possible that some PRs that have been 15616 fixed are not listed here). 15617 15618 GCC 4.0.2 15619 15620 This is the [13]list of problem reports (PRs) from GCC's bug tracking 15621 system that are known to be fixed in the 4.0.2 release. This list might 15622 not be complete (that is, it is possible that some PRs that have been 15623 fixed are not listed here). 15624 15625 Unfortunately, due to a release engineering failure, this release has a 15626 regression on Solaris that will affect some C++ programs. We suggest 15627 that Solaris users apply a [14]patch that corrects the problem. Users 15628 who do not wish to apply the patch should explicitly link C++ programs 15629 with the -pthreads option, even if they do not use threads. This 15630 problem has been corrected in the current 4.0 branch sources and will 15631 not be present in GCC 4.0.3. 15632 15633 GCC 4.0.3 15634 15635 Starting with this release, the function getcontext is recognized by 15636 the compiler as having the same semantics as the setjmp function. In 15637 particular, the compiler will ensure that all registers are dead before 15638 calling such a function and will emit a warning about the variables 15639 that may be clobbered after the second return from the function. 15640 15641 GCC 4.0.4 15642 15643 This is the [15]list of problem reports (PRs) from GCC's bug tracking 15644 system that are known to be fixed in the 4.0.4 release. This list might 15645 not be complete (that is, it is possible that some PRs that have been 15646 fixed are not listed here). 15647 15648 The 4.0.4 release is provided for those that require a high degree of 15649 binary compatibility with previous 4.0.x releases. For most users, the 15650 GCC team recommends that version 4.1.1 or later be used instead." 15651 15652 15653 For questions related to the use of GCC, please consult these web 15654 pages and the [16]GCC manuals. If that fails, the 15655 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15656 web pages and the development of GCC are welcome on our developer 15657 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 15658 archives. 15659 15660 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 15661 distribution of this entire article is permitted in any medium, 15662 provided this notice is preserved. 15663 15664 These pages are [21]maintained by the GCC team. Last modified 15665 2022-10-26. 15666 15667 References 15668 15669 1. https://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 15670 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 15671 3. https://gcc.gnu.org/projects/tree-ssa/ 15672 4. https://gcc.gnu.org/projects/tree-ssa/vectorization.html 15673 5. https://gcc.gnu.org/news/sms.html 15674 6. https://www.akkadia.org/drepper/dsohowto.pdf 15675 7. https://gcc.gnu.org/gcc-4.0/changes.html#visibility 15676 8. https://itanium-cxx-abi.github.io/cxx-abi/ 15677 9. https://gcc.gnu.org/fortran/ 15678 10. https://gcc.gnu.org/install/ 15679 11. https://gcc.gnu.org/wiki/Visibility 15680 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 15681 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 15682 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 15683 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 15684 16. https://gcc.gnu.org/onlinedocs/ 15685 17. mailto:gcc-help (a] gcc.gnu.org 15686 18. mailto:gcc (a] gcc.gnu.org 15687 19. https://gcc.gnu.org/lists.html 15688 20. https://www.fsf.org/ 15689 21. https://gcc.gnu.org/about.html 15690 ====================================================================== 15691 http://gcc.gnu.org/gcc-3.4/index.html 15692 GCC 3.4 Release Series 15693 15694 (This release series is no longer supported.) 15695 15696 May 26, 2006 15697 15698 The [1]GNU project and the GCC developers are pleased to announce the 15699 release of GCC 3.4.6. 15700 15701 This release is a bug-fix release, containing fixes for regressions in 15702 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 15703 3.4.x series. 15704 15705 The GCC 3.4 release series includes numerous [2]new features, 15706 improvements, bug fixes, and other changes, thanks to an [3]amazing 15707 group of volunteers. 15708 15709 Release History 15710 15711 GCC 3.4.6 15712 March 6, 2006 ([4]changes) 15713 15714 GCC 3.4.5 15715 November 30, 2005 ([5]changes) 15716 15717 GCC 3.4.4 15718 May 18, 2005 ([6]changes) 15719 15720 GCC 3.4.3 15721 November 4, 2004 ([7]changes) 15722 15723 GCC 3.4.2 15724 September 6, 2004 ([8]changes) 15725 15726 GCC 3.4.1 15727 July 1, 2004 ([9]changes) 15728 15729 GCC 3.4.0 15730 April 18, 2004 ([10]changes) 15731 15732 References and Acknowledgements 15733 15734 GCC used to stand for the GNU C Compiler, but since the compiler 15735 supports several other languages aside from C, it now stands for the 15736 GNU Compiler Collection. 15737 15738 A list of [11]successful builds is updated as new information becomes 15739 available. 15740 15741 The GCC developers would like to thank the numerous people that have 15742 contributed new features, improvements, bug fixes, and other changes as 15743 well as test results to GCC. This [12]amazing group of volunteers is 15744 what makes GCC successful. 15745 15746 For additional information about GCC please refer to the [13]GCC 15747 project web site or contact the [14]GCC development mailing list. 15748 15749 To obtain GCC please use [15]our mirror sites, or [16]our version 15750 control system. 15751 15752 15753 For questions related to the use of GCC, please consult these web 15754 pages and the [17]GCC manuals. If that fails, the 15755 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15756 web pages and the development of GCC are welcome on our developer 15757 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 15758 archives. 15759 15760 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 15761 distribution of this entire article is permitted in any medium, 15762 provided this notice is preserved. 15763 15764 These pages are [22]maintained by the GCC team. Last modified 15765 2022-10-26. 15766 15767 References 15768 15769 1. http://www.gnu.org/ 15770 2. https://gcc.gnu.org/gcc-3.4/changes.html 15771 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15772 4. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 15773 5. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 15774 6. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 15775 7. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 15776 8. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 15777 9. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 15778 10. https://gcc.gnu.org/gcc-3.4/changes.html 15779 11. https://gcc.gnu.org/gcc-3.4/buildstat.html 15780 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15781 13. https://gcc.gnu.org/index.html 15782 14. mailto:gcc (a] gcc.gnu.org 15783 15. https://gcc.gnu.org/mirrors.html 15784 16. https://gcc.gnu.org/git.html 15785 17. https://gcc.gnu.org/onlinedocs/ 15786 18. mailto:gcc-help (a] gcc.gnu.org 15787 19. mailto:gcc (a] gcc.gnu.org 15788 20. https://gcc.gnu.org/lists.html 15789 21. https://www.fsf.org/ 15790 22. https://gcc.gnu.org/about.html 15791 ====================================================================== 15792 http://gcc.gnu.org/gcc-3.4/changes.html 15793 GCC 3.4 Release Series 15794 Changes, New Features, and Fixes 15795 15796 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 15797 is now closed. 15798 15799 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 15800 a bug, please make sure it's really GCC, and not your code, that is 15801 broken. 15802 15803 Caveats 15804 15805 * GNU Make is now required to build GCC. 15806 * With -nostdinc the preprocessor used to ignore both standard 15807 include paths and include paths contained in environment variables. 15808 It was neither documented nor intended that environment variable 15809 paths be ignored, so this has been corrected. 15810 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 15811 -fvolatile-static. It is unlikely that they worked correctly in any 15812 3.x release. 15813 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 15814 * Support for all the systems [3]obsoleted in GCC 3.3 has been 15815 removed from GCC 3.4. See below for a [4]list of systems which are 15816 obsoleted in this release. 15817 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 15818 compilers will not work. 15819 * The implementation of the [5]MIPS ABIs has changed. As a result, 15820 the code generated for certain MIPS targets will not be binary 15821 compatible with earlier releases. 15822 * In previous releases, the MIPS port had a fake "hilo" register with 15823 the user-visible name accum. This register has been removed. 15824 * The implementation of the [6]SPARC ABIs has changed. As a result, 15825 the code generated will not be binary compatible with earlier 15826 releases in certain cases. 15827 * The configure option --enable-threads=pthreads has been removed; 15828 use --enable-threads=posix instead, which should have the same 15829 effect. 15830 * Code size estimates used by inlining heuristics for C, Objective-C, 15831 C++ and Java have been redesigned significantly. As a result the 15832 parameters of -finline-insns, --param max-inline-insns-single and 15833 --param max-inline-insns-auto need to be reconsidered. 15834 * --param max-inline-slope and --param min-inline-insns have been 15835 removed; they are not needed for the new bottom-up inlining 15836 heuristics. 15837 * The new unit-at-a-time compilation scheme has several compatibility 15838 issues: 15839 + The order in which functions, variables, and top-level asm 15840 statements are emitted may have changed. Code relying on some 15841 particular ordering needs to be updated. The majority of such 15842 top-level asm statements can be replaced by section 15843 attributes. 15844 + Unreferenced static variables and functions are removed. This 15845 may result in undefined references when an asm statement 15846 refers to the variable/function directly. In that case either 15847 the variable/function shall be listed in asm statement operand 15848 or in the case of top-level asm statements the attribute used 15849 shall be used to force function/variable to be always output 15850 and considered as a possibly used by unknown code. 15851 For variables the attribute is accepted only by GCC 3.4 and 15852 newer, while for earlier versions it is sufficient to use 15853 unused to silence warnings about the variables not being 15854 referenced. To keep code portable across different GCC 15855 versions, you can use appropriate preprocessor conditionals. 15856 + Static functions now can use non-standard passing conventions 15857 that may break asm statements calling functions directly. 15858 Again the attribute used shall be used to prevent this 15859 behavior. 15860 As a temporary workaround, -fno-unit-at-a-time can be used, but 15861 this scheme may not be supported by future releases of GCC. 15862 * GCC 3.4 automatically places zero-initialized variables in the .bss 15863 section on some operating systems. Versions of GNU Emacs up to (and 15864 including) 21.3 will not work correctly when using this 15865 optimization; you can use -fno-zero-initialized-in-bss to disable 15866 it. 15867 * If GCC 3.4 is configured with --enable-threads=posix (the default 15868 on most targets that support pthreads) then _REENTRANT will be 15869 defined unconditionally by some libstdc++ headers. C++ code which 15870 relies on that macro to detect whether multi-threaded code is being 15871 compiled might change in meaning, possibly resulting in linker 15872 errors for single-threaded programs. Affected users of [7]Boost 15873 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 15874 See Bugzilla for [8]more information. 15875 15876 General Optimizer Improvements 15877 15878 * Usability of the profile feedback and coverage testing has been 15879 improved. 15880 + Performance of profiled programs has been improved by faster 15881 profile merging code. 15882 + Better use of the profile feedback for optimization (loop 15883 unrolling and loop peeling). 15884 + File locking support allowing fork() calls and parallel runs 15885 of profiled programs. 15886 + Coverage file format has been redesigned. 15887 + gcov coverage tool has been improved. 15888 + make profiledbootstrap available to build a faster compiler. 15889 Experiments made on i386 hardware showed an 11% speedup on -O0 15890 and a 7.5% speedup on -O2 compilation of a [9]large C++ 15891 testcase. 15892 + New value profiling pass enabled via -fprofile-values 15893 + New value profile transformations pass enabled via -fvpt aims 15894 to optimize some code sequences by exploiting knowledge about 15895 value ranges or other properties of the operands. At the 15896 moment a conversion of expensive divisions into cheaper 15897 operations has been implemented. 15898 + New -fprofile-generate and -fprofile-use command-line options 15899 to simplify the use of profile feedback. 15900 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 15901 Java which is enabled via -funit-at-a-time (and implied by -O2). In 15902 this scheme a whole file is parsed first and optimized later. The 15903 following basic inter-procedural optimizations are implemented: 15904 + Removal of unreachable functions and variables 15905 + Discovery of local functions (functions with static linkage 15906 whose address is never taken) 15907 + On i386, these local functions use register parameter passing 15908 conventions. 15909 + Reordering of functions in topological order of the call graph 15910 to enable better propagation of optimizing hints (such as the 15911 stack alignments needed by functions) in the back end. 15912 + Call graph based out-of-order inlining heuristics which allows 15913 to limit overall compilation unit growth (--param 15914 inline-unit-growth). 15915 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 15916 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 15917 CPU). 15918 * More realistic code size estimates used by inlining for C, 15919 Objective-C, C++ and Java. The growth of large functions can now be 15920 limited via --param large-function-insns and --param 15921 large-function-growth. 15922 * A new cfg-level loop optimizer pass replaces the old loop unrolling 15923 pass and adds two other loop transformations -- loop peeling and 15924 loop unswitching -- and also uses the profile feedback to limit 15925 code growth. (The three optimizations are enabled by 15926 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 15927 respectively). 15928 The old loop unroller still can be enabled by -fold-unroll-loops 15929 and may produce better code in some cases, especially when the 15930 webizer optimization pass is not run. 15931 * A new web construction pass enabled via -fweb (and implied by -O3) 15932 improves the quality of register allocation, CSE, first scheduling 15933 pass and some other optimization passes by avoiding re-use of 15934 pseudo registers with non-overlapping live ranges. The pass almost 15935 always improves code quality but does make debugging difficult and 15936 thus is not enabled by default by -O2 15937 The pass is especially effective as cleanup after code duplication 15938 passes, such as the loop unroller or the tracer. 15939 * Experimental implementations of superblock or trace scheduling in 15940 the second scheduling pass can be enabled via 15941 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 15942 15943 New Languages and Language specific improvements 15944 15945 Ada 15946 15947 * The Ada front end has been updated to include numerous bug fixes 15948 and enhancements. These include: 15949 + Improved project file support 15950 + Additional set of warnings about potential wrong code 15951 + Improved error messages 15952 + Improved code generation 15953 + Improved cross reference information 15954 + Improved inlining 15955 + Better run-time check elimination 15956 + Better error recovery 15957 + More efficient implementation of unbounded strings 15958 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 15959 ... 15960 + New GNAT.xxxx packages (e.g. GNAT.Strings, 15961 GNAT.Exception_Action) 15962 + New pragmas 15963 + New -gnatS switch replacing gnatpsta 15964 + Implementation of new Ada features (in particular limited 15965 with, limited aggregates) 15966 15967 C/Objective-C/C++ 15968 15969 * Precompiled headers are now supported. Precompiled headers can 15970 dramatically speed up compilation of some projects. There are some 15971 known defects in the current precompiled header implementation that 15972 will result in compiler crashes in relatively rare situations. 15973 Therefore, precompiled headers should be considered a "technology 15974 preview" in this release. Read the manual for details about how to 15975 use precompiled headers. 15976 * File handling in the preprocessor has been rewritten. GCC no longer 15977 gets confused by symlinks and hardlinks, and now has a correct 15978 implementation of #import and #pragma once. These two directives 15979 have therefore been un-deprecated. 15980 * The undocumented extension that allowed C programs to have a label 15981 at the end of a compound statement, which has been deprecated since 15982 GCC 3.0, has been removed. 15983 * The cast-as-lvalue extension has been removed for C++ and 15984 deprecated for C and Objective-C. In particular, code like this: 15985 int i; 15986 (char) i = 5; 15987 15988 or this: 15989 char *p; 15990 ((int *) p)++; 15991 15992 is no longer accepted for C++ and will not be accepted for C and 15993 Objective-C in a future version. 15994 * The conditional-expression-as-lvalue extension has been deprecated 15995 for C and Objective-C. In particular, code like this: 15996 int a, b, c; 15997 (a ? b : c) = 2; 15998 15999 will not be accepted for C and Objective-C in a future version. 16000 * The compound-expression-as-lvalue extension has been deprecated for 16001 C and Objective-C. In particular, code like this: 16002 int a, b; 16003 (a, b) = 2; 16004 16005 will not be accepted for C and Objective-C in a future version. A 16006 possible non-intrusive workaround is the following: 16007 (*(a, &b)) = 2; 16008 16009 * Several [10]built-in functions such as __builtin_popcount for 16010 counting bits, finding the highest and lowest bit in a word, and 16011 parity have been added. 16012 * The -fwritable-strings option has been deprecated and will be 16013 removed. 16014 * Many C math library functions are now recognized as built-ins and 16015 optimized. 16016 * The C, C++, and Objective-C compilers can now handle source files 16017 written in any character encoding supported by the host C library. 16018 The default input character set is taken from the current locale, 16019 and may be overridden with the -finput-charset command line option. 16020 In the future we will add support for inline encoding markers. 16021 16022 C++ 16023 16024 * G++ is now much closer to full conformance to the ISO/ANSI C++ 16025 standard. This means, among other things, that a lot of invalid 16026 constructs which used to be accepted in previous versions will now 16027 be rejected. It is very likely that existing C++ code will need to 16028 be fixed. This document lists some of the most common issues. 16029 * A hand-written recursive-descent C++ parser has replaced the 16030 YACC-derived C++ parser from previous GCC releases. The new parser 16031 contains much improved infrastructure needed for better parsing of 16032 C++ source codes, handling of extensions, and clean separation 16033 (where possible) between proper semantics analysis and parsing. The 16034 new parser fixes many bugs that were found in the old parser. 16035 * You must now use the typename and template keywords to disambiguate 16036 dependent names, as required by the C++ standard. 16037 struct K { 16038 typedef int mytype_t; 16039 }; 16040 16041 template <class T1> struct A { 16042 template <class T2> struct B { 16043 void callme(void); 16044 }; 16045 16046 template <int N> void bar(void) 16047 { 16048 // Use 'typename' to tell the parser that T1::mytype_t names 16049 // a type. This is needed because the name is dependent (in 16050 // this case, on template parameter T1). 16051 typename T1::mytype_t x; 16052 x = 0; 16053 } 16054 }; 16055 16056 template <class T> void template_func(void) 16057 { 16058 // Use 'template' to prefix member templates within 16059 // dependent types (a has type A<T>, which depends on 16060 // the template parameter T). 16061 A<T> a; 16062 a.template bar<0>(); 16063 16064 // Use 'template' to tell the parser that B is a nested 16065 // template class (dependent on template parameter T), and 16066 // 'typename' because the whole A<T>::B<int> is 16067 // the name of a type (again, dependent). 16068 typename A<T>::template B<int> b; 16069 b.callme(); 16070 } 16071 16072 void non_template_func(void) 16073 { 16074 // Outside of any template class or function, no names can be 16075 // dependent, so the use of the keyword 'typename' and 'template' 16076 // is not needed (and actually forbidden). 16077 A<K> a; 16078 a.bar<0>(); 16079 A<K>::B<float> b; 16080 b.callme(); 16081 } 16082 * In a template definition, unqualified names will no longer find 16083 members of a dependent base (as specified by [temp.dep]/3 in the 16084 C++ standard). For example, 16085 template <typename T> struct B { 16086 int m; 16087 int n; 16088 int f (); 16089 int g (); 16090 }; 16091 int n; 16092 int g (); 16093 template <typename T> struct C : B<T> { 16094 void h () 16095 { 16096 m = 0; // error 16097 f (); // error 16098 n = 0; // ::n is modified 16099 g (); // ::g is called 16100 } 16101 }; 16102 You must make the names dependent, e.g. by prefixing them with 16103 this->. Here is the corrected definition of C<T>::h, 16104 template <typename T> void C<T>::h () 16105 { 16106 this->m = 0; 16107 this->f (); 16108 this->n = 0 16109 this->g (); 16110 } 16111 As an alternative solution (unfortunately not backwards compatible 16112 with GCC 3.3), you may use using declarations instead of this->: 16113 template <typename T> struct C : B<T> { 16114 using B<T>::m; 16115 using B<T>::f; 16116 using B<T>::n; 16117 using B<T>::g; 16118 void h () 16119 { 16120 m = 0; 16121 f (); 16122 n = 0; 16123 g (); 16124 } 16125 }; 16126 * In templates, all non-dependent names are now looked up and bound 16127 at definition time (while parsing the code), instead of later when 16128 the template is instantiated. For instance: 16129 void foo(int); 16130 16131 template <int> struct A { 16132 static void bar(void){ 16133 foo('a'); 16134 } 16135 }; 16136 16137 void foo(char); 16138 16139 int main() 16140 { 16141 A<0>::bar(); // Calls foo(int), used to call foo(char). 16142 } 16143 16144 * In an explicit instantiation of a class template, you must use 16145 class or struct before the template-id: 16146 template <int N> 16147 class A {}; 16148 16149 template A<0>; // error, not accepted anymore 16150 template class A<0>; // OK 16151 * The "named return value" and "implicit typename" extensions have 16152 been removed. 16153 * Default arguments in function types have been deprecated and will 16154 be removed. 16155 * ARM-style name-injection of friend declarations has been deprecated 16156 and will be removed. For example: struct S { friend void f(); }; 16157 void g() { f(); } will not be accepted by future versions of G++; 16158 instead a declaration of "f" will need to be present outside of the 16159 scope of "S". 16160 * Covariant returns are implemented for all but varadic functions 16161 that require an adjustment. 16162 * When -pedantic is used, G++ now issues errors about spurious 16163 semicolons. For example, 16164 namespace N {}; // Invalid semicolon. 16165 void f() {}; // Invalid semicolon. 16166 * G++ no longer accepts attributes for a declarator after the 16167 initializer associated with that declarator. For example, 16168 X x(1) __attribute__((...)); 16169 is no longer accepted. Instead, use: 16170 X x __attribute__((...)) (1); 16171 * Inside the scope of a template class, the name of the class itself 16172 can be treated as either a class or a template. So GCC used to 16173 accept the class name as argument of type template, and template 16174 template parameter. However this is not C++ standard compliant. Now 16175 the name is not treated as a valid template template argument 16176 unless you qualify the name by its scope. For example, the code 16177 below no longer compiles. 16178 template <template <class> class TT> class X {}; 16179 template <class T> class Y { 16180 X<Y> x; // Invalid, Y is always a type template parameter. 16181 }; 16182 The valid code for the above example is 16183 X< ::Y> x; // Valid. 16184 (Notice the space between < and : to prevent GCC to interpret this 16185 as a digraph for [.) 16186 * Friend declarations that refer to template specializations are 16187 rejected if the template has not already been declared. For 16188 example, 16189 template <typename T> 16190 class C { 16191 friend void f<> (C&); 16192 }; 16193 is rejected. You must first declare f as a template, 16194 template <typename T> 16195 void f(T); 16196 * In case of friend declarations, every name used in the friend 16197 declaration must be accessible at the point of that declaration. 16198 Previous versions of G++ used to be less strict about this and 16199 allowed friend declarations for private class members, for example. 16200 See the ISO C++ Standard Committee's [11]defect report #209 for 16201 details. 16202 * Declaration of member functions of class templates as friends are 16203 supported. For example, 16204 template <typename T> struct A { 16205 void f(); 16206 }; 16207 class C { 16208 template <typename T> friend void A<T>::f(); 16209 }; 16210 * You must use template <> to introduce template specializations, as 16211 required by the standard. For example, 16212 template <typename T> 16213 struct S; 16214 16215 struct S<int> { }; 16216 is rejected. You must write, 16217 template <> struct S<int> {}; 16218 * G++ used to accept code like this, 16219 struct S { 16220 int h(); 16221 void f(int i = g()); 16222 int g(int i = h()); 16223 }; 16224 This behavior is not mandated by the standard. Now G++ issues an 16225 error about this code. To avoid the error, you must move the 16226 declaration of g before the declaration of f. The default arguments 16227 for g must be visible at the point where it is called. 16228 * The C++ ABI Section 3.3.3 specifications for the array construction 16229 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 16230 NULL when the allocator argument returns NULL. These changes are 16231 incorporated into the libstdc++ runtime library. 16232 * Using a name introduced by a typedef in a friend declaration or in 16233 an explicit instantiation is now rejected, as specified by the ISO 16234 C++ standard. 16235 class A; 16236 typedef A B; 16237 class C { 16238 friend class B; // error, no typedef name here 16239 friend B; // error, friend always needs class/struct/enum 16240 friend class A; // OK 16241 }; 16242 16243 template <int> class Q {}; 16244 typedef Q<0> R; 16245 template class R; // error, no typedef name here 16246 template class Q<0>; // OK 16247 * When allocating an array with a new expression, GCC used to allow 16248 parentheses around the type name. This is actually ill-formed and 16249 it is now rejected: 16250 int* a = new (int)[10]; // error, not accepted anymore 16251 int* a = new int[10]; // OK 16252 * When binding an rvalue of class type to a reference, the copy 16253 constructor of the class must be accessible. For instance, consider 16254 the following code: 16255 class A 16256 { 16257 public: 16258 A(); 16259 16260 private: 16261 A(const A&); // private copy ctor 16262 }; 16263 16264 A makeA(void); 16265 void foo(const A&); 16266 16267 void bar(void) 16268 { 16269 foo(A()); // error, copy ctor is not accessible 16270 foo(makeA()); // error, copy ctor is not accessible 16271 16272 A a1; 16273 foo(a1); // OK, a1 is a lvalue 16274 } 16275 This might be surprising at first sight, especially since most 16276 popular compilers do not correctly implement this rule ([12]further 16277 details). 16278 * When forming a pointer to member or a pointer to member function, 16279 access checks for class visibility (public, protected, private) are 16280 now performed using the qualifying scope of the name itself. This 16281 is better explained with an example: 16282 class A 16283 { 16284 public: 16285 void pub_func(); 16286 protected: 16287 void prot_func(); 16288 private: 16289 void priv_func(); 16290 }; 16291 16292 class B : public A 16293 { 16294 public: 16295 void foo() 16296 { 16297 &A::pub_func; // OK, pub_func is accessible through A 16298 &A::prot_func; // error, cannot access prot_func through A 16299 &A::priv_func; // error, cannot access priv_func through A 16300 16301 &B::pub_func; // OK, pub_func is accessible through B 16302 &B::prot_func; // OK, can access prot_func through B (within B) 16303 &B::priv_func; // error, cannot access priv_func through B 16304 } 16305 }; 16306 16307 Runtime Library (libstdc++) 16308 16309 * Optimization work: 16310 + Streamlined streambuf, filebuf, separate synched with C 16311 Standard I/O streambuf. 16312 + All formatted I/O now uses cached locale information. 16313 + STL optimizations (memory/speed for list, red-black trees as 16314 used by sets and maps). 16315 + More use of GCC builtins. 16316 + String optimizations (avoid contention on 16317 increment/decrement-and-test of the reference count in the 16318 empty-string object, constructor from input_iterators 16319 speedup). 16320 * Static linkage size reductions. 16321 * Large File Support (files larger than 2 GB on 32-bit systems). 16322 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 16323 * Generic character traits. 16324 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 16325 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 16326 * The allocator class is now standard-conformant, and two additional 16327 extension allocators have been added, mt_alloc and 16328 bitmap_allocator. 16329 * PCH support: -include bits/stdc++.h (2x compile speedup). 16330 * Rewrote __cxa_demangle with support for C++ style allocators. 16331 * New debug modes for STL containers and iterators. 16332 * Testsuite rewrite: five times as many tests, plus increasingly 16333 sophisticated tests, including I/O, MT, multi-locale, wide and 16334 narrow characters. 16335 * Use current versions of GNU "autotools" for build/configuration. 16336 16337 Objective-C 16338 16339 * The Objective-C front end has been updated to include the numerous 16340 bug fixes and enhancements previously available only in Apple's 16341 version of GCC. These include: 16342 + Structured exception (@try... @catch... @finally, @throw) and 16343 synchronization (@synchronized) support. These are accessible 16344 via the -fobjc-exceptions switch; as of this writing, they may 16345 only be used in conjunction with -fnext-runtime on Mac OS X 16346 10.3 and later. See [13]Options Controlling Objective-C 16347 Dialect for more information. 16348 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 16349 may now be encoded as 'B'. In addition, the back-end/codegen 16350 dependencies have been removed. 16351 + An overhaul of message dispatch construction, ensuring that 16352 the various receiver types (and casts thereof) are handled 16353 properly, and that correct diagnostics are issued. 16354 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 16355 (-freplace-objc-classes) debugging modes, currently available 16356 on Mac OS X 10.3 and later. See [14]Options Controlling 16357 Objective-C Dialect for more information. 16358 + Access to optimized runtime entry points (-fno-nil-receivers ) 16359 on the assumption that message receivers are never nil. This 16360 is currently available on Mac OS X 10.3 and later. See 16361 [15]Options Controlling Objective-C Dialect for more 16362 information. 16363 16364 Java 16365 16366 * Compiling a .jar file will now cause non-.class entries to be 16367 automatically compiled as resources. 16368 * libgcj has been ported to Darwin. 16369 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 16370 to gcj. 16371 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 16372 code from shared libraries. 16373 * libgcj has been much more completely merged with [16]GNU Classpath. 16374 * Class loading is now much more correct; in particular the caller's 16375 class loader is now used when that is required. 16376 * [17]Eclipse 2.x will run out of the box using gij. 16377 * Parts of java.nio have been implemented. Direct and indirect 16378 buffers work, as do fundamental file and socket operations. 16379 * java.awt has been improved, though it is still not ready for 16380 general use. 16381 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 16382 method. 16383 * The MinGW port has matured. Enhancements include socket timeout 16384 support, thread interruption, improved Runtime.exec() handling and 16385 support for accented characters in filenames. 16386 16387 Fortran 16388 16389 * Fortran improvements are listed in the [18]Fortran documentation. 16390 16391 New Targets and Target Specific Improvements 16392 16393 Alpha 16394 16395 * Several [19]built-in functions have been added such as 16396 __builtin_alpha_zap to allow utilizing the more obscure 16397 instructions of the CPU. 16398 * Parameter passing of complex arguments has changed to match the 16399 ABI. This change is incompatible with previous GCC versions, but 16400 does fix compatibility with the Tru64 compiler and several corner 16401 cases where GCC was incompatible with itself. 16402 16403 ARM 16404 16405 * Nicolas Pitre has contributed his hand-coded floating-point support 16406 code for ARM. It is both significantly smaller and faster than the 16407 existing C-based implementation, even when building applications 16408 for Thumb. The arm-elf configuration has been converted to use the 16409 new code. 16410 * Support for the Intel's iWMMXt architecture, a second generation 16411 XScale processor, has been added. Enabled at run time with the 16412 -mcpu=iwmmxt command line switch. 16413 * A new ARM target has been added: arm-wince-pe. This is similar to 16414 the arm-pe target, but it defaults to using the APCS32 ABI. 16415 * The existing ARM pipeline description has been converted to the use 16416 the [20]DFA processor pipeline model. There is not much change in 16417 code performance, but the description is now [21]easier to 16418 understand. 16419 * Support for the Cirrus EP9312 Maverick floating point co-processor 16420 added. Enabled at run time with the -mcpu=ep9312 command line 16421 switch. Note however that the multilibs to support this chip are 16422 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 16423 enable their production you will have to uncomment the entries in 16424 that file. 16425 16426 H8/300 16427 16428 * Support for long long has been added. 16429 * Support for saveall attribute has been added. 16430 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 16431 for H8/300H and H8S, which is much faster than the previous 16432 implementation. 16433 * A lot of small performance improvements. 16434 16435 IA-32/AMD64 (x86-64) 16436 16437 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 16438 -march=k8 and -mcpu=k8. 16439 * Scalar SSE code generation carefully avoids reformatting penalties, 16440 hidden dependencies and minimizes the number of uops generated on 16441 both Intel and AMD CPUs. 16442 * Vector MMX and SSE operands are now passed in registers to improve 16443 performance and match the argument passing convention used by the 16444 Intel C++ Compiler. As a result it is not possible to call 16445 functions accepting vector arguments compiled by older GCC version. 16446 * Conditional jump elimination is now more aggressive on modern CPUs. 16447 * The Athlon ports has been converted to use the DFA processor 16448 pipeline description. 16449 * Optimization of indirect tail calls is now possible in a similar 16450 fashion as direct sibcall optimization. 16451 * Further small performance improvements. 16452 * -m128bit-long-double is now less buggy. 16453 * __float128 support in 64-bit compilation. 16454 * Support for data structures exceeding 2GB in 64-bit mode. 16455 * -mcpu has been renamed to -mtune. 16456 16457 IA-64 16458 16459 * Tuning code for the Itanium 2 processor has been added. The 16460 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 16461 enabled by default now. To generate code tuned for Itanium 1 the 16462 option -mtune=itanium1 should be used. 16463 * [22]DFA processor pipeline descriptions for the IA-64 processors 16464 have been added. This resulted in about 3% improvement on the 16465 SPECInt2000 benchmark for Itanium 2. 16466 * Instruction bundling for the IA-64 processors has been rewritten 16467 using the DFA pipeline hazard recognizer. It resulted in about 60% 16468 compiler speedup on the SPECInt2000 C programs. 16469 16470 M32R 16471 16472 * Support for the M32R/2 processor has been added by Renesas. 16473 * Support for an M32R GNU/Linux target and PIC code generation has 16474 been added by Renesas. 16475 16476 M68000 16477 16478 * Bernardo Innocenti (Develer S.r.l.) has contributed the 16479 m68k-uclinux target, based on former work done by Paul Dale 16480 (SnapGear Inc.). Code generation for the ColdFire processors family 16481 has been enhanced and extended to support the MCF 53xx and MCF 54xx 16482 cores, integrating former work done by Peter Barada (Motorola). 16483 16484 MIPS 16485 16486 Processor-specific changes 16487 16488 * Support for the RM7000 and RM9000 processors has been added. It can 16489 be selected using the -march compiler option and should work with 16490 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 16491 * Support for revision 2 of the MIPS32 ISA has been added. It can be 16492 selected with the command-line option -march=mips32r2. 16493 * There is a new option, -mfix-sb1, to work around certain SB-1 16494 errata. 16495 16496 Configuration 16497 16498 * It is possible to customize GCC using the following configure-time 16499 options: 16500 + --with-arch, which specifies the default value of the -march 16501 option. 16502 + --with-tune, which specifies the default value of the -mtune 16503 option. 16504 + --with-abi, which specifies the default ABI. 16505 + --with-float=soft, which tells GCC to use software floating 16506 point by default. 16507 + --with-float=hard, which tells GCC to use hardware floating 16508 point by default. 16509 * A 64-bit GNU/Linux port has been added. The associated 16510 configurations are mips64-linux-gnu and mips64el-linux-gnu. 16511 * The 32-bit GNU/Linux port now supports Java. 16512 * The IRIX 6 configuration now supports the o32 ABI and will build 16513 o32 multilibs by default. This support is compatible with both 16514 binutils and the SGI tools, but note that several features, 16515 including debugging information and DWARF2 exception handling, are 16516 only available when using the GNU assembler. Use of the GNU 16517 assembler and linker (version 2.15 or above) is strongly 16518 recommended. 16519 * The IRIX 6 configuration now supports 128-bit long doubles. 16520 * There are two new RTEMS-specific configurations, mips-rtems and 16521 mipsel-rtems. 16522 * There are two new *-elf configurations, mipsisa32r2-elf and 16523 mipsisa32r2el-elf. 16524 16525 General 16526 16527 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 16528 will break binary compatibility with earlier releases. 16529 * GCC can now use explicit relocation operators when generating 16530 -mabicalls code. This behavior is controlled by -mexplicit-relocs 16531 and can have several performance benefits. For example: 16532 + It allows for more optimization of GOT accesses, including 16533 better scheduling and redundancy elimination. 16534 + It allows sibling calls to be implemented as jumps. 16535 + n32 and n64 leaf functions can use a call-clobbered global 16536 pointer instead of $28. 16537 + The code to set up $gp can be removed from functions that 16538 don't need it. 16539 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 16540 option is equivalent to the assembler's -xgot option and should be 16541 used instead of -Wa,-xgot. 16542 * Frame pointer elimination is now supported when generating 64-bit 16543 MIPS16 code. 16544 * Inline block moves have been optimized to take more account of 16545 alignment information. 16546 * Many internal changes have been made to the MIPS port, mostly aimed 16547 at reducing the reliance on assembler macros. 16548 16549 PowerPC 16550 16551 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 16552 [24]ABI incompatibilities regarding the way parameters are passed 16553 during functions calls. These changes may result in incompatibility 16554 between code compiled with GCC 3.3 and GCC 3.4. 16555 16556 PowerPC Darwin 16557 16558 * Support for shared/dylib gcc libraries has been added. It is 16559 enabled by default on powerpc-apple-darwin7.0.0 and up. 16560 * Libgcj is enabled by default. On systems older than 16561 powerpc-apple-darwin7.0.0 you need to install dlcompat. 16562 * 128-bit IBM extended precision format support added for long 16563 double. 16564 16565 PowerPC64 GNU/Linux 16566 16567 * By default, PowerPC64 GNU/Linux now uses natural alignment of 16568 structure elements. The old four byte alignment for double, with 16569 special rules for a struct starting with a double, can be chosen 16570 with -malign-power. This change may result in incompatibility 16571 between code compiled with GCC 3.3 and GCC 3.4. 16572 * -mabi=altivec is now the default rather than -mabi=no-altivec. 16573 * 128-bit IBM extended precision format support added for long 16574 double. 16575 16576 S/390 and zSeries 16577 16578 * New command-line options allow to specify the intended execution 16579 environment for generated code: 16580 + -mesa/-mzarch allows to specify whether to generate code 16581 running in ESA/390 mode or in z/Architecture mode (this is 16582 applicable to 31-bit code only). 16583 + -march allows to specify a minimum processor architecture 16584 level (g5, g6, z900, or z990). 16585 + -mtune allows to specify which processor to tune for. 16586 * It is possible to customize GCC using the following configure-time 16587 options: 16588 + --with-mode, which specifies whether to default to assuming 16589 ESA/390 or z/Architecture mode. 16590 + --with-arch, which specifies the default value of the -march 16591 option. 16592 + --with-tune, which specifies the default value of the -mtune 16593 option. 16594 * Support for the z990 processor has been added, and can be selected 16595 using -march=z990 or -mtune=z990. This includes instruction 16596 scheduling tuned for the superscalar instruction pipeline of the 16597 z990 processor as well as support for all new instructions provided 16598 by the long-displacement facility. 16599 * Support to generate 31-bit code optimized for zSeries processors 16600 (running in ESA/390 or in z/Architecture mode) has been added. This 16601 can be selected using -march=z900 and -mzarch respectively. 16602 * Instruction scheduling for the z900 and z990 processors now uses 16603 the DFA pipeline hazard recognizer. 16604 * GCC no longer generates code to maintain a stack backchain, 16605 previously used to generate stack backtraces for debugging 16606 purposes. As replacement that does not incur runtime overhead, 16607 DWARF-2 call frame information is provided by GCC; this is 16608 supported by GDB 6.1. The old behavior can be restored using the 16609 -mbackchain option. 16610 * The stack frame size of functions may now exceed 2 GB in 64-bit 16611 code. 16612 * A port for the 64-bit IBM TPF operating system has been added; the 16613 configuration is s390x-ibm-tpf. This configuration is supported as 16614 cross-compilation target only. 16615 * Various changes to improve the generated code have been 16616 implemented, including: 16617 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 16618 instructions to significantly speed up many floating-point 16619 applications. 16620 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 16621 WITH BORROW instructions to speed up long long arithmetic. 16622 + GCC now uses the SEARCH STRING instruction to implement 16623 strlen(). 16624 + In many cases, function call overhead for 31-bit code has been 16625 reduced by placing the literal pool after the function code 16626 instead of after the function prolog. 16627 + Register 14 is no longer reserved in 64-bit code. 16628 + Handling of global register variables has been improved. 16629 16630 SPARC 16631 16632 * The option -mflat is deprecated. 16633 * Support for large (> 2GB) frames has been added to the 64-bit port. 16634 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 16635 will break binary compatibility with earlier releases. 16636 * The default debugging format has been switched from STABS to 16637 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 16638 the default debugging format for 64-bit code on Solaris. 16639 16640 SuperH 16641 16642 * Support for the SH2E processor has been added. Enabled at run time 16643 with the -m2e command line switch, or at configure time by 16644 specifying sh2e as the machine part of the target triple. 16645 16646 V850 16647 16648 * Support for the Mitsubishi V850E1 processor has been added. This is 16649 a variant of the V850E processor with some additional debugging 16650 instructions. 16651 16652 Xtensa 16653 16654 * Several ABI bugs have been fixed. Unfortunately, these changes 16655 break binary compatibility with earlier releases. 16656 + For big-endian processors, the padding of aggregate return 16657 values larger than a word has changed. If the size of an 16658 aggregate return value is not a multiple of 32 bits, previous 16659 versions of GCC inserted padding in the most-significant bytes 16660 of the first return value register. Aggregates larger than a 16661 word are now padded in the least-significant bytes of the last 16662 return value register used. Aggregates smaller than a word are 16663 still padded in the most-significant bytes. The return value 16664 padding has not changed for little-endian processors. 16665 + Function arguments with 16-byte alignment are now properly 16666 aligned. 16667 + The implementation of the va_list type has changed. A va_list 16668 value created by va_start from a previous release cannot be 16669 used with va_arg from this release, or vice versa. 16670 * More processor configuration options for Xtensa processors are 16671 supported: 16672 + the ABS instruction is now optional; 16673 + the ADDX* and SUBX* instructions are now optional; 16674 + an experimental CONST16 instruction can be used to synthesize 16675 constants instead of loading them from constant pools. 16676 These and other Xtensa processor configuration options can no 16677 longer be enabled or disabled by command-line options; the 16678 processor configuration must be specified by the xtensa-config.h 16679 header file when building GCC. Additionally, the 16680 -mno-serialize-volatile option is no longer supported. 16681 16682 Obsolete Systems 16683 16684 Support for a number of older systems has been declared obsolete in GCC 16685 3.4. Unless there is activity to revive them, the next release of GCC 16686 will have their sources permanently removed. 16687 16688 All configurations of the following processor architectures have been 16689 declared obsolete: 16690 * Mitsubishi D30V, d30v-* 16691 * AT&T DSP1600 and DSP1610, dsp16xx-* 16692 * Intel 80960, i960 16693 16694 Also, some individual systems have been obsoleted: 16695 * ARM Family 16696 + Support for generating code for operation in APCS/26 mode 16697 (-mapcs-26). 16698 * IBM ESA/390 16699 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 16700 maintained and supported.) 16701 * Intel 386 family 16702 + MOSS, i?86-moss-msdos and i?86-*-moss* 16703 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 16704 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 16705 i?86-*-freebsd2* 16706 + GNU/Linux with a.out object format, i?86-linux*aout* 16707 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 16708 + Interix versions before Interix 3, i?86-*-interix 16709 + Mach microkernel, i?86-mach* 16710 + SCO UnixWare with UDK, i?86-*-udk* 16711 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 16712 + VSTa microkernel, i386-*-vsta 16713 * Motorola M68000 family 16714 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 16715 + NetBSD with a.out object format (before NetBSD 1.4), 16716 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 16717 + Generic System V r.4, m68k-*-sysv4* 16718 * VAX 16719 + Generic VAX, vax-*-* (This is generic VAX only; we have not 16720 obsoleted any VAX triples for specific operating systems.) 16721 16722 Documentation improvements 16723 16724 Other significant improvements 16725 16726 * The build system has undergone several significant cleanups. 16727 Subdirectories will only be configured if they are being built, and 16728 all subdirectory configures are run from the make command. The top 16729 level has been autoconfiscated. 16730 * Building GCC no longer writes to its source directory. This should 16731 help those wishing to share a read-only source directory over NFS 16732 or build from a CD. The exceptions to this feature are if you 16733 configure with either --enable-maintainer-mode or 16734 --enable-generated-files-in-srcdir. 16735 * The -W warning option has been renamed to -Wextra, which is more 16736 easily understood. The older spelling will be retained for 16737 backwards compatibility. 16738 * Substantial improvements in compile time have been made, 16739 particularly for non-optimizing compilations. 16740 __________________________________________________________________ 16741 16742 GCC 3.4.0 16743 16744 Bug Fixes 16745 16746 A vast number of bugs have been fixed in 3.4.0, too many to publish a 16747 complete list here. [26]Follow this link to query the Bugzilla database 16748 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 16749 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 16750 regressions. 16751 __________________________________________________________________ 16752 16753 GCC 3.4.1 16754 16755 Bug Fixes 16756 16757 This section lists the problem reports (PRs) from GCC's bug tracking 16758 system that are known to be fixed in the 3.4.1 release. This list might 16759 not be complete (that is, it is possible that some PRs that have been 16760 fixed are not listed here). 16761 16762 Bootstrap failures 16763 16764 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 16765 emitted - PIC related 16766 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 16767 * [29]14760 A bug in configure.in prevents using both 16768 --program-suffix and --program-prefix 16769 * [30]14671 [hppa64] bootstrap fails: ICE in 16770 save_call_clobbered_regs, in caller_save.c 16771 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 16772 Alpha 16773 * [32]15178 Solaris 9/x86 fails linking after stage 3 16774 16775 Multi-platform internal compiler errors (ICEs) 16776 16777 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 16778 input 16779 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 16780 * [35]14810 (c++) tree check failures with invalid code involving 16781 templates 16782 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 16783 cp/parser.c 16784 * [37]15044 (c++) ICE on syntax error, template header 16785 * [38]15057 (c++) Compiling of conditional value throw constructs 16786 cause a segmentation violation 16787 * [39]15064 (c++) typeid of template parameter gives ICE 16788 * [40]15142 (c++) ICE when passing a string where a char* is expected 16789 in a throw statement 16790 * [41]15159 ICE in rtl_verify_flow_info_1 16791 * [42]15165 (c++) ICE in instantiate_template 16792 * [43]15193 Unary minus using pointer to V4SF vector causes 16793 -fforce-mem to exhaust all memory 16794 * [44]15209 (c++) Runs out of memory with packed structs 16795 * [45]15227 (c++) Trouble with invalid function definition 16796 * [46]15285 (c++) instantiate_type ICE when forming pointer to 16797 template function 16798 * [47]15299 (c++) ICE in resolve_overloaded_unification 16799 * [48]15329 (c++) ICE on constructor of member template 16800 * [49]15550 ICE in extract_insn, in recog.c 16801 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 16802 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 16803 cp/name-lookup.c 16804 * [52]15666 [unit-at-a-time] Gcc abort on valid code 16805 * [53]15696 (c++) ICE with bad pointer-to-member code 16806 * [54]15701 (c++) ICE with friends and template template parameter 16807 * [55]15761 ICE in do_SUBST, in combine.c 16808 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 16809 16810 Ada 16811 16812 * [57]14538 All RTEMS targets broken for gnat 16813 16814 C front end 16815 16816 * [58]12391 missing warning about assigning to an incomplete type 16817 * [59]14649 atan(1.0) should not be a constant expression 16818 * [60]15004 [unit-at-a-time] no warning for unused paramater in 16819 static function 16820 * [61]15749 --pedantic-errors behaves differently from --pedantic 16821 with C-compiler on GNU/Linux 16822 16823 C++ compiler and library 16824 16825 * [62]10646 non-const reference is incorrectly matched in a "const T" 16826 partial specialization 16827 * [63]12077 wcin.rdbuf()->in_avail() return value too high 16828 * [64]13598 enc_filebuf doesn't work 16829 * [65]14211 const_cast returns lvalue but should be rvalue 16830 * [66]14220 num_put::do_put() undesired float/double behavior 16831 * [67]14245 problem with user-defined allocators in std::basic_string 16832 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 16833 const_iterator 16834 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 16835 FILE* 16836 * [70]14668 no warning anymore for reevaluation of declaration 16837 * [71]14775 LFS (large file support) tests missing 16838 * [72]14821 Duplicate namespace alias declaration should not conflict 16839 * [73]14930 Friend declaration ignored 16840 * [74]14932 cannot use offsetof to get offsets of array elements in 16841 g++ 3.4.0 16842 * [75]14950 [non unit-at-a-time] always_inline does not mix with 16843 templates and -O0 16844 * [76]14962 g++ ignores #pragma redefine_extname 16845 * [77]14975 Segfault on low-level write error during imbue 16846 * [78]15002 Linewise stream input is unusably slow (std::string slow) 16847 * [79]15025 compiler accepts redeclaration of template as 16848 non-template 16849 * [80]15046 [arm] Math functions misdetected by cross configuration 16850 * [81]15069 a bit test on a variable of enum type is miscompiled 16851 * [82]15074 g++ -lsupc++ still links against libstdc++ 16852 * [83]15083 spurious "statement has no effect" warning 16853 * [84]15096 parse error with templates and pointer to const member 16854 * [85]15287 combination of operator[] and operator .* fails in 16855 templates 16856 * [86]15317 __attribute__ unused in first parameter of constructor 16857 gives error 16858 * [87]15337 sizeof on incomplete type diagnostic 16859 * [88]15361 bitset<>::_Find_next fails 16860 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 16861 namespaces 16862 * [90]15427 valid code results in incomplete type error 16863 * [91]15471 Incorrect member pointer offsets in anonymous 16864 structs/unions 16865 * [92]15503 nested template problem 16866 * [93]15507 compiler hangs while laying out union 16867 * [94]15542 operator & and template definitions 16868 * [95]15565 SLES9: leading + sign for unsigned int with showpos 16869 * [96]15625 friend defined inside a template fails to find static 16870 function 16871 * [97]15629 Function templates, overloads, and friend name injection 16872 * [98]15742 'noreturn' attribute ignored in method of template 16873 functions. 16874 * [99]15775 Allocator::pointer consistently ignored 16875 * [100]15821 Duplicate namespace alias within namespace rejected 16876 * [101]15862 'enum yn' fails (confict with undeclared builtin) 16877 * [102]15875 rejects pointer to member in template 16878 * [103]15877 valid code using templates and anonymous enums is 16879 rejected 16880 * [104]15947 Puzzling error message for wrong destructor declaration 16881 in template class 16882 * [105]16020 cannot copy __gnu_debug::bitset 16883 * [106]16154 input iterator concept too restrictive 16884 * [107]16174 deducing top-level consts 16885 16886 Java 16887 16888 * [108]14315 Java compiler is not parallel make safe 16889 16890 Fortran 16891 16892 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 16893 16894 Objective-C 16895 16896 * [110]7993 private variables cannot be shadowed in subclasses 16897 16898 Optimization bugs 16899 16900 * [111]15228 useless copies of floating point operands 16901 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 16902 functions not optimized away 16903 * [113]15945 Incorrect floating point optimization 16904 * [114]15526 ftrapv aborts on 0 * (-1) 16905 * [115]14690 Miscompiled POOMA tests 16906 * [116]15112 GCC generates code to write to unchanging memory 16907 16908 Preprocessor 16909 16910 * [117]15067 Minor glitch in the source of cpp 16911 16912 Main driver program bugs 16913 16914 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 16915 ldstyle_liblookup 16916 16917 x86-specific (Intel/AMD) 16918 16919 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 16920 section} 16921 16922 HPPA-specific 16923 16924 * [120]14782 GCC produces an unaligned data access at -O2 16925 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 16926 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 16927 16928 IA64-specific 16929 16930 * [123]14610 __float80 constants incorrectly emitted 16931 * [124]14813 init_array sections are initialized in the wrong order 16932 * [125]14857 GCC segfault on duplicated asm statement 16933 * [126]15598 Gcc 3.4 ICE on valid code 16934 * [127]15653 Gcc 3.4 ICE on valid code 16935 16936 MIPS-specific 16937 16938 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 16939 -mno-split-addresses -mno-explicit-relocs 16940 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 16941 2.14.91 16942 * [130]16144 Bogus reference to __divdf3 when -O1 16943 * [131]16176 Miscompilation of unaligned data in MIPS backend 16944 16945 PowerPC-specific 16946 16947 * [132]11591 ICE in gcc.dg/altivec-5.c 16948 * [133]12028 powerpc-eabispe produces bad sCOND operation 16949 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 16950 * [135]14567 long double and va_arg complex args 16951 * [136]14715 Altivec stack layout may overlap gpr save with stack 16952 temps 16953 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 16954 option is used. 16955 * [138]14924 Compiler ICE on valid code 16956 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 16957 * [140]15106 vector varargs failure passing from altivec to 16958 non-altivec code for -m32 16959 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 16960 half-word operation 16961 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 16962 and stvx 16963 * [143]15662 Segmentation fault when an exception is thrown - even if 16964 try and catch are specified 16965 16966 s390-specific 16967 16968 * [144]15054 Bad code due to overlapping stack temporaries 16969 16970 SPARC-specific 16971 16972 * [145]15783 ICE with union assignment in 64-bit mode 16973 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 16974 R_SPARC_UA32" 16975 16976 x86-64-specific 16977 16978 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 16979 * [148]14723 Backported -march=nocona from mainline 16980 * [149]15290 __float128 failed to pass to function properly 16981 16982 Cygwin/Mingw32-specific 16983 16984 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 16985 conformant to MS layout 16986 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 16987 worker on windows32 targets 16988 16989 Bugs specific to embedded processors 16990 16991 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 16992 varaible on stack 16993 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 16994 gcc continues to use the register as if it was not clobbered 16995 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 16996 TARGET_COLDFIRE 16997 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 16998 * [156]14457 [m6811hc] ICE with simple c++ source 16999 * [157]14542 [m6811hc] ICE on simple source 17000 * [158]15100 [SH] cc1plus got hang-up on 17001 libstdc++-v3/testsuite/abi_check.cc 17002 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 17003 cris-* 17004 * [160]15396 [SH] ICE with -O2 -fPIC 17005 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 17006 ColdFire 17007 17008 Testsuite problems (compiler not affected) 17009 17010 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 17011 * [163]15488 (libstdc++) possibly insufficient file permissions for 17012 executing test suite 17013 * [164]15489 (libstdc++) testsuite_files determined incorrectly 17014 17015 Documentation bugs 17016 17017 * [165]13928 (libstdc++) no whatis info in some man pages generated 17018 by doxygen 17019 * [166]14150 Ada documentation out of date 17020 * [167]14949 (c++) Need to document method visibility changes 17021 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 17022 __________________________________________________________________ 17023 17024 GCC 3.4.2 17025 17026 Bug Fixes 17027 17028 This section lists the problem reports (PRs) from GCC's bug tracking 17029 system that are known to be fixed in the 3.4.2 release. This list might 17030 not be complete (that is, it is possible that some PRs that have been 17031 fixed are not listed here). 17032 17033 Bootstrap failures and issues 17034 17035 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 17036 libstdc++-v3/testsuite 17037 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 17038 profiledbootstrap does not work with the built compiler 17039 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 17040 17041 Multi-platform internal compiler errors (ICEs) 17042 17043 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 17044 cp_parser_class_specifier, in cp/parser.c 17045 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 17046 * [174]15461 (c++) ICE due to NRV and inlining 17047 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 17048 * [176]16180 ICE: segmentation fault in RTL optimization 17049 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 17050 * [178]16408 ICE: in delete_insn, in cfgrtl.c 17051 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 17052 the name of any other entity 17053 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 17054 * [181]16706 (c++) ICE in finish_member_declaration, in 17055 cp/semantics.c 17056 * [182]16810 (c++) Legal C++ program with cast gives ICE in 17057 build_ptrmemfunc 17058 * [183]16851 (c++) ICE when throwing a comma expression 17059 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 17060 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 17061 cp/typeck.c 17062 * [186]16905 (c++) ICE (segfault) with exceptions 17063 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 17064 redefinition 17065 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 17066 (identifier_node) in dependent_template_p, in cp/pt.c 17067 17068 Preprocessor bugs 17069 17070 * [189]16366 Preprocessor option -remap causes memory corruption 17071 17072 Optimization 17073 17074 * [190]15345 unreferenced nested inline functions not optimized away 17075 * [191]16590 Incorrect execution when compiling with -O2 17076 * [192]16693 Bitwise AND is lost when used within a cast to an enum 17077 of the same precision 17078 * [193]17078 Jump into if(0) substatement fails 17079 17080 Problems in generated debug information 17081 17082 * [194]13956 incorrect stabs for nested local variables 17083 17084 C front end bugs 17085 17086 * [195]16684 GCC should not warn about redundant redeclarations of 17087 built-ins 17088 17089 C++ compiler and library 17090 17091 * [196]12658 Thread safety problems in locale::global() and 17092 locale::locale() 17093 * [197]13092 g++ accepts invalid pointer-to-member conversion 17094 * [198]15320 Excessive memory consumption 17095 * [199]16246 Incorrect template argument deduction 17096 * [200]16273 Memory exhausted when using nested classes and virtual 17097 functions 17098 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 17099 * [202]16411 undefined reference to 17100 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 17101 >::file() 17102 * [203]16489 G++ incorrectly rejects use of a null constant integral 17103 expression as a null constant pointer 17104 * [204]16618 offsetof fails with constant member 17105 * [205]16637 syntax error reported for valid input code 17106 * [206]16717 __attribute__((constructor)) broken in C++ 17107 * [207]16813 compiler error in DEBUG version of range insertion 17108 std::map::insert 17109 * [208]16853 pointer-to-member initialization from incompatible one 17110 accepted 17111 * [209]16889 ambiguity is not detected 17112 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 17113 17114 Java compiler and library 17115 17116 * [211]7587 direct threaded interpreter not thread-safe 17117 * [212]16473 ServerSocket accept() leaks file descriptors 17118 * [213]16478 Hash synchronization deadlock with finalizers 17119 17120 Alpha-specific 17121 17122 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 17123 * [215]16974 could not split insn (ice in final_scan_insn, in 17124 final.c) 17125 17126 x86-specific 17127 17128 * [216]16298 ICE in output_operand 17129 * [217]17113 ICE with SSE2 intrinsics 17130 17131 x86-64 specific 17132 17133 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 17134 17135 MIPS-specific 17136 17137 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 17138 * [220]16325 [mips64] value profiling clobbers gp on mips 17139 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 17140 char[]s 17141 * [222]16380 [mips64] Use of uninitialised register after dbra 17142 conversion 17143 * [223]16407 [mips64] Unaligned access to local variables 17144 * [224]16643 [mips64] verify_local_live_at_start ICE after 17145 crossjumping & cfgcleanup 17146 17147 ARM-specific 17148 17149 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 17150 off by 1 17151 * [226]15948 THUMB: ICE with non-commutative cbranch 17152 * [227]17019 THUMB: bad switch statement in md code for 17153 addsi3_cbranch_scratch 17154 17155 IA64-specific 17156 17157 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 17158 (-mtune=merced) 17159 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 17160 (-mtune=itanium) 17161 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 17162 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 17163 result 17164 * [232]16445 ICE on valid code: don't count ignored insns 17165 * [233]16490 ICE (segfault) while compiling with -fprofile-use 17166 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 17167 17168 PowerPC-specific 17169 17170 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 17171 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 17172 issue) 17173 17174 SPARC-specific 17175 17176 * [237]16199 ICE while compiling apache 2.0.49 17177 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 17178 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 17179 17180 Bugs specific to embedded processors 17181 17182 * [240]16379 [m32r] can't output large model function call of memcpy 17183 * [241]17093 [m32r] ICE with -msdata=use -O0 17184 * [242]17119 [m32r] ICE at switch case 0x8000 17185 17186 DJGPP-specific 17187 17188 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 17189 17190 Alpha Tru64-specific 17191 17192 * [244]16210 libstdc++ gratuitously omits "long long" I/O 17193 17194 Testsuite, documentation issues (compiler is not affected): 17195 17196 * [245]15488 (libstdc++) possibly insufficient file permissions for 17197 executing test suite 17198 * [246]16250 ada/doctools runs makeinfo even in release tarball 17199 __________________________________________________________________ 17200 17201 GCC 3.4.3 17202 17203 This is the [247]list of problem reports (PRs) from GCC's bug tracking 17204 system that are known to be fixed in the 3.4.3 release. This list might 17205 not be complete (that is, it is possible that some PRs that have been 17206 fixed are not listed here). 17207 17208 Bootstrap failures 17209 17210 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 17211 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 17212 when undeclared 17213 17214 Internal compiler errors (ICEs) affecting multiple platforms 17215 17216 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 17217 .class files 17218 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 17219 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 17220 directive 17221 * [253]16566 ICE with flexible arrays 17222 * [254]17023 ICE with nested functions in parameter declaration 17223 * [255]17027 ICE with noreturn function in loop at -O2 17224 * [256]17524 ICE in grokdeclarator, in cp/decl.c 17225 * [257]17826 (c++) ICE in cp_tree_equal 17226 17227 C and optimization bugs 17228 17229 * [258]15526 -ftrapv aborts on 0 * (-1) 17230 * [259]16999 #ident stopped working 17231 * [260]17503 quadratic behaviour in invalid_mode_change_p 17232 * [261]17581 Long long arithmetic fails inside a switch/case 17233 statement when compiled with -O2 17234 * [262]18129 -fwritable-strings doesn't work 17235 17236 C++ compiler and library bugs 17237 17238 * [263]10975 incorrect initial ostringstream::tellp() 17239 * [264]11722 Unbuffered filebuf::sgetn is slow 17240 * [265]14534 Unrecognizing static function as a template parameter 17241 when its return value is also templated 17242 * [266]15172 Copy constructor optimization in aggregate 17243 initialization 17244 * [267]15786 Bad error message for frequently occuring error. 17245 * [268]16162 Rejects valid member-template-definition 17246 * [269]16612 empty basic_strings can't live in shared memory 17247 * [270]16715 std::basic_iostream is instantiated when used, even 17248 though instantiations are already contained in libstdc++ 17249 * [271]16848 code in /ext/demangle.h appears broken 17250 * [272]17132 GCC fails to eliminate function template specialization 17251 when argument deduction fails 17252 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 17253 in ropeimpl.h 17254 * [274]17327 use of `enumeral_type' in template type unification 17255 * [275]17393 "unused variable '._0'" warning with -Wall 17256 * [276]17501 Confusion with member templates 17257 * [277]17537 g++ not passing -lstdc++ to linker when all command line 17258 arguments are libraries 17259 * [278]17585 usage of unqualified name of static member from within 17260 class not allowed 17261 * [279]17821 Poor diagnostic for using "." instead of "->" 17262 * [280]17829 wrong error: call of overloaded function is ambiguous 17263 * [281]17851 Misleading diagnostic for invalid function declarations 17264 with undeclared types 17265 * [282]17976 Destructor is called twice 17266 * [283]18020 rejects valid definition of enum value in template 17267 * [284]18093 bogus conflict in namespace aliasing 17268 * [285]18140 C++ parser bug when using >> in templates 17269 17270 Fortran 17271 17272 * [286]17541 data statements with double precision constants fail 17273 17274 x86-specific 17275 17276 * [287]17853 -O2 ICE for MMX testcase 17277 17278 SPARC-specific 17279 17280 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 17281 17282 Darwin-specific 17283 17284 * [289]17167 FATAL:Symbol L_foo$stub already defined. 17285 17286 AIX-specific 17287 17288 * [290]17277 could not catch an exception when specified -maix64 17289 17290 Solaris-specific 17291 17292 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 17293 missing from system libraries 17294 17295 HP/UX specific: 17296 17297 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 17298 17299 ARM-specific 17300 17301 * [293]17384 ICE with mode attribute on structures 17302 17303 MIPS-specific 17304 17305 * [294]17770 No NOP after LWL with -mips1 17306 17307 Other embedded target specific 17308 17309 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 17310 * [296]14064 [avr-elf] -fdata-sections triggers ICE 17311 * [297]14678 [m68hc11-elf] gcc ICE 17312 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 17313 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 17314 target 17315 * [300]15886 [SH] Miscompilation with -O2 -fPIC 17316 * [301]16884 [avr-elf] [fweb related] bug while initializing 17317 variables 17318 17319 Bugs relating to debugger support 17320 17321 * [302]13841 missing debug info for _Complex function arguments 17322 * [303]15860 [big-endian targets] No DW_AT_location debug info is 17323 emitted for formal arguments to a function that uses "register" 17324 qualifiers 17325 17326 Testsuite issues (compiler not affected) 17327 17328 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 17329 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 17330 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 17331 testsuite 17332 17333 Documentation 17334 17335 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 17336 should be en_GB 17337 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 17338 document broken shell 17339 * [309]16406 USE_LD_AS_NEEDED undocumented 17340 __________________________________________________________________ 17341 17342 GCC 3.4.4 17343 17344 This is the [310]list of problem reports (PRs) from GCC's bug tracking 17345 system that are known to be fixed in the 3.4.4 release. This list might 17346 not be complete (that is, it is possible that some PRs that have been 17347 fixed are not listed here). 17348 __________________________________________________________________ 17349 17350 GCC 3.4.5 17351 17352 This is the [311]list of problem reports (PRs) from GCC's bug tracking 17353 system that are known to be fixed in the 3.4.5 release. This list might 17354 not be complete (that is, it is possible that some PRs that have been 17355 fixed are not listed here). 17356 17357 Bootstrap issues 17358 17359 * [312]24688 sco_math fixincl breaks math.h 17360 17361 C compiler bugs 17362 17363 * [313]17188 struct Foo { } redefinition 17364 * [314]20187 wrong code for ((unsigned char)(unsigned long 17365 long)((a?a:1)&(a*b)))?0:1) 17366 * [315]21873 infinite warning loop on bad array initializer 17367 * [316]21899 enum definition accepts values to be overriden 17368 * [317]22061 ICE in find_function_data, in function.c 17369 * [318]22308 Failure to diagnose violation of constraint 6.516p2 17370 * [319]22458 ICE on missing brace 17371 * [320]22589 ICE casting to long long 17372 * [321]24101 Segfault with preprocessed source 17373 17374 C++ compiler and library bugs 17375 17376 * [322]10611 operations on vector mode not recognized in C++ 17377 * [323]13377 unexpected behavior of namespace usage directive 17378 * [324]16002 Strange error message with new parser 17379 * [325]17413 local classes as template argument 17380 * [326]17609 spurious error message after using keyword 17381 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 17382 * [328]18124 ICE with invalid template template parameter 17383 * [329]18155 typedef in template declaration not rejected 17384 * [330]18177 ICE with const_cast for undeclared variable 17385 * [331]18368 C++ error message regression 17386 * [332]16378 ICE when returning a copy of a packed member 17387 * [333]18466 int ::i; accepted 17388 * [334]18512 ICE on invalid usage of template base class 17389 * [335]18454 ICE when returning undefined type 17390 * [336]18738 typename not allowed with non-dependent qualified name 17391 * [337]18803 rejects access to operator() in template 17392 * [338]19004 ICE in uses_template_parms, in cp/pt.c 17393 * [339]19208 Spurious error about variably modified type 17394 * [340]18253 bad error message / ICE for invalid template parameter 17395 * [341]19608 ICE after friend function definition in local class 17396 * [342]19884 ICE on explicit instantiation of a non-template 17397 constructor 17398 * [343]20153 ICE when C++ template function contains anonymous union 17399 * [344]20563 Infinite loop in diagnostic (and ice after error 17400 message) 17401 * [345]20789 ICE with incomplete type in template 17402 * [346]21336 Internal compiler error when using custom new operators 17403 * [347]21768 ICE in error message due to violation of coding 17404 conventions 17405 * [348]21853 constness of pointer to data member ignored 17406 * [349]21903 Default argument of template function causes a 17407 compile-time error 17408 * [350]21983 multiple diagnostics 17409 * [351]21987 New testsuite failure 17410 g++.dg/warn/conversion-function-1.C 17411 * [352]22153 ICE on invalid template specialization 17412 * [353]22172 Internal compiler error, seg fault. 17413 * [354]21286 filebuf::xsgetn vs pipes 17414 * [355]22233 ICE with wrong number of template parameters 17415 * [356]22508 ICE after invalid operator new 17416 * [357]22545 ICE with pointer to class member & user defined 17417 conversion operator 17418 * [358]23528 Wrong default allocator in ext/hash_map 17419 * [359]23550 char_traits requirements/1.cc test bad math 17420 * [360]23586 Bad diagnostic for invalid namespace-name 17421 * [361]23624 ICE in invert_truthvalue, in fold-const.c 17422 * [362]23639 Bad error message: not a member of '<declaration error>' 17423 * [363]23797 ICE on typename outside template 17424 * [364]23965 Bogus error message: no matching function for call to 17425 'foo(<type error>)' 17426 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 17427 error> 17428 * [366]24580 virtual base class cause exception not to be caught 17429 17430 Problems in generated debug information 17431 17432 * [367]24267 Bad DWARF for altivec vectors 17433 17434 Optimizations issues 17435 17436 * [368]17810 ICE in verify_local_live_at_start 17437 * [369]17860 Wrong generated code for loop with varying bound 17438 * [370]21709 ICE on compile-time complex NaN 17439 * [371]21964 broken tail call at -O2 or more 17440 * [372]22167 Strange optimization bug when using -Os 17441 * [373]22619 Compilation failure for real_const_1.f and 17442 real_const_2.f90 17443 * [374]23241 Invalid code generated for comparison of uchar to 255 17444 * [375]23478 Miscompilation due to reloading of a var that is also 17445 used in EH pad 17446 * [376]24470 segmentation fault in cc1plus when compiling with -O 17447 * [377]24950 ICE in operand_subword_force 17448 17449 Precompiled headers problems 17450 17451 * [378]14400 Cannot compile qt-x11-free-3.3.0 17452 * [379]14940 PCH largefile test fails on various platforms 17453 17454 Preprocessor bugs 17455 17456 * [380]20239 ICE on empty preprocessed input 17457 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 17458 source directory 17459 17460 Testsuite issues 17461 17462 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 17463 i686-pc-linux-gnu 17464 17465 Alpha specific 17466 17467 * [383]21888 bootstrap failure with linker relaxation enabled 17468 17469 ARM specific 17470 17471 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 17472 * [385]23985 Memory aliasing information incorrect in inlined memcpy 17473 17474 ColdFile specific 17475 17476 * [386]16719 Illegal move of byte into address register causes 17477 compiler to ICE 17478 17479 HPPA specific 17480 17481 * [387]21723 ICE while building libgfortran 17482 * [388]21841 -mhp-ld/-mgnu-ld documentation 17483 17484 IA-64 specific 17485 17486 * [389]23644 IA-64 hardware models and configuration options 17487 documentation error 17488 * [390]24718 Shared libgcc not used for linking by default 17489 17490 M68000 specific 17491 17492 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 17493 17494 MIPS specific 17495 17496 * [392]20621 ICE in change_address_1, in emit-rtl.c 17497 17498 PowerPC and PowerPC64 specific 17499 17500 * [393]18583 error on valid code: const 17501 __attribute__((altivec(vector__))) doesn't work in arrays 17502 * [394]20191 ICE in reload_cse_simplify_operands 17503 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 17504 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 17505 * [397]23404 gij trashes args of functions with more than 8 fp args 17506 * [398]23539 C & C++ compiler generating misaligned references 17507 regardless of compiler flags 17508 * [399]24102 floatdisf2_internal2 broken 17509 * [400]24465 -mminimal-toc miscompilation of __thread vars 17510 17511 Solaris specific 17512 17513 * [401]19933 Problem with define of HUGE_VAL in math_c99 17514 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 17515 symbols 17516 17517 SPARC specific 17518 17519 * [403]19300 PCH failures on sparc-linux 17520 * [404]20301 Assembler labels have a leading "-" 17521 * [405]20673 C PCH testsuite assembly comparison failure 17522 17523 x86 and x86_64 specific 17524 17525 * [406]18582 ICE with arrays of type V2DF 17526 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 17527 -fsched2-use-traces 17528 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 17529 * [409]24315 amd64 fails -fpeephole2 17530 __________________________________________________________________ 17531 17532 GCC 3.4.6 17533 17534 This is the [410]list of problem reports (PRs) from GCC's bug tracking 17535 system that are known to be fixed in the 3.4.6 release. This list might 17536 not be complete (that is, it is possible that some PRs that have been 17537 fixed are not listed here). 17538 17539 17540 For questions related to the use of GCC, please consult these web 17541 pages and the [411]GCC manuals. If that fails, the 17542 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17543 web pages and the development of GCC are welcome on our developer 17544 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 17545 archives. 17546 17547 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 17548 distribution of this entire article is permitted in any medium, 17549 provided this notice is preserved. 17550 17551 These pages are [416]maintained by the GCC team. Last modified 17552 2023-01-27. 17553 17554 References 17555 17556 1. https://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 17557 2. https://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 17558 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 17559 4. https://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 17560 5. https://gcc.gnu.org/gcc-3.4/mips-abi.html 17561 6. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 17562 7. https://www.boost.org/ 17563 8. https://gcc.gnu.org/PR11953 17564 9. https://gcc.gnu.org/PR8361 17565 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins 17566 11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 17567 12. https://gcc.gnu.org/bugs/#cxx_rvalbind 17568 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 17569 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 17570 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 17571 16. http://www.gnu.org/software/classpath/ 17572 17. https://www.eclipse.org/ 17573 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 17574 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 17575 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 17576 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 17577 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 17578 23. https://gcc.gnu.org/gcc-3.4/mips-abi.html 17579 24. https://gcc.gnu.org/gcc-3.4/powerpc-abi.html 17580 25. https://gcc.gnu.org/gcc-3.4/sparc-abi.html 17581 26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=\[3\.4.*[Rr]egression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 17582 27. https://gcc.gnu.org/PR10129 17583 28. https://gcc.gnu.org/PR14576 17584 29. https://gcc.gnu.org/PR14760 17585 30. https://gcc.gnu.org/PR14671 17586 31. https://gcc.gnu.org/PR15093 17587 32. https://gcc.gnu.org/PR15178 17588 33. https://gcc.gnu.org/PR12753 17589 34. https://gcc.gnu.org/PR13985 17590 35. https://gcc.gnu.org/PR14810 17591 36. https://gcc.gnu.org/PR14883 17592 37. https://gcc.gnu.org/PR15044 17593 38. https://gcc.gnu.org/PR15057 17594 39. https://gcc.gnu.org/PR15064 17595 40. https://gcc.gnu.org/PR15142 17596 41. https://gcc.gnu.org/PR15159 17597 42. https://gcc.gnu.org/PR15165 17598 43. https://gcc.gnu.org/PR15193 17599 44. https://gcc.gnu.org/PR15209 17600 45. https://gcc.gnu.org/PR15227 17601 46. https://gcc.gnu.org/PR15285 17602 47. https://gcc.gnu.org/PR15299 17603 48. https://gcc.gnu.org/PR15329 17604 49. https://gcc.gnu.org/PR15550 17605 50. https://gcc.gnu.org/PR15554 17606 51. https://gcc.gnu.org/PR15640 17607 52. https://gcc.gnu.org/PR15666 17608 53. https://gcc.gnu.org/PR15696 17609 54. https://gcc.gnu.org/PR15701 17610 55. https://gcc.gnu.org/PR15761 17611 56. https://gcc.gnu.org/PR15829 17612 57. https://gcc.gnu.org/PR14538 17613 58. https://gcc.gnu.org/PR12391 17614 59. https://gcc.gnu.org/PR14649 17615 60. https://gcc.gnu.org/PR15004 17616 61. https://gcc.gnu.org/PR15749 17617 62. https://gcc.gnu.org/PR10646 17618 63. https://gcc.gnu.org/PR12077 17619 64. https://gcc.gnu.org/PR13598 17620 65. https://gcc.gnu.org/PR14211 17621 66. https://gcc.gnu.org/PR14220 17622 67. https://gcc.gnu.org/PR14245 17623 68. https://gcc.gnu.org/PR14340 17624 69. https://gcc.gnu.org/PR14600 17625 70. https://gcc.gnu.org/PR14668 17626 71. https://gcc.gnu.org/PR14775 17627 72. https://gcc.gnu.org/PR14821 17628 73. https://gcc.gnu.org/PR14930 17629 74. https://gcc.gnu.org/PR14932 17630 75. https://gcc.gnu.org/PR14950 17631 76. https://gcc.gnu.org/PR14962 17632 77. https://gcc.gnu.org/PR14975 17633 78. https://gcc.gnu.org/PR15002 17634 79. https://gcc.gnu.org/PR15025 17635 80. https://gcc.gnu.org/PR15046 17636 81. https://gcc.gnu.org/PR15069 17637 82. https://gcc.gnu.org/PR15074 17638 83. https://gcc.gnu.org/PR15083 17639 84. https://gcc.gnu.org/PR15096 17640 85. https://gcc.gnu.org/PR15287 17641 86. https://gcc.gnu.org/PR15317 17642 87. https://gcc.gnu.org/PR15337 17643 88. https://gcc.gnu.org/PR15361 17644 89. https://gcc.gnu.org/PR15412 17645 90. https://gcc.gnu.org/PR15427 17646 91. https://gcc.gnu.org/PR15471 17647 92. https://gcc.gnu.org/PR15503 17648 93. https://gcc.gnu.org/PR15507 17649 94. https://gcc.gnu.org/PR15542 17650 95. https://gcc.gnu.org/PR15565 17651 96. https://gcc.gnu.org/PR15625 17652 97. https://gcc.gnu.org/PR15629 17653 98. https://gcc.gnu.org/PR15742 17654 99. https://gcc.gnu.org/PR15775 17655 100. https://gcc.gnu.org/PR15821 17656 101. https://gcc.gnu.org/PR15862 17657 102. https://gcc.gnu.org/PR15875 17658 103. https://gcc.gnu.org/PR15877 17659 104. https://gcc.gnu.org/PR15947 17660 105. https://gcc.gnu.org/PR16020 17661 106. https://gcc.gnu.org/PR16154 17662 107. https://gcc.gnu.org/PR16174 17663 108. https://gcc.gnu.org/PR14315 17664 109. https://gcc.gnu.org/PR15151 17665 110. https://gcc.gnu.org/PR7993 17666 111. https://gcc.gnu.org/PR15228 17667 112. https://gcc.gnu.org/PR15345 17668 113. https://gcc.gnu.org/PR15945 17669 114. https://gcc.gnu.org/PR15526 17670 115. https://gcc.gnu.org/PR14690 17671 116. https://gcc.gnu.org/PR15112 17672 117. https://gcc.gnu.org/PR15067 17673 118. https://gcc.gnu.org/PR1963 17674 119. https://gcc.gnu.org/PR15717 17675 120. https://gcc.gnu.org/PR14782 17676 121. https://gcc.gnu.org/PR14828 17677 122. https://gcc.gnu.org/PR15202 17678 123. https://gcc.gnu.org/PR14610 17679 124. https://gcc.gnu.org/PR14813 17680 125. https://gcc.gnu.org/PR14857 17681 126. https://gcc.gnu.org/PR15598 17682 127. https://gcc.gnu.org/PR15653 17683 128. https://gcc.gnu.org/PR15189 17684 129. https://gcc.gnu.org/PR15331 17685 130. https://gcc.gnu.org/PR16144 17686 131. https://gcc.gnu.org/PR16176 17687 132. https://gcc.gnu.org/PR11591 17688 133. https://gcc.gnu.org/PR12028 17689 134. https://gcc.gnu.org/PR14478 17690 135. https://gcc.gnu.org/PR14567 17691 136. https://gcc.gnu.org/PR14715 17692 137. https://gcc.gnu.org/PR14902 17693 138. https://gcc.gnu.org/PR14924 17694 139. https://gcc.gnu.org/PR14960 17695 140. https://gcc.gnu.org/PR15106 17696 141. https://gcc.gnu.org/PR16026 17697 142. https://gcc.gnu.org/PR15191 17698 143. https://gcc.gnu.org/PR15662 17699 144. https://gcc.gnu.org/PR15054 17700 145. https://gcc.gnu.org/PR15783 17701 146. https://gcc.gnu.org/PR15626 17702 147. https://gcc.gnu.org/PR14326 17703 148. https://gcc.gnu.org/PR14723 17704 149. https://gcc.gnu.org/PR15290 17705 150. https://gcc.gnu.org/PR15250 17706 151. https://gcc.gnu.org/PR15551 17707 152. https://gcc.gnu.org/PR8309 17708 153. https://gcc.gnu.org/PR13250 17709 154. https://gcc.gnu.org/PR13803 17710 155. https://gcc.gnu.org/PR14093 17711 156. https://gcc.gnu.org/PR14457 17712 157. https://gcc.gnu.org/PR14542 17713 158. https://gcc.gnu.org/PR15100 17714 159. https://gcc.gnu.org/PR15296 17715 160. https://gcc.gnu.org/PR15396 17716 161. https://gcc.gnu.org/PR15782 17717 162. https://gcc.gnu.org/PR11610 17718 163. https://gcc.gnu.org/PR15488 17719 164. https://gcc.gnu.org/PR15489 17720 165. https://gcc.gnu.org/PR13928 17721 166. https://gcc.gnu.org/PR14150 17722 167. https://gcc.gnu.org/PR14949 17723 168. https://gcc.gnu.org/PR15123 17724 169. https://gcc.gnu.org/PR16469 17725 170. https://gcc.gnu.org/PR16344 17726 171. https://gcc.gnu.org/PR16842 17727 172. https://gcc.gnu.org/PR12608 17728 173. https://gcc.gnu.org/PR14492 17729 174. https://gcc.gnu.org/PR15461 17730 175. https://gcc.gnu.org/PR15890 17731 176. https://gcc.gnu.org/PR16180 17732 177. https://gcc.gnu.org/PR16224 17733 178. https://gcc.gnu.org/PR16408 17734 179. https://gcc.gnu.org/PR16529 17735 180. https://gcc.gnu.org/PR16698 17736 181. https://gcc.gnu.org/PR16706 17737 182. https://gcc.gnu.org/PR16810 17738 183. https://gcc.gnu.org/PR16851 17739 184. https://gcc.gnu.org/PR16870 17740 185. https://gcc.gnu.org/PR16904 17741 186. https://gcc.gnu.org/PR16905 17742 187. https://gcc.gnu.org/PR16964 17743 188. https://gcc.gnu.org/PR17068 17744 189. https://gcc.gnu.org/PR16366 17745 190. https://gcc.gnu.org/PR15345 17746 191. https://gcc.gnu.org/PR16590 17747 192. https://gcc.gnu.org/PR16693 17748 193. https://gcc.gnu.org/PR17078 17749 194. https://gcc.gnu.org/PR13956 17750 195. https://gcc.gnu.org/PR16684 17751 196. https://gcc.gnu.org/PR12658 17752 197. https://gcc.gnu.org/PR13092 17753 198. https://gcc.gnu.org/PR15320 17754 199. https://gcc.gnu.org/PR16246 17755 200. https://gcc.gnu.org/PR16273 17756 201. https://gcc.gnu.org/PR16401 17757 202. https://gcc.gnu.org/PR16411 17758 203. https://gcc.gnu.org/PR16489 17759 204. https://gcc.gnu.org/PR16618 17760 205. https://gcc.gnu.org/PR16637 17761 206. https://gcc.gnu.org/PR16717 17762 207. https://gcc.gnu.org/PR16813 17763 208. https://gcc.gnu.org/PR16853 17764 209. https://gcc.gnu.org/PR16889 17765 210. https://gcc.gnu.org/PR16959 17766 211. https://gcc.gnu.org/PR7587 17767 212. https://gcc.gnu.org/PR16473 17768 213. https://gcc.gnu.org/PR16478 17769 214. https://gcc.gnu.org/PR10695 17770 215. https://gcc.gnu.org/PR16974 17771 216. https://gcc.gnu.org/PR16298 17772 217. https://gcc.gnu.org/PR17113 17773 218. https://gcc.gnu.org/PR14697 17774 219. https://gcc.gnu.org/PR15869 17775 220. https://gcc.gnu.org/PR16325 17776 221. https://gcc.gnu.org/PR16357 17777 222. https://gcc.gnu.org/PR16380 17778 223. https://gcc.gnu.org/PR16407 17779 224. https://gcc.gnu.org/PR16643 17780 225. https://gcc.gnu.org/PR15927 17781 226. https://gcc.gnu.org/PR15948 17782 227. https://gcc.gnu.org/PR17019 17783 228. https://gcc.gnu.org/PR16130 17784 229. https://gcc.gnu.org/PR16142 17785 230. https://gcc.gnu.org/PR16278 17786 231. https://gcc.gnu.org/PR16414 17787 232. https://gcc.gnu.org/PR16445 17788 233. https://gcc.gnu.org/PR16490 17789 234. https://gcc.gnu.org/PR16683 17790 235. https://gcc.gnu.org/PR16195 17791 236. https://gcc.gnu.org/PR16239 17792 237. https://gcc.gnu.org/PR16199 17793 238. https://gcc.gnu.org/PR16416 17794 239. https://gcc.gnu.org/PR16430 17795 240. https://gcc.gnu.org/PR16379 17796 241. https://gcc.gnu.org/PR17093 17797 242. https://gcc.gnu.org/PR17119 17798 243. https://gcc.gnu.org/PR15928 17799 244. https://gcc.gnu.org/PR16210 17800 245. https://gcc.gnu.org/PR15488 17801 246. https://gcc.gnu.org/PR16250 17802 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 17803 248. https://gcc.gnu.org/PR17369 17804 249. https://gcc.gnu.org/PR17850 17805 250. https://gcc.gnu.org/PR13948 17806 251. https://gcc.gnu.org/PR14492 17807 252. https://gcc.gnu.org/PR16301 17808 253. https://gcc.gnu.org/PR16566 17809 254. https://gcc.gnu.org/PR17023 17810 255. https://gcc.gnu.org/PR17027 17811 256. https://gcc.gnu.org/PR17524 17812 257. https://gcc.gnu.org/PR17826 17813 258. https://gcc.gnu.org/PR15526 17814 259. https://gcc.gnu.org/PR16999 17815 260. https://gcc.gnu.org/PR17503 17816 261. https://gcc.gnu.org/PR17581 17817 262. https://gcc.gnu.org/PR18129 17818 263. https://gcc.gnu.org/PR10975 17819 264. https://gcc.gnu.org/PR11722 17820 265. https://gcc.gnu.org/PR14534 17821 266. https://gcc.gnu.org/PR15172 17822 267. https://gcc.gnu.org/PR15786 17823 268. https://gcc.gnu.org/PR16162 17824 269. https://gcc.gnu.org/PR16612 17825 270. https://gcc.gnu.org/PR16715 17826 271. https://gcc.gnu.org/PR16848 17827 272. https://gcc.gnu.org/PR17132 17828 273. https://gcc.gnu.org/PR17259 17829 274. https://gcc.gnu.org/PR17327 17830 275. https://gcc.gnu.org/PR17393 17831 276. https://gcc.gnu.org/PR17501 17832 277. https://gcc.gnu.org/PR17537 17833 278. https://gcc.gnu.org/PR17585 17834 279. https://gcc.gnu.org/PR17821 17835 280. https://gcc.gnu.org/PR17829 17836 281. https://gcc.gnu.org/PR17851 17837 282. https://gcc.gnu.org/PR17976 17838 283. https://gcc.gnu.org/PR18020 17839 284. https://gcc.gnu.org/PR18093 17840 285. https://gcc.gnu.org/PR18140 17841 286. https://gcc.gnu.org/PR17541 17842 287. https://gcc.gnu.org/PR17853 17843 288. https://gcc.gnu.org/PR17245 17844 289. https://gcc.gnu.org/PR17167 17845 290. https://gcc.gnu.org/PR17277 17846 291. https://gcc.gnu.org/PR17505 17847 292. https://gcc.gnu.org/PR17684 17848 293. https://gcc.gnu.org/PR17384 17849 294. https://gcc.gnu.org/PR17770 17850 295. https://gcc.gnu.org/PR11476 17851 296. https://gcc.gnu.org/PR14064 17852 297. https://gcc.gnu.org/PR14678 17853 298. https://gcc.gnu.org/PR15583 17854 299. https://gcc.gnu.org/PR15790 17855 300. https://gcc.gnu.org/PR15886 17856 301. https://gcc.gnu.org/PR16884 17857 302. https://gcc.gnu.org/PR13841 17858 303. https://gcc.gnu.org/PR15860 17859 304. https://gcc.gnu.org/PR17465 17860 305. https://gcc.gnu.org/PR17469 17861 306. https://gcc.gnu.org/PR18138 17862 307. https://gcc.gnu.org/PR15498 17863 308. https://gcc.gnu.org/PR15747 17864 309. https://gcc.gnu.org/PR16406 17865 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 17866 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 17867 312. https://gcc.gnu.org/PR24688 17868 313. https://gcc.gnu.org/PR17188 17869 314. https://gcc.gnu.org/PR20187 17870 315. https://gcc.gnu.org/PR21873 17871 316. https://gcc.gnu.org/PR21899 17872 317. https://gcc.gnu.org/PR22061 17873 318. https://gcc.gnu.org/PR22208 17874 319. https://gcc.gnu.org/PR22458 17875 320. https://gcc.gnu.org/PR22589 17876 321. https://gcc.gnu.org/PR24101 17877 322. https://gcc.gnu.org/PR10611 17878 323. https://gcc.gnu.org/PR13377 17879 324. https://gcc.gnu.org/PR16002 17880 325. https://gcc.gnu.org/PR17413 17881 326. https://gcc.gnu.org/PR17609 17882 327. https://gcc.gnu.org/PR17618 17883 328. https://gcc.gnu.org/PR18124 17884 329. https://gcc.gnu.org/PR18155 17885 330. https://gcc.gnu.org/PR18177 17886 331. https://gcc.gnu.org/PR18368 17887 332. https://gcc.gnu.org/PR18378 17888 333. https://gcc.gnu.org/PR18466 17889 334. https://gcc.gnu.org/PR18512 17890 335. https://gcc.gnu.org/PR18545 17891 336. https://gcc.gnu.org/PR18738 17892 337. https://gcc.gnu.org/PR18803 17893 338. https://gcc.gnu.org/PR19004 17894 339. https://gcc.gnu.org/PR19208 17895 340. https://gcc.gnu.org/PR19253 17896 341. https://gcc.gnu.org/PR19608 17897 342. https://gcc.gnu.org/PR19884 17898 343. https://gcc.gnu.org/PR20153 17899 344. https://gcc.gnu.org/PR20563 17900 345. https://gcc.gnu.org/PR20789 17901 346. https://gcc.gnu.org/PR21336 17902 347. https://gcc.gnu.org/PR21768 17903 348. https://gcc.gnu.org/PR21853 17904 349. https://gcc.gnu.org/PR21903 17905 350. https://gcc.gnu.org/PR21983 17906 351. https://gcc.gnu.org/PR21987 17907 352. https://gcc.gnu.org/PR22153 17908 353. https://gcc.gnu.org/PR22172 17909 354. https://gcc.gnu.org/PR21286 17910 355. https://gcc.gnu.org/PR22233 17911 356. https://gcc.gnu.org/PR22508 17912 357. https://gcc.gnu.org/PR22545 17913 358. https://gcc.gnu.org/PR23528 17914 359. https://gcc.gnu.org/PR23550 17915 360. https://gcc.gnu.org/PR23586 17916 361. https://gcc.gnu.org/PR23624 17917 362. https://gcc.gnu.org/PR23639 17918 363. https://gcc.gnu.org/PR23797 17919 364. https://gcc.gnu.org/PR23965 17920 365. https://gcc.gnu.org/PR24052 17921 366. https://gcc.gnu.org/PR24580 17922 367. https://gcc.gnu.org/PR24267 17923 368. https://gcc.gnu.org/PR17810 17924 369. https://gcc.gnu.org/PR17860 17925 370. https://gcc.gnu.org/PR21709 17926 371. https://gcc.gnu.org/PR21964 17927 372. https://gcc.gnu.org/PR22167 17928 373. https://gcc.gnu.org/PR22619 17929 374. https://gcc.gnu.org/PR23241 17930 375. https://gcc.gnu.org/PR23478 17931 376. https://gcc.gnu.org/PR24470 17932 377. https://gcc.gnu.org/PR24950 17933 378. https://gcc.gnu.org/PR14400 17934 379. https://gcc.gnu.org/PR14940 17935 380. https://gcc.gnu.org/PR20239 17936 381. https://gcc.gnu.org/PR15220 17937 382. https://gcc.gnu.org/PR19275 17938 383. https://gcc.gnu.org/PR21888 17939 384. https://gcc.gnu.org/PR15342 17940 385. https://gcc.gnu.org/PR23985 17941 386. https://gcc.gnu.org/PR16719 17942 387. https://gcc.gnu.org/PR21723 17943 388. https://gcc.gnu.org/PR21841 17944 389. https://gcc.gnu.org/PR23644 17945 390. https://gcc.gnu.org/PR24718 17946 391. https://gcc.gnu.org/PR18421 17947 392. https://gcc.gnu.org/PR20621 17948 393. https://gcc.gnu.org/PR18583 17949 394. https://gcc.gnu.org/PR20191 17950 395. https://gcc.gnu.org/PR22083 17951 396. https://gcc.gnu.org/PR23070 17952 397. https://gcc.gnu.org/PR23404 17953 398. https://gcc.gnu.org/PR23539 17954 399. https://gcc.gnu.org/PR24102 17955 400. https://gcc.gnu.org/PR24465 17956 401. https://gcc.gnu.org/PR19933 17957 402. https://gcc.gnu.org/PR21889 17958 403. https://gcc.gnu.org/PR19300 17959 404. https://gcc.gnu.org/PR20301 17960 405. https://gcc.gnu.org/PR20673 17961 406. https://gcc.gnu.org/PR18582 17962 407. https://gcc.gnu.org/PR19340 17963 408. https://gcc.gnu.org/PR21716 17964 409. https://gcc.gnu.org/PR24315 17965 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 17966 411. https://gcc.gnu.org/onlinedocs/ 17967 412. mailto:gcc-help (a] gcc.gnu.org 17968 413. mailto:gcc (a] gcc.gnu.org 17969 414. https://gcc.gnu.org/lists.html 17970 415. https://www.fsf.org/ 17971 416. https://gcc.gnu.org/about.html 17972 ====================================================================== 17973 http://gcc.gnu.org/gcc-3.3/index.html 17974 GCC 3.3 Release Series 17975 17976 (This release series is no longer supported.) 17977 17978 May 03, 2005 17979 17980 The [1]GNU project and the GCC developers are pleased to announce the 17981 release of GCC 3.3.6. 17982 17983 This release is a bug-fix release, containing fixes for regressions in 17984 GCC 3.3.5 relative to previous releases of GCC. 17985 17986 This release is the last of the series 3.3.x. 17987 17988 The GCC 3.3 release series includes numerous [2]new features, 17989 improvements, bug fixes, and other changes, thanks to an [3]amazing 17990 group of volunteers. 17991 17992 Release History 17993 17994 GCC 3.3.6 17995 May 3, 2005 ([4]changes) 17996 17997 GCC 3.3.5 17998 September 30, 2004 ([5]changes) 17999 18000 GCC 3.3.4 18001 May 31, 2004 ([6]changes) 18002 18003 GCC 3.3.3 18004 February 14, 2004 ([7]changes) 18005 18006 GCC 3.3.2 18007 October 16, 2003 ([8]changes) 18008 18009 GCC 3.3.1 18010 August 8, 2003 ([9]changes) 18011 18012 GCC 3.3 18013 May 14, 2003 ([10]changes) 18014 18015 References and Acknowledgements 18016 18017 GCC used to stand for the GNU C Compiler, but since the compiler 18018 supports several other languages aside from C, it now stands for the 18019 GNU Compiler Collection. 18020 18021 A list of [11]successful builds is updated as new information becomes 18022 available. 18023 18024 The GCC developers would like to thank the numerous people that have 18025 contributed new features, improvements, bug fixes, and other changes as 18026 well as test results to GCC. This [12]amazing group of volunteers is 18027 what makes GCC successful. 18028 18029 For additional information about GCC please refer to the [13]GCC 18030 project web site or contact the [14]GCC development mailing list. 18031 18032 To obtain GCC please use [15]our mirror sites, or our CVS server. 18033 18034 18035 For questions related to the use of GCC, please consult these web 18036 pages and the [16]GCC manuals. If that fails, the 18037 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18038 web pages and the development of GCC are welcome on our developer 18039 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 18040 archives. 18041 18042 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 18043 distribution of this entire article is permitted in any medium, 18044 provided this notice is preserved. 18045 18046 These pages are [21]maintained by the GCC team. Last modified 18047 2022-10-26. 18048 18049 References 18050 18051 1. http://www.gnu.org/ 18052 2. https://gcc.gnu.org/gcc-3.3/changes.html 18053 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18054 4. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 18055 5. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 18056 6. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 18057 7. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 18058 8. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 18059 9. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 18060 10. https://gcc.gnu.org/gcc-3.3/changes.html 18061 11. https://gcc.gnu.org/gcc-3.3/buildstat.html 18062 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18063 13. https://gcc.gnu.org/index.html 18064 14. mailto:gcc (a] gcc.gnu.org 18065 15. https://gcc.gnu.org/mirrors.html 18066 16. https://gcc.gnu.org/onlinedocs/ 18067 17. mailto:gcc-help (a] gcc.gnu.org 18068 18. mailto:gcc (a] gcc.gnu.org 18069 19. https://gcc.gnu.org/lists.html 18070 20. https://www.fsf.org/ 18071 21. https://gcc.gnu.org/about.html 18072 ====================================================================== 18073 http://gcc.gnu.org/gcc-3.3/changes.html 18074 GCC 3.3 Release Series 18075 Changes, New Features, and Fixes 18076 18077 The latest release in the 3.3 release series is [1]GCC 3.3.6. 18078 18079 Caveats 18080 18081 * The preprocessor no longer accepts multi-line string literals. They 18082 were deprecated in 3.0, 3.1, and 3.2. 18083 * The preprocessor no longer supports the -A- switch when appearing 18084 alone. -A- followed by an assertion is still supported. 18085 * Support for all the systems [2]obsoleted in GCC 3.1 has been 18086 removed from GCC 3.3. See below for a [3]list of systems which are 18087 obsoleted in this release. 18088 * Checking for null format arguments has been decoupled from the rest 18089 of the format checking mechanism. Programs which use the format 18090 attribute may regain this functionality by using the new [4]nonnull 18091 function attribute. Note that all functions for which GCC has a 18092 built-in format attribute, an appropriate built-in nonnull 18093 attribute is also applied. 18094 * The DWARF (version 1) debugging format has been deprecated and will 18095 be removed in a future version of GCC. Version 2 of the DWARF 18096 debugging format will continue to be supported for the foreseeable 18097 future. 18098 * The C and Objective-C compilers no longer accept the "Naming Types" 18099 extension (typedef foo = bar); it was already unavailable in C++. 18100 Code which uses it will need to be changed to use the "typeof" 18101 extension instead: typedef typeof(bar) foo. (We have removed this 18102 extension without a period of deprecation because it has caused the 18103 compiler to crash since version 3.0 and no one noticed until very 18104 recently. Thus we conclude it is not in widespread use.) 18105 * The -traditional C compiler option has been removed. It was 18106 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 18107 available.) The <varargs.h> header, used for writing variadic 18108 functions in traditional C, still exists but will produce an error 18109 message if used. 18110 * GCC 3.3.1 automatically places zero-initialized variables in the 18111 .bss section on some operating systems. Versions of GNU Emacs up to 18112 (and including) 21.3 will not work correctly when using this 18113 optimization; you can use -fno-zero-initialized-in-bss to disable 18114 it. 18115 18116 General Optimizer Improvements 18117 18118 * A new scheme for accurately describing processor pipelines, the 18119 [5]DFA scheduler, has been added. 18120 * Pavel Nejedly, Charles University Prague, has contributed new file 18121 format used by the edge coverage profiler (-fprofile-arcs). 18122 The new format is robust and diagnoses common mistakes where 18123 profiles from different versions (or compilations) of the program 18124 are combined resulting in nonsensical profiles and slow code to 18125 produced with profile feedback. Additionally this format allows 18126 extra data to be gathered. Currently, overall statistics are 18127 produced helping optimizers to identify hot spots of a program 18128 globally replacing the old intra-procedural scheme and resulting in 18129 better code. Note that the gcov tool from older GCC versions will 18130 not be able to parse the profiles generated by GCC 3.3 and vice 18131 versa. 18132 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 18133 pass enabled using -ftracer. This pass simplifies the control flow 18134 of functions allowing other optimizations to do better job. 18135 He also contributed the function reordering pass 18136 (-freorder-functions) to optimize function placement using profile 18137 feedback. 18138 18139 New Languages and Language specific improvements 18140 18141 C/ObjC/C++ 18142 18143 * The preprocessor now accepts directives within macro arguments. It 18144 processes them just as if they had not been within macro arguments. 18145 * The separate ISO and traditional preprocessors have been completely 18146 removed. The front end handles either type of preprocessed output 18147 if necessary. 18148 * In C99 mode preprocessor arithmetic is done in the precision of the 18149 target's intmax_t, as required by that standard. 18150 * The preprocessor can now copy comments inside macros to the output 18151 file when the macro is expanded. This feature, enabled using the 18152 -CC option, is intended for use by applications which place 18153 metadata or directives inside comments, such as lint. 18154 * The method of constructing the list of directories to be searched 18155 for header files has been revised. If a directory named by a -I 18156 option is a standard system include directory, the option is 18157 ignored to ensure that the default search order for system 18158 directories and the special treatment of system header files are 18159 not defeated. 18160 * A few more [6]ISO C99 features now work correctly. 18161 * A new function attribute, nonnull, has been added which allows 18162 pointer arguments to functions to be specified as requiring a 18163 non-null value. The compiler currently uses this information to 18164 issue a warning when it detects a null value passed in such an 18165 argument slot. 18166 * A new type attribute, may_alias, has been added. Accesses to 18167 objects with types with this attribute are not subjected to 18168 type-based alias analysis, but are instead assumed to be able to 18169 alias any other type of objects, just like the char type. 18170 18171 C++ 18172 18173 * Type based alias analysis has been implemented for C++ aggregate 18174 types. 18175 18176 Objective-C 18177 18178 * Generate an error if Objective-C objects are passed by value in 18179 function and method calls. 18180 * When -Wselector is used, check the whole list of selectors at the 18181 end of compilation, and emit a warning if a @selector() is not 18182 known. 18183 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 18184 * No longer need to include objc/objc-class.h to compile self calls 18185 in class methods (NeXT runtime only). 18186 * New -Wundeclared-selector option. 18187 * Removed selector bloating which was causing object files to be 10% 18188 bigger on average (GNU runtime only). 18189 * Using at run time @protocol() objects has been fixed in certain 18190 situations (GNU runtime only). 18191 * Type checking has been fixed and improved in many situations 18192 involving protocols. 18193 18194 Java 18195 18196 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 18197 1.4) API. 18198 * The JDK 1.4 assert facility has been implemented. 18199 * The bytecode interpreter is now direct threaded and thus faster. 18200 18201 Fortran 18202 18203 * Fortran improvements are listed in [7]the Fortran documentation. 18204 18205 Ada 18206 18207 * Ada tasking now works with glibc 2.3.x threading libraries. 18208 18209 New Targets and Target Specific Improvements 18210 18211 * The following changes have been made to the HP-PA port: 18212 + The port now defaults to scheduling for the PA8000 series of 18213 processors. 18214 + Scheduling support for the PA7300 processor has been added. 18215 + The 32-bit port now supports weak symbols under HP-UX 11. 18216 + The handling of initializers and finalizers has been improved 18217 under HP-UX 11. The 64-bit port no longer uses collect2. 18218 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 18219 + ABI fixes to correct the passing of small structures by value. 18220 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 18221 use the DFA processor pipeline description. 18222 * The following NetBSD configurations for the SuperH processor family 18223 have been added: 18224 + SH3, big-endian, sh-*-netbsdelf* 18225 + SH3, little-endian, shle-*-netbsdelf* 18226 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 18227 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 18228 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 18229 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 18230 * The following changes have been made to the IA-32/x86-64 port: 18231 + SSE2 and 3dNOW! intrinsics are now supported. 18232 + Support for thread local storage has been added to the IA-32 18233 and x86-64 ports. 18234 + The x86-64 port has been significantly improved. 18235 * The following changes have been made to the MIPS port: 18236 + All configurations now accept the -mabi switch. Note that you 18237 will need appropriate multilibs for this option to work 18238 properly. 18239 + ELF configurations will always pass an ABI flag to the 18240 assembler, except when the MIPS EABI is selected. 18241 + -mabi=64 no longer selects MIPS IV code. 18242 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 18243 been removed from this release. 18244 + -march now changes the core ISA level. In previous releases, 18245 it would change the use of processor-specific extensions, but 18246 would leave the core ISA unchanged. For example, mips64-elf 18247 -march=r8000 will now generate MIPS IV code. 18248 + Under most configurations, -mipsN now acts as a synonym for 18249 -march. 18250 + There are some new preprocessor macros to describe the -march 18251 and -mtune settings. See the documentation of those options 18252 for details. 18253 + Support for the NEC VR-Series processors has been added. This 18254 includes the 54xx, 5500, and 41xx series. 18255 + Support for the Sandcraft sr71k processor has been added. 18256 * The following changes have been made to the S/390 port: 18257 + Support to build the Java runtime libraries has been added. 18258 Java is now enabled by default on s390-*-linux* and 18259 s390x-*-linux* targets. 18260 + Multilib support for the s390x-*-linux* target has been added; 18261 this allows to build 31-bit binaries using the -m31 option. 18262 + Support for thread local storage has been added. 18263 + Inline assembler code may now use the 'Q' constraint to 18264 specify memory operands without index register. 18265 + Various platform-specific performance improvements have been 18266 implemented; in particular, the compiler now uses the BRANCH 18267 ON COUNT family of instructions and makes more frequent use of 18268 the TEST UNDER MASK family of instructions. 18269 * The following changes have been made to the PowerPC port: 18270 + Support for IBM Power4 processor added. 18271 + Support for Motorola e500 SPE added. 18272 + Support for AIX 5.2 added. 18273 + Function and Data sections now supported on AIX. 18274 + Sibcall optimizations added. 18275 * The support for H8 Tiny is added to the H8/300 port with -mn. 18276 18277 Obsolete Systems 18278 18279 Support for a number of older systems has been declared obsolete in GCC 18280 3.3. Unless there is activity to revive them, the next release of GCC 18281 will have their sources permanently removed. 18282 18283 All configurations of the following processor architectures have been 18284 declared obsolete: 18285 * Matsushita MN10200, mn10200-*-* 18286 * Motorola 88000, m88k-*-* 18287 * IBM ROMP, romp-*-* 18288 18289 Also, some individual systems have been obsoleted: 18290 * Alpha 18291 + Interix, alpha*-*-interix* 18292 + Linux libc1, alpha*-*-linux*libc1* 18293 + Linux ECOFF, alpha*-*-linux*ecoff* 18294 * ARM 18295 + Generic a.out, arm*-*-aout* 18296 + Conix, arm*-*-conix* 18297 + "Old ABI," arm*-*-oabi 18298 + StrongARM/COFF, strongarm-*-coff* 18299 * HPPA (PA-RISC) 18300 + Generic OSF, hppa1.0-*-osf* 18301 + Generic BSD, hppa1.0-*-bsd* 18302 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 18303 + HiUX, hppa*-*-hiux* 18304 + Mach Lites, hppa*-*-lites* 18305 * Intel 386 family 18306 + Windows NT 3.x, i?86-*-win32 18307 * MC68000 family 18308 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 18309 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 18310 m68k-sun-mach* 18311 + AT&T systems, m68000-att-sysv* 18312 + Atari systems, m68k-atari-sysv* 18313 + Motorola systems, m68k-motorola-sysv* 18314 + NCR systems, m68k-ncr-sysv* 18315 + Plexus systems, m68k-plexus-sysv* 18316 + Commodore systems, m68k-cbm-sysv* 18317 + Citicorp TTI, m68k-tti-* 18318 + Unos, m68k-crds-unos* 18319 + Concurrent RTU, m68k-ccur-rtu* 18320 + Linux a.out, m68k-*-linux*aout* 18321 + Linux libc1, m68k-*-linux*libc1* 18322 + pSOS, m68k-*-psos* 18323 * MIPS 18324 + Generic ECOFF, mips*-*-ecoff* 18325 + SINIX, mips-sni-sysv4 18326 + Orion RTEMS, mips64orion-*-rtems* 18327 * National Semiconductor 32000 18328 + OpenBSD, ns32k-*-openbsd* 18329 * POWER (aka RS/6000) and PowerPC 18330 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 18331 + Bull BOSX, rs6000-bull-bosx 18332 + Generic Mach, rs6000-*-mach* 18333 + Generic SysV, powerpc*-*-sysv* 18334 + Linux libc1, powerpc*-*-linux*libc1* 18335 * Sun SPARC 18336 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 18337 sparclite-*-aout*, and sparc86x-*-aout* 18338 + NetBSD a.out, sparc-*-netbsd*aout* 18339 + Generic BSD, sparc-*-bsd* 18340 + ChorusOS, sparc-*-chorusos* 18341 + Linux a.out, sparc-*-linux*aout* 18342 + Linux libc1, sparc-*-linux*libc1* 18343 + LynxOS, sparc-*-lynxos* 18344 + Solaris on HAL hardware, sparc-hal-solaris2* 18345 + SunOS versions 3 and 4, sparc-*-sunos[34]* 18346 * NEC V850 18347 + RTEMS, v850-*-rtems* 18348 * VAX 18349 + VMS, vax-*-vms* 18350 18351 Documentation improvements 18352 18353 Other significant improvements 18354 18355 * Almost all front-end dependencies in the compiler have been 18356 separated out into a set of language hooks. This should make adding 18357 a new front end clearer and easier. 18358 * One effect of removing the separate preprocessor is a small 18359 increase in the robustness of the compiler in general, and the 18360 maintainability of target descriptions. Previously target-specific 18361 built-in macros and others, such as __FAST_MATH__, had to be 18362 handled with so-called specs that were hard to maintain. Often they 18363 would fail to behave properly when conflicting options were 18364 supplied on the command line, and define macros in the user's 18365 namespace even when strict ISO compliance was requested. 18366 Integrating the preprocessor has cleanly solved these issues. 18367 * The Makefile suite now supports redirection of make install by 18368 means of the variable DESTDIR. 18369 __________________________________________________________________ 18370 18371 GCC 3.3 18372 18373 Detailed release notes for the GCC 3.3 release follow. 18374 18375 Bug Fixes 18376 18377 bootstrap failures 18378 18379 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 18380 [9]10198,[10]10338) 18381 18382 Internal compiler errors (multi-platform) 18383 18384 * [11]3581 large string causes segmentation fault in cc1 18385 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 18386 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 18387 init, invalid_op) 18388 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 18389 * [15]6412 (c++) ICE in retrieve_specialization 18390 * [16]6620 (c++) partial template specialization causes an ICE 18391 (segmentation fault) 18392 * [17]6663 (c++) ICE with attribute aligned 18393 * [18]7068 ICE with incomplete types 18394 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 18395 * [20]7647 (c++) ICE when data member has the name of the enclosing 18396 class 18397 * [21]7675 ICE in fixup_var_refs_1 18398 * [22]7718 'complex' template instantiation causes ICE 18399 * [23]8116 (c++) ICE in member template function 18400 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 18401 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 18402 fault 18403 * [26]8564 (c++) ICE in find_function_data, in function.c 18404 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 18405 * [28]8766 (c++) ICE after failed initialization of static template 18406 variable 18407 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 18408 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 18409 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 18410 definition 18411 * [32]9216 (c++) ICE on missing template parameter 18412 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 18413 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 18414 loop 18415 * [35]9429 (c++) ICE in template instantiation with a pointered new 18416 operator 18417 * [36]9516 Internal error when using a big array 18418 * [37]9600 (c++) ICE with typedefs in template class 18419 * [38]9629 (c++) virtual inheritance segfault 18420 * [39]9672 (c++) ICE: Error reporting routines re-entered 18421 * [40]9749 (c++) ICE in write_expression on invalid function 18422 prototype 18423 * [41]9794 (fortran) ICE: floating point exception during constant 18424 folding 18425 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 18426 * [43]9916 (c++) ICE with noreturn function in ?: statement 18427 * [44]9936 ICE with local function and variable-length 2d array 18428 * [45]10262 (c++) cc1plus crashes with large generated code 18429 * [46]10278 (c++) ICE in parser for invalid code 18430 * [47]10446 (c++) ICE on definition of nonexistent member function of 18431 nested class in a class template 18432 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 18433 declaration 18434 * [49]10506 (c++) ICE in build_new at cp/init.c with 18435 -fkeep-inline-functions and multiple inheritance 18436 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 18437 precision of the declared type 18438 18439 Optimization bugs 18440 18441 * [51]2001 Inordinately long compile times in reload CSE regs 18442 * [52]2391 Exponential compilation time explosion in combine 18443 * [53]2960 Duplicate loop conditions even with -Os 18444 * [54]4046 redundant conditional branch 18445 * [55]6405 Loop-unrolling related performance regressions 18446 * [56]6798 very long compile time with large case-statement 18447 * [57]6871 const objects shouldn't be moved to .bss 18448 * [58]6909 problem w/ -Os on modified loop-2c.c test case 18449 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 18450 non-void function'' warning 18451 * [60]7642 optimization problem with signbit() 18452 * [61]8634 incorrect code for inlining of memcpy under -O2 18453 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 18454 regular function call 18455 18456 C front end 18457 18458 * [63]2161 long if-else cascade overflows parser stack 18459 * [64]4319 short accepted on typedef'd char 18460 * [65]8602 incorrect line numbers in warning messages when using 18461 inline functions 18462 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 18463 AST nodes and breaks debugging dumps 18464 * [67]9853 miscompilation of non-constant structure initializer 18465 18466 c++ compiler and library 18467 18468 * [68]45 legal template specialization code is rejected (DUP: 18469 [69]3784) 18470 * [70]764 lookup failure: friend operator and dereferencing a pointer 18471 and templates (DUP: [71]5116) 18472 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 18473 2863) 18474 * [73]3663 G++ doesn't check access control during template 18475 instantiation 18476 * [74]3797 gcc fails to emit explicit specialization of a template 18477 member 18478 * [75]3948 Two destructors are called when no copy destructor is 18479 defined (ABI change) 18480 * [76]4137 Conversion operator within template is not accepted 18481 * [77]4361 bogus ambiguity taking the address of a member template 18482 * [78]4802 g++ accepts illegal template code (access to private 18483 member; DUP: [79]5837) 18484 * [80]4803 inline function is used but never defined, and g++ does 18485 not object 18486 * [81]5094 Partial specialization cannot be friend? 18487 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 18488 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 18489 time 18490 * [84]7015 certain __asm__ constructs rejected 18491 * [85]7086 compile time regression (quadratic behavior in 18492 fixup_var_refs) 18493 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 18494 std::abort 18495 * [87]7247 copy constructor missing when inlining enabled (invalid 18496 optimization?) 18497 * [88]7441 string array initialization compilation time regression 18498 from seconds to minutes 18499 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 18500 * [90]7804 bad printing of floating point constant in warning message 18501 * [91]8099 Friend classes and template specializations 18502 * [92]8117 member function pointers and multiple inheritance 18503 * [93]8205 using declaration and multiple inheritance 18504 * [94]8645 unnecessary non-zero checks in stl_tree.h 18505 * [95]8724 explicit destructor call for incomplete class allowed 18506 * [96]8805 compile time regression with many member variables 18507 * [97]8691 -O3 and -fno-implicit-templates are incompatible 18508 * [98]8700 unhelpful error message for binding temp to reference 18509 * [99]8724 explicit destructor call for incomplete class allowed 18510 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 18511 * [101]9016 Failure to consistently constant fold "constant" C++ 18512 objects 18513 * [102]9053 g++ confused about ambiguity of overloaded function 18514 templates 18515 * [103]9152 undefined virtual thunks 18516 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 18517 * [105]9297 data corruption due to codegen bug (when copying.) 18518 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 18519 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 18520 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 18521 local classes 18522 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 18523 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 18524 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 18525 and <iostream.h> 18526 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 18527 [114][DR 231] 18528 * [115]9555 ostream inserters fail to set badbit on exception 18529 * [116]9561 ostream inserters rethrow exception of wrong type 18530 * [117]9563 ostream::sentry returns true after a failed preparation 18531 * [118]9582 one-definition rule violation in std::allocator 18532 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 18533 * [120]9683 bug in initialization chains for static const variables 18534 from template classes 18535 * [121]9791 -Woverloaded-virtual reports hiding of destructor 18536 * [122]9817 collate::compare doesn't handle nul characters 18537 * [123]9825 filebuf::sputbackc breaks sbumpc 18538 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 18539 with custom traits 18540 * [125]9924 Multiple using statements for builtin functions not 18541 allowed 18542 * [126]9946 destructor is not called for temporary object 18543 * [127]9964 filebuf::close() sometimes fails to close file 18544 * [128]9988 filebuf::overflow writes EOF to file 18545 * [129]10033 optimization breaks polymorphic references w/ typeid 18546 operator 18547 * [130]10097 filebuf::underflow drops characters 18548 * [131]10132 filebuf destructor can throw exceptions 18549 * [132]10180 gcc fails to warn about non-inlined function 18550 * [133]10199 method parametrized by template does not work everywhere 18551 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 18552 * [135]10427 Stack corruption with variable-length automatic arrays 18553 and virtual destructors 18554 * [136]10503 Compilation never stops in fixed_type_or_null 18555 18556 Objective-C 18557 18558 * [137]5956 selectors aren't matched properly when added to the 18559 selector table 18560 18561 Fortran compiler and library 18562 18563 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 18564 detect 18565 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 18566 info requested 18567 * [140]5634 doc: explain that configure --prefix=~/... does not work 18568 * [141]6367 multiple repeat counts confuse namelist read into array 18569 * [142]6491 Logical operations error on logicals when using 18570 -fugly-logint 18571 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 18572 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 18573 on irix6.5 18574 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 18575 assume a direct access file 18576 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 18577 -fno-automatic) 18578 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 18579 * [148]7388 Incorrect output with 0-based array of characters 18580 * [149]8587 Double complex zero ** double precision number -> NaN 18581 instead of zero 18582 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 18583 unknown register name line-length-none 18584 * [151]10197 Direct access files not unformatted by default 18585 18586 Java compiler and library 18587 18588 * [152]6005 gcj fails to build rhug on alpha 18589 * [153]6389 System.getProperty("") should always throw an 18590 IllegalArgumentException 18591 * [154]6576 java.util.ResourceBundle.getResource ignores locale 18592 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 18593 * [156]7060 getMethod() doesn't search super interface 18594 * [157]7073 bytecode interpreter gives wrong answer for interface 18595 getSuperclass() 18596 * [158]7180 possible bug in 18597 javax.naming.spi.NamingManager.getPlusPath() 18598 * [159]7416 java.security startup refs "GNU libgcj.security" 18599 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 18600 env (DUP: [161]7578) 18601 * [162]7611 Internal error while compiling libjava with -O 18602 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 18603 * [164]7766 ZipInputStream.available returns 0 immediately after 18604 construction 18605 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 18606 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 18607 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 18608 * [168]8234 ZipInputStream chokes when InputStream.read() returns 18609 small chunks 18610 * [169]8415 reflection bug: exception info for Method 18611 * [170]8481 java.Random.nextInt(int) may return negative 18612 * [171]8593 Error reading GZIPped files with BufferedReader 18613 * [172]8759 java.beans.Introspector has no flushCaches() or 18614 flushFromCaches() methods 18615 * [173]8997 spin() calls Thread.sleep 18616 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 18617 instead of the root content of C: 18618 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 18619 wrong return codes 18620 * [176]9271 Severe bias in java.security.SecureRandom 18621 18622 Ada compiler and library 18623 18624 * [177]6767 make gnatlib-shared fails on -laddr2line 18625 * [178]9911 gnatmake fails to link when GCC configured with 18626 --with-sjlj-exceptions=yes 18627 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 18628 * [180]10546 Ada tasking not working on Red Hat 9 18629 18630 preprocessor 18631 18632 * [181]7029 preprocessor should ignore #warning with -M 18633 18634 ARM-specific 18635 18636 * [182]2903 [arm] Optimization bug with long long arithmetic 18637 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 18638 18639 FreeBSD-specific 18640 18641 * [184]7680 float functions undefined in math.h/cmath with #define 18642 _XOPEN_SOURCE 18643 18644 HP-UX or HP-PA-specific 18645 18646 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 18647 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 18648 fputc_unlocked 18649 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 18650 18651 m68hc11-specific 18652 18653 * [188]6744 Bad assembler code generated: reference to pseudo 18654 register z 18655 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 18656 in reload1.c 18657 18658 MIPS-specific 18659 18660 * [190]9496 [mips-linux] bug in optimizer? 18661 18662 PowerPC-specific 18663 18664 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 18665 space 18666 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 18667 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 18668 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 18669 18670 SPARC-specific 18671 18672 * [195]10267 (documentation) Wrong build instructions for 18673 *-*-solaris2* 18674 18675 x86-specific (Intel/AMD) 18676 18677 * [196]7916 ICE in instantiate_virtual_register_1 18678 * [197]7926 (c++) i486 instructions in header files make c++ programs 18679 crash on i386 18680 * [198]8555 ICE in gen_split_1231 18681 * [199]8994 ICE with -O -march=pentium4 18682 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 18683 * [201]9806 ICE in inline assembly with -fPIC flag 18684 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 18685 regs 18686 * [203]10233 64-bit comparison only comparing bottom 32-bits 18687 * [204]10286 type-punning doesn't work with __m64 and -O 18688 * [205]10308 [x86] ICE with -O -fgcse or -O2 18689 __________________________________________________________________ 18690 18691 GCC 3.3.1 18692 18693 Bug Fixes 18694 18695 This section lists the problem reports (PRs) from GCC's bug tracking 18696 system that are known to be fixed in the 3.3.1 release. This list might 18697 not be complete (that is, it is possible that some PRs that have been 18698 fixed are not listed here). 18699 18700 Bootstrap failures 18701 18702 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 18703 18704 Internal compiler errors (multi-platform) 18705 18706 * [207]5754 ICE on invalid nested template class 18707 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 18708 and --enable-checking 18709 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 18710 * [210]7053 (c++) ICE when declaring a function already defined as a 18711 friend method of a template class 18712 * [211]8164 (c++) ICE when using different const expressions as 18713 template parameter 18714 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 18715 * [213]9559 (c++) ICE with invalid initialization of a static const 18716 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 18717 when redeclaring a static member variable 18718 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 18719 dwarfout.c with -g -O -finline-functions 18720 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 18721 * [217]10475 ICE in subreg_highpart_offset for code with long long 18722 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 18723 from a void pointer 18724 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 18725 instantiating static member variables 18726 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 18727 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 18728 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 18729 MAX_INT_64BIT 18730 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 18731 sched.c 18732 * [224]10939 (c++) ICE with template code 18733 * [225]10956 (c++) ICE when specializing a template member function 18734 of a template class, in tsubst, in cp/pt.c 18735 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 18736 defined) 18737 * [227]11059 (c++) ICE with empty union 18738 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 18739 -O2 -fnon-call-exceptions 18740 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 18741 * [230]11149 (c++) ICE on error when instantiation with call function 18742 of a base type 18743 * [231]11228 (c++) ICE on new-expression using array operator new and 18744 default-initialization 18745 * [232]11282 (c++) Infinite memory usage after syntax error 18746 * [233]11301 (fortran) ICE with -fno-globals 18747 * [234]11308 (c++) ICE when using an enum type name as if it were a 18748 class or namespace 18749 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 18750 an empty struct 18751 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 18752 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 18753 template member functions 18754 18755 Optimization bugs 18756 18757 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 18758 problem) 18759 * [239]11304 Wrong code production with -fomit-frame-pointer 18760 * [240]11381 volatile memory access optimized away 18761 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 18762 * [242]11557 constant folding bug generates wrong code 18763 18764 C front end 18765 18766 * [243]5897 No warning for statement after return 18767 * [244]11279 DWARF-2 output mishandles large enums 18768 18769 Preprocessor bugs 18770 18771 * [245]11022 no warning for non-compatible macro redefinition 18772 18773 C++ compiler and library 18774 18775 * [246]2330 static_cast<>() to a private base is allowed 18776 * [247]5388 Incorrect message "operands to ?: have different types" 18777 * [248]5390 Libiberty fails to demangle multi-digit template 18778 parameters 18779 * [249]7877 Incorrect parameter passing to specializations of member 18780 function templates 18781 * [250]9393 Anonymous namespaces and compiling the same file twice 18782 * [251]10032 -pedantic converts some errors to warnings 18783 * [252]10468 const typeof(x) is non-const, but only in templates 18784 * [253]10527 confused error message with "new int()" parameter 18785 initializer 18786 * [254]10679 parameter MIN_INLINE_INSNS is not honored 18787 * [255]10682 gcc chokes on a typedef for an enum inside a class 18788 template 18789 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 18790 0. 18791 * [257]10845 template member function (with nested template as 18792 parameter) cannot be called anymore if another unrelated template 18793 member function is defined 18794 * [258]10849 Cannot define an out-of-class specialization of a 18795 private nested template class 18796 * [259]10888 Suppress -Winline warnings for system headers 18797 * [260]10929 -Winline warns about functions for which no definition 18798 is visible 18799 * [261]10931 valid conversion static_cast<const unsigned 18800 int&>(lvalue-of-type-int) is rejected 18801 * [262]10940 Bad code with explicit specialization 18802 * [263]10968 If member function implicitly instantiated, explicit 18803 instantiation of class fails to instantiate it 18804 * [264]10990 Cannot convert with dynamic_cast<> to a private base 18805 class from within a member function 18806 * [265]11039 Bad interaction between implicit typename deprecation 18807 and friendship 18808 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 18809 "__unused__" instead 18810 * [267]11095 C++ iostream manipulator causes segfault when called 18811 with negative argument 18812 * [268]11098 g++ doesn't emit complete debugging information for 18813 local variables in destructors 18814 * [269]11137 GNU/Linux shared library constructors not called unless 18815 there's one global object 18816 * [270]11154 spurious ambiguity report for template class 18817 specialization 18818 * [271]11329 Compiler cannot find user defined implicit typecast 18819 * [272]11332 Spurious error with casts in ?: expression 18820 * [273]11431 static_cast behavior with subclasses when default 18821 constructor available 18822 * [274]11528 money_get facet does not accept "$.00" as valid 18823 * [275]11546 Type lookup problems in out-of-line definition of a 18824 class doubly nested from a template class 18825 * [276]11567 C++ code containing templated member function with same 18826 name as pure virtual member function results in linking failure 18827 * [277]11645 Failure to deal with using and private inheritance 18828 18829 Java compiler and library 18830 18831 * [278]5179 Qualified static field access doesn't initialize its 18832 class 18833 * [279]8204 gcj -O2 to native reorders certain instructions 18834 improperly 18835 * [280]10838 java.io.ObjectInputStream syntax error 18836 * [281]10886 The RMI registry that comes with GCJ does not work 18837 correctly 18838 * [282]11349 JNDI URL context factories not located correctly 18839 18840 x86-specific (Intel/AMD) 18841 18842 * [283]4823 ICE on inline assembly code 18843 * [284]8878 miscompilation with -O and SSE 18844 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 18845 -masm=intel 18846 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 18847 in reload1.c 18848 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 18849 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 18850 * [289]11044 [x86] out of range loop instructions for FP code on K6 18851 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 18852 built-ins 18853 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 18854 is used 18855 18856 SPARC- or Solaris- specific 18857 18858 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 18859 * [293]10142 [SPARC64] gcc produces wrong code when passing 18860 structures by value 18861 * [294]10663 New configure check aborts with Sun tools. 18862 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 18863 * [296]10876 ICE in calculate_giv_inc when building KDE 18864 * [297]10955 wrong code at -O3 for structure argument in context of 18865 structure return 18866 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 18867 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 18868 Linux kernel 18869 18870 ia64 specific 18871 18872 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 18873 * [301]11320 scheduler bug (in machine depended reorganization pass) 18874 * [302]11599 bug with conditional and __builtin_prefetch 18875 18876 PowerPC specific 18877 18878 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 18879 during loop) 18880 * [304]10871 error in rs6000_stack_info save_size computation 18881 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 18882 cures it 18883 18884 m68k-specific 18885 18886 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 18887 * [307]10557 [m68k] ICE in subreg_offset_representable_p 18888 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 18889 18890 ARM-specific 18891 18892 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 18893 functions with __attribute__ ((interrupt ("IRQ"))) 18894 * [310]10842 [arm] Clobbered link register is copied to pc under 18895 certain circumstances 18896 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 18897 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 18898 (3.4) 18899 18900 MIPS-specific 18901 18902 * [313]11084 ICE in propagate_one_insn, in flow.c 18903 18904 SH-specific 18905 18906 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 18907 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 18908 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 18909 C++ files 18910 18911 GNU/Linux (or Hurd?) specific 18912 18913 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 18914 18915 UnixWare specific 18916 18917 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 18918 7.1.1 18919 18920 Cygwin (or mingw) specific 18921 18922 * [319]5287 ICE with dllimport attribute 18923 * [320]10148 [MingW/CygWin] Compiler dumps core 18924 18925 DJGPP specific 18926 18927 * [321]8787 GCC fails to emit .intel_syntax when invoked with 18928 -masm=intel on DJGPP 18929 18930 Darwin (and MacOS X) specific 18931 18932 * [322]10900 trampolines crash 18933 18934 Documentation 18935 18936 * [323]1607 (c++) Format attributes on methods undocumented 18937 * [324]4252 Invalid option `-fdump-translation-unit' 18938 * [325]4490 Clarify restrictions on -m96bit-long-double, 18939 -m128bit-long-double 18940 * [326]10355 document an issue with regparm attribute on some systems 18941 (e.g. Solaris) 18942 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 18943 (Unix)" is wrong 18944 * [328]10805 document bug in old version of Sun assembler 18945 * [329]10815 warn against GNU binutils on AIX 18946 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 18947 * [331]11280 Manual incorrect with respect to -freorder-blocks 18948 * [332]11466 Document -mlittle-endian and its restrictions for the 18949 sparc64 port 18950 18951 Testsuite bugs (compiler itself is not affected) 18952 18953 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 18954 report failure 18955 * [334]10810 gcc-3.3 fails make check: buffer overrun in 18956 test_demangle.c 18957 __________________________________________________________________ 18958 18959 GCC 3.3.2 18960 18961 Bug Fixes 18962 18963 This section lists the problem reports (PRs) from GCC's bug tracker 18964 that are known to be fixed in the 3.3.2 release. This list might not be 18965 complete (that is, it is possible that some PRs that have been fixed 18966 are not listed here). 18967 18968 Bootstrap failures and problems 18969 18970 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 18971 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 18972 --enable-threads=posix 18973 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 18974 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 18975 7.1.1) 18976 * [339]11687 xstormy16-elf build fails in libf2c 18977 * [340]12263 [SGI IRIX] bootstrap fails during compile of 18978 libf2c/libI77/backspace.c 18979 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 18980 fix-header processing) 18981 18982 Internal compiler errors (multi-platform) 18983 18984 * [342]7277 Casting integers to vector types causes ICE 18985 * [343]7939 (c++) ICE on invalid function template specialization 18986 * [344]11063 (c++) ICE on parsing initialization list of const array 18987 member 18988 * [345]11207 ICE with negative index in array element designator 18989 * [346]11522 (fortran) g77 dwarf-2 ICE in 18990 add_abstract_origin_attribute 18991 * [347]11595 (c++) ICE on duplicate label definition 18992 * [348]11646 (c++) ICE in commit_one_edge_insertion with 18993 -fnon-call-exceptions -fgcse -O 18994 * [349]11665 ICE in struct initializer when taking address 18995 * [350]11852 (c++) ICE with bad struct initializer. 18996 * [351]11878 (c++) ICE in cp_expr_size 18997 * [352]11883 ICE with any -O on mercury-generated C code 18998 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 18999 cp/typeck2.c when applying typeid operator to template template 19000 parameter 19001 * [354]12146 ICE in lookup_template_function, in cp/pt.c 19002 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 19003 -fno-gcse -O2 19004 * [356]12369 (c++) ICE with templates and friends 19005 * [357]12446 ICE in emit_move_insn on complicated array reference 19006 * [358]12510 ICE in final_scan_insn 19007 * [359]12544 ICE with large parameters used in nested functions 19008 19009 C and optimization bugs 19010 19011 * [360]9862 spurious warnings with -W -finline-functions 19012 * [361]10962 lookup_field is a linear search on a linked list (can be 19013 slow if large struct) 19014 * [362]11370 -Wunreachable-code gives false complaints 19015 * [363]11637 invalid assembly with -fnon-call-exceptions 19016 * [364]11885 Problem with bitfields in packed structs 19017 * [365]12082 Inappropriate unreachable code warnings 19018 * [366]12180 Inline optimization fails for variadic function 19019 * [367]12340 loop unroller + gcse produces wrong code 19020 19021 C++ compiler and library 19022 19023 * [368]3907 nested template parameter collides with member name 19024 * [369]5293 confusing message when binding a temporary to a reference 19025 * [370]5296 [DR115] Pointers to functions and to template functions 19026 behave differently in deduction 19027 * [371]7939 ICE on function template specialization 19028 * [372]8656 Unable to assign function with __attribute__ and pointer 19029 return type to an appropriate variable 19030 * [373]10147 Confusing error message for invalid template function 19031 argument 19032 * [374]11400 std::search_n() makes assumptions about Size parameter 19033 * [375]11409 issues with using declarations, overloading, and 19034 built-in functions 19035 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 19036 multiple bits in mask 19037 * [377]11786 operator() call on variable in other namespace not 19038 recognized 19039 * [378]11867 static_cast ignores ambiguity 19040 * [379]11928 bug with conversion operators that are typedefs 19041 * [380]12114 Uninitialized memory accessed in dtor 19042 * [381]12163 static_cast + explicit constructor regression 19043 * [382]12181 Wrong code with comma operator and c++ 19044 * [383]12236 regparm and fastcall messes up parameters 19045 * [384]12266 incorrect instantiation of unneeded template during 19046 overload resolution 19047 * [385]12296 istream::peek() doesn't set eofbit 19048 * [386]12298 [sjlj exceptions] Stack unwind destroys 19049 not-yet-constructed object 19050 * [387]12369 ICE with templates and friends 19051 * [388]12337 apparently infinite loop in g++ 19052 * [389]12344 stdcall attribute ignored if function returns a pointer 19053 * [390]12451 missing(late) class forward declaration in cxxabi.h 19054 * [391]12486 g++ accepts invalid use of a qualified name 19055 19056 x86 specific (Intel/AMD) 19057 19058 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 19059 builtins 19060 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 19061 -O2 19062 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 19063 * [395]12116 [k6] Invalid assembly output values with X-MAME code 19064 * [396]12070 ICE converting between double and long double with 19065 -msoft-float 19066 19067 ia64-specific 19068 19069 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 19070 * [398]11535 __builtin_return_address may not work on ia64 19071 * [399]11693 [ia64] ICE in gen_nop_type 19072 * [400]12224 [ia64] Thread-local storage doesn't work 19073 19074 PowerPC-specific 19075 19076 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 19077 kernel 19078 * [402]11319 loop miscompiled on ppc32 19079 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 19080 19081 SPARC-specific 19082 19083 * [404]11662 wrong code for expr. with cast to long long and 19084 exclusive or 19085 * [405]11965 invalid assembler code for a shift < 32 operation 19086 * [406]12301 (c++) stack corruption when a returned expression throws 19087 an exception 19088 19089 Alpha-specific 19090 19091 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 19092 kernel 2.4.22-pre8 19093 19094 HPUX-specific 19095 19096 * [408]11313 problem with #pragma weak and static inline functions 19097 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 19098 19099 Solaris specific 19100 19101 * [410]12166 Profiled programs crash if PROFDIR is set 19102 19103 Solaris-x86 specific 19104 19105 * [411]12101 i386 Solaris no longer works with GNU as? 19106 19107 Miscellaneous embedded target-specific bugs 19108 19109 * [412]10988 [m32r-elf] wrong blockmove code with -O3 19110 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 19111 -O2 19112 * [414]11902 [sh4] spec file improperly inserts rpath even when none 19113 needed 19114 * [415]11903 [sh4] -pthread fails to link due to error in spec file 19115 on sh4 19116 __________________________________________________________________ 19117 19118 GCC 3.3.3 19119 19120 Minor features 19121 19122 In addition to the bug fixes documented below, this release contains 19123 few minor features such as: 19124 * Support for --with-sysroot 19125 * Support for automatic detection of executable stacks 19126 * Support for SSE3 instructions 19127 * Support for thread local storage debugging under GDB on S390 19128 19129 Bug Fixes 19130 19131 This section lists the problem reports (PRs) from GCC's bug tracker 19132 that are known to be fixed in the 3.3.3 release. This list might not be 19133 complete (that is, it is possible that some PRs that have been fixed 19134 are not listed here). 19135 19136 Bootstrap failures and issues 19137 19138 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 19139 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 19140 unable to infer tagged configuration 19141 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 19142 subdirectories properly 19143 19144 Internal compiler errors (multi-platform) 19145 19146 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 19147 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 19148 * [420]10555 (c++) ICE on undefined template argument 19149 * [421]10706 (c++) ICE in mangle_class_name_for_template 19150 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 19151 active 19152 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 19153 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 19154 * [425]12632 (fortran) -fbounds-check ICE 19155 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 19156 3.3.2 19157 * [427]12726 (c++) ICE (segfault) on trivial code 19158 * [428]12890 (c++) ICE on compilation of class with throwing method 19159 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 19160 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 19161 correct code with -O2 -fno-force-mem 19162 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 19163 template 19164 * [432]13318 ICE: floating point exception in the loop optimizer 19165 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 19166 except.c 19167 * [434]13574 (c++) invalid array default initializer in class lets 19168 gcc consume all memory and die 19169 * [435]13475 ICE on SIMD variables with partial value initialization 19170 * [436]13797 (c++) ICE on invalid template parameter 19171 * [437]13824 (java) gcj SEGV with simple .java program 19172 19173 C and optimization bugs 19174 19175 * [438]8776 loop invariants are not removed (most likely) 19176 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 19177 strncmp by memcmp 19178 * [440]11350 undefined labels with -Os -fPIC 19179 * [441]12826 Optimizer removes reference through volatile pointer 19180 * [442]12500 stabs debug info: void no longer a predefined / builtin 19181 type 19182 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 19183 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 19184 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 19185 * [446]13507 spurious printf format warning 19186 * [447]13382 Type information for const pointer disappears during 19187 optimization. 19188 * [448]13394 noreturn attribute ignored on recursive invokation 19189 * [449]13400 Compiled code crashes storing to read-only location 19190 * [450]13521 Endless loop in calculate_global_regs_live 19191 19192 C++ compiler and library 19193 19194 Some of the bug fixes in this list were made to implement decisions 19195 that the ISO C++ standards committee has made concerning several defect 19196 reports (DRs). Links in the list below point to detailed discussion of 19197 the relevant defect report. 19198 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 19199 unification 19200 * [452]2294 using declaration confusion 19201 * [453]5050 template instantiation depth exceeds limit: recursion 19202 problem? 19203 * [454]9371 Bad exception handling in 19204 i/ostream::operator>>/<<(streambuf*) 19205 * [455]9546 bad exception handling in ostream members 19206 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 19207 face of unknown locales 19208 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 19209 * [459]10095 istream::operator>>(int&) sets ios::badbit when 19210 ios::failbit is set. 19211 * [460]11554 Warning about reordering of initializers doesn't mention 19212 location of constructor 19213 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 19214 * [462]12352 Exception safety problems in src/localename.cc 19215 * [463]12438 Memory leak in locale::combine() 19216 * [464]12540 Memory leak in locale::locale(const char*) 19217 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 19218 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 19219 * [470]12696 memory eating infinite loop in diagnostics (error 19220 recovery problem) 19221 * [471]12815 Code compiled with optimization behaves unexpectedly 19222 * [472]12862 Conflicts between typedefs/enums and namespace member 19223 declarations 19224 * [473]12926 Wrong value after assignment in initialize list using 19225 bit-fields 19226 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 19227 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 19228 * [478]13007 basic_streambuf::pubimbue, imbue wrong 19229 * [479]13009 Implicitly-defined assignment operator writes to wrong 19230 memory 19231 * [480]13057 regparm attribute not applied to destructor 19232 * [481]13070 -Wformat option ignored in g++ 19233 * [482]13081 forward template declarations in <complex> let inlining 19234 fail 19235 * [483]13239 Assertion does not seem to work correctly anymore 19236 * [484]13262 "xxx is private within this context" when initializing a 19237 self-contained template class 19238 * [485]13290 simple typo in concept checking for std::generate_n 19239 * [486]13323 Template code does not compile in presence of typedef 19240 * [487]13369 __verify_grouping (and __add_grouping?) not correct 19241 * [488]13371 infinite loop with packed struct and inlining 19242 * [489]13445 Template argument replacement "dereferences" a typedef 19243 * [490]13461 Fails to access protected-ctor from public constant 19244 * [491]13462 Non-standard-conforming type set::pointer 19245 * [492]13478 gcc uses wrong constructor to initialize a const 19246 reference 19247 * [493]13544 "conflicting types" for enums in different scopes 19248 * [494]13650 string::compare should not (always) use 19249 traits_type::length() 19250 * [495]13683 bogus warning about passing non-PODs through ellipsis 19251 * [496]13688 Derived class is denied access to protected base class 19252 member class 19253 * [497]13774 Member variable cleared in virtual multiple inheritance 19254 class 19255 * [498]13884 Protect sstream.tcc from extern template use 19256 19257 Java compiler and library 19258 19259 * [499]10746 [win32] garbage collection crash in GCJ 19260 19261 Objective-C compiler and library 19262 19263 * [500]11433 Crash due to dereferencing null pointer when querying 19264 protocol 19265 19266 Fortran compiler and library 19267 19268 * [501]12633 logical expression gives incorrect result with 19269 -fugly-logint option 19270 * [502]13037 [gcse-lm] g77 generates incorrect code 19271 * [503]13213 Hex constant problem when compiling with -fugly-logint 19272 and -ftypeless-boz 19273 19274 x86-specific (Intel/AMD) 19275 19276 * [504]4490 ICE with -m128bit-long-double 19277 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 19278 `reg' in make_field_assignment, in combine.c 19279 * [506]12441 ICE: can't find a register to spill 19280 * [507]12943 array static-init failure under -fpic, -fPIC 19281 * [508]13608 Incorrect code with -O3 -ffast-math 19282 19283 PowerPC-specific 19284 19285 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 19286 __attribute__((aligned(16))) 19287 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 19288 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 19289 altivec.md) 19290 * [512]12537 g++ generates writeable text sections 19291 19292 SPARC-specific 19293 19294 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 19295 -m64 19296 * [514]12865 mprotect call to make trampoline executable may fail 19297 * [515]13354 ICE in sparc_emit_set_const32 19298 19299 ARM-specific 19300 19301 * [516]10467 [arm] ICE in pre_insert_copy_insn, 19302 19303 ia64-specific 19304 19305 * [517]11226 ICE passing struct arg with two floats 19306 * [518]11227 ICE for _Complex float, _Complex long double args 19307 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 19308 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 19309 * Various fixes for libunwind 19310 19311 Alpha-specific 19312 19313 * [521]12654 Incorrect comparison code generated for Alpha 19314 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 19315 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 19316 19317 HPPA-specific 19318 19319 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 19320 * [525]12158 [hppa] compilation does not terminate at -O1 19321 19322 S390-specific 19323 19324 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 19325 (1<<24)-1 possible for CLCL-Instruction 19326 19327 SH-specific 19328 19329 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 19330 * [528]10392 optimizer generates faulty array indexing 19331 * [529]11322 SH profiler outputs multiple definitions of symbol 19332 * [530]13069 gcc/config/sh/rtems.h broken 19333 * [531]13302 Putting a va_list in a struct causes seg fault 19334 * [532]13585 Incorrect optimization of call to sfunc 19335 * Fix inappropriately exported libgcc functions from the shared 19336 library 19337 19338 Other embedded target specific 19339 19340 * [533]8916 [mcore] unsigned char assign gets hosed. 19341 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 19342 * [535]13122 [h8300] local variable gets corrupted by function call 19343 when -fomit-frame-pointer is given 19344 * [536]13256 [cris] strict_low_part mistreated in delay slots 19345 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 19346 -fexpensive-optimizations produces wrong code on mcore 19347 19348 GNU HURD-specific 19349 19350 * [538]12561 gcc/config/t-gnu needs updating to work with 19351 --with-sysroot 19352 19353 Tru64 Unix specific 19354 19355 * [539]6243 testsuite fails almost all tests due to no libintl in 19356 LD_LIBRARY_PATH during test. 19357 * [540]11397 weak aliases broken on Tru64 UNIX 19358 19359 AIX-specific 19360 19361 * [541]12505 build failure due to defines of uchar in cpphash.h and 19362 sys/types.h 19363 * [542]13150 WEAK symbols not exported by collect2 19364 19365 IRIX-specific 19366 19367 * [543]12666 fixincludes problem on IRIX 6.5.19m 19368 19369 Solaris-specific 19370 19371 * [544]12969 Including sys/byteorder.h breaks configure checks 19372 19373 Testsuite problems (compiler is not affected) 19374 19375 * [545]10819 testsuite creates CR+LF on compiler version lines in 19376 test summary files 19377 * [546]11612 abi_check not finding correct libgcc_s.so.1 19378 19379 Miscellaneous 19380 19381 * [547]13211 using -###, incorrect warnings about unused linker file 19382 are produced 19383 __________________________________________________________________ 19384 19385 GCC 3.3.4 19386 19387 This is the [548]list of problem reports (PRs) from GCC's bug tracking 19388 system that are known to be fixed in the 3.3.4 release. This list might 19389 not be complete (that is, it is possible that some PRs that have been 19390 fixed are not listed here). 19391 __________________________________________________________________ 19392 19393 GCC 3.3.5 19394 19395 This is the [549]list of problem reports (PRs) from GCC's bug tracking 19396 system that are known to be fixed in the 3.3.5 release. This list might 19397 not be complete (that is, it is possible that some PRs that have been 19398 fixed are not listed here). 19399 __________________________________________________________________ 19400 19401 GCC 3.3.6 19402 19403 This is the [550]list of problem reports (PRs) from GCC's bug tracking 19404 system that are known to be fixed in the 3.3.6 release. This list might 19405 not be complete (that is, it is possible that some PRs that have been 19406 fixed are not listed here). 19407 19408 19409 For questions related to the use of GCC, please consult these web 19410 pages and the [551]GCC manuals. If that fails, the 19411 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 19412 web pages and the development of GCC are welcome on our developer 19413 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 19414 archives. 19415 19416 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 19417 distribution of this entire article is permitted in any medium, 19418 provided this notice is preserved. 19419 19420 These pages are [556]maintained by the GCC team. Last modified 19421 2023-01-19. 19422 19423 References 19424 19425 1. https://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 19426 2. https://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 19427 3. https://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 19428 4. https://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 19429 5. https://gcc.gnu.org/news/dfa.html 19430 6. https://gcc.gnu.org/c99status.html 19431 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 19432 8. https://gcc.gnu.org/PR10140 19433 9. https://gcc.gnu.org/PR10198 19434 10. https://gcc.gnu.org/PR10338 19435 11. https://gcc.gnu.org/PR3581 19436 12. https://gcc.gnu.org/PR4382 19437 13. https://gcc.gnu.org/PR5533 19438 14. https://gcc.gnu.org/PR6387 19439 15. https://gcc.gnu.org/PR6412 19440 16. https://gcc.gnu.org/PR6620 19441 17. https://gcc.gnu.org/PR6663 19442 18. https://gcc.gnu.org/PR7068 19443 19. https://gcc.gnu.org/PR7083 19444 20. https://gcc.gnu.org/PR7647 19445 21. https://gcc.gnu.org/PR7675 19446 22. https://gcc.gnu.org/PR7718 19447 23. https://gcc.gnu.org/PR8116 19448 24. https://gcc.gnu.org/PR8358 19449 25. https://gcc.gnu.org/PR8511 19450 26. https://gcc.gnu.org/PR8564 19451 27. https://gcc.gnu.org/PR8660 19452 28. https://gcc.gnu.org/PR8766 19453 29. https://gcc.gnu.org/PR8803 19454 30. https://gcc.gnu.org/PR8846 19455 31. https://gcc.gnu.org/PR8906 19456 32. https://gcc.gnu.org/PR9216 19457 33. https://gcc.gnu.org/PR9261 19458 34. https://gcc.gnu.org/PR9263 19459 35. https://gcc.gnu.org/PR9429 19460 36. https://gcc.gnu.org/PR9516 19461 37. https://gcc.gnu.org/PR9600 19462 38. https://gcc.gnu.org/PR9629 19463 39. https://gcc.gnu.org/PR9672 19464 40. https://gcc.gnu.org/PR9749 19465 41. https://gcc.gnu.org/PR9794 19466 42. https://gcc.gnu.org/PR9829 19467 43. https://gcc.gnu.org/PR9916 19468 44. https://gcc.gnu.org/PR9936 19469 45. https://gcc.gnu.org/PR10262 19470 46. https://gcc.gnu.org/PR10278 19471 47. https://gcc.gnu.org/PR10446 19472 48. https://gcc.gnu.org/PR10451 19473 49. https://gcc.gnu.org/PR10506 19474 50. https://gcc.gnu.org/PR10549 19475 51. https://gcc.gnu.org/PR2001 19476 52. https://gcc.gnu.org/PR2391 19477 53. https://gcc.gnu.org/PR2960 19478 54. https://gcc.gnu.org/PR4046 19479 55. https://gcc.gnu.org/PR6405 19480 56. https://gcc.gnu.org/PR6798 19481 57. https://gcc.gnu.org/PR6871 19482 58. https://gcc.gnu.org/PR6909 19483 59. https://gcc.gnu.org/PR7189 19484 60. https://gcc.gnu.org/PR7642 19485 61. https://gcc.gnu.org/PR8634 19486 62. https://gcc.gnu.org/PR8750 19487 63. https://gcc.gnu.org/PR2161 19488 64. https://gcc.gnu.org/PR4319 19489 65. https://gcc.gnu.org/PR8602 19490 66. https://gcc.gnu.org/PR9177 19491 67. https://gcc.gnu.org/PR9853 19492 68. https://gcc.gnu.org/PR45 19493 69. https://gcc.gnu.org/PR3784 19494 70. https://gcc.gnu.org/PR764 19495 71. https://gcc.gnu.org/PR5116 19496 72. https://gcc.gnu.org/PR2862 19497 73. https://gcc.gnu.org/PR3663 19498 74. https://gcc.gnu.org/PR3797 19499 75. https://gcc.gnu.org/PR3948 19500 76. https://gcc.gnu.org/PR4137 19501 77. https://gcc.gnu.org/PR4361 19502 78. https://gcc.gnu.org/PR4802 19503 79. https://gcc.gnu.org/PR5837 19504 80. https://gcc.gnu.org/PR4803 19505 81. https://gcc.gnu.org/PR5094 19506 82. https://gcc.gnu.org/PR5730 19507 83. https://gcc.gnu.org/PR6713 19508 84. https://gcc.gnu.org/PR7015 19509 85. https://gcc.gnu.org/PR7086 19510 86. https://gcc.gnu.org/PR7099 19511 87. https://gcc.gnu.org/PR7247 19512 88. https://gcc.gnu.org/PR7441 19513 89. https://gcc.gnu.org/PR7768 19514 90. https://gcc.gnu.org/PR7804 19515 91. https://gcc.gnu.org/PR8099 19516 92. https://gcc.gnu.org/PR8117 19517 93. https://gcc.gnu.org/PR8205 19518 94. https://gcc.gnu.org/PR8645 19519 95. https://gcc.gnu.org/PR8724 19520 96. https://gcc.gnu.org/PR8805 19521 97. https://gcc.gnu.org/PR8691 19522 98. https://gcc.gnu.org/PR8700 19523 99. https://gcc.gnu.org/PR8724 19524 100. https://gcc.gnu.org/PR8949 19525 101. https://gcc.gnu.org/PR9016 19526 102. https://gcc.gnu.org/PR9053 19527 103. https://gcc.gnu.org/PR9152 19528 104. https://gcc.gnu.org/PR9182 19529 105. https://gcc.gnu.org/PR9297 19530 106. https://gcc.gnu.org/PR9318 19531 107. https://gcc.gnu.org/PR9320 19532 108. https://gcc.gnu.org/PR9400 19533 109. https://gcc.gnu.org/PR9424 19534 110. https://gcc.gnu.org/PR9425 19535 111. https://gcc.gnu.org/PR9439 19536 112. https://gcc.gnu.org/PR9474 19537 113. https://gcc.gnu.org/PR9548 19538 114. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 19539 115. https://gcc.gnu.org/PR9555 19540 116. https://gcc.gnu.org/PR9561 19541 117. https://gcc.gnu.org/PR9563 19542 118. https://gcc.gnu.org/PR9582 19543 119. https://gcc.gnu.org/PR9622 19544 120. https://gcc.gnu.org/PR9683 19545 121. https://gcc.gnu.org/PR9791 19546 122. https://gcc.gnu.org/PR9817 19547 123. https://gcc.gnu.org/PR9825 19548 124. https://gcc.gnu.org/PR9826 19549 125. https://gcc.gnu.org/PR9924 19550 126. https://gcc.gnu.org/PR9946 19551 127. https://gcc.gnu.org/PR9964 19552 128. https://gcc.gnu.org/PR9988 19553 129. https://gcc.gnu.org/PR10033 19554 130. https://gcc.gnu.org/PR10097 19555 131. https://gcc.gnu.org/PR10132 19556 132. https://gcc.gnu.org/PR10180 19557 133. https://gcc.gnu.org/PR10199 19558 134. https://gcc.gnu.org/PR10300 19559 135. https://gcc.gnu.org/PR10427 19560 136. https://gcc.gnu.org/PR10503 19561 137. https://gcc.gnu.org/PR5956 19562 138. https://gcc.gnu.org/PR1832 19563 139. https://gcc.gnu.org/PR3924 19564 140. https://gcc.gnu.org/PR5634 19565 141. https://gcc.gnu.org/PR6367 19566 142. https://gcc.gnu.org/PR6491 19567 143. https://gcc.gnu.org/PR6742 19568 144. https://gcc.gnu.org/PR7113 19569 145. https://gcc.gnu.org/PR7236 19570 146. https://gcc.gnu.org/PR7278 19571 147. https://gcc.gnu.org/PR7384 19572 148. https://gcc.gnu.org/PR7388 19573 149. https://gcc.gnu.org/PR8587 19574 150. https://gcc.gnu.org/PR9038 19575 151. https://gcc.gnu.org/PR10197 19576 152. https://gcc.gnu.org/PR6005 19577 153. https://gcc.gnu.org/PR6389 19578 154. https://gcc.gnu.org/PR6576 19579 155. https://gcc.gnu.org/PR6652 19580 156. https://gcc.gnu.org/PR7060 19581 157. https://gcc.gnu.org/PR7073 19582 158. https://gcc.gnu.org/PR7180 19583 159. https://gcc.gnu.org/PR7416 19584 160. https://gcc.gnu.org/PR7570 19585 161. https://gcc.gnu.org/PR7578 19586 162. https://gcc.gnu.org/PR7611 19587 163. https://gcc.gnu.org/PR7709 19588 164. https://gcc.gnu.org/PR7766 19589 165. https://gcc.gnu.org/PR7785 19590 166. https://gcc.gnu.org/PR7786 19591 167. https://gcc.gnu.org/PR8142 19592 168. https://gcc.gnu.org/PR8234 19593 169. https://gcc.gnu.org/PR8415 19594 170. https://gcc.gnu.org/PR8481 19595 171. https://gcc.gnu.org/PR8593 19596 172. https://gcc.gnu.org/PR8759 19597 173. https://gcc.gnu.org/PR8997 19598 174. https://gcc.gnu.org/PR9253 19599 175. https://gcc.gnu.org/PR9254 19600 176. https://gcc.gnu.org/PR9271 19601 177. https://gcc.gnu.org/PR6767 19602 178. https://gcc.gnu.org/PR9911 19603 179. https://gcc.gnu.org/PR10020 19604 180. https://gcc.gnu.org/PR10546 19605 181. https://gcc.gnu.org/PR7029 19606 182. https://gcc.gnu.org/PR2903 19607 183. https://gcc.gnu.org/PR7873 19608 184. https://gcc.gnu.org/PR7680 19609 185. https://gcc.gnu.org/PR8705 19610 186. https://gcc.gnu.org/PR9986 19611 187. https://gcc.gnu.org/PR10056 19612 188. https://gcc.gnu.org/PR6744 19613 189. https://gcc.gnu.org/PR7361 19614 190. https://gcc.gnu.org/PR9496 19615 191. https://gcc.gnu.org/PR7067 19616 192. https://gcc.gnu.org/PR8480 19617 193. https://gcc.gnu.org/PR8784 19618 194. https://gcc.gnu.org/PR10315 19619 195. https://gcc.gnu.org/PR10267 19620 196. https://gcc.gnu.org/PR7916 19621 197. https://gcc.gnu.org/PR7926 19622 198. https://gcc.gnu.org/PR8555 19623 199. https://gcc.gnu.org/PR8994 19624 200. https://gcc.gnu.org/PR9426 19625 201. https://gcc.gnu.org/PR9806 19626 202. https://gcc.gnu.org/PR10077 19627 203. https://gcc.gnu.org/PR10233 19628 204. https://gcc.gnu.org/PR10286 19629 205. https://gcc.gnu.org/PR10308 19630 206. https://gcc.gnu.org/PR11272 19631 207. https://gcc.gnu.org/PR5754 19632 208. https://gcc.gnu.org/PR6597 19633 209. https://gcc.gnu.org/PR6949 19634 210. https://gcc.gnu.org/PR7053 19635 211. https://gcc.gnu.org/PR8164 19636 212. https://gcc.gnu.org/PR8384 19637 213. https://gcc.gnu.org/PR9559 19638 214. https://gcc.gnu.org/PR9649 19639 215. https://gcc.gnu.org/PR9864 19640 216. https://gcc.gnu.org/PR10432 19641 217. https://gcc.gnu.org/PR10475 19642 218. https://gcc.gnu.org/PR10635 19643 219. https://gcc.gnu.org/PR10661 19644 220. https://gcc.gnu.org/PR10700 19645 221. https://gcc.gnu.org/PR10712 19646 222. https://gcc.gnu.org/PR10796 19647 223. https://gcc.gnu.org/PR10890 19648 224. https://gcc.gnu.org/PR10939 19649 225. https://gcc.gnu.org/PR10956 19650 226. https://gcc.gnu.org/PR11041 19651 227. https://gcc.gnu.org/PR11059 19652 228. https://gcc.gnu.org/PR11083 19653 229. https://gcc.gnu.org/PR11105 19654 230. https://gcc.gnu.org/PR11149 19655 231. https://gcc.gnu.org/PR11228 19656 232. https://gcc.gnu.org/PR11282 19657 233. https://gcc.gnu.org/PR11301 19658 234. https://gcc.gnu.org/PR11308 19659 235. https://gcc.gnu.org/PR11473 19660 236. https://gcc.gnu.org/PR11503 19661 237. https://gcc.gnu.org/PR11513 19662 238. https://gcc.gnu.org/PR11198 19663 239. https://gcc.gnu.org/PR11304 19664 240. https://gcc.gnu.org/PR11381 19665 241. https://gcc.gnu.org/PR11536 19666 242. https://gcc.gnu.org/PR11557 19667 243. https://gcc.gnu.org/PR5897 19668 244. https://gcc.gnu.org/PR11279 19669 245. https://gcc.gnu.org/PR11022 19670 246. https://gcc.gnu.org/PR2330 19671 247. https://gcc.gnu.org/PR5388 19672 248. https://gcc.gnu.org/PR5390 19673 249. https://gcc.gnu.org/PR7877 19674 250. https://gcc.gnu.org/PR9393 19675 251. https://gcc.gnu.org/PR10032 19676 252. https://gcc.gnu.org/PR10468 19677 253. https://gcc.gnu.org/PR10527 19678 254. https://gcc.gnu.org/PR10679 19679 255. https://gcc.gnu.org/PR10682 19680 256. https://gcc.gnu.org/PR10689 19681 257. https://gcc.gnu.org/PR10845 19682 258. https://gcc.gnu.org/PR10849 19683 259. https://gcc.gnu.org/PR10888 19684 260. https://gcc.gnu.org/PR10929 19685 261. https://gcc.gnu.org/PR10931 19686 262. https://gcc.gnu.org/PR10940 19687 263. https://gcc.gnu.org/PR10968 19688 264. https://gcc.gnu.org/PR10990 19689 265. https://gcc.gnu.org/PR11039 19690 266. https://gcc.gnu.org/PR11062 19691 267. https://gcc.gnu.org/PR11095 19692 268. https://gcc.gnu.org/PR11098 19693 269. https://gcc.gnu.org/PR11137 19694 270. https://gcc.gnu.org/PR11154 19695 271. https://gcc.gnu.org/PR11329 19696 272. https://gcc.gnu.org/PR11332 19697 273. https://gcc.gnu.org/PR11431 19698 274. https://gcc.gnu.org/PR11528 19699 275. https://gcc.gnu.org/PR11546 19700 276. https://gcc.gnu.org/PR11567 19701 277. https://gcc.gnu.org/PR11645 19702 278. https://gcc.gnu.org/PR5179 19703 279. https://gcc.gnu.org/PR8204 19704 280. https://gcc.gnu.org/PR10838 19705 281. https://gcc.gnu.org/PR10886 19706 282. https://gcc.gnu.org/PR11349 19707 283. https://gcc.gnu.org/PR4823 19708 284. https://gcc.gnu.org/PR8878 19709 285. https://gcc.gnu.org/PR9815 19710 286. https://gcc.gnu.org/PR10402 19711 287. https://gcc.gnu.org/PR10504 19712 288. https://gcc.gnu.org/PR10673 19713 289. https://gcc.gnu.org/PR11044 19714 290. https://gcc.gnu.org/PR11089 19715 291. https://gcc.gnu.org/PR11420 19716 292. https://gcc.gnu.org/PR9362 19717 293. https://gcc.gnu.org/PR10142 19718 294. https://gcc.gnu.org/PR10663 19719 295. https://gcc.gnu.org/PR10835 19720 296. https://gcc.gnu.org/PR10876 19721 297. https://gcc.gnu.org/PR10955 19722 298. https://gcc.gnu.org/PR11018 19723 299. https://gcc.gnu.org/PR11556 19724 300. https://gcc.gnu.org/PR10907 19725 301. https://gcc.gnu.org/PR11320 19726 302. https://gcc.gnu.org/PR11599 19727 303. https://gcc.gnu.org/PR9745 19728 304. https://gcc.gnu.org/PR10871 19729 305. https://gcc.gnu.org/PR11440 19730 306. https://gcc.gnu.org/PR7594 19731 307. https://gcc.gnu.org/PR10557 19732 308. https://gcc.gnu.org/PR11054 19733 309. https://gcc.gnu.org/PR10834 19734 310. https://gcc.gnu.org/PR10842 19735 311. https://gcc.gnu.org/PR11052 19736 312. https://gcc.gnu.org/PR11183 19737 313. https://gcc.gnu.org/PR11084 19738 314. https://gcc.gnu.org/PR10331 19739 315. https://gcc.gnu.org/PR10413 19740 316. https://gcc.gnu.org/PR11096 19741 317. https://gcc.gnu.org/PR2873 19742 318. https://gcc.gnu.org/PR3163 19743 319. https://gcc.gnu.org/PR5287 19744 320. https://gcc.gnu.org/PR10148 19745 321. https://gcc.gnu.org/PR8787 19746 322. https://gcc.gnu.org/PR10900 19747 323. https://gcc.gnu.org/PR1607 19748 324. https://gcc.gnu.org/PR4252 19749 325. https://gcc.gnu.org/PR4490 19750 326. https://gcc.gnu.org/PR10355 19751 327. https://gcc.gnu.org/PR10726 19752 328. https://gcc.gnu.org/PR10805 19753 329. https://gcc.gnu.org/PR10815 19754 330. https://gcc.gnu.org/PR10877 19755 331. https://gcc.gnu.org/PR11280 19756 332. https://gcc.gnu.org/PR11466 19757 333. https://gcc.gnu.org/PR10737 19758 334. https://gcc.gnu.org/PR10810 19759 335. https://gcc.gnu.org/PR8336 19760 336. https://gcc.gnu.org/PR9330 19761 337. https://gcc.gnu.org/PR9631 19762 338. https://gcc.gnu.org/PR9877 19763 339. https://gcc.gnu.org/PR11687 19764 340. https://gcc.gnu.org/PR12263 19765 341. https://gcc.gnu.org/PR12490 19766 342. https://gcc.gnu.org/PR7277 19767 343. https://gcc.gnu.org/PR7939 19768 344. https://gcc.gnu.org/PR11063 19769 345. https://gcc.gnu.org/PR11207 19770 346. https://gcc.gnu.org/PR11522 19771 347. https://gcc.gnu.org/PR11595 19772 348. https://gcc.gnu.org/PR11646 19773 349. https://gcc.gnu.org/PR11665 19774 350. https://gcc.gnu.org/PR11852 19775 351. https://gcc.gnu.org/PR11878 19776 352. https://gcc.gnu.org/PR11883 19777 353. https://gcc.gnu.org/PR11991 19778 354. https://gcc.gnu.org/PR12146 19779 355. https://gcc.gnu.org/PR12215 19780 356. https://gcc.gnu.org/PR12369 19781 357. https://gcc.gnu.org/PR12446 19782 358. https://gcc.gnu.org/PR12510 19783 359. https://gcc.gnu.org/PR12544 19784 360. https://gcc.gnu.org/PR9862 19785 361. https://gcc.gnu.org/PR10962 19786 362. https://gcc.gnu.org/PR11370 19787 363. https://gcc.gnu.org/PR11637 19788 364. https://gcc.gnu.org/PR11885 19789 365. https://gcc.gnu.org/PR12082 19790 366. https://gcc.gnu.org/PR12180 19791 367. https://gcc.gnu.org/PR12340 19792 368. https://gcc.gnu.org/PR3907 19793 369. https://gcc.gnu.org/PR5293 19794 370. https://gcc.gnu.org/PR5296 19795 371. https://gcc.gnu.org/PR7939 19796 372. https://gcc.gnu.org/PR8656 19797 373. https://gcc.gnu.org/PR10147 19798 374. https://gcc.gnu.org/PR11400 19799 375. https://gcc.gnu.org/PR11409 19800 376. https://gcc.gnu.org/PR11740 19801 377. https://gcc.gnu.org/PR11786 19802 378. https://gcc.gnu.org/PR11867 19803 379. https://gcc.gnu.org/PR11928 19804 380. https://gcc.gnu.org/PR12114 19805 381. https://gcc.gnu.org/PR12163 19806 382. https://gcc.gnu.org/PR12181 19807 383. https://gcc.gnu.org/PR12236 19808 384. https://gcc.gnu.org/PR12266 19809 385. https://gcc.gnu.org/PR12296 19810 386. https://gcc.gnu.org/PR12298 19811 387. https://gcc.gnu.org/PR12369 19812 388. https://gcc.gnu.org/PR12337 19813 389. https://gcc.gnu.org/PR12344 19814 390. https://gcc.gnu.org/PR12451 19815 391. https://gcc.gnu.org/PR12486 19816 392. https://gcc.gnu.org/PR8869 19817 393. https://gcc.gnu.org/PR9786 19818 394. https://gcc.gnu.org/PR11689 19819 395. https://gcc.gnu.org/PR12116 19820 396. https://gcc.gnu.org/PR12070 19821 397. https://gcc.gnu.org/PR11184 19822 398. https://gcc.gnu.org/PR11535 19823 399. https://gcc.gnu.org/PR11693 19824 400. https://gcc.gnu.org/PR12224 19825 401. https://gcc.gnu.org/PR11087 19826 402. https://gcc.gnu.org/PR11319 19827 403. https://gcc.gnu.org/PR11949 19828 404. https://gcc.gnu.org/PR11662 19829 405. https://gcc.gnu.org/PR11965 19830 406. https://gcc.gnu.org/PR12301 19831 407. https://gcc.gnu.org/PR11717 19832 408. https://gcc.gnu.org/PR11313 19833 409. https://gcc.gnu.org/PR11712 19834 410. https://gcc.gnu.org/PR12166 19835 411. https://gcc.gnu.org/PR12101 19836 412. https://gcc.gnu.org/PR10988 19837 413. https://gcc.gnu.org/PR11805 19838 414. https://gcc.gnu.org/PR11902 19839 415. https://gcc.gnu.org/PR11903 19840 416. https://gcc.gnu.org/PR11890 19841 417. https://gcc.gnu.org/PR12399 19842 418. https://gcc.gnu.org/PR13068 19843 419. https://gcc.gnu.org/PR10060 19844 420. https://gcc.gnu.org/PR10555 19845 421. https://gcc.gnu.org/PR10706 19846 422. https://gcc.gnu.org/PR11496 19847 423. https://gcc.gnu.org/PR11741 19848 424. https://gcc.gnu.org/PR12440 19849 425. https://gcc.gnu.org/PR12632 19850 426. https://gcc.gnu.org/PR12712 19851 427. https://gcc.gnu.org/PR12726 19852 428. https://gcc.gnu.org/PR12890 19853 429. https://gcc.gnu.org/PR12900 19854 430. https://gcc.gnu.org/PR13060 19855 431. https://gcc.gnu.org/PR13289 19856 432. https://gcc.gnu.org/PR13318 19857 433. https://gcc.gnu.org/PR13392 19858 434. https://gcc.gnu.org/PR13574 19859 435. https://gcc.gnu.org/PR13475 19860 436. https://gcc.gnu.org/PR13797 19861 437. https://gcc.gnu.org/PR13824 19862 438. https://gcc.gnu.org/PR8776 19863 439. https://gcc.gnu.org/PR10339 19864 440. https://gcc.gnu.org/PR11350 19865 441. https://gcc.gnu.org/PR12826 19866 442. https://gcc.gnu.org/PR12500 19867 443. https://gcc.gnu.org/PR12941 19868 444. https://gcc.gnu.org/PR12953 19869 445. https://gcc.gnu.org/PR13041 19870 446. https://gcc.gnu.org/PR13507 19871 447. https://gcc.gnu.org/PR13382 19872 448. https://gcc.gnu.org/PR13394 19873 449. https://gcc.gnu.org/PR13400 19874 450. https://gcc.gnu.org/PR13521 19875 451. https://gcc.gnu.org/PR2094 19876 452. https://gcc.gnu.org/PR2294 19877 453. https://gcc.gnu.org/PR5050 19878 454. https://gcc.gnu.org/PR9371 19879 455. https://gcc.gnu.org/PR9546 19880 456. https://gcc.gnu.org/PR10081 19881 457. https://gcc.gnu.org/PR10093 19882 458. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 19883 459. https://gcc.gnu.org/PR10095 19884 460. https://gcc.gnu.org/PR11554 19885 461. https://gcc.gnu.org/PR12297 19886 462. https://gcc.gnu.org/PR12352 19887 463. https://gcc.gnu.org/PR12438 19888 464. https://gcc.gnu.org/PR12540 19889 465. https://gcc.gnu.org/PR12594 19890 466. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 19891 467. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 19892 468. https://gcc.gnu.org/PR12657 19893 469. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 19894 470. https://gcc.gnu.org/PR12696 19895 471. https://gcc.gnu.org/PR12815 19896 472. https://gcc.gnu.org/PR12862 19897 473. https://gcc.gnu.org/PR12926 19898 474. https://gcc.gnu.org/PR12967 19899 475. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 19900 476. https://gcc.gnu.org/PR12971 19901 477. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 19902 478. https://gcc.gnu.org/PR13007 19903 479. https://gcc.gnu.org/PR13009 19904 480. https://gcc.gnu.org/PR13057 19905 481. https://gcc.gnu.org/PR13070 19906 482. https://gcc.gnu.org/PR13081 19907 483. https://gcc.gnu.org/PR13239 19908 484. https://gcc.gnu.org/PR13262 19909 485. https://gcc.gnu.org/PR13290 19910 486. https://gcc.gnu.org/PR13323 19911 487. https://gcc.gnu.org/PR13369 19912 488. https://gcc.gnu.org/PR13371 19913 489. https://gcc.gnu.org/PR13445 19914 490. https://gcc.gnu.org/PR13461 19915 491. https://gcc.gnu.org/PR13462 19916 492. https://gcc.gnu.org/PR13478 19917 493. https://gcc.gnu.org/PR13544 19918 494. https://gcc.gnu.org/PR13650 19919 495. https://gcc.gnu.org/PR13683 19920 496. https://gcc.gnu.org/PR13688 19921 497. https://gcc.gnu.org/PR13774 19922 498. https://gcc.gnu.org/PR13884 19923 499. https://gcc.gnu.org/PR10746 19924 500. https://gcc.gnu.org/PR11433 19925 501. https://gcc.gnu.org/PR12633 19926 502. https://gcc.gnu.org/PR13037 19927 503. https://gcc.gnu.org/PR13213 19928 504. https://gcc.gnu.org/PR4490 19929 505. https://gcc.gnu.org/PR12292 19930 506. https://gcc.gnu.org/PR12441 19931 507. https://gcc.gnu.org/PR12943 19932 508. https://gcc.gnu.org/PR13608 19933 509. https://gcc.gnu.org/PR11598 19934 510. https://gcc.gnu.org/PR11793 19935 511. https://gcc.gnu.org/PR12467 19936 512. https://gcc.gnu.org/PR12537 19937 513. https://gcc.gnu.org/PR12496 19938 514. https://gcc.gnu.org/PR12865 19939 515. https://gcc.gnu.org/PR13354 19940 516. https://gcc.gnu.org/PR10467 19941 517. https://gcc.gnu.org/PR11226 19942 518. https://gcc.gnu.org/PR11227 19943 519. https://gcc.gnu.org/PR12644 19944 520. https://gcc.gnu.org/PR13149 19945 521. https://gcc.gnu.org/PR12654 19946 522. https://gcc.gnu.org/PR12965 19947 523. https://gcc.gnu.org/PR13031 19948 524. https://gcc.gnu.org/PR11634 19949 525. https://gcc.gnu.org/PR12158 19950 526. https://gcc.gnu.org/PR11992 19951 527. https://gcc.gnu.org/PR9365 19952 528. https://gcc.gnu.org/PR10392 19953 529. https://gcc.gnu.org/PR11322 19954 530. https://gcc.gnu.org/PR13069 19955 531. https://gcc.gnu.org/PR13302 19956 532. https://gcc.gnu.org/PR13585 19957 533. https://gcc.gnu.org/PR8916 19958 534. https://gcc.gnu.org/PR11576 19959 535. https://gcc.gnu.org/PR13122 19960 536. https://gcc.gnu.org/PR13256 19961 537. https://gcc.gnu.org/PR13373 19962 538. https://gcc.gnu.org/PR12561 19963 539. https://gcc.gnu.org/PR6243 19964 540. https://gcc.gnu.org/PR11397 19965 541. https://gcc.gnu.org/PR12505 19966 542. https://gcc.gnu.org/PR13150 19967 543. https://gcc.gnu.org/PR12666 19968 544. https://gcc.gnu.org/PR12969 19969 545. https://gcc.gnu.org/PR10819 19970 546. https://gcc.gnu.org/PR11612 19971 547. https://gcc.gnu.org/PR13211 19972 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 19973 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 19974 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 19975 551. https://gcc.gnu.org/onlinedocs/ 19976 552. mailto:gcc-help (a] gcc.gnu.org 19977 553. mailto:gcc (a] gcc.gnu.org 19978 554. https://gcc.gnu.org/lists.html 19979 555. https://www.fsf.org/ 19980 556. https://gcc.gnu.org/about.html 19981 ====================================================================== 19982 http://gcc.gnu.org/gcc-3.2/index.html 19983 GCC 3.2 Release Series 19984 19985 (This release series is no longer supported.) 19986 19987 April 25, 2003 19988 19989 The [1]GNU project and the GCC developers are pleased to announce the 19990 release of GCC 3.2.3. 19991 19992 The purpose of the GCC 3.2 release series is to provide a stable 19993 platform for OS distributors to use building their next releases. A 19994 primary objective was to stabilize the C++ ABI; we believe that the 19995 interface to the compiler and the C++ standard library are now 19996 relatively stable. 19997 19998 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 19999 interoperate with code compiled by GCC 3.1.1 or earlier. 20000 20001 Please refer to our [2]detailed list of news, caveats, and bug-fixes 20002 for further information. 20003 20004 Release History 20005 20006 GCC 3.2.3 20007 April 25, 2003 ([3]changes) 20008 20009 GCC 3.2.2 20010 February 5, 2003 ([4]changes) 20011 20012 GCC 3.2.1 20013 November 19, 2002 ([5]changes) 20014 20015 GCC 3.2 20016 August 14, 2002 ([6]changes) 20017 20018 References and Acknowledgements 20019 20020 GCC used to stand for the GNU C Compiler, but since the compiler 20021 supports several other languages aside from C, it now stands for the 20022 GNU Compiler Collection. 20023 20024 A list of [7]successful builds is updated as new information becomes 20025 available. 20026 20027 The GCC developers would like to thank the numerous people that have 20028 contributed new features, improvements, bug fixes, and other changes as 20029 well as test results to GCC. This [8]amazing group of volunteers is 20030 what makes GCC successful. 20031 20032 For additional information about GCC please refer to the [9]GCC project 20033 web site or contact the [10]GCC development mailing list. 20034 20035 To obtain GCC please use [11]our mirror sites, or our CVS server. 20036 20037 20038 For questions related to the use of GCC, please consult these web 20039 pages and the [12]GCC manuals. If that fails, the 20040 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20041 web pages and the development of GCC are welcome on our developer 20042 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 20043 archives. 20044 20045 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 20046 distribution of this entire article is permitted in any medium, 20047 provided this notice is preserved. 20048 20049 These pages are [17]maintained by the GCC team. Last modified 20050 2022-10-26. 20051 20052 References 20053 20054 1. http://www.gnu.org/ 20055 2. https://gcc.gnu.org/gcc-3.2/changes.html 20056 3. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 20057 4. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 20058 5. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 20059 6. https://gcc.gnu.org/gcc-3.2/changes.html#3.2 20060 7. https://gcc.gnu.org/gcc-3.2/buildstat.html 20061 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20062 9. https://gcc.gnu.org/index.html 20063 10. mailto:gcc (a] gcc.gnu.org 20064 11. https://gcc.gnu.org/mirrors.html 20065 12. https://gcc.gnu.org/onlinedocs/ 20066 13. mailto:gcc-help (a] gcc.gnu.org 20067 14. mailto:gcc (a] gcc.gnu.org 20068 15. https://gcc.gnu.org/lists.html 20069 16. https://www.fsf.org/ 20070 17. https://gcc.gnu.org/about.html 20071 ====================================================================== 20072 http://gcc.gnu.org/gcc-3.2/changes.html 20073 GCC 3.2 Release Series 20074 Changes, New Features, and Fixes 20075 20076 The latest release in the 3.2 release series is [1]GCC 3.2.3. 20077 20078 Caveats and New Features 20079 20080 Caveats 20081 20082 * The C++ compiler does not correctly zero-initialize 20083 pointers-to-data members. You must explicitly initialize them. For 20084 example: int S::*m(0); will work, but depending on 20085 default-initialization to zero will not work. This bug cannot be 20086 fixed in GCC 3.2 without inducing unacceptable risks. It will be 20087 fixed in GCC 3.3. 20088 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 20089 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 20090 a number of C++ ABI fixes which make its C++ compiler generate 20091 binary code which is incompatible with the C++ compilers found in 20092 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 20093 20094 Frontend Enhancements 20095 20096 C/C++/Objective-C 20097 20098 * The method of constructing the list of directories to be searched 20099 for header files has been revised. If a directory named by a -I 20100 option is a standard system include directory, the option is 20101 ignored to ensure that the default search order for system 20102 directories and the special treatment of system header files are 20103 not defeated. 20104 * The C and Objective-C compilers no longer accept the "Naming Types" 20105 extension (typedef foo = bar); it was already unavailable in C++. 20106 Code which uses it will need to be changed to use the "typeof" 20107 extension instead: typedef typeof(bar) foo. (We have removed this 20108 extension without a period of deprecation because it has caused the 20109 compiler to crash since version 3.0 and no one noticed until very 20110 recently. Thus we conclude it is not in widespread use.) 20111 20112 C++ 20113 20114 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 20115 in GCC and the multi-vendor standard, but more have been found 20116 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 20117 code which is affected by these bugs. We will fix these bugs in 20118 some future release, once we are confident that all have been 20119 found; until then, it is our intention to make changes to the ABI 20120 only if they are necessary for correct compilation of C++, as 20121 opposed to conformance to the ABI documents. 20122 * For details on how to build an ABI compliant compiler for GNU/Linux 20123 systems, check the [3]common C++ ABI page. 20124 20125 New Targets and Target Specific Improvements 20126 20127 IA-32 20128 20129 * Fixed a number of bugs in SSE and MMX intrinsics. 20130 * Fixed common compiler crashes with SSE instruction set enabled 20131 (implied by -march=pentium3, pentium4, athlon-xp) 20132 * __m128 and __m128i is not 128bit aligned when used in structures. 20133 20134 x86-64 20135 20136 * A bug whereby the compiler could generate bad code for bzero has 20137 been fixed. 20138 * ABI fixes (implying ABI incompatibilities with previous version in 20139 some corner cases) 20140 * Fixed prefetch code generation 20141 __________________________________________________________________ 20142 20143 GCC 3.2.3 20144 20145 3.2.3 is a bug fix release only; there are no new features that were 20146 not present in GCC 3.2.2. 20147 20148 Bug Fixes 20149 20150 This section lists the problem reports (PRs) from GCC's bug tracking 20151 system that are known to be fixed in the 3.2.3 release. This list might 20152 not be complete (that is, it is possible that some PRs that have been 20153 fixed are not listed here), and some of the titles have been changed to 20154 make them more clear. 20155 20156 Internal Compiler Errors (multi-platform) 20157 20158 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 20159 cc1plus 20160 * [5]6440: (c++) template specializations cause ICE 20161 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 20162 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 20163 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 20164 * [9]8068: exceedingly high (infinite) memory usage 20165 * [10]8178: ICE with __builtin_ffs 20166 * [11]8396: ICE in copy_to_mode_reg, in explow.c 20167 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 20168 * [13]9768: ICE when optimizing inline code at -O2 20169 * [14]9798: (c++) Infinite recursion (segfault) in 20170 cp/decl.c:push_using_directive with recursive using directives 20171 * [15]9799: mismatching structure initializer with nested flexible 20172 array member: ICE 20173 * [16]9928: ICE on duplicate enum declaration 20174 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 20175 sparc, alpha) 20176 * [18]10352: ICE in find_reloads_toplev 20177 * [19]10336: ICE with -Wunreachable-code 20178 20179 C/optimizer bugs: 20180 20181 * [20]8224: Incorrect joining of signed and unsigned division 20182 * [21]8613: -O2 produces wrong code with builtin strlen and 20183 postincrements 20184 * [22]8828: gcc reports some code is unreachable when it is not 20185 * [23]9226: GCSE breaking argument passing 20186 * [24]9853: miscompilation of non-constant structure initializer 20187 * [25]9797: C99-style struct initializers are miscompiled 20188 * [26]9967: Some standard C function calls should not be replaced 20189 when optimizing for size 20190 * [27]10116: ce2: invalid merge of join_bb in the context of switch 20191 statements 20192 * [28]10171: wrong code for inlined function 20193 * [29]10175: -Wunreachable-code doesn't work for single lines 20194 20195 C++ compiler and library: 20196 20197 * [30]8316: Confusing diagnostic for code that misuses conversion 20198 operators 20199 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 20200 * [32]9420: incomplete type incorrectly reported 20201 * [33]9459: typeof in return type specification of template not 20202 supported 20203 * [34]9507: filebuf::open handles ios_base::ate incorrectly 20204 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 20205 * [36]9602: Total confusion about template/friend/virtual/abstract 20206 * [37]9993: destructor not called for local object created within and 20207 returned from infinite loop 20208 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 20209 system 20210 20211 Java compiler and library: 20212 20213 * [39]9652: libgcj build fails on irix6.5.1[78] 20214 * [40]10144: gas on solaris complains about bad .stabs lines for 20215 java, native as unaffected 20216 20217 x86-specific (Intel/AMD): 20218 20219 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 20220 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 20221 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 20222 failed 20223 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 20224 failed 20225 20226 SPARC-specific: 20227 20228 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 20229 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 20230 unroll.c 20231 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 20232 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 20233 execute/loop-2d.c 20234 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 20235 * [50]9414: Scheduling bug on Ultrasparc 20236 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 20237 20238 m68k-specific: 20239 20240 * [52]7248: broken "inclusive or" code 20241 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 20242 20243 PowerPC-specific: 20244 20245 * [54]9732: Wrong code with -O2 -fPIC 20246 * [55]10073: ICE: powerpc cannot split insn 20247 20248 Alpha-specific: 20249 20250 * [56]7702: optimization problem on a DEC alpha under OSF1 20251 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 20252 20253 HP-specific: 20254 20255 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 20256 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 20257 (missing symbol) 20258 * [60]10271: Floating point args don't get reloaded across function 20259 calls with -O2 20260 20261 MIPS specific: 20262 20263 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 20264 compile/920501-4.c 20265 20266 CRIS specific: 20267 20268 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 20269 20270 Miscellaneous and minor bugs: 20271 20272 * [63]6955: collect2 says "core dumped" when there is no core 20273 __________________________________________________________________ 20274 20275 GCC 3.2.2 20276 20277 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 20278 install by means of the DESTDIR variable. Parts of the GCC tree have 20279 featured that support long before, but now it is available even from 20280 the top level. 20281 20282 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 20283 features that were not present in GCC 3.2.1. 20284 20285 Bug Fixes 20286 20287 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 20288 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 20289 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 20290 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 20291 change, and thus restores ABI-compatibility with previous releases 20292 (except GCC 3.2.1) on these platforms. 20293 20294 This section lists the problem reports (PRs) from GCC's bug tracking 20295 system that are known to be fixed in the 3.2.2 release. This list might 20296 not be complete (that is, it is possible that some PRs that have been 20297 fixed are not listed here) and some of the titles have been changed to 20298 make them more clear. 20299 20300 Internal Compiler Errors (multi-platform) 20301 20302 * [64]5919: (c++) ICE when passing variable array to template 20303 function 20304 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 20305 * [66]7507: ICE with -O2 when address of called function is a 20306 complicated expression 20307 * [67]7622: ICE with nested inline functions if function's address is 20308 taken 20309 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 20310 [69]9258) 20311 * [70]8031: (c++) ICE in code comparing typeids and casting from 20312 virtual base 20313 * [71]8275: ICE in simplify_subreg 20314 * [72]8332: (c++) builtin strlen/template interaction causes ICE 20315 * [73]8372: (c++) ICE on explicit call of destructor 20316 * [74]8439: (c, not c++) empty struct causes ICE 20317 * [75]8442: (c++) ICE with nested template classes 20318 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 20319 * [77]8615: (c++) ICE with out-of-range character constant template 20320 argument 20321 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 20322 * [79]8799: (c++) ICE: error reporting routines re-entered 20323 * [80]9328: (c++) ICE with typeof(X) for overloaded X 20324 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 20325 20326 C++ (compiler and library) bugs 20327 20328 * [82]47: scoping in nested classes is broken 20329 * [83]6745: problems with iostream rdbuf() member function 20330 * [84]8214: conversion from const char* const to char* sometimes 20331 accepted illegally 20332 * [85]8493: builtin strlen and overload resolution (same bug as 20333 [86]8332) 20334 * [87]8503: strange behaviour of function types 20335 * [88]8727: compiler confused by inheritance from an anonymous struct 20336 * [89]7445: poor performance of std::locale::classic() in 20337 multi-threaded applications 20338 * [90]8230: mishandling of overflow in vector<T>::resize 20339 * [91]8399: sync_with_stdio(false) breaks unformatted input 20340 * [92]8662: illegal access of private member of unnamed class is 20341 accepted 20342 * [93]8707: "make distclean" fails in libstdc++-v3 directory 20343 * [94]8708: __USE_MALLOC doesn't work 20344 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 20345 * [96]8887: Bug in date formats with --enable-clocale=generic 20346 * [97]9076: Call Frame Instructions are not handled correctly during 20347 unwind operation 20348 * [98]9151: std::setprecision limited to 16 digits when outputting a 20349 double to a stream 20350 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 20351 * [100]9269: libstdc++ headers: explicit specialization of function 20352 must precede its first use 20353 * [101]9322: return value of basic_streambuf<>::getloc affected by 20354 locale::global 20355 * [102]9433: segfault in runtime support for dynamic_cast 20356 20357 C and optimizer bugs 20358 20359 * [103]8032: GCC incorrectly initializes static structs that have 20360 flexible arrays 20361 * [104]8639: simple arithmetic expression broken 20362 * [105]8794: optimization improperly eliminates certain expressions 20363 * [106]8832: traditional "asm volatile" code is illegally optimized 20364 * [107]8988: loop optimizer bug: with -O2, code is generated that 20365 segfaults (found on i386, bug present for all platforms) 20366 * [108]9492: structure copy clobbers subsequent stores to structure 20367 20368 Objective-C bugs 20369 20370 * [109]9267: Objective-C parser won't build with newer bison versions 20371 (e.g. 1.875) 20372 20373 Ada bugs 20374 20375 * [110]8344: Ada build problem due to conflict between gcc/final.o, 20376 gcc/ada/final.o 20377 20378 Preprocessor bugs 20379 20380 * [111]8524: _Pragma within macros is improperly expanded 20381 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 20382 -fshort-wchar 20383 20384 ARM-specific 20385 20386 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 20387 20388 x86-specific (Intel/AMD) 20389 20390 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 20391 * [115]8599: loop unroll bug with -march=k6-3 20392 * [116]9506: ABI breakage in structure return (affects BSD and 20393 Cygwin, but not GNU/Linux) 20394 20395 FreeBSD 5.0 specific 20396 20397 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 20398 20399 RTEMS-specific 20400 20401 * [118]9292: hppa1.1-rtems configurery problems 20402 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 20403 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 20404 * [121]9296: gthr-rtems regression 20405 * [122]9316: powerpc-rtems: extending multilibs 20406 20407 HP-PA specific 20408 20409 * [123]9493: ICE with -O2 when building a simple function 20410 20411 Documentation 20412 20413 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 20414 * [125]8947: Please add a warning about "-malign-double" in docs 20415 * [126]7448, [127]8882: typo cleanups 20416 __________________________________________________________________ 20417 20418 GCC 3.2.1 20419 20420 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 20421 generates code that is known not to be binary-compatible with the 20422 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 20423 in the distribution, for details. 20424 20425 This release also removes an old GCC extension, "naming types", and the 20426 documentation now directs users to use a different GCC extension, 20427 __typeof__, instead. The feature had evidently been broken for a while. 20428 20429 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 20430 the new warning there are no new features that were not present in GCC 20431 3.2. 20432 20433 In addition, the previous fix for [128]PR 7445 (poor performance of 20434 std::locale::classic() in multi-threaded applications) was reverted 20435 ("unfixed"), because the "fix" was not thread-safe. 20436 20437 Bug Fixes 20438 20439 This section lists the problem reports (PRs) from GCC's bug tracking 20440 system that are known to be fixed in the 3.2.1 release. This list might 20441 not be complete (that is, it is possible that some PRs that have been 20442 fixed are not listed here). As you can see, the number of bug fixes is 20443 quite large, so it is strongly recommended that users of earlier GCC 20444 3.x releases upgrade to GCC 3.2.1. 20445 20446 Internal Compiler Errors (multi-platform) 20447 20448 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 20449 * [130]5661: (c++) ICE instantiating template on array of unknown 20450 size (bad code) 20451 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 20452 64-bit platforms 20453 * [132]6994: (c++) ICE in find_function_data 20454 * [133]7150: preprocessor: GCC -dM -E gives an ICE 20455 * [134]7160: ICE when optimizing branches without a return value 20456 * [135]7228: (c++) ICE when using member template and template 20457 function 20458 * [136]7266: (c++) ICE with -pedantic on missing typename 20459 * [137]7353: ICE from use of "Naming Types" extension, see above 20460 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 20461 * [139]7478: (c++) ICE on static_cast inside template 20462 * [140]7526: preprocessor core dump when _Pragma implies #pragma 20463 dependency 20464 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 20465 is a duplicate) 20466 * [143]7754: (c++) ICE on union with template parameter 20467 * [144]7788: (c++) redeclaring a definition as an incomplete class 20468 causes ICE 20469 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 20470 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 20471 kernel 20472 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 20473 variables 20474 * [148]8134: (c++) ICE in force_store_init_value on legal code 20475 * [149]8149: (c++) ICE on incomplete type 20476 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 20477 initialization 20478 20479 C++ (compiler and library) bugs 20480 20481 * [151]5607: No pointer adjustment in covariant return types 20482 * [152]6579: Infinite loop with statement expressions in member 20483 initialization 20484 * [153]6803: Default copy constructor bug in GCC 3.1 20485 * [154]7176: g++ confused by friend and static member with same name 20486 * [155]7188: Segfault with template class and recursive (incorrect) 20487 initializer list 20488 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 20489 inheritance if a method has a variable number of arguments 20490 * [157]7461: ctype<char>::classic_table() returns offset array on 20491 Cygwin 20492 * [158]7524: f(const float arg[3]) fails 20493 * [159]7584: Erroneous ambiguous base error on using declaration 20494 * [160]7676: Member template overloading problem 20495 * [161]7679: infinite loop when a right parenthesis is missing 20496 * [162]7811: default locale not taken from environment 20497 * [163]7961: compare( char *) implemented incorrectly in 20498 basic_string<> 20499 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 20500 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 20501 [166]6745) 20502 * [167]8096: deque::at() throws std::range_error instead of 20503 std::out_of_range 20504 * [168]8127: cout << cin.rdbuf() infinite loop 20505 * [169]8218: Excessively large memory consumed for classes with large 20506 array members 20507 * [170]8287: GCC 3.2: Destructor called for non-constructed local 20508 object 20509 * [171]8347: empty vector range used in string construction causes 20510 core dump 20511 * [172]8348: fail() flag is set in istringstream when eof() flag is 20512 set 20513 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 20514 20515 C and optimizer bugs 20516 20517 * [174]6627: -fno-align-functions doesn't seem to disable function 20518 alignment 20519 * [175]6631: life_analysis misoptimizes code to initialize fields of 20520 a structure 20521 * [176]7102: unsigned char division results in floating exception 20522 * [177]7120: Run once loop should *always* be unrolled 20523 (pessimization) 20524 * [178]7209: Bug involving array referencing and ?: operator 20525 * [179]7515: invalid inlining of global function with -O3 20526 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 20527 * [181]8467: bug in sibling call optimization 20528 20529 Preprocessor bugs 20530 20531 * [182]4890: incorrect line markers from the traditional preprocessor 20532 * [183]7357: -M option omits system headers files (making it the same 20533 as -MM) 20534 * [184]7358: Changes to Sun's make Dependencies 20535 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 20536 C headers 20537 * [186]7862: Interrupting GCC -MD removes .d file but not .o 20538 * [187]8190: Failed compilation deletes -MD dependency file 20539 * [188]8524: _Pragma within macro is improperly expanded 20540 20541 x86 specific (Intel/AMD) 20542 20543 * [189]5351: (i686-only) function pass-by-value structure copy 20544 corrupts stack ([190]7591 is a duplicate) 20545 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 20546 -march=pentium3/pentium2/athlon (these are all the same underlying 20547 bug, in MMX register use) 20548 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 20549 as above?) 20550 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 20551 * [199]6981: wrong code in 64-bit manipulation on x86 20552 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 20553 macro 20554 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 20555 intrinsics are broken 20556 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 20557 -march=pentium4 20558 * [203]7693: Typo in i386 mmintrin.h header 20559 * [204]7723: ICE - Pentium3 sse - GCC 3.2 20560 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 20561 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 20562 20563 PowerPC specific 20564 20565 * [207]5967: GCC bug when profiling nested functions on powerpc 20566 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 20567 loop on PowerPC 20568 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 20569 * [210]7130: miscompiled code for GCC-3.1 on 20570 powerpc-unknown-linux-gnu with -funroll-all-loops 20571 * [211]7133: PowerPC ICE: unrecognizable insn 20572 * [212]7380: ICE in extract_insn, at recog.c:2148 20573 * [213]8252: ICE on Altivec code with optimization turned on 20574 * [214]8451: Altivec ICE in GCC 3.2 20575 20576 HP/PA specific 20577 20578 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 20579 20580 SPARC specific 20581 20582 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 20583 in the wrong place on sparc-solaris 20584 * [217]7151: ICE when compiling for UltraSPARC 20585 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 20586 double and -O1 20587 * [219]7842: [REGRESSION] SPARC code gen bug 20588 20589 ARM specific 20590 20591 * [220]7856: [arm] invalid offset in constant pool reference 20592 * [221]7967: optimization produces wrong code (ARM) 20593 20594 Alpha specific 20595 20596 * [222]7374: __builtin_fabsl broken on alpha 20597 20598 IBM s390 specific 20599 20600 * [223]7370: ICE in fixup_var_refs_1 on s390x 20601 * [224]7409: loop optimization bug on s390x-linux-gnu 20602 * [225]8232: s390x: ICE when using bcmp with int length argument 20603 20604 SCO specific 20605 20606 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 20607 symbol: BITS_PER_UNIT 20608 20609 m68k/Coldfire specific 20610 20611 * [227]8314: crtbegin, crtend need to be multilib'ed for this 20612 platform 20613 20614 Documentation 20615 20616 * [228]761: Document some undocumented options 20617 * [229]5610: Fix documentation about invoking SSE instructions 20618 (-mfpmath=sse) 20619 * [230]7484: List -Wmissing-declarations as C-only option 20620 * [231]7531: -mcmodel not documented for x86-64 20621 * [232]8120: Update documentation of bad use of ## 20622 __________________________________________________________________ 20623 20624 GCC 3.2 20625 20626 3.2 is a small bug fix release, but there is a change to the 20627 application binary interface (ABI), hence the change to the second part 20628 of the version number. 20629 20630 The main purpose of the 3.2 release is to correct a couple of problems 20631 in the C++ ABI, with the intention of providing a stable interface 20632 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 20633 20634 Bug Fixes 20635 20636 C++ 20637 20638 * [233]7320: g++ 3.2 relocation problem 20639 * [234]7470: vtable: virtual function pointers not in declaration 20640 order 20641 20642 libstdc++ 20643 20644 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 20645 * [236]6503, [237]6642, [238]7186: Problems with comparing or 20646 subtracting various types of const and non-const iterators 20647 * [239]7216: ambiguity with basic_iostream::traits_type 20648 * [240]7220: problem with basic_istream::ignore(0,delimiter) 20649 * [241]7222: locale::operator==() doesn't work on std::locale("") 20650 * [242]7286: placement operator delete issue 20651 * [243]7442: cxxabi.h does not match the C++ ABI 20652 * [244]7445: poor performance of std::locale::classic() in 20653 multi-threaded applications 20654 20655 x86-64 specific 20656 20657 * [245]7291: off-by-one in generated inline bzero code for x86-64 20658 20659 20660 For questions related to the use of GCC, please consult these web 20661 pages and the [246]GCC manuals. If that fails, the 20662 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20663 web pages and the development of GCC are welcome on our developer 20664 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 20665 archives. 20666 20667 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 20668 distribution of this entire article is permitted in any medium, 20669 provided this notice is preserved. 20670 20671 These pages are [251]maintained by the GCC team. Last modified 20672 2022-10-26. 20673 20674 References 20675 20676 1. https://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 20677 2. https://gcc.gnu.org/gcc-3.1/changes.html 20678 3. https://gcc.gnu.org/gcc-3.2/c++-abi.html 20679 4. https://gcc.gnu.org/PR3782 20680 5. https://gcc.gnu.org/PR6440 20681 6. https://gcc.gnu.org/PR7050 20682 7. https://gcc.gnu.org/PR7741 20683 8. https://gcc.gnu.org/PR7982 20684 9. https://gcc.gnu.org/PR8068 20685 10. https://gcc.gnu.org/PR8178 20686 11. https://gcc.gnu.org/PR8396 20687 12. https://gcc.gnu.org/PR8674 20688 13. https://gcc.gnu.org/PR9768 20689 14. https://gcc.gnu.org/PR9798 20690 15. https://gcc.gnu.org/PR9799 20691 16. https://gcc.gnu.org/PR9928 20692 17. https://gcc.gnu.org/PR10114 20693 18. https://gcc.gnu.org/PR10352 20694 19. https://gcc.gnu.org/PR10336 20695 20. https://gcc.gnu.org/PR8224 20696 21. https://gcc.gnu.org/PR8613 20697 22. https://gcc.gnu.org/PR8828 20698 23. https://gcc.gnu.org/PR9226 20699 24. https://gcc.gnu.org/PR9853 20700 25. https://gcc.gnu.org/PR9797 20701 26. https://gcc.gnu.org/PR9967 20702 27. https://gcc.gnu.org/PR10116 20703 28. https://gcc.gnu.org/PR10171 20704 29. https://gcc.gnu.org/PR10175 20705 30. https://gcc.gnu.org/PR8316 20706 31. https://gcc.gnu.org/PR9169 20707 32. https://gcc.gnu.org/PR9420 20708 33. https://gcc.gnu.org/PR9459 20709 34. https://gcc.gnu.org/PR9507 20710 35. https://gcc.gnu.org/PR9538 20711 36. https://gcc.gnu.org/PR9602 20712 37. https://gcc.gnu.org/PR9993 20713 38. https://gcc.gnu.org/PR10167 20714 39. https://gcc.gnu.org/PR9652 20715 40. https://gcc.gnu.org/PR10144 20716 41. https://gcc.gnu.org/PR8746 20717 42. https://gcc.gnu.org/PR9888 20718 43. https://gcc.gnu.org/PR9638 20719 44. https://gcc.gnu.org/PR9954 20720 45. https://gcc.gnu.org/PR7784 20721 46. https://gcc.gnu.org/PR7796 20722 47. https://gcc.gnu.org/PR8281 20723 48. https://gcc.gnu.org/PR8366 20724 49. https://gcc.gnu.org/PR8726 20725 50. https://gcc.gnu.org/PR9414 20726 51. https://gcc.gnu.org/PR10067 20727 52. https://gcc.gnu.org/PR7248 20728 53. https://gcc.gnu.org/PR8343 20729 54. https://gcc.gnu.org/PR9732 20730 55. https://gcc.gnu.org/PR10073 20731 56. https://gcc.gnu.org/PR7702 20732 57. https://gcc.gnu.org/PR9671 20733 58. https://gcc.gnu.org/PR8694 20734 59. https://gcc.gnu.org/PR9953 20735 60. https://gcc.gnu.org/PR10271 20736 61. https://gcc.gnu.org/PR6362 20737 62. https://gcc.gnu.org/PR10377 20738 63. https://gcc.gnu.org/PR6955 20739 64. https://gcc.gnu.org/PR5919 20740 65. https://gcc.gnu.org/PR7129 20741 66. https://gcc.gnu.org/PR7507 20742 67. https://gcc.gnu.org/PR7622 20743 68. https://gcc.gnu.org/PR7681 20744 69. https://gcc.gnu.org/PR9528 20745 70. https://gcc.gnu.org/PR8031 20746 71. https://gcc.gnu.org/PR8275 20747 72. https://gcc.gnu.org/PR8332 20748 73. https://gcc.gnu.org/PR8372 20749 74. https://gcc.gnu.org/PR8439 20750 75. https://gcc.gnu.org/PR8442 20751 76. https://gcc.gnu.org/PR8518 20752 77. https://gcc.gnu.org/PR8615 20753 78. https://gcc.gnu.org/PR8663 20754 79. https://gcc.gnu.org/PR8799 20755 80. https://gcc.gnu.org/PR9328 20756 81. https://gcc.gnu.org/PR9465 20757 82. https://gcc.gnu.org/PR47 20758 83. https://gcc.gnu.org/PR6745 20759 84. https://gcc.gnu.org/PR8214 20760 85. https://gcc.gnu.org/PR8493 20761 86. https://gcc.gnu.org/PR8332 20762 87. https://gcc.gnu.org/PR8503 20763 88. https://gcc.gnu.org/PR8727 20764 89. https://gcc.gnu.org/PR7445 20765 90. https://gcc.gnu.org/PR8230 20766 91. https://gcc.gnu.org/PR8399 20767 92. https://gcc.gnu.org/PR8662 20768 93. https://gcc.gnu.org/PR8707 20769 94. https://gcc.gnu.org/PR8708 20770 95. https://gcc.gnu.org/PR8790 20771 96. https://gcc.gnu.org/PR8887 20772 97. https://gcc.gnu.org/PR9076 20773 98. https://gcc.gnu.org/PR9151 20774 99. https://gcc.gnu.org/PR9168 20775 100. https://gcc.gnu.org/PR9269 20776 101. https://gcc.gnu.org/PR9322 20777 102. https://gcc.gnu.org/PR9433 20778 103. https://gcc.gnu.org/PR8032 20779 104. https://gcc.gnu.org/PR8639 20780 105. https://gcc.gnu.org/PR8794 20781 106. https://gcc.gnu.org/PR8832 20782 107. https://gcc.gnu.org/PR8988 20783 108. https://gcc.gnu.org/PR9492 20784 109. https://gcc.gnu.org/PR9267 20785 110. https://gcc.gnu.org/PR8344 20786 111. https://gcc.gnu.org/PR8524 20787 112. https://gcc.gnu.org/PR8880 20788 113. https://gcc.gnu.org/PR9090 20789 114. https://gcc.gnu.org/PR8588 20790 115. https://gcc.gnu.org/PR8599 20791 116. https://gcc.gnu.org/PR9506 20792 117. https://gcc.gnu.org/PR9484 20793 118. https://gcc.gnu.org/PR9292 20794 119. https://gcc.gnu.org/PR9293 20795 120. https://gcc.gnu.org/PR9295 20796 121. https://gcc.gnu.org/PR9296 20797 122. https://gcc.gnu.org/PR9316 20798 123. https://gcc.gnu.org/PR9493 20799 124. https://gcc.gnu.org/PR7341 20800 125. https://gcc.gnu.org/PR8947 20801 126. https://gcc.gnu.org/PR7448 20802 127. https://gcc.gnu.org/PR8882 20803 128. https://gcc.gnu.org/PR7445 20804 129. https://gcc.gnu.org/PR2521 20805 130. https://gcc.gnu.org/PR5661 20806 131. https://gcc.gnu.org/PR6419 20807 132. https://gcc.gnu.org/PR6994 20808 133. https://gcc.gnu.org/PR7150 20809 134. https://gcc.gnu.org/PR7160 20810 135. https://gcc.gnu.org/PR7228 20811 136. https://gcc.gnu.org/PR7266 20812 137. https://gcc.gnu.org/PR7353 20813 138. https://gcc.gnu.org/PR7411 20814 139. https://gcc.gnu.org/PR7478 20815 140. https://gcc.gnu.org/PR7526 20816 141. https://gcc.gnu.org/PR7721 20817 142. https://gcc.gnu.org/PR7803 20818 143. https://gcc.gnu.org/PR7754 20819 144. https://gcc.gnu.org/PR7788 20820 145. https://gcc.gnu.org/PR8031 20821 146. https://gcc.gnu.org/PR8055 20822 147. https://gcc.gnu.org/PR8067 20823 148. https://gcc.gnu.org/PR8134 20824 149. https://gcc.gnu.org/PR8149 20825 150. https://gcc.gnu.org/PR8160 20826 151. https://gcc.gnu.org/PR5607 20827 152. https://gcc.gnu.org/PR6579 20828 153. https://gcc.gnu.org/PR6803 20829 154. https://gcc.gnu.org/PR7176 20830 155. https://gcc.gnu.org/PR7188 20831 156. https://gcc.gnu.org/PR7306 20832 157. https://gcc.gnu.org/PR7461 20833 158. https://gcc.gnu.org/PR7524 20834 159. https://gcc.gnu.org/PR7584 20835 160. https://gcc.gnu.org/PR7676 20836 161. https://gcc.gnu.org/PR7679 20837 162. https://gcc.gnu.org/PR7811 20838 163. https://gcc.gnu.org/PR7961 20839 164. https://gcc.gnu.org/PR8071 20840 165. https://gcc.gnu.org/PR8127 20841 166. https://gcc.gnu.org/PR6745 20842 167. https://gcc.gnu.org/PR8096 20843 168. https://gcc.gnu.org/PR8127 20844 169. https://gcc.gnu.org/PR8218 20845 170. https://gcc.gnu.org/PR8287 20846 171. https://gcc.gnu.org/PR8347 20847 172. https://gcc.gnu.org/PR8348 20848 173. https://gcc.gnu.org/PR8391 20849 174. https://gcc.gnu.org/PR6627 20850 175. https://gcc.gnu.org/PR6631 20851 176. https://gcc.gnu.org/PR7102 20852 177. https://gcc.gnu.org/PR7120 20853 178. https://gcc.gnu.org/PR7209 20854 179. https://gcc.gnu.org/PR7515 20855 180. https://gcc.gnu.org/PR7814 20856 181. https://gcc.gnu.org/PR8467 20857 182. https://gcc.gnu.org/PR4890 20858 183. https://gcc.gnu.org/PR7357 20859 184. https://gcc.gnu.org/PR7358 20860 185. https://gcc.gnu.org/PR7602 20861 186. https://gcc.gnu.org/PR7862 20862 187. https://gcc.gnu.org/PR8190 20863 188. https://gcc.gnu.org/PR8524 20864 189. https://gcc.gnu.org/PR5351 20865 190. https://gcc.gnu.org/PR7591 20866 191. https://gcc.gnu.org/PR6845 20867 192. https://gcc.gnu.org/PR7034 20868 193. https://gcc.gnu.org/PR7124 20869 194. https://gcc.gnu.org/PR7174 20870 195. https://gcc.gnu.org/PR7134 20871 196. https://gcc.gnu.org/PR7375 20872 197. https://gcc.gnu.org/PR7390 20873 198. https://gcc.gnu.org/PR6890 20874 199. https://gcc.gnu.org/PR6981 20875 200. https://gcc.gnu.org/PR7242 20876 201. https://gcc.gnu.org/PR7396 20877 202. https://gcc.gnu.org/PR7630 20878 203. https://gcc.gnu.org/PR7693 20879 204. https://gcc.gnu.org/PR7723 20880 205. https://gcc.gnu.org/PR7951 20881 206. https://gcc.gnu.org/PR8146 20882 207. https://gcc.gnu.org/PR5967 20883 208. https://gcc.gnu.org/PR6984 20884 209. https://gcc.gnu.org/PR7114 20885 210. https://gcc.gnu.org/PR7130 20886 211. https://gcc.gnu.org/PR7133 20887 212. https://gcc.gnu.org/PR7380 20888 213. https://gcc.gnu.org/PR8252 20889 214. https://gcc.gnu.org/PR8451 20890 215. https://gcc.gnu.org/PR7250 20891 216. https://gcc.gnu.org/PR6668 20892 217. https://gcc.gnu.org/PR7151 20893 218. https://gcc.gnu.org/PR7335 20894 219. https://gcc.gnu.org/PR7842 20895 220. https://gcc.gnu.org/PR7856 20896 221. https://gcc.gnu.org/PR7967 20897 222. https://gcc.gnu.org/PR7374 20898 223. https://gcc.gnu.org/PR7370 20899 224. https://gcc.gnu.org/PR7409 20900 225. https://gcc.gnu.org/PR8232 20901 226. https://gcc.gnu.org/PR7623 20902 227. https://gcc.gnu.org/PR8314 20903 228. https://gcc.gnu.org/PR761 20904 229. https://gcc.gnu.org/PR5610 20905 230. https://gcc.gnu.org/PR7484 20906 231. https://gcc.gnu.org/PR7531 20907 232. https://gcc.gnu.org/PR8120 20908 233. https://gcc.gnu.org/PR7320 20909 234. https://gcc.gnu.org/PR7470 20910 235. https://gcc.gnu.org/PR6410 20911 236. https://gcc.gnu.org/PR6503 20912 237. https://gcc.gnu.org/PR6642 20913 238. https://gcc.gnu.org/PR7186 20914 239. https://gcc.gnu.org/PR7216 20915 240. https://gcc.gnu.org/PR7220 20916 241. https://gcc.gnu.org/PR7222 20917 242. https://gcc.gnu.org/PR7286 20918 243. https://gcc.gnu.org/PR7442 20919 244. https://gcc.gnu.org/PR7445 20920 245. https://gcc.gnu.org/PR7291 20921 246. https://gcc.gnu.org/onlinedocs/ 20922 247. mailto:gcc-help (a] gcc.gnu.org 20923 248. mailto:gcc (a] gcc.gnu.org 20924 249. https://gcc.gnu.org/lists.html 20925 250. https://www.fsf.org/ 20926 251. https://gcc.gnu.org/about.html 20927 ====================================================================== 20928 http://gcc.gnu.org/gcc-3.1/index.html 20929 GCC 3.1 20930 20931 (This release series is no longer supported.) 20932 20933 July 27, 2002 20934 20935 The [1]GNU project and the GCC developers are pleased to announce the 20936 release of GCC 3.1.1. 20937 20938 The links below still apply to GCC 3.1.1. 20939 20940 May 15, 2002 20941 20942 The [2]GNU project and the GCC developers are pleased to announce the 20943 release of GCC 3.1. 20944 20945 GCC used to stand for the GNU C Compiler, but since the compiler 20946 supports several other languages aside from C, it now stands for the 20947 GNU Compiler Collection. 20948 20949 A list of [3]successful builds is updated as new information becomes 20950 available. 20951 20952 The GCC developers would like to thank the numerous people that have 20953 contributed [4]new features, improvements, bug fixes, and other changes 20954 as well as test results to GCC. This [5]amazing group of volunteers is 20955 what makes GCC successful. 20956 20957 For additional information about GCC please refer to the [6]GCC project 20958 web site or contact the [7]GCC development mailing list. 20959 20960 To obtain GCC please use [8]our mirror sites, or our CVS server. 20961 __________________________________________________________________ 20962 20963 20964 For questions related to the use of GCC, please consult these web 20965 pages and the [9]GCC manuals. If that fails, the 20966 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 20967 web pages and the development of GCC are welcome on our developer 20968 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 20969 archives. 20970 20971 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 20972 distribution of this entire article is permitted in any medium, 20973 provided this notice is preserved. 20974 20975 These pages are [14]maintained by the GCC team. Last modified 20976 2022-10-26. 20977 20978 References 20979 20980 1. http://www.gnu.org/ 20981 2. http://www.gnu.org/ 20982 3. https://gcc.gnu.org/gcc-3.1/buildstat.html 20983 4. https://gcc.gnu.org/gcc-3.1/changes.html 20984 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20985 6. https://gcc.gnu.org/index.html 20986 7. mailto:gcc (a] gcc.gnu.org 20987 8. https://gcc.gnu.org/mirrors.html 20988 9. https://gcc.gnu.org/onlinedocs/ 20989 10. mailto:gcc-help (a] gcc.gnu.org 20990 11. mailto:gcc (a] gcc.gnu.org 20991 12. https://gcc.gnu.org/lists.html 20992 13. https://www.fsf.org/ 20993 14. https://gcc.gnu.org/about.html 20994 ====================================================================== 20995 http://gcc.gnu.org/gcc-3.1/changes.html 20996 GCC 3.1 Release Series 20997 Changes, New Features, and Fixes 20998 20999 Additional changes in GCC 3.1.1 21000 21001 * A bug related to how structures and unions are returned has been 21002 fixed for powerpc-*-netbsd*. 21003 * An important bug in the implementation of -fprefetch-loop-arrays 21004 has been fixed. Previously the optimization prefetched random 21005 blocks of memory for most targets except for i386. 21006 * The Java compiler now compiles Java programs much faster and also 21007 works with parallel make. 21008 * Nested functions have been fixed for mips*-*-netbsd*. 21009 * Some missing floating point support routines have beed added for 21010 mips*-*-netbsd*. 21011 * This [1]message gives additional information about the bugs fixed 21012 in this release. 21013 21014 Caveats 21015 21016 * The -traditional C compiler option has been deprecated and will be 21017 removed in GCC 3.3. (It remains possible to preprocess non-C code 21018 with the traditional preprocessor.) 21019 * The default debugging format for most ELF platforms (including 21020 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 21021 from stabs to DWARF2. This requires GDB 5.1.1 or later. 21022 21023 General Optimizer Improvements 21024 21025 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 21026 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 21027 for profile driven optimizations. 21028 Options -fprofile-arcs and -fbranch-probabilities can now be used 21029 to improve speed of the generated code by profiling the actual 21030 program behaviour on typical runs. In the absence of profile info 21031 the compiler attempts to guess the profile statically. 21032 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 21033 monitor performance of the generated code. 21034 According to the SPECInt2000 results on an AMD Athlon CPU, the code 21035 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 21036 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 21037 is about 2.1% faster compared to 2.95.3. Tests were done using the 21038 -O2 -march=athlon command-line options. 21039 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 21040 infrastructure developed by CodeSourcery, LLC for the C++ front 21041 end, so that it is now used in the C front end too. Inlining 21042 functions as trees exposes them earlier to the compiler, giving it 21043 more opportunities for optimization. 21044 * Support for data prefetching instructions has been added to the GCC 21045 back end and several targets. A new __builtin_prefetch intrinsic is 21046 available to explicitly insert prefetch instructions and 21047 experimental support for loop array prefetching has been added (see 21048 -fprefetch-loop-array documentation). 21049 * Support for emitting debugging information for macros has been 21050 added for DWARF2. It is activated using -g3. 21051 21052 New Languages and Language specific improvements 21053 21054 C/C++ 21055 21056 * A few more [4]ISO C99 features. 21057 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 21058 * The preprocessor's symbol table has been merged with the symbol 21059 table of the C, C++ and Objective-C front ends. 21060 * The preprocessor consumes less memory than the preprocessor in GCC 21061 3.0, often significantly so. On normal input files, it typically 21062 consumes less memory than pre-3.0 cccp-based GCC, too. 21063 21064 C++ 21065 21066 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 21067 was a workaround to allow std compliant code to work with the 21068 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 21069 * The C++ ABI has been fixed so that void (A::*)() const is mangled 21070 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 21071 affects pointer to cv-qualified member function types. 21072 * The C++ ABI has been changed to correctly handle this code: 21073 struct A { 21074 void operator delete[] (void *, size_t); 21075 }; 21076 21077 struct B : public A { 21078 }; 21079 21080 new B[10]; 21081 21082 The amount of storage allocated for the array will be greater than 21083 it was in 3.0, in order to store the number of elements in the 21084 array, so that the correct size can be passed to operator delete[] 21085 when the array is deleted. Previously, the value passed to operator 21086 delete[] was unpredictable. 21087 This change will only affect code that declares a two-argument 21088 operator delete[] with a second parameter of type size_t in a base 21089 class, and does not override that definition in a derived class. 21090 * The C++ ABI has been changed so that: 21091 struct A { 21092 void operator delete[] (void *, size_t); 21093 void operator delete[] (void *); 21094 }; 21095 21096 does not cause unnecessary storage to be allocated when an array of 21097 A objects is allocated. 21098 This change will only affect code that declares both of these forms 21099 of operator delete[], and declared the two-argument form before the 21100 one-argument form. 21101 * The C++ ABI has been changed so that when a parameter is passed by 21102 value, any cleanup for that parameter is performed in the caller, 21103 as specified by the ia64 C++ ABI, rather than the called function 21104 as before. As a result, classes with a non-trivial destructor but a 21105 trivial copy constructor will be passed and returned by invisible 21106 reference, rather than by bitwise copy as before. 21107 * G++ now supports the "named return value optimization": for code 21108 like 21109 A f () { 21110 A a; 21111 ... 21112 return a; 21113 } 21114 21115 G++ will allocate a in the return value slot, so that the return 21116 becomes a no-op. For this to work, all return statements in the 21117 function must return the same variable. 21118 * Improvements to the C++ library are listed in [5]the libstdc++-v3 21119 FAQ. 21120 21121 Objective-C 21122 21123 * Annoying linker warnings (due to incorrect code being generated) 21124 have been fixed. 21125 * If a class method cannot be found, the compiler no longer issues a 21126 warning if a corresponding instance method exists in the root 21127 class. 21128 * Forward @protocol declarations have been fixed. 21129 * Loading of categories has been fixed in certain situations (GNU run 21130 time only). 21131 * The class lookup in the run-time library has been rewritten so that 21132 class method dispatch is more than twice as fast as it used to be 21133 (GNU run time only). 21134 21135 Java 21136 21137 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 21138 javax.transaction. 21139 * Property files and other system resources can be compiled into 21140 executables which use libgcj using the new gcj --resource feature. 21141 * libgcj has been ported to more platforms. In particular there is 21142 now a mostly-functional mingw32 (Windows) target port. 21143 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 21144 Java code can now be called from a C/C++ application. 21145 * gcj can now use builtin functions for certain known methods, for 21146 instance Math.cos. 21147 * gcj can now automatically remove redundant array-store checks in 21148 some common cases. 21149 * The --no-store-checks optimization option was added. This can be 21150 used to omit runtime store checks for code which is known not to 21151 throw ArrayStoreException 21152 * The following third party interface standards were added to libgcj: 21153 org.w3c.dom and org.xml.sax. 21154 * java.security has been merged with GNU Classpath. The new package 21155 is now JDK 1.2 compliant, and much more complete. 21156 * A bytecode verifier was added to the libgcj interpreter. 21157 * java.lang.Character was rewritten to comply with the Unicode 3.0 21158 standard, and improve performance. 21159 * Partial support for many more locales was added to libgcj. 21160 * Socket timeouts have been implemented. 21161 * libgcj has been merged into a single shared library. There are no 21162 longer separate shared libraries for the garbage collector and 21163 zlib. 21164 * Several performance improvements were made to gcj and libgcj: 21165 + Hash synchronization (thin locks) 21166 + A special allocation path for finalizer-free objects 21167 + Thread-local allocation 21168 + Parallel GC, and other GC tweaks 21169 21170 Fortran 21171 21172 Fortran improvements are listed in [6]the Fortran documentation. 21173 21174 Ada 21175 21176 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 21177 tools. The GNAT compiler fully implements the Ada language as defined 21178 by the ISO/IEC 8652 standard. 21179 21180 Please note that the integration of the Ada front end is still work in 21181 progress. 21182 21183 New Targets and Target Specific Improvements 21184 21185 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 21186 architecture used in new editions of Donald E. Knuth's The Art of 21187 Computer Programming. 21188 * Axis Communications has contributed its port to the CRIS CPU 21189 architecture, used in the ETRAX system-on-a-chip series. 21190 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 21191 SH5 64-bit RISC microprocessor architecture, extending the existing 21192 SH port. 21193 * UltraSPARC is fully supported in 64-bit mode. The option -m64 21194 enables it. 21195 * For compatibility with the Sun compiler #pragma redefine_extname 21196 has been implemented on Solaris. 21197 * The x86 back end has had some noticeable work done to it. 21198 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 21199 Jaeger have contributed a port to the AMD x86-64 architecture. 21200 For more information on x86-64 see http://www.x86-64.org. 21201 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 21202 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 21203 enable the respective instruction sets. Intel C++ compatible 21204 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 21205 will be added in next major release. 21206 + Following those improvements, targets for Pentium MMX, K6-2, 21207 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 21208 added. Refer to the documentation on -march= and -mcpu= 21209 options for details. 21210 + For those targets that support it, -mfpmath=sse will cause the 21211 compiler to generate SSE/SSE2 instructions for floating point 21212 math instead of x87 instructions. Usually, this will lead to 21213 quicker code especially on the Pentium 4. Note that only 21214 scalar floating point instructions are used and GCC does not 21215 exploit SIMD features yet. 21216 + Prefetch support has been added to the Pentium III, Pentium 4, 21217 K6-2, K6-3, and Athlon series. 21218 + Code generated for floating point to integer conversions has 21219 been improved leading to better performance of many 3D 21220 applications. 21221 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 21222 * C++ support for AIX has been improved. 21223 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 21224 PowerPC port supporting the AltiVec programming model (SIMD). The 21225 support, though presently useful, is experimental and is expected 21226 to stabilize for 3.2. The support is written to conform to 21227 Motorola's AltiVec specs. See -maltivec. 21228 21229 Obsolete Systems 21230 21231 Support for a number of older systems has been declared obsolete in GCC 21232 3.1. Unless there is activity to revive them, the next release of GCC 21233 will have their sources permanently removed. 21234 21235 All configurations of the following processor architectures have been 21236 declared obsolete: 21237 * MIL-STD-1750A, 1750a-*-* 21238 * AMD A29k, a29k-*-* 21239 * Convex, c*-convex-* 21240 * Clipper, clipper-*-* 21241 * Elxsi, elxsi-*-* 21242 * Intel i860, i860-*-* 21243 * Sun picoJava, pj-*-* and pjl-*-* 21244 * Western Electric 32000, we32k-*-* 21245 21246 Most configurations of the following processor architectures have been 21247 declared obsolete, but we are preserving a few systems which may have 21248 active developers. It is unlikely that the remaining systems will 21249 survive much longer unless we see definite signs of port activity. 21250 * Motorola 88000 except 21251 + Generic a.out, m88k-*-aout* 21252 + Generic SVR4, m88k-*-sysv4 21253 + OpenBSD, m88k-*-openbsd* 21254 * NS32k except 21255 + NetBSD, ns32k-*-netbsd* 21256 + OpenBSD, ns32k-*-openbsd*. 21257 * ROMP except 21258 + OpenBSD, romp-*-openbsd*. 21259 21260 Finally, only some configurations of these processor architectures are 21261 being obsoleted. 21262 * Alpha: 21263 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 21264 alpha*-*-osf[45], are still supported.) 21265 * ARM: 21266 + RISCiX, arm-*-riscix*. 21267 * i386: 21268 + 386BSD, i?86-*-bsd* 21269 + Chorus, i?86-*-chorusos* 21270 + DG/UX, i?86-*-dgux* 21271 + FreeBSD 1.x, i?86-*-freebsd1.* 21272 + IBM AIX, i?86-*-aix* 21273 + ISC UNIX, i?86-*-isc* 21274 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 21275 + NEXTstep, i?86-next-* 21276 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 21277 + RTEMS/coff, i?86-*-rtemscoff* 21278 + RTEMS/go32, i?86-go32-rtems* 21279 + Sequent/BSD, i?86-sequent-bsd* 21280 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 21281 i?86-sequent-sysv3* 21282 + SunOS, i?86-*-sunos* 21283 * Motorola 68000: 21284 + Altos, m68[k0]*-altos-* 21285 + Apollo, m68[k0]*-apollo-* 21286 + Apple A/UX, m68[k0]*-apple-* 21287 + Bull, m68[k0]*-bull-* 21288 + Convergent, m68[k0]*-convergent-* 21289 + Generic SVR3, m68[k0]*-*-sysv3* 21290 + ISI, m68[k0]*-isi-* 21291 + LynxOS, m68[k0]*-*-lynxos* 21292 + NEXT, m68[k0]*-next-* 21293 + RTEMS/coff, m68[k0]*-*-rtemscoff* 21294 + Sony, m68[k0]*-sony-* 21295 * MIPS: 21296 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 21297 + Generic BSD, mips-*-bsd* 21298 + Generic System V, mips-*-sysv* 21299 + IRIX before version 5, mips-sgi-irix[1234]* 21300 + RiscOS, mips-*-riscos* 21301 + Sony, mips-sony-* 21302 + Tandem, mips-tandem-* 21303 * SPARC: 21304 + RTEMS/a.out, sparc-*-rtemsaout*. 21305 21306 Documentation improvements 21307 21308 * The old manual ("Using and Porting the GNU Compiler Collection") 21309 has been replaced by a users manual ("Using the GNU Compiler 21310 Collection") and a separate internals reference manual ("GNU 21311 Compiler Collection Internals"). 21312 * More complete and much improved documentation about GCC's internal 21313 representation used by the C and C++ front ends. 21314 * Many cleanups and improvements in general. 21315 21316 21317 For questions related to the use of GCC, please consult these web 21318 pages and the [8]GCC manuals. If that fails, the 21319 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21320 web pages and the development of GCC are welcome on our developer 21321 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 21322 archives. 21323 21324 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 21325 distribution of this entire article is permitted in any medium, 21326 provided this notice is preserved. 21327 21328 These pages are [13]maintained by the GCC team. Last modified 21329 2022-10-26. 21330 21331 References 21332 21333 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 21334 2. https://gcc.gnu.org/news/profiledriven.html 21335 3. https://gcc.gnu.org/benchmarks/ 21336 4. https://gcc.gnu.org/c99status.html 21337 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 21338 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 21339 7. https://www.adacore.com/ 21340 8. https://gcc.gnu.org/onlinedocs/ 21341 9. mailto:gcc-help (a] gcc.gnu.org 21342 10. mailto:gcc (a] gcc.gnu.org 21343 11. https://gcc.gnu.org/lists.html 21344 12. https://www.fsf.org/ 21345 13. https://gcc.gnu.org/about.html 21346 ====================================================================== 21347 http://gcc.gnu.org/gcc-3.0/index.html 21348 GCC 3.0.4 21349 21350 (This release series is no longer supported.) 21351 21352 February 20, 2002 21353 21354 The [1]GNU project and the GCC developers are pleased to announce the 21355 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 21356 series. 21357 21358 GCC used to stand for the GNU C Compiler, but since the compiler 21359 supports several other languages aside from C, it now stands for the 21360 GNU Compiler Collection. 21361 21362 GCC 3.0.x has several new optimizations, new targets, new languages and 21363 many other new features, relative to GCC 2.95.x. See the [2]new 21364 features page for a more complete list. 21365 21366 A list of [3]successful builds is updated as new information becomes 21367 available. 21368 21369 The GCC developers would like to thank the numerous people that have 21370 contributed new features, test results, bug fixes, etc to GCC. This 21371 [4]amazing group of volunteers is what makes GCC successful. 21372 21373 And finally, we can't in good conscience fail to mention some 21374 [5]caveats to using GCC 3.0.x. 21375 21376 For additional information about GCC please refer to the [6]GCC project 21377 web site or contact the [7]GCC development mailing list. 21378 21379 To obtain GCC please use [8]our mirror sites, or our CVS server. 21380 __________________________________________________________________ 21381 21382 Previous 3.0.x Releases 21383 21384 December 20, 2001: GCC 3.0.3 has been released. 21385 October 25, 2001: GCC 3.0.2 has been released. 21386 August 20, 2001: GCC 3.0.1 has been released. 21387 June 18, 2001: GCC 3.0 has been released. 21388 21389 21390 For questions related to the use of GCC, please consult these web 21391 pages and the [9]GCC manuals. If that fails, the 21392 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21393 web pages and the development of GCC are welcome on our developer 21394 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 21395 archives. 21396 21397 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 21398 distribution of this entire article is permitted in any medium, 21399 provided this notice is preserved. 21400 21401 These pages are [14]maintained by the GCC team. Last modified 21402 2022-10-26. 21403 21404 References 21405 21406 1. http://www.gnu.org/ 21407 2. https://gcc.gnu.org/gcc-3.0/features.html 21408 3. https://gcc.gnu.org/gcc-3.0/buildstat.html 21409 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 21410 5. https://gcc.gnu.org/gcc-3.0/caveats.html 21411 6. https://gcc.gnu.org/index.html 21412 7. mailto:gcc (a] gcc.gnu.org 21413 8. https://gcc.gnu.org/mirrors.html 21414 9. https://gcc.gnu.org/onlinedocs/ 21415 10. mailto:gcc-help (a] gcc.gnu.org 21416 11. mailto:gcc (a] gcc.gnu.org 21417 12. https://gcc.gnu.org/lists.html 21418 13. https://www.fsf.org/ 21419 14. https://gcc.gnu.org/about.html 21420 ====================================================================== 21421 http://gcc.gnu.org/gcc-3.0/features.html 21422 GCC 3.0 New Features 21423 21424 Additional changes in GCC 3.0.4 21425 21426 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 21427 system, which use the ELF object file format, on x86 processors. 21428 * Correct debugging information is generated from functions that have 21429 lines from multiple files (e.g. yacc output). 21430 * A fix for whitespace handling in the -traditional preprocessor, 21431 which can affect Fortran. 21432 * Fixes to the exception handling runtime. 21433 * More fixes for bad code generation in C++. 21434 * A fix for shared library generation under AIX 4.3. 21435 * Documentation updates. 21436 * Port of GCC to Tensilica's Xtensa processor contributed. 21437 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 21438 21439 Additional changes in GCC 3.0.3 21440 21441 * A fix to correct an accidental change to the PowerPC ABI. 21442 * Fixes for bad code generation on a variety of architectures. 21443 * Improvements to the debugging information generated for C++ 21444 classes. 21445 * Fixes for bad code generation in C++. 21446 * A fix to avoid crashes in the C++ demangler. 21447 * A fix to the C++ standard library to avoid buffer overflows. 21448 * Miscellaneous improvements for a variety of architectures. 21449 21450 Additional changes in GCC 3.0.2 21451 21452 * Fixes for bad code generation during loop unrolling. 21453 * Fixes for bad code generation by the sibling call optimization. 21454 * Minor improvements to x86 code generation. 21455 * Implementation of function descriptors in C++ vtables for IA64. 21456 * Numerous minor bug-fixes. 21457 21458 Additional changes in GCC 3.0.1 21459 21460 * C++ fixes for incorrect code-generation. 21461 * Improved cross-compiling support for the C++ standard library. 21462 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 21463 in GCC 3.0. 21464 * Fixes for various exception-handling bugs. 21465 * A port to the S/390 architecture. 21466 21467 General Optimizer Improvements 21468 21469 * [2]Basic block reordering pass. 21470 * New if-conversion pass with support for conditional (predicated) 21471 execution. 21472 * New tail call and sibling call elimination optimizations. 21473 * New register renaming pass. 21474 * New (experimental) [3]static single assignment (SSA) representation 21475 support. 21476 * New dead-code elimination pass implemented using the SSA 21477 representation. 21478 * [4]Global null pointer test elimination. 21479 * [5]Global code hoisting/unification. 21480 * More builtins and optimizations for stdio.h, string.h and old BSD 21481 functions, as well as for ISO C99 functions. 21482 * New builtin __builtin_expect for giving hints to the branch 21483 predictor. 21484 21485 New Languages and Language specific improvements 21486 21487 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 21488 and supported, including the run-time library containing most 21489 common non-GUI Java classes, a bytecode interpreter, and the Boehm 21490 conservative garbage collector. Many bugs have been fixed. GCJ can 21491 compile Java source or Java bytecodes to either native code or Java 21492 class files, and supports native methods written in either the 21493 standard JNI or the more efficient and convenient CNI. 21494 * Here is a [6]partial list of C++ improvements, both new features 21495 and those no longer supported. 21496 * New C++ ABI. On the IA-64 platform GCC is capable of 21497 inter-operating with other IA-64 compilers. 21498 * The new ABI also significantly reduces the size of symbol and debug 21499 information. 21500 * New C++ support library and many C++ bug fixes, vastly improving 21501 our conformance to the ISO C++ standard. 21502 * New [7]inliner for C++. 21503 * Rewritten C preprocessor, integrated into the C, C++ and Objective 21504 C compilers, with very many improvements including ISO C99 support 21505 and [8]improvements to dependency generation. 21506 * Support for more [9]ISO C99 features. 21507 * Many improvements to support for checking calls to format functions 21508 such as printf and scanf, including support for ISO C99 format 21509 features, extensions from the Single Unix Specification and GNU 21510 libc 2.2, checking of strfmon formats and features to assist in 21511 auditing for format string security bugs. 21512 * New warnings for C code that may have undefined semantics because 21513 of violations of sequence point rules in the C standard (such as a 21514 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 21515 * Additional warning option -Wfloat-equal. 21516 * Improvements to -Wtraditional. 21517 * Fortran improvements are listed in [10]the Fortran documentation. 21518 21519 New Targets and Target Specific Improvements 21520 21521 * New x86 back end, generating much improved code. 21522 * Support for a generic i386-elf target contributed. 21523 * New option to emit x86 assembly code using Intel style syntax 21524 (-mintel-syntax). 21525 * HPUX 11 support contributed. 21526 * Improved PowerPC code generation, including scheduled prologue and 21527 epilogue. 21528 * Port of GCC to Intel's IA-64 processor contributed. 21529 * Port of GCC to Motorola's MCore 210 and 340 contributed. 21530 * New unified back-end for Arm, Thumb and StrongArm contributed. 21531 * Port of GCC to Intel's XScale processor contributed. 21532 * Port of GCC to Atmel's AVR microcontrollers contributed. 21533 * Port of GCC to Mitsubishi's D30V processor contributed. 21534 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 21535 processor family) contributed. 21536 * Port of GCC to Fujitsu's FR30 processor contributed. 21537 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 21538 * Port of GCC to Sun's picoJava processor core contributed. 21539 21540 Documentation improvements 21541 21542 * Substantially rewritten and improved C preprocessor manual. 21543 * Many improvements to other documentation. 21544 * Manpages for gcc, cpp and gcov are now generated automatically from 21545 the master Texinfo manual, eliminating the problem of manpages 21546 being out of date. (The generated manpages are only extracts from 21547 the full manual, which is provided in Texinfo form, from which 21548 info, HTML, other formats and a printed manual can be generated.) 21549 * Generated info files are included in the release tarballs alongside 21550 their Texinfo sources, avoiding problems on some platforms with 21551 building makeinfo as part of the GCC distribution. 21552 21553 Other significant improvements 21554 21555 * Garbage collection used internally by the compiler for most memory 21556 allocation instead of obstacks. 21557 * Lengauer and Tarjan algorithm used for computing dominators in the 21558 CFG. This algorithm can be significantly faster and more space 21559 efficient than our older algorithm. 21560 * gccbug script provided to assist in submitting bug reports to our 21561 bug tracking system. (Bug reports previously submitted directly to 21562 our mailing lists, for which you received no bug tracking number, 21563 should be submitted again using gccbug if you can reproduce the 21564 problem with GCC 3.0.) 21565 * The internal libgcc library is [11]built as a shared library on 21566 systems that support it. 21567 * Extensive testsuite included with GCC, with many new tests. In 21568 addition to tests for GCC bugs that have been fixed, many tests 21569 have been added for language features, compiler warnings and 21570 builtin functions. 21571 * Additional language-independent warning options -Wpacked, -Wpadded, 21572 -Wunreachable-code and -Wdisabled-optimization. 21573 * Target-independent options -falign-functions, -falign-loops and 21574 -falign-jumps. 21575 21576 Plus a great many bug fixes and almost all the [12]features found in 21577 GCC 2.95. 21578 21579 21580 For questions related to the use of GCC, please consult these web 21581 pages and the [13]GCC manuals. If that fails, the 21582 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21583 web pages and the development of GCC are welcome on our developer 21584 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 21585 archives. 21586 21587 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 21588 distribution of this entire article is permitted in any medium, 21589 provided this notice is preserved. 21590 21591 These pages are [18]maintained by the GCC team. Last modified 21592 2022-10-26. 21593 21594 References 21595 21596 1. http://www.netbsd.org/ 21597 2. https://gcc.gnu.org/news/reorder.html 21598 3. https://gcc.gnu.org/news/ssa.html 21599 4. https://gcc.gnu.org/news/null.html 21600 5. https://gcc.gnu.org/news/unify.html 21601 6. https://gcc.gnu.org/gcc-3.0/c++features.html 21602 7. https://gcc.gnu.org/news/inlining.html 21603 8. https://gcc.gnu.org/news/dependencies.html 21604 9. https://gcc.gnu.org/c99status.html 21605 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 21606 11. https://gcc.gnu.org/gcc-3.0/libgcc.html 21607 12. https://gcc.gnu.org/gcc-2.95/features.html 21608 13. https://gcc.gnu.org/onlinedocs/ 21609 14. mailto:gcc-help (a] gcc.gnu.org 21610 15. mailto:gcc (a] gcc.gnu.org 21611 16. https://gcc.gnu.org/lists.html 21612 17. https://www.fsf.org/ 21613 18. https://gcc.gnu.org/about.html 21614 ====================================================================== 21615 http://gcc.gnu.org/gcc-3.0/caveats.html 21616 GCC 3.0 Caveats 21617 21618 * -fstrict-aliasing is now part of -O2 and higher optimization 21619 levels. This allows the compiler to assume the strictest aliasing 21620 rules applicable to the language being compiled. For C and C++, 21621 this activates optimizations based on the type of expressions. This 21622 optimization may thus break old, non-compliant code. 21623 * Enumerations are now properly promoted to int in function 21624 parameters and function returns. Normally this change is not 21625 visible, but when using -fshort-enums this is an ABI change. 21626 * The undocumented extension that allowed C programs to have a label 21627 at the end of a compound statement has been deprecated and may be 21628 removed in a future version. Programs that now generate a warning 21629 about this may be fixed by adding a null statement (a single 21630 semicolon) after the label. 21631 * The poorly documented extension that allowed string constants in C, 21632 C++ and Objective C to contain unescaped newlines has been 21633 deprecated and may be removed in a future version. Programs using 21634 this extension may be fixed in several ways: the bare newline may 21635 be replaced by \n, or preceded by \n\, or string concatenation may 21636 be used with the bare newline preceded by \n" and " placed at the 21637 start of the next line. 21638 * The Chill compiler is not included in GCC 3.0, because of the lack 21639 of a volunteer to convert it to use garbage collection. 21640 * Certain non-standard iostream methods from earlier versions of 21641 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 21642 ostream::form, and istream::gets. 21643 * The new C++ ABI is not yet fully supported by current (as of 21644 2001-07-01) releases and development versions of GDB, or any 21645 earlier versions. There is a problem setting breakpoints by line 21646 number, and other related issues that have been fixed in GCC 3.0 21647 but not yet handled in GDB: 21648 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 21649 21650 21651 For questions related to the use of GCC, please consult these web 21652 pages and the [2]GCC manuals. If that fails, the 21653 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21654 web pages and the development of GCC are welcome on our developer 21655 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 21656 21657 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 21658 distribution of this entire article is permitted in any medium, 21659 provided this notice is preserved. 21660 21661 These pages are [7]maintained by the GCC team. Last modified 21662 2022-10-26. 21663 21664 References 21665 21666 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 21667 2. https://gcc.gnu.org/onlinedocs/ 21668 3. mailto:gcc-help (a] gcc.gnu.org 21669 4. mailto:gcc (a] gcc.gnu.org 21670 5. https://gcc.gnu.org/lists.html 21671 6. https://www.fsf.org/ 21672 7. https://gcc.gnu.org/about.html 21673 ====================================================================== 21674 http://gcc.gnu.org/gcc-2.95/index.html 21675 GCC 2.95 21676 21677 (This release series is no longer supported.) 21678 21679 March 16, 2001: The GNU project and the GCC developers are pleased to 21680 announce the release of GCC version 2.95.3. 21681 21682 Release History 21683 21684 GCC 2.95.3 21685 March 16, 2001 21686 21687 GCC 2.95.2 21688 October 27, 1999 21689 21690 GCC 2.95.1 21691 August 19, 1999 21692 21693 GCC 2.95 21694 July 31, 1999. This is the first release of GCC since the April 21695 1999 GCC/EGCS reunification and includes nearly a year's worth 21696 of new development and bugfixes. 21697 21698 References and Acknowledgements 21699 21700 GCC used to stand for the GNU C Compiler, but since the compiler 21701 supports several other languages aside from C, it now stands for the 21702 GNU Compiler Collection. 21703 21704 The whole suite has been extensively [1]regression tested and 21705 [2]package tested. It should be reliable and suitable for widespread 21706 use. 21707 21708 The compiler has several new optimizations, new targets, new languages 21709 and other new features. See the [3]new features page for a more 21710 complete list of new features found in the GCC 2.95 releases. 21711 21712 The sources include installation instructions in both HTML and 21713 plaintext forms in the install directory in the distribution. However, 21714 the most up to date installation instructions and [4]build/test status 21715 are on the web pages. We will update those pages as new information 21716 becomes available. 21717 21718 The GCC developers would like to thank the numerous people that have 21719 contributed new features, test results, bugfixes, etc to GCC. This 21720 [5]amazing group of volunteers is what makes GCC successful. 21721 21722 And finally, we can't in good conscience fail to mention some 21723 [6]caveats to using GCC 2.95. 21724 21725 Download GCC 2.95 from one of our many [7]mirror sites. 21726 21727 For additional information about GCC please see the [8]GCC project web 21728 server or contact the [9]GCC development mailing list. 21729 21730 21731 For questions related to the use of GCC, please consult these web 21732 pages and the [10]GCC manuals. If that fails, the 21733 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21734 web pages and the development of GCC are welcome on our developer 21735 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 21736 archives. 21737 21738 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 21739 distribution of this entire article is permitted in any medium, 21740 provided this notice is preserved. 21741 21742 These pages are [15]maintained by the GCC team. Last modified 21743 2022-10-26. 21744 21745 References 21746 21747 1. https://gcc.gnu.org/gcc-2.95/regress.html 21748 2. https://gcc.gnu.org/gcc-2.95/othertest.html 21749 3. https://gcc.gnu.org/gcc-2.95/features.html 21750 4. https://gcc.gnu.org/gcc-2.95/buildstat.html 21751 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 21752 6. https://gcc.gnu.org/gcc-2.95/caveats.html 21753 7. https://gcc.gnu.org/mirrors.html 21754 8. https://gcc.gnu.org/index.html 21755 9. mailto:gcc (a] gcc.gnu.org 21756 10. https://gcc.gnu.org/onlinedocs/ 21757 11. mailto:gcc-help (a] gcc.gnu.org 21758 12. mailto:gcc (a] gcc.gnu.org 21759 13. https://gcc.gnu.org/lists.html 21760 14. https://www.fsf.org/ 21761 15. https://gcc.gnu.org/about.html 21762 ====================================================================== 21763 http://gcc.gnu.org/gcc-2.95/features.html 21764 GCC 2.95 New Features 21765 21766 * General Optimizer Improvements: 21767 + [1]Localized register spilling to improve speed and code 21768 density especially on small register class machines. 21769 + [2]Global CSE using lazy code motion algorithms. 21770 + [3]Improved global constant/copy propagation. 21771 + [4]Improved control flow graph analysis and manipulation. 21772 + [5]Local dead store elimination. 21773 + [6]Memory Load hoisting/store sinking in loops. 21774 + [7]Type based alias analysis is enabled by default. Note this 21775 feature will expose bugs in the Linux kernel. Please refer to 21776 the FAQ (as shipped with GCC 2.95) for additional information 21777 on this issue. 21778 + Major revamp of GIV detection, combination and simplification 21779 to improve loop performance. 21780 + Major improvements to register allocation and reloading. 21781 * New Languages and Language specific improvements 21782 + [8]Many C++ improvements. 21783 + [9]Many Fortran improvements. 21784 + [10]Java front-end has been integrated. A [11]runtime library 21785 is available separately. 21786 + [12]ISO C99 support 21787 + [13]Chill front-end and runtime has been integrated. 21788 + Boehm garbage collector support in libobjc. 21789 + More support for various pragmas which appear in vendor 21790 include files 21791 * New Targets and Target Specific Improvements 21792 + [14]SPARC back end rewrite. 21793 + -mschedule=8000 will optimize code for PA8000 class 21794 processors; -mpa-risc-2-0 will generate code for PA2.0 21795 processors 21796 + Various micro-optimizations for the ia32 port. K6 21797 optimizations 21798 + Compiler will attempt to align doubles in the stack on the 21799 ia32 port 21800 + Alpha EV6 support 21801 + PowerPC 750 21802 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 21803 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 21804 + c3x, c4x 21805 + HyperSPARC 21806 + SparcLite86x 21807 + sh4 21808 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 21809 arm-linux) 21810 + vxWorks targets include support for vxWorks threads 21811 + StrongARM 110 and ARM9 support added. ARM Scheduling 21812 parameters rewritten. 21813 + Various changes to the MIPS port to avoid assembler macros, 21814 which in turn improves performance 21815 + Various performance improvements to the i960 port. 21816 + Major rewrite of ns32k port 21817 * Other significant improvements 21818 + [15]Ability to dump cfg information and display it using vcg. 21819 + The new faster scheme for fixing vendor header files is 21820 enabled by default. 21821 + Experimental internationalization support. 21822 + multibyte character support 21823 + Some compile-time speedups for pathological problems 21824 + Better support for complex types 21825 * Plus the usual mountain of bugfixes 21826 * Core compiler is based on the gcc2 development tree from Sept 30, 21827 1998, so we have all of the [16]features found in GCC 2.8. 21828 21829 Additional Changes in GCC 2.95.1 21830 21831 * Generic bugfixes and improvements 21832 + Various documentation fixes related to the GCC/EGCS merger. 21833 + Fix memory management bug which could lead to spurious aborts, 21834 core dumps or random parsing errors in the compiler. 21835 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 21836 support. 21837 + Fix infinite loop in the CSE optimizer. 21838 + Avoid undefined behavior in compiler FP emulation code 21839 + Fix install problem when prefix is overridden on the make 21840 install command. 21841 + Fix problem with unwanted installation of assert.h on some 21842 systems. 21843 + Fix problem with finding the wrong assembler in a single tree 21844 build. 21845 + Avoid increasing the known alignment of a register that is 21846 already known to be a pointer. 21847 * Platform specific bugfixes and improvements 21848 + Codegen bugfix for prologue/epilogue for cpu32 target. 21849 + Fix long long code generation bug for the Coldfire target. 21850 + Fix various aborts in the SH compiler. 21851 + Fix bugs in libgcc support library for the SH. 21852 + Fix alpha ev6 code generation bug. 21853 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 21854 AIX platforms. 21855 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 21856 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 21857 targets. 21858 + Fix weak symbol handling for rs6000/ppc svr4 targets. 21859 + Fix various problems with 64bit code generation for the 21860 rs6000/ppc port. 21861 + Fix codegen bug which caused tetex to be mis-compiled on the 21862 x86. 21863 + Fix compiler abort in new cfg code exposed by x86 port. 21864 + Fix out of range array reference in code convert flat 21865 registers to the x87 stacked FP register file. 21866 + Fix minor vxworks configuration bug. 21867 + Fix return type of bsearch for SunOS 4.x. 21868 * Language & Runtime specific fixes. 21869 + The G++ signature extension has been deprecated. It will be 21870 removed in the next major release of G++. Use of signatures 21871 will result in a warning from the compiler. 21872 + Several bugs relating to templates and namespaces were fixed. 21873 + A bug that caused crashes when combining templates with -g on 21874 DWARF1 platforms was fixed. 21875 + Pointers-to-members, virtual functions, and multiple 21876 inheritance should now work together correctly. 21877 + Some code-generation bugs relating to function try blocks were 21878 fixed. 21879 + G++ is a little bit more lenient with certain archaic 21880 constructs than in GCC 2.95. 21881 + Fix to prevent shared library version #s from bring truncated 21882 to 1 digit 21883 + Fix missing std:: in the libstdc++ library. 21884 + Fix stream locking problems in libio. 21885 + Fix problem in java compiler driver. 21886 21887 Additional Changes in GCC 2.95.2 21888 21889 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 21890 the optimizations performed by -fstrict-aliasing are valid according to 21891 the C and C++ standards, the optimization have caused some problems, 21892 particularly with old non-conforming code. 21893 21894 The GCC developers are experimenting with ways to warn users about code 21895 which violates the C/C++ standards, but those warnings are not ready 21896 for widespread use at this time. Rather than wait for those warnings 21897 the GCC developers have chosen to disable -fstrict-aliasing by default 21898 for the GCC 2.95.2 release. 21899 21900 We strongly encourage developers to find and fix code which violates 21901 the C/C++ standards as -fstrict-aliasing may be enabled by default in 21902 future releases. Use the option -fstrict-aliasing to re-enable these 21903 optimizations. 21904 * Generic bugfixes and improvements 21905 + Fix incorrectly optimized memory reference in global common 21906 subexpression elimination (GCSE) optimization pass. 21907 + Fix code generation bug in regmove.c in which it could 21908 incorrectly change a "const" value. 21909 + Fix bug in optimization of conditionals involving volatile 21910 memory references. 21911 + Avoid over-allocation of stack space for some procedures. 21912 + Fixed bug in the compiler which caused incorrect optimization 21913 of an obscure series of bit manipulations, shifts and 21914 arithmetic. 21915 + Fixed register allocator bug which caused teTeX to be 21916 mis-compiled on SPARC targets. 21917 + Avoid incorrect optimization of degenerate case statements for 21918 certain targets such as the ARM. 21919 + Fix out of range memory reference in the jump optimizer. 21920 + Avoid dereferencing null pointer in fix-header. 21921 + Fix test for GCC specific features so that it is possible to 21922 bootstrap with gcc-2.6.2 and older versions of GCC. 21923 + Fix typo in scheduler which could potentially cause out of 21924 range memory accesses. 21925 + Avoid incorrect loop reversal which caused incorrect code for 21926 certain loops on PowerPC targets. 21927 + Avoid incorrect optimization of switch statements on certain 21928 targets (for example the ARM). 21929 * Platform specific bugfixes and improvements 21930 + Work around bug in Sun V5.0 compilers which caused bootstrap 21931 comparison failures on SPARC targets. 21932 + Fix SPARC back end bug which caused aborts in final.c. 21933 + Fix sparc-hal-solaris2* configuration fragments. 21934 + Fix bug in sparc block profiling. 21935 + Fix obscure code generation bug for the PARISC targets. 21936 + Define __STDC_EXT__ for HPUX configurations. 21937 + Various POWERPC64 code generation bugfixes. 21938 + Fix abort for PPC targets using ELF (ex GNU/Linux). 21939 + Fix collect2 problems for AIX targets. 21940 + Correct handling of .file directive for PPC targets. 21941 + Fix bug in fix_trunc x86 patterns. 21942 + Fix x86 port to correctly pop the FP stack for functions that 21943 return structures in memory. 21944 + Fix minor bug in strlen x86 pattern. 21945 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 21946 + Fix template repository code to handle leading underscore in 21947 mangled names. 21948 + Fix weak/weak alias support for OpenBSD. 21949 + GNU/Linux for the ARM has C++ compatible include files. 21950 * Language & Runtime specific fixes. 21951 + Fix handling of constructor attribute in the C front-end which 21952 caused problems building the Chill runtime library on some 21953 targets. 21954 + Fix minor problem merging type qualifiers in the C front-end. 21955 + Fix aliasing bug for pointers and references (C/C++). 21956 + Fix incorrect "non-constant initializer bug" when -traditional 21957 or -fwritable-strings is enabled. 21958 + Fix build error for Chill front-end on SunOS. 21959 + Do not complain about duplicate instantiations when using 21960 -frepo (C++). 21961 + Fix array bounds handling in C++ front-end which caused 21962 problems with dwarf debugging information in some 21963 circumstances. 21964 + Fix minor namespace problem. 21965 + Fix problem linking java programs. 21966 21967 Additional Changes in GCC 2.95.3 21968 21969 * Generic bugfixes and improvements 21970 + Fix numerous problems that caused incorrect optimization in 21971 the register reloading code. 21972 + Fix numerous problems that caused incorrect optimization in 21973 the loop optimizer. 21974 + Fix aborts in the functions build_insn_chain and scan_loops 21975 under some circumstances. 21976 + Fix an alias analysis bug. 21977 + Fix an infinite compilation bug in the combiner. 21978 + A few problems with complex number support have been fixed. 21979 + It is no longer possible for gcc to act as a fork bomb when 21980 installed incorrectly. 21981 + The -fpack-struct option should be recognized now. 21982 + Fixed a bug that caused incorrect code to be generated due to 21983 a lost stack adjustment. 21984 * Platform specific bugfixes and improvements 21985 + Support building ARM toolchains hosted on Windows. 21986 + Fix attribute calculations in ARM toolchains. 21987 + arm-linux support has been improved. 21988 + Fix a PIC failure on sparc targets. 21989 + On ix86 targets, the regparm attribute should now work 21990 reliably. 21991 + Several updates for the h8300 port. 21992 + Fix problem building libio with glibc 2.2. 21993 21994 21995 For questions related to the use of GCC, please consult these web 21996 pages and the [17]GCC manuals. If that fails, the 21997 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 21998 web pages and the development of GCC are welcome on our developer 21999 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 22000 archives. 22001 22002 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 22003 distribution of this entire article is permitted in any medium, 22004 provided this notice is preserved. 22005 22006 These pages are [22]maintained by the GCC team. Last modified 22007 2023-09-02. 22008 22009 References 22010 22011 1. https://gcc.gnu.org/news/spill.html 22012 2. https://gcc.gnu.org/news/lcm.html 22013 3. https://gcc.gnu.org/news/cprop.html 22014 4. https://gcc.gnu.org/news/cfg.html 22015 5. https://gcc.gnu.org/news/dse.html 22016 6. https://gcc.gnu.org/news/hoist.html 22017 7. https://gcc.gnu.org/news/alias.html 22018 8. https://gcc.gnu.org/gcc-2.95/c++features.html 22019 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 22020 10. https://gcc.gnu.org/news/gcj-announce.txt 22021 11. https://gcc.gnu.org/news/javaannounce.html 22022 12. https://gcc.gnu.org/c99status.html 22023 13. https://gcc.gnu.org/news/chill.html 22024 14. https://gcc.gnu.org/news/sparc.html 22025 15. https://gcc.gnu.org/news/egcs-vcg.html 22026 16. https://gcc.gnu.org/egcs-1.0/features-2.8.html 22027 17. https://gcc.gnu.org/onlinedocs/ 22028 18. mailto:gcc-help (a] gcc.gnu.org 22029 19. mailto:gcc (a] gcc.gnu.org 22030 20. https://gcc.gnu.org/lists.html 22031 21. https://www.fsf.org/ 22032 22. https://gcc.gnu.org/about.html 22033 ====================================================================== 22034 http://gcc.gnu.org/gcc-2.95/caveats.html 22035 GCC 2.95 Caveats 22036 22037 * GCC 2.95 will issue an error for invalid asm statements that had 22038 been silently accepted by earlier versions of the compiler. This is 22039 particularly noticeable when compiling older versions of the Linux 22040 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 22041 for more information on this issue. 22042 * GCC 2.95 implements type based alias analysis to disambiguate 22043 memory references. Some programs, particularly the Linux kernel 22044 violate ANSI/ISO aliasing rules and therefore may not operate 22045 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 22046 shipped with GCC 2.95) for more information on this issue. 22047 * GCC 2.95 has a known bug in its handling of complex variables for 22048 64bit targets. Instead of silently generating incorrect code, GCC 22049 2.95 will issue a fatal error for situations it can not handle. 22050 This primarily affects the Fortran community as Fortran makes more 22051 use of complex variables than C or C++. 22052 * GCC 2.95 has an integrated libstdc++, but does not have an 22053 integrated libg++. Furthermore old libg++ releases will not work 22054 with GCC 2.95. You can retrieve a recent copy of libg++ from the 22055 [1]GCC ftp server. 22056 Note most C++ programs only need libstdc++. 22057 * Exception handling may not work with shared libraries, particularly 22058 on alphas, hppas, rs6000/powerpc and mips based platforms. 22059 Exception handling is known to work on x86 GNU/Linux platforms with 22060 shared libraries. 22061 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 22062 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 22063 or EGCS 1.1. As a result it may be necessary to fix C++ code before 22064 it will compile with GCC 2.95. 22065 * G++ is also converting toward the ISO C++ standard; as a result 22066 code which was previously valid (and thus accepted by other 22067 compilers and older versions of g++) may no longer be accepted. The 22068 flag -fpermissive may allow some non-conforming code to compile 22069 with GCC 2.95. 22070 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 22071 1.1.x, EGCS 1.0.x or GCC 2.8.x. 22072 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 22073 between Sept 30, 1998 and April 30, 1999 (the official end of the 22074 GCC 2.8 project). Future GCC releases will include all the changes 22075 from the defunct GCC 2.8 sources. 22076 22077 22078 For questions related to the use of GCC, please consult these web 22079 pages and the [2]GCC manuals. If that fails, the 22080 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22081 web pages and the development of GCC are welcome on our developer 22082 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 22083 22084 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 22085 distribution of this entire article is permitted in any medium, 22086 provided this notice is preserved. 22087 22088 These pages are [7]maintained by the GCC team. Last modified 22089 2022-10-26. 22090 22091 References 22092 22093 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 22094 2. https://gcc.gnu.org/onlinedocs/ 22095 3. mailto:gcc-help (a] gcc.gnu.org 22096 4. mailto:gcc (a] gcc.gnu.org 22097 5. https://gcc.gnu.org/lists.html 22098 6. https://www.fsf.org/ 22099 7. https://gcc.gnu.org/about.html 22100 ====================================================================== 22101 http://gcc.gnu.org/egcs-1.1/index.html 22102 EGCS 1.1 22103 22104 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 22105 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 22106 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 22107 22108 EGCS is a free software project to further the development of the GNU 22109 compilers using an open development environment. 22110 22111 EGCS 1.1 is a major new release of the EGCS compiler system. It has 22112 been [1]extensively tested and is believed to be stable and suitable 22113 for widespread use. 22114 22115 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 22116 development sources; it contains all of the new features found in GCC 22117 2.8.1 as well as all new development from GCC up to June 6, 1998. 22118 22119 EGCS 1.1 also contains many improvements and features not found in GCC 22120 or in older versions of EGCS: 22121 * Global common subexpression elimination and global constant/copy 22122 propagation (aka [2]gcse) 22123 * Ongoing improvements to the [3]alias analysis support to allow for 22124 better optimizations throughout the compiler. 22125 * Vastly improved [4]C++ compiler and integrated C++ runtime 22126 libraries. 22127 * Fixes for the /tmp symlink race security problems. 22128 * New targets including mips16, arm-thumb and 64-bit PowerPC. 22129 * Improvements to GNU Fortran (g77) compiler and runtime library made 22130 since g77 version 0.5.23. 22131 22132 See the [5]new features page for a more complete list of new features 22133 found in EGCS 1.1 releases. 22134 22135 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 22136 1.1: 22137 * General improvements and fixes 22138 + Avoid some stack overflows when compiling large functions. 22139 + Avoid incorrect loop invariant code motions. 22140 + Fix some core dumps on Linux kernel code. 22141 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 22142 + Fix code generation problem in gcse. 22143 + Various documentation related fixes. 22144 * g++/libstdc++ improvements and fixes 22145 + MT safe EH fix for setjmp/longjmp based exception handling. 22146 + Fix a few bad interactions between optimization and exception 22147 handling. 22148 + Fixes for demangling of template names starting with "__". 22149 + Fix a bug that would fail to run destructors in some cases 22150 with -O2. 22151 + Fix 'new' of classes with virtual bases. 22152 + Fix crash building Qt on the Alpha. 22153 + Fix failure compiling WIFEXITED macro on GNU/Linux. 22154 + Fix some -frepo failures. 22155 * g77 and libf2c improvements and fixes 22156 + Various documentation fixes. 22157 + Avoid compiler crash on RAND intrinsic. 22158 + Fix minor bugs in makefiles exposed by BSD make programs. 22159 + Define _XOPEN_SOURCE for libI77 build to avoid potential 22160 problems on some 64-bit systems. 22161 + Fix problem with implicit endfile on rewind. 22162 + Fix spurious recursive I/O errors. 22163 * platform specific improvements and fixes 22164 + Match all versions of UnixWare7. 22165 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 22166 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 22167 from unsigned ints to double precision floats. 22168 + Fix ARM ABI issue with NetBSD. 22169 + Fix a few arm code generation bugs. 22170 + Fixincludes will fix additional broken SCO OpenServer header 22171 files. 22172 + Fix a m68k back end bug which caused invalid offsets in reg+d 22173 addresses. 22174 + Fix problems with 64bit AIX 4.3 support. 22175 + Fix handling of long longs for varargs/stdarg functions on the 22176 ppc. 22177 + Minor fixes to CPP predefines for Windows. 22178 + Fix code generation problems with gpr<->fpr copies for 64bit 22179 ppc. 22180 + Fix a few coldfire code generation bugs. 22181 + Fix some more header file problems on SunOS 4.x. 22182 + Fix assert.h handling for RTEMS. 22183 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 22184 + Fix x86 compiler abort in reg-stack pass. 22185 + Fix cygwin/windows problem with section attributes. 22186 + Fix Alpha code generation problem exposed by SMP Linux 22187 kernels. 22188 + Fix typo in m68k 32->64bit integer conversion. 22189 + Make sure target libraries build with -fPIC for PPC & Alpha 22190 targets. 22191 22192 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 22193 1.1.1: 22194 * General improvements and fixes 22195 + Fix bug in loop optimizer which caused the SPARC (and 22196 potentially other) ports to segfault. 22197 + Fix infinite recursion in alias analysis and combiner code. 22198 + Fix bug in regclass preferencing. 22199 + Fix incorrect loop reversal which caused incorrect code to be 22200 generated for several targets. 22201 + Fix return value for builtin memcpy. 22202 + Reduce compile time for certain loops which exposed quadratic 22203 behavior in the loop optimizer. 22204 + Fix bug which caused volatile memory to be written multiple 22205 times when only one write was needed/desired. 22206 + Fix compiler abort in caller-save.c 22207 + Fix combiner bug which caused incorrect code generation for 22208 certain division by constant operations. 22209 + Fix incorrect code generation due to a bug in range check 22210 optimizations. 22211 + Fix incorrect code generation due to mis-handling of clobbered 22212 values in CSE. 22213 + Fix compiler abort/segfault due to incorrect register 22214 splitting when unrolling loops. 22215 + Fix code generation involving autoincremented addresses with 22216 ternary operators. 22217 + Work around bug in the scheduler which caused qt to be 22218 mis-compiled on some platforms. 22219 + Fix code generation problems with -fshort-enums. 22220 + Tighten security for temporary files. 22221 + Improve compile time for codes which make heavy use of 22222 overloaded functions. 22223 + Fix multiply defined constructor/destructor symbol problems. 22224 + Avoid setting bogus RPATH environment variable during 22225 bootstrap. 22226 + Avoid GNU-make dependencies in the texinfo subdir. 22227 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 22228 --enable-cpp=<dirname> can be used to specify an additional 22229 install directory for the cpp wrapper script. 22230 + Fix CSE bug which caused incorrect label-label refs to appear 22231 on some platforms. 22232 + Avoid linking in EH routines from libgcc if they are not 22233 needed. 22234 + Avoid obscure bug in aliasing code. 22235 + Fix bug in weak symbol handling. 22236 * Platform-specific improvements and fixes 22237 + Fix detection of PPro/PII on Unixware 7. 22238 + Fix compiler segfault when building spec99 and other programs 22239 for SPARC targets. 22240 + Fix code-generation bugs for integer and floating point 22241 conditional move instructions on the PPro/PII. 22242 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 22243 + Fix build failure for the arc port. 22244 + Fix floating point format configuration for i?86-gnu port. 22245 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 22246 threads are enabled. 22247 + Fix coldfire code generation bugs. 22248 + Fix "unrecognized insn" problems for Alpha and PPC ports. 22249 + Fix h8/300 code generation problem with floating point values 22250 in memory. 22251 + Fix unrecognized insn problems for the m68k port. 22252 + Fix namespace-pollution problem for the x86 port. 22253 + Fix problems with old assembler on x86 NeXT systems. 22254 + Fix PIC code-generation problems for the SPARC port. 22255 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 22256 + Fix minor ISO namespace violation in Alpha varargs/stdarg 22257 support. 22258 + Fix incorrect "braf" instruction usage for the SH port. 22259 + Fix minor bug in va-sh which prevented its use with -ansi. 22260 + Fix problems recognizing and supporting FreeBSD. 22261 + Handle OpenBSD systems correctly. 22262 + Minor fixincludes fix for Digital UNIX 4.0B. 22263 + Fix problems with ctors/dtors in SCO shared libraries. 22264 + Abort instead of generating incorrect code for PPro/PII 22265 floating point conditional moves. 22266 + Avoid multiply defined symbols on GNU/Linux systems using 22267 libc-5.4.xx. 22268 + Fix abort in alpha compiler. 22269 * Fortran-specific fixes 22270 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 22271 is in the documented, non-Y2K-compliant range of 0-99, instead 22272 of being returned as 100 in the year 2000. 22273 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 22274 milliseconds value properly in Values(8). 22275 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 22276 information properly in SArray(7). 22277 22278 Each release includes installation instructions in both HTML and 22279 plaintext forms (see the INSTALL directory in the toplevel directory of 22280 the distribution). However, we also keep the most up to date 22281 installation instructions and [6]build/test status on our web page. We 22282 will update those pages as new information becomes available. 22283 22284 The EGCS project would like to thank the numerous people that have 22285 contributed new features, test results, bugfixes, etc. This [7]amazing 22286 group of volunteers is what makes EGCS successful. 22287 22288 And finally, we can't in good conscience fail to mention some 22289 [8]caveats to using EGCS 1.1. 22290 22291 Download EGCS from egcs.cygnus.com (USA California). 22292 22293 The EGCS 1.1 release is also available on many mirror sites. 22294 [9]Goto mirror list to find a closer site. 22295 22296 22297 For questions related to the use of GCC, please consult these web 22298 pages and the [10]GCC manuals. If that fails, the 22299 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22300 web pages and the development of GCC are welcome on our developer 22301 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 22302 archives. 22303 22304 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 22305 distribution of this entire article is permitted in any medium, 22306 provided this notice is preserved. 22307 22308 These pages are [15]maintained by the GCC team. Last modified 22309 2024-05-18. 22310 22311 References 22312 22313 1. https://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 22314 2. https://gcc.gnu.org/news/gcse.html 22315 3. https://gcc.gnu.org/news/alias.html 22316 4. https://gcc.gnu.org/egcs-1.1/c++features.html 22317 5. https://gcc.gnu.org/egcs-1.1/features.html 22318 6. https://gcc.gnu.org/egcs-1.1/buildstat.html 22319 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 22320 8. https://gcc.gnu.org/egcs-1.1/caveats.html 22321 9. https://gcc.gnu.org/mirrors.html 22322 10. https://gcc.gnu.org/onlinedocs/ 22323 11. mailto:gcc-help (a] gcc.gnu.org 22324 12. mailto:gcc (a] gcc.gnu.org 22325 13. https://gcc.gnu.org/lists.html 22326 14. https://www.fsf.org/ 22327 15. https://gcc.gnu.org/about.html 22328 ====================================================================== 22329 http://gcc.gnu.org/egcs-1.1/features.html 22330 EGCS 1.1 new features 22331 22332 * Integrated GNU Fortran (g77) compiler and runtime library with 22333 improvements, based on g77 version 0.5.23. 22334 * Vast improvements in the C++ compiler; so many they have [1]page of 22335 their own! 22336 * Compiler implements [2]global common subexpression elimination and 22337 global copy/constant propagation. 22338 * More major improvements in the [3]alias analysis code. 22339 * More major improvements in the exception handling code to improve 22340 performance, lower static overhead and provide the infrastructure 22341 for future improvements. 22342 * The infamous /tmp symlink race security problems have been fixed. 22343 * The regmove optimization pass has been nearly completely rewritten 22344 to improve performance of generated code. 22345 * The compiler now recomputes register usage information before local 22346 register allocation. By providing more accurate information to the 22347 priority based allocator, we get better register allocation. 22348 * The register reloading phase of the compiler optimizes spill code 22349 much better than in previous releases. 22350 * Some bad interactions between the register allocator and 22351 instruction scheduler have been fixed, resulting in much better 22352 code for certain programs. Additionally, we have tuned the 22353 scheduler in various ways to improve performance of generated code 22354 for some architectures. 22355 * The compiler's branch shortening algorithms have been significantly 22356 improved to work better on targets which align jump targets. 22357 * The compiler now supports -Os to prefer optimizing for code space 22358 over optimizing for code speed. 22359 * The compiler will now totally eliminate library calls which compute 22360 constant values. This primarily helps targets with no integer 22361 div/mul support and targets without floating point support. 22362 * The compiler now supports an extensive "--help" option. 22363 * cpplib has been greatly improved and may be suitable for limited 22364 use. 22365 * Memory footprint for the compiler has been significantly reduced 22366 for some pathological cases. 22367 * The time to build EGCS has been improved for certain targets 22368 (particularly the alpha and mips platforms). 22369 * Many infrastructure improvements throughout the compiler, plus the 22370 usual mountain of bugfixes and minor improvements. 22371 * Target dependent improvements: 22372 + SPARC port now includes V8 plus and V9 support as well as 22373 performance tuning for Ultra class machines. The SPARC port 22374 now uses the Haifa scheduler. 22375 + Alpha port has been tuned for the EV6 processor and has an 22376 optimized expansion of memcpy/bzero. The Alpha port now uses 22377 the Haifa scheduler. 22378 + RS6000/PowerPC: support for the Power64 architecture and AIX 22379 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 22380 + x86: Alignment of static store data and jump targets is per 22381 Intel recommendations now. Various improvements throughout the 22382 x86 port to improve performance on Pentium processors 22383 (including improved epilogue sequences for Pentium chips and 22384 back end improvements which should help register allocation on 22385 all x86 variants. Conditional move support has been fixed and 22386 enabled for PPro processors. The x86 port also better supports 22387 64bit operations now. Unixware 7, a System V Release 5 target, 22388 is now supported and SCO OpenServer targets can support GAS. 22389 + MIPS has improved multiply/multiply-add support and now 22390 includes mips16 ISA support. 22391 + M68k has many micro-optimizations and Coldfire fixes. 22392 * Core compiler is based on the GCC development tree from June 9, 22393 1998, so we have all of the [4]features found in GCC 2.8. 22394 22395 22396 For questions related to the use of GCC, please consult these web 22397 pages and the [5]GCC manuals. If that fails, the 22398 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22399 web pages and the development of GCC are welcome on our developer 22400 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 22401 22402 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 22403 distribution of this entire article is permitted in any medium, 22404 provided this notice is preserved. 22405 22406 These pages are [10]maintained by the GCC team. Last modified 22407 2023-09-02. 22408 22409 References 22410 22411 1. https://gcc.gnu.org/egcs-1.1/c++features.html 22412 2. https://gcc.gnu.org/news/gcse.html 22413 3. https://gcc.gnu.org/news/alias.html 22414 4. https://gcc.gnu.org/egcs-1.0/features-2.8.html 22415 5. https://gcc.gnu.org/onlinedocs/ 22416 6. mailto:gcc-help (a] gcc.gnu.org 22417 7. mailto:gcc (a] gcc.gnu.org 22418 8. https://gcc.gnu.org/lists.html 22419 9. https://www.fsf.org/ 22420 10. https://gcc.gnu.org/about.html 22421 ====================================================================== 22422 http://gcc.gnu.org/egcs-1.1/caveats.html 22423 EGCS 1.1 Caveats 22424 22425 * EGCS has an integrated libstdc++, but does not have an integrated 22426 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 22427 Lu has made a libg++-2.8.1.2 snapshot available which may work with 22428 EGCS. 22429 Note most C++ programs only need libstdc++. 22430 * Exception handling may not work with shared libraries, particularly 22431 on alphas, hppas, rs6000/powerpc and mips based platforms. 22432 Exception handling is known to work on x86-linux platforms with 22433 shared libraries. 22434 * Some versions of the Linux kernel have bugs which prevent them from 22435 being compiled or from running when compiled by EGCS. See the FAQ 22436 (as shipped with EGCS 1.1) for additional information. 22437 * In general, EGCS is more rigorous about rejecting invalid C++ code 22438 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 22439 a result it may be necessary to fix C++ code before it will compile 22440 with EGCS. 22441 * G++ is also converting toward the ISO C++ standard; as a result 22442 code which was previously valid (and thus accepted by other 22443 compilers and older versions of g++) may no longer be accepted. 22444 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 22445 or GCC 2.8.x due to changes necessary to support thread safe 22446 exception handling. 22447 22448 22449 For questions related to the use of GCC, please consult these web 22450 pages and the [1]GCC manuals. If that fails, the 22451 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22452 web pages and the development of GCC are welcome on our developer 22453 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 22454 22455 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 22456 distribution of this entire article is permitted in any medium, 22457 provided this notice is preserved. 22458 22459 These pages are [6]maintained by the GCC team. Last modified 22460 2022-10-26. 22461 22462 References 22463 22464 1. https://gcc.gnu.org/onlinedocs/ 22465 2. mailto:gcc-help (a] gcc.gnu.org 22466 3. mailto:gcc (a] gcc.gnu.org 22467 4. https://gcc.gnu.org/lists.html 22468 5. https://www.fsf.org/ 22469 6. https://gcc.gnu.org/about.html 22470 ====================================================================== 22471 http://gcc.gnu.org/egcs-1.0/index.html 22472 EGCS 1.0 22473 22474 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 22475 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 22476 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 22477 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 22478 22479 EGCS is a collaborative effort involving several groups of hackers 22480 using an open development model to accelerate development and testing 22481 of GNU compilers and runtime libraries. 22482 22483 An important goal of EGCS is to allow wide scale testing of 22484 experimental features and optimizations; therefore, EGCS contains some 22485 features and optimizations which are still under development. However, 22486 EGCS has been carefully tested and should be comparable in quality to 22487 most GCC releases. 22488 22489 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 22490 development sources; it contains nearly all of the new features found 22491 in GCC 2.8. 22492 22493 EGCS 1.0 also contains many improvements and features not found in GCC 22494 2.7 and even the GCC 2.8 series (which was released after the original 22495 EGCS 1.0 release). 22496 * Integrated C++ runtime libraries, including support for most major 22497 GNU/Linux systems! 22498 * The integrated libstdc++ library includes a verbatim copy of SGI's 22499 STL release. 22500 * Integrated GNU Fortran compiler. 22501 * New instruction scheduler. 22502 * New alias analysis code. 22503 22504 See the [1]new features page for a more complete list of new features. 22505 22506 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 22507 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 22508 EGCS 1.0 release: 22509 * Add support for Red Hat 5.0 Linux and better support for Linux 22510 systems using glibc2. 22511 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 22512 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 22513 fix these problems. 22514 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 22515 handling interfaces. 22516 To avoid future compatibility problems, we strongly urge anyone who 22517 is planning on distributing shared libraries that contain C++ code 22518 to upgrade to EGCS 1.0.1 first. 22519 Soon after EGCS 1.0 was released, the GCC developers made some 22520 incompatible changes in libgcc's exception handling interfaces. 22521 These changes were needed to solve problems on some platforms. This 22522 means that GCC 2.8.0, when released, will not be seamlessly 22523 compatible with shared libraries built by EGCS 1.0. The reason is 22524 that the libgcc.a in GCC 2.8.0 will not contain a function needed 22525 by the old interface. 22526 The result of this is that there may be compatibility problems with 22527 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 22528 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 22529 and libgcc.a has the support routines for both the old and the new 22530 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 22531 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 22532 The maintainers of GCC 2.x have decided against including seamless 22533 support for the old interface in 2.8.0, since it was never 22534 "official", so to avoid future compatibility problems we recommend 22535 against distributing any shared libraries built by EGCS 1.0 that 22536 contain C++ code (upgrade to 1.0.1 and use that). 22537 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 22538 The x86 changes fix code generation errors exposed when building 22539 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 22540 The hppa change fixes a compiler abort when configured for use with 22541 RTEMS. 22542 The MIPS changes fix problems with the definition of LONG_MAX on 22543 newer systems, allow for command line selection of the target ABI, 22544 and fix one code generation problem. 22545 The rs6000/ppc change fixes some problems with passing structures 22546 to varargs/stdarg functions. 22547 * A few machine independent bugfixes, mostly to fix code generation 22548 errors when building Linux kernels or glibc. 22549 * Fix a few critical exception handling and template bugs in the C++ 22550 compiler. 22551 * Fix Fortran namelist bug on alphas. 22552 * Fix build problems on x86-solaris systems. 22553 22554 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 22555 serious problems in EGCS 1.0.1. 22556 * General improvements and fixes 22557 + Memory consumption significantly reduced, especially for 22558 templates and inline functions. 22559 + Fix various problems with glibc2.1. 22560 + Fix loop optimization bug exposed by rs6000/ppc port. 22561 + Fix to avoid potential code generation problems in jump.c. 22562 + Fix some undefined symbol problems in dwarf1 debug support. 22563 * g++/libstdc++ improvements and fixes 22564 + libstdc++ in the EGCS release has been updated and should be 22565 link compatible with libstdc++-2.8. 22566 + Various fixes in libio/libstdc++ to work better on GNU/Linux 22567 systems. 22568 + Fix problems with duplicate symbols on systems that do not 22569 support weak symbols. 22570 + Memory corruption bug and undefined symbols in bastring have 22571 been fixed. 22572 + Various exception handling fixes. 22573 + Fix compiler abort for very long thunk names. 22574 * g77 improvements and fixes 22575 + Fix compiler crash for omitted bound in Fortran CASE 22576 statement. 22577 + Add missing entries to g77 lang-options. 22578 + Fix problem with -fpedantic in the g77 compiler. 22579 + Fix "backspace" problem with g77 on alphas. 22580 + Fix x86 back end problem with Fortran literals and -fpic. 22581 + Fix some of the problems with negative subscripts for g77 on 22582 alphas. 22583 + Fixes for Fortran builds on cygwin32/mingw32. 22584 * platform specific improvements and fixes 22585 + Fix long double problems on x86 (exposed by glibc). 22586 + x86 ports define i386 again to keep imake happy. 22587 + Fix exception handling support on NetBSD ports. 22588 + Several changes to collect2 to fix many problems with AIX. 22589 + Define __ELF__ for GNU/Linux on rs6000. 22590 + Fix -mcall-linux problem on GNU/Linux on rs6000. 22591 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 22592 + Allow autoconf to select a proper install problem on AIX 3.1. 22593 + m68k port support includes -mcpu32 option as well as cpu32 22594 multilibs. 22595 + Fix stdarg bug for irix6. 22596 + Allow EGCS to build on irix5 without the gnu assembler. 22597 + Fix problem with static linking on sco5. 22598 + Fix bootstrap on sco5 with native compiler. 22599 + Fix for abort building newlib on H8 target. 22600 + Fix fixincludes handling of math.h on SunOS. 22601 + Minor fix for Motorola 3300 m68k systems. 22602 22603 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 22604 problems reported by Red Hat for builds of Red Hat 5.1. 22605 * Generic bugfixes: 22606 + Fix a typo in the libio library which resulted in incorrect 22607 behavior of istream::get. 22608 + Fix the Fortran negative array index problem. 22609 + Fix a major problem with the ObjC runtime thread support 22610 exposed by glibc2. 22611 + Reduce memory consumption of the Haifa scheduler. 22612 * Target specific bugfixes: 22613 + Fix one x86 floating point code generation bug exposed by 22614 glibc2 builds. 22615 + Fix one x86 internal compiler error exposed by glibc2 builds. 22616 + Fix profiling bugs on the Alpha. 22617 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 22618 + Fix rs6000/ppc bug when converting values from integer types 22619 to floating point types. 22620 22621 The EGCS 1.0 releases include installation instructions in both HTML 22622 and plaintext forms (see the INSTALL directory in the toplevel 22623 directory of the distribution). However, we also keep the most up to 22624 date installation instructions and [2]build/test status on our web 22625 page. We will update those pages as new information becomes available. 22626 22627 And, we can't in good conscience fail to mention some [3]caveats to 22628 using EGCS. 22629 22630 Update: Big thanks to Stanford for providing a high speed link for 22631 downloading EGCS (go.cygnus.com)! 22632 22633 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 22634 (USA California -- High speed link provided by Stanford). 22635 22636 The EGCS 1.0 release is also available many mirror sites. 22637 [4]Goto mirror list to find a closer site 22638 22639 We'd like to thank the numerous people that have contributed new 22640 features, test results, bugfixes, etc. Unfortunately, they're far too 22641 numerous to mention by name. 22642 22643 22644 For questions related to the use of GCC, please consult these web 22645 pages and the [5]GCC manuals. If that fails, the 22646 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22647 web pages and the development of GCC are welcome on our developer 22648 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 22649 22650 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 22651 distribution of this entire article is permitted in any medium, 22652 provided this notice is preserved. 22653 22654 These pages are [10]maintained by the GCC team. Last modified 22655 2023-09-02. 22656 22657 References 22658 22659 1. https://gcc.gnu.org/egcs-1.0/features.html 22660 2. https://gcc.gnu.org/egcs-1.0/buildstat.html 22661 3. https://gcc.gnu.org/egcs-1.0/caveats.html 22662 4. https://gcc.gnu.org/mirrors.html 22663 5. https://gcc.gnu.org/onlinedocs/ 22664 6. mailto:gcc-help (a] gcc.gnu.org 22665 7. mailto:gcc (a] gcc.gnu.org 22666 8. https://gcc.gnu.org/lists.html 22667 9. https://www.fsf.org/ 22668 10. https://gcc.gnu.org/about.html 22669 ====================================================================== 22670 http://gcc.gnu.org/egcs-1.0/features.html 22671 EGCS 1.0 features 22672 22673 * Core compiler is based on the gcc2 development tree from Aug 2, 22674 1997, so we have most of the [1]features found in GCC 2.8. 22675 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 22676 * Vast improvements in the C++ compiler; so many they have [2]page of 22677 their own! 22678 * Integrated C++ runtime libraries, including support for most major 22679 GNU/Linux systems! 22680 * New instruction scheduler from IBM Haifa which includes support for 22681 function wide instruction scheduling as well as superscalar 22682 scheduling. 22683 * Significantly improved alias analysis code. 22684 * Improved register allocation for two address machines. 22685 * Significant code generation improvements for Fortran code on 22686 Alphas. 22687 * Various optimizations from the g77 project as well as improved loop 22688 optimizations. 22689 * Dwarf2 debug format support for some targets. 22690 * egcs libstdc++ includes the SGI STL implementation without changes. 22691 * As a result of these and other changes, egcs libstc++ is not binary 22692 compatible with previous releases of libstdc++. 22693 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 22694 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 22695 1.1), Support for RTEMS on several embedded targets, Support for 22696 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 22697 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 22698 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 22699 * RS6000/PowerPC ports generate code which can run on all 22700 RS6000/PowerPC variants by default. 22701 * -mcpu= and -march= switches for the x86 port to allow better 22702 control over how the x86 port generates code. 22703 * Includes the template repository patch (aka repo patch); note the 22704 new template code makes repo obsolete for ELF systems using gnu-ld 22705 such as GNU/Linux. 22706 * Plus the usual assortment of bugfixes and improvements. 22707 22708 22709 For questions related to the use of GCC, please consult these web 22710 pages and the [3]GCC manuals. If that fails, the 22711 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22712 web pages and the development of GCC are welcome on our developer 22713 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 22714 22715 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 22716 distribution of this entire article is permitted in any medium, 22717 provided this notice is preserved. 22718 22719 These pages are [8]maintained by the GCC team. Last modified 22720 2022-10-26. 22721 22722 References 22723 22724 1. https://gcc.gnu.org/egcs-1.0/features-2.8.html 22725 2. https://gcc.gnu.org/egcs-1.0/c++features.html 22726 3. https://gcc.gnu.org/onlinedocs/ 22727 4. mailto:gcc-help (a] gcc.gnu.org 22728 5. mailto:gcc (a] gcc.gnu.org 22729 6. https://gcc.gnu.org/lists.html 22730 7. https://www.fsf.org/ 22731 8. https://gcc.gnu.org/about.html 22732 ====================================================================== 22733 http://gcc.gnu.org/egcs-1.0/caveats.html 22734 EGCS 1.0 Caveats 22735 22736 * EGCS has an integrated libstdc++, but does not have an integrated 22737 libg++. Furthermore old libg++ releases will not work with egc; HJ 22738 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 22739 Note most C++ programs only need libstdc++. 22740 * Note that using -pedantic or -Wreturn-type can cause an explosion 22741 in the amount of memory needed for template-heavy C++ code, such as 22742 code that uses STL. Also note that -Wall includes -Wreturn-type, so 22743 if you use -Wall you will need to specify -Wno-return-type to turn 22744 it off. 22745 * Exception handling may not work with shared libraries, particularly 22746 on alphas, hppas, and mips based platforms. Exception handling is 22747 known to work on x86-linux platforms with shared libraries. 22748 * Some versions of the Linux kernel have bugs which prevent them from 22749 being compiled or from running when compiled by EGCS. See the FAQ 22750 (as shipped with EGCS 1.0) for additional information. 22751 * In general, EGCS is more rigorous about rejecting invalid C++ code 22752 or deprecated C++ constructs than G++ 2.7. As a result it may be 22753 necessary to fix C++ code before it will compile with EGCS. 22754 * G++ is also aggressively tracking the C++ standard; as a result 22755 code which was previously valid (and thus accepted by other 22756 compilers and older versions of G++) may no longer be accepted. 22757 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 22758 1.0.x and later releases should work with Red Hat Linux 5.0. 22759 22760 22761 For questions related to the use of GCC, please consult these web 22762 pages and the [1]GCC manuals. If that fails, the 22763 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 22764 web pages and the development of GCC are welcome on our developer 22765 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 22766 22767 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 22768 distribution of this entire article is permitted in any medium, 22769 provided this notice is preserved. 22770 22771 These pages are [6]maintained by the GCC team. Last modified 22772 2022-10-26. 22773 22774 References 22775 22776 1. https://gcc.gnu.org/onlinedocs/ 22777 2. mailto:gcc-help (a] gcc.gnu.org 22778 3. mailto:gcc (a] gcc.gnu.org 22779 4. https://gcc.gnu.org/lists.html 22780 5. https://www.fsf.org/ 22781 6. https://gcc.gnu.org/about.html 22782 ====================================================================== 22783
Indexes created Tue Oct 27 00:25:24 UTC 2026