NEWS revision 1.7
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-6/index.html 9 10 GCC 6 Release Series 11 12 July 4, 2017 13 14 The [1]GNU project and the GCC developers are pleased to announce the 15 release of GCC 6.4. 16 17 This release is a bug-fix release, containing fixes for regressions in 18 GCC 6.3 relative to previous releases of GCC. 19 20 Release History 21 22 GCC 6.4 23 July 4, 2017 ([2]changes, [3]documentation) 24 25 GCC 6.3 26 December 21, 2016 ([4]changes, [5]documentation) 27 28 GCC 6.2 29 August 22, 2016 ([6]changes, [7]documentation) 30 31 GCC 6.1 32 April 27, 2016 ([8]changes, [9]documentation) 33 34 References and Acknowledgements 35 36 GCC used to stand for the GNU C Compiler, but since the compiler 37 supports several other languages aside from C, it now stands for the 38 GNU Compiler Collection. 39 40 A list of [10]successful builds is updated as new information becomes 41 available. 42 43 The GCC developers would like to thank the numerous people that have 44 contributed new features, improvements, bug fixes, and other changes as 45 well as test results to GCC. This [11]amazing group of volunteers is 46 what makes GCC successful. 47 48 For additional information about GCC please refer to the [12]GCC 49 project web site or contact the [13]GCC development mailing list. 50 51 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 52 53 54 For questions related to the use of GCC, please consult these web 55 pages and the [16]GCC manuals. If that fails, the 56 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 57 web pages and the development of GCC are welcome on our developer 58 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 59 archives. 60 61 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 62 distribution of this entire article is permitted in any medium, 63 provided this notice is preserved. 64 65 These pages are [21]maintained by the GCC team. Last modified 66 2017-07-04[22]. 67 68 References 69 70 1. http://www.gnu.org/ 71 2. http://gcc.gnu.org/gcc-6/changes.html 72 3. http://gcc.gnu.org/onlinedocs/6.4.0/ 73 4. http://gcc.gnu.org/gcc-6/changes.html 74 5. http://gcc.gnu.org/onlinedocs/6.3.0/ 75 6. http://gcc.gnu.org/gcc-6/changes.html 76 7. http://gcc.gnu.org/onlinedocs/6.2.0/ 77 8. http://gcc.gnu.org/gcc-6/changes.html 78 9. http://gcc.gnu.org/onlinedocs/6.1.0/ 79 10. http://gcc.gnu.org/gcc-6/buildstat.html 80 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 81 12. http://gcc.gnu.org/index.html 82 13. mailto:gcc (a] gcc.gnu.org 83 14. http://gcc.gnu.org/mirrors.html 84 15. http://gcc.gnu.org/svn.html 85 16. https://gcc.gnu.org/onlinedocs/ 86 17. mailto:gcc-help (a] gcc.gnu.org 87 18. mailto:gcc (a] gcc.gnu.org 88 19. https://gcc.gnu.org/lists.html 89 20. http://www.fsf.org/ 90 21. https://gcc.gnu.org/about.html 91 22. http://validator.w3.org/check/referer 92 ====================================================================== 93 http://gcc.gnu.org/gcc-6/changes.html 94 95 GCC 6 Release Series 96 Changes, New Features, and Fixes 97 98 This page is a brief summary of some of the huge number of improvements 99 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 100 the [2]full GCC documentation. 101 102 Caveats 103 104 * The default mode for C++ is now -std=gnu++14 instead of 105 -std=gnu++98. 106 * Support for a number of older systems and recently unmaintained or 107 untested target ports of GCC has been declared obsolete in GCC 6. 108 Unless there is activity to revive them, the next release of GCC 109 will have their sources permanently removed. 110 The following ports for individual systems on particular 111 architectures have been obsoleted: 112 + SH5 / SH64 (sh64-*-*) as announced [3]here. 113 * The AVR port requires binutils version 2.26.1 or later for the fix 114 for [4]PR71151 to work. 115 116 General Optimizer Improvements 117 118 * UndefinedBehaviorSanitizer gained a new sanitization option, 119 -fsanitize=bounds-strict, which enables strict checking of array 120 bounds. In particular, it enables -fsanitize=bounds as well as 121 instrumentation of flexible array member-like arrays. 122 * Type-based alias analysis now disambiguates accesses to different 123 pointers. This improves precision of the alias oracle by about 124 20-30% on higher-level C++ programs. Programs doing invalid type 125 punning of pointer types may now need -fno-strict-aliasing to work 126 correctly. 127 * Alias analysis now correctly supports the weakref and alias 128 attributes. This allows accessing both a variable and its alias in 129 one translation unit which is common with link-time optimization. 130 * Value range propagation now assumes that the this pointer in C++ 131 member functions is non-null. This eliminates common null pointer 132 checks but also breaks some non-conforming code-bases (such as 133 Qt-5, Chromium, KDevelop). As a temporary work-around 134 -fno-delete-null-pointer-checks can be used. Wrong code can be 135 identified by using -fsanitize=undefined. 136 * Link-time optimization improvements: 137 + warning and error attributes are now correctly preserved by 138 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 139 supported with -flto. 140 + Type merging was fixed to handle C and Fortran 141 interoperability rules as defined by the Fortran 2008 language 142 standard. 143 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 144 with char in all cases because it is an array while char is 145 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 146 general, this inter-operability cannot be implemented, for 147 example on targets where the argument passing convention for 148 arrays differs from scalars. 149 + More type information is now preserved at link time, reducing 150 the loss of accuracy of the type-based alias analysis compared 151 to builds without link-time optimization. 152 + Invalid type punning on global variables and declarations is 153 now reported with -Wodr-type-mismatch. 154 + The size of LTO object files was reduced by about 11% 155 (measured by compiling Firefox 46.0). 156 + Link-time parallelization (enabled using -flto=n) was 157 significantly improved by decreasing the size of streamed data 158 when partitioning programs. The size of streamed IL while 159 compiling Firefox 46.0 was reduced by 66%. 160 + The linker plugin was extended to pass information about the 161 type of binary produced to the GCC back end. (That can also be 162 controlled manually by -flinker-output.) This makes it 163 possible to properly configure the code generator and support 164 incremental linking. Incremental linking of LTO objects by gcc 165 -r is now supported for plugin-enabled setups. 166 There are two ways to perform incremental linking: 167 1. Linking by ld -r will result in an object file with all 168 sections from individual object files mechanically 169 merged. This delays the actual link-time optimization to 170 the final linking step and thus permits whole program 171 optimization. Linking the final binary with such object 172 files is however slower. 173 2. Linking by gcc -r will lead to link-time optimization and 174 emit the final binary into the object file. Linking such 175 an object file is fast but avoids any benefits from whole 176 program optimization. 177 GCC 7 will support incremental link-time optimization with gcc 178 -r. 179 * Inter-procedural optimization improvements: 180 + Basic jump threading is now performed before profile 181 construction and inline analysis, resulting in more realistic 182 size and time estimates that drive the heuristics of the 183 inliner and function cloning passes. 184 + Function cloning now more aggressively eliminates unused 185 function parameters. 186 187 New Languages and Language specific improvements 188 189 Compared to GCC 5, the GCC 6 release series includes a much improved 190 implementation of the [5]OpenACC 2.0a specification. Highlights are: 191 * In addition to single-threaded host-fallback execution, offloading 192 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 193 little-endian GNU/Linux host systems. For nvptx offloading, with 194 the OpenACC parallel construct, the execution model allows for an 195 arbitrary number of gangs, up to 32 workers, and 32 vectors. 196 * Initial support for parallelized execution of OpenACC kernels 197 constructs: 198 + Parallelization of a kernels region is switched on by 199 -fopenacc combined with -O2 or higher. 200 + Code is offloaded onto multiple gangs, but executes with just 201 one worker, and a vector length of 1. 202 + Directives inside a kernels region are not supported. 203 + Loops with reductions can be parallelized. 204 + Only kernels regions with one loop nest are parallelized. 205 + Only the outer-most loop of a loop nest can be parallelized. 206 + Loop nests containing sibling loops are not parallelized. 207 Typically, using the OpenACC parallel construct gives much better 208 performance, compared to the initial support of the OpenACC kernels 209 construct. 210 * The device_type clause is not supported. The bind and nohost 211 clauses are not supported. The host_data directive is not supported 212 in Fortran. 213 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 214 * Usage of OpenACC constructs inside multithreaded contexts (such as 215 created by OpenMP, or pthread programming) is not supported. 216 * If a call to the acc_on_device function has a compile-time constant 217 argument, the function call evaluates to a compile-time constant 218 value only for C and C++ but not for Fortran. 219 220 See the [6]OpenACC and [7]Offloading wiki pages for further 221 information. 222 223 C family 224 225 * Version 4.5 of the [8]OpenMP specification is now supported in the 226 C and C++ compilers. 227 * The C and C++ compilers now support attributes on enumerators. For 228 instance, it is now possible to mark enumerators as deprecated: 229 230 enum { 231 newval, 232 oldval __attribute__ ((deprecated ("too old"))) 233 }; 234 235 * Source locations for the C and C++ compilers are now tracked as 236 ranges, rather than just points, making it easier to identify the 237 subexpression of interest within a complicated expression. For 238 example: 239 240 test.cc: In function 'int test(int, int, foo, int, int)': 241 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 242 ) 243 return p + q * r * s + t; 244 ~~^~~ 245 246 In addition, there is now initial support for precise diagnostic 247 locations within strings: 248 249 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 250 t' argument [-Wformat=] 251 printf("%*d"); 252 ^ 253 254 * Diagnostics can now contain "fix-it hints", which are displayed in 255 context underneath the relevant source code. For example: 256 257 fixits.c: In function 'bad_deref': 258 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 259 return ptr.x; 260 ^ 261 -> 262 263 * The C and C++ compilers now offer suggestions for misspelled field 264 names: 265 266 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 267 you mean 'color'? 268 return ptr->colour; 269 ^~~~~~ 270 271 * New command-line options have been added for the C and C++ 272 compilers: 273 + -Wshift-negative-value warns about left shifting a negative 274 value. 275 + -Wshift-overflow warns about left shift overflows. This 276 warning is enabled by default. -Wshift-overflow=2 also warns 277 about left-shifting 1 into the sign bit. 278 + -Wtautological-compare warns if a self-comparison always 279 evaluates to true or false. This warning is enabled by -Wall. 280 + -Wnull-dereference warns if the compiler detects paths that 281 trigger erroneous or undefined behavior due to dereferencing a 282 null pointer. This option is only active when 283 -fdelete-null-pointer-checks is active, which is enabled by 284 optimizations in most targets. The precision of the warnings 285 depends on the optimization options used. 286 + -Wduplicated-cond warns about duplicated conditions in an 287 if-else-if chain. 288 + -Wmisleading-indentation warns about places where the 289 indentation of the code gives a misleading idea of the block 290 structure of the code to a human reader. For example, given 291 [9]CVE-2014-1266: 292 293 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 294 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 295 g-indentation] 296 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 297 ^~ 298 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 299 indented as if it is guarded by the 'if' 300 goto fail; 301 ^~~~ 302 303 This warning is enabled by -Wall. 304 * The C and C++ compilers now emit saner error messages if 305 merge-conflict markers are present in a source file. 306 307 test.c:3:1: error: version control conflict marker in file 308 <<<<<<< HEAD 309 ^~~~~~~ 310 311 C 312 313 * It is possible to disable warnings when an initialized field of a 314 structure or a union with side effects is being overridden when 315 using designated initializers via a new warning option 316 -Woverride-init-side-effects. 317 * A new type attribute scalar_storage_order applying to structures 318 and unions has been introduced. It specifies the storage order (aka 319 endianness) in memory of scalar fields in structures or unions. 320 321 C++ 322 323 * The default mode has been changed to -std=gnu++14. 324 * [10]C++ Concepts are now supported when compiling with -fconcepts. 325 * -flifetime-dse is more aggressive in dead-store elimination in 326 situations where a memory store to a location precedes a 327 constructor to that memory location. 328 * G++ now supports [11]C++17 fold expressions, u8 character literals, 329 extended static_assert, and nested namespace definitions. 330 * G++ now allows constant evaluation for all non-type template 331 arguments. 332 * G++ now supports C++ Transactional Memory when compiling with 333 -fgnu-tm. 334 335 Runtime Library (libstdc++) 336 337 * Extensions to the C++ Library to support mathematical special 338 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 339 * Experimental support for C++17, including the following new 340 features: 341 + std::uncaught_exceptions function (this is also available for 342 -std=gnu++NN modes); 343 + new member functions try_emplace and insert_or_assign for 344 unique_key maps; 345 + non-member functions std::size, std::empty, and std::data for 346 accessing containers and arrays; 347 + std::invoke; 348 + std::shared_mutex; 349 + std::void_t and std::bool_constant metaprogramming utilities. 350 Thanks to Ville Voutilainen for contributing many of the C++17 351 features. 352 * An experimental implementation of the File System TS. 353 * Experimental support for most features of the second version of the 354 Library Fundamentals TS. This includes polymorphic memory resources 355 and array support in shared_ptr, thanks to Fan You. 356 * Some assertions checked by Debug Mode can now also be enabled by 357 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 358 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 359 don't affect the library ABI, so can be enabled per-translation 360 unit. 361 * Timed mutex types are supported on more targets, including Darwin. 362 * Improved std::locale support for DragonFly and FreeBSD, thanks to 363 John Marino and Andreas Tobler. 364 365 Fortran 366 367 * Fortran 2008 SUBMODULE support. 368 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 369 support. 370 * Improved support for Fortran 2003 deferred-length character 371 variables. 372 * Improved support for OpenMP and OpenACC. 373 * The MATMUL intrinsic is now inlined for straightforward cases if 374 front-end optimization is active. The maximum size for inlining can 375 be set to n with the -finline-matmul-limit=n option and turned off 376 with -finline-matmul-limit=0. 377 * The -Wconversion-extra option will warn about REAL constants which 378 have excess precision for their kind. 379 * The -Winteger-division option has been added, which warns about 380 divisions of integer constants which are truncated. This option is 381 included in -Wall by default. 382 383 libgccjit 384 385 * The driver code is now run in-process within libgccjit, providing a 386 small speed-up of the compilation process. 387 * The API has gained entrypoints for 388 + [12]timing how long was spent in different parts of code, 389 + [13]creating switch statements, 390 + [14]allowing unreachable basic blocks in a function, and 391 + [15]adding arbitrary command-line options to a compilation. 392 393 New Targets and Target Specific Improvements 394 395 AArch64 396 397 * A number of AArch64-specific options have been added. The most 398 important ones are summarised in this section; for more detailed 399 information please refer to the documentation. 400 * The command-line options -march=native, -mcpu=native and 401 -mtune=native are now available on native AArch64 GNU/Linux 402 systems. Specifying these options causes GCC to auto-detect the 403 host CPU and choose the optimal setting for that system. 404 * -fpic is now supported when generating code for the small code 405 model (-mcmodel=small). The size of the global offset table (GOT) 406 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 407 ILP32 SysV ABI. 408 * The AArch64 port now supports target attributes and pragmas. Please 409 refer to the [16]documentation for details of available attributes 410 and pragmas as well as usage instructions. 411 * Link-time optimization across translation units with different 412 target-specific options is now supported. 413 * The option -mtls-size= is now supported. It can be used to specify 414 the bit size of TLS offsets, allowing GCC to generate better TLS 415 instruction sequences. 416 * The option -fno-plt is now fully functional. 417 * The ARMv8.1-A architecture and the Large System Extensions are now 418 supported. They can be used by specifying the -march=armv8.1-a 419 option. Additionally, the +lse option extension can be used in a 420 similar fashion to other option extensions. The Large System 421 Extensions introduce new instructions that are used in the 422 implementation of atomic operations. 423 * The ACLE half-precision floating-point type __fp16 is now supported 424 in the C and C++ languages. 425 * The ARM Cortex-A35 processor is now supported via the 426 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 427 equivalent target attributes and pragmas. 428 * The Qualcomm QDF24xx processor is now supported via the 429 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 430 target attributes and pragmas. 431 * Code generation for the ARM Cortex-A57 processor is improved. Among 432 general code generation improvements, a better algorithm is added 433 for allocating registers to floating-point multiply-accumulate 434 instructions offering increased performance when compiling with 435 -mcpu=cortex-a57 or -mtune=cortex-a57. 436 * Code generation for the ARM Cortex-A53 processor is improved. A 437 more accurate instruction scheduling model for the processor is now 438 used, and a number of compiler tuning parameters have been set to 439 offer increased performance when compiling with -mcpu=cortex-a53 or 440 -mtune=cortex-a53. 441 * Code generation for the Samsung Exynos M1 processor is improved. A 442 more accurate instruction scheduling model for the processor is now 443 used, and a number of compiler tuning parameters have been set to 444 offer increased performance when compiling with -mcpu=exynos-m1 or 445 -mtune=exynos-m1. 446 * Improvements in the generation of conditional branches and literal 447 pools allow the compiler to compile functions of a large size. 448 Constant pools are now placed into separate rodata sections. The 449 new option -mpc-relative-literal-loads generates per-function 450 literal pools, limiting the maximum size of functions to 1MiB. 451 * Several correctness issues generating Advanced SIMD instructions 452 for big-endian targets have been fixed resulting in improved code 453 generation for ACLE intrinsics with -mbig-endian. 454 455 ARM 456 457 * Support for revisions of the ARM architecture prior to ARMv4t has 458 been deprecated and will be removed in a future GCC release. The 459 -mcpu and -mtune values that are deprecated are: arm2, arm250, 460 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 461 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 462 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 463 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 464 arm7tdmi is still supported. The values of -march that are 465 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 466 * The ARM port now supports target attributes and pragmas. Please 467 refer to the [17]documentation for details of available attributes 468 and pragmas as well as usage instructions. 469 * Support has been added for the following processors (GCC 470 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 471 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 472 identifiers can be used as arguments to the -mcpu or -mtune 473 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 474 475 Heterogeneous Systems Architecture 476 477 * GCC can now generate HSAIL (Heterogeneous System Architecture 478 Intermediate Language) for simple OpenMP device constructs if 479 configured with --enable-offload-targets=hsa. A new libgomp plugin 480 then runs the HSA GPU kernels implementing these constructs on HSA 481 capable GPUs via a standard HSA run time. 482 If the HSA compilation back end determines it cannot output HSAIL 483 for a particular input, it gives a warning by default. These 484 warnings can be suppressed with -Wno-hsa. To give a few examples, 485 the HSA back end does not implement compilation of code using 486 function pointers, automatic allocation of variable sized arrays, 487 functions with variadic arguments as well as a number of other less 488 common programming constructs. 489 When compilation for HSA is enabled, the compiler attempts to 490 compile composite OpenMP constructs 491 492 #pragma omp target teams distribute parallel for 493 494 into parallel HSA GPU kernels. 495 496 IA-32/x86-64 497 498 * GCC now supports the Intel CPU named Skylake with AVX-512 499 extensions through -march=skylake-avx512. The switch enables the 500 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 501 AVX-512DQ. 502 * Support for new AMD instructions monitorx and mwaitx has been 503 added. This includes new intrinsic and built-in support. It is 504 enabled through option -mmwaitx. The instructions monitorx and 505 mwaitx implement the same functionality as the old monitor and 506 mwait instructions. In addition mwaitx adds a configurable timer. 507 The timer value is received as third argument and stored in 508 register %ebx. 509 * x86-64 targets now allow stack realignment from a word-aligned 510 stack pointer using the command-line option -mstackrealign or 511 __attribute__ ((force_align_arg_pointer)). This allows functions 512 compiled with a vector-aligned stack to be invoked from objects 513 that keep only word-alignment. 514 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 515 can be used to access data via the %fs and %gs segments without 516 having to resort to inline assembly. Please refer to the 517 [18]documentation for usage instructions. 518 * Support for AMD Zen (family 17h) processors is now available 519 through the -march=znver1 and -mtune=znver1 options. 520 521 MeP 522 523 * Support for the MeP (mep-elf) architecture has been deprecated and 524 will be removed in a future GCC release. 525 526 MSP430 527 528 * The MSP430 compiler now has the ability to automatically distribute 529 code and data between low memory (addresses below 64K) and high 530 memory. This only applies to parts that actually have both memory 531 regions and only if the linker script for the part has been 532 specifically set up to support this feature. 533 A new attribute of either can be applied to both functions and 534 data, and this tells the compiler to place the object into low 535 memory if there is room and into high memory otherwise. Two other 536 new attributes - lower and upper - can be used to explicitly state 537 that an object should be placed in the specified memory region. If 538 there is not enough left in that region the compilation will fail. 539 Two new command-line options - -mcode-region=[lower|upper|either] 540 and -mdata-region=[lower|upper|either] - can be used to tell the 541 compiler what to do with objects that do not have one of these new 542 attributes. 543 544 PowerPC / PowerPC64 / RS6000 545 546 * PowerPC64 now supports IEEE 128-bit floating-point using the 547 __float128 data type. In GCC 6, this is not enabled by default, but 548 you can enable it with -mfloat128. The IEEE 128-bit floating-point 549 support requires the use of the VSX instruction set. IEEE 128-bit 550 floating-point values are passed and returned as a single vector 551 value. The software emulator for IEEE 128-bit floating-point 552 support is only built on PowerPC GNU/Linux systems where the 553 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 554 and later), you will be able to use the -mfloat128-hardware option 555 to use the ISA 3.0 instructions that support IEEE 128-bit 556 floating-point. An additional type (__ibm128) has been added to 557 refer to the IBM extended double type that normally implements long 558 double. This will allow for a future transition to implementing 559 long double with IEEE 128-bit floating-point. 560 * Basic support has been added for POWER9 hardware that will use the 561 recently published OpenPOWER ISA 3.0 instructions. The following 562 new switches are available: 563 + -mcpu=power9: Implement all of the ISA 3.0 instructions 564 supported by the compiler. 565 + -mtune=power9: In the future, apply tuning for POWER9 systems. 566 Currently, POWER8 tunings are used. 567 + -mmodulo: Generate code using the ISA 3.0 integer instructions 568 (modulus, count trailing zeros, array index support, integer 569 multiply/add). 570 + -mpower9-fusion: Generate code to suitably fuse instruction 571 sequences for a POWER9 system. 572 + -mpower9-dform: Generate code to use the new D-form 573 (register+offset) memory instructions for the vector 574 registers. 575 + -mpower9-vector: Generate code using the new ISA 3.0 vector 576 (VSX or Altivec) instructions. 577 + -mpower9-minmax: Reserved for future development. 578 + -mtoc-fusion: Keep TOC entries together to provide more fusion 579 opportunities. 580 * New constraints have been added to support IEEE 128-bit 581 floating-point and ISA 3.0 instructions: 582 + wb: Altivec register if -mpower9-dform is enabled. 583 + we: VSX register if -mpower9-vector is enabled for 64-bit code 584 generation. 585 + wo: VSX register if -mpower9-vector is enabled. 586 + wp: Reserved for future use if long double is implemented with 587 IEEE 128-bit floating-point instead of IBM extended double. 588 + wq: VSX register if -mfloat128 is enabled. 589 + wF: Memory operand suitable for POWER9 fusion load/store. 590 + wG: Memory operand suitable for TOC fusion memory references. 591 + wL: Integer constant identifying the element number mfvsrld 592 accesses within a vector. 593 * Support has been added for __builtin_cpu_is() and 594 __builtin_cpu_supports(), allowing for very fast access to 595 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 596 glibc 2.23 or later. 597 * All hardware transactional memory builtins now correctly behave as 598 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 599 determine whether their "old" compiler treats the builtins as 600 barriers. 601 * Split-stack support has been added for gccgo on PowerPC64 for both 602 big- and little-endian (but not for 32-bit). The gold linker from 603 at least binutils 2.25.1 must be available in the PATH when 604 configuring and building gccgo to enable split stack. (The 605 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 606 split-stack feature allows a small initial stack size to be 607 allocated for each goroutine, which increases as needed. 608 * GCC on PowerPC now supports the standard lround function. 609 * A new configuration option ---with-advance-toolchain=at was added 610 for PowerPC 64-bit GNU/Linux systems to use the header files, 611 library files, and the dynamic linker from a specific Advance 612 Toolchain release instead of the default versions that are provided 613 by the GNU/Linux distribution. In general, this option is intended 614 for the developers of GCC, and it is not intended for general use. 615 * The "q", "S", "T", and "t" asm-constraints have been removed. 616 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 617 618 S/390, System z, IBM z Systems 619 620 * Support for the IBM z13 processor has been added. When using the 621 -march=z13 option, the compiler will generate code making use of 622 the new instructions and registers introduced with the vector 623 extension facility. The -mtune=z13 option enables z13 specific 624 instruction scheduling without making use of new instructions. 625 Compiling code with -march=z13 reduces the default alignment of 626 vector types bigger than 8 bytes to 8. This is an ABI change and 627 care must be taken when linking modules compiled with different 628 arch levels which interchange variables containing vector type 629 values. For newly compiled code the GNU linker will emit a warning. 630 * The -mzvector option enables a C/C++ language extension. This 631 extension provides a new keyword vector which can be used to define 632 vector type variables. (Note: This is not available when enforcing 633 strict standard compliance e.g. with -std=c99. Either enable GNU 634 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 635 Additionally a set of overloaded builtins is provided which is 636 partially compatible to the PowerPC Altivec builtins. In order to 637 make use of these builtins the vecintrin.h header file needs to be 638 included. 639 * The new command-line options -march=native, and -mtune=native are 640 now available on native IBM z Systems. Specifying these options 641 causes GCC to auto-detect the host CPU and choose the optimal 642 setting for that system. If GCC is unable to detect the host CPU 643 these options have no effect. 644 * The IBM z Systems port now supports target attributes and pragmas. 645 Please refer to the [19]documentation for details of available 646 attributes and pragmas as well as usage instructions. 647 * -fsplit-stack is now supported as part of the IBM z Systems port. 648 This feature requires a recent gold linker to be used. 649 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 650 been deprecated and will be removed in a future GCC release. -m31 651 from now on defaults to -march=z900 if not specified otherwise. 652 -march=native on a g5/g6 machine will default to -march=z900. 653 654 SH 655 656 * Support for SH5 / SH64 has been declared obsolete and will be 657 removed in future releases. 658 * Support for the FDPIC ABI has been added. It can be enabled using 659 the new -mfdpic target option and --enable-fdpic configure option. 660 661 SPARC 662 663 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 664 change will break binary compatibility with earlier releases for 665 code it affects, but this should be pretty rare in practice. The 666 conditions are: a 16-byte structure containing a double or a 8-byte 667 vector in the second half is passed to a subprogram in slot #15, 668 for example as 16th parameter if the first 15 ones have at most 8 669 bytes. The double or vector was wrongly passed in floating-point 670 register %d32 in lieu of on the stack as per the SPARC calling 671 conventions. 672 673 Operating Systems 674 675 AIX 676 677 * DWARF debugging support for AIX 7.1 has been enabled as an optional 678 debugging format. A more recent Technology Level (TL) and GCC built 679 with that level are required for full exploitation of DWARF 680 debugging capabilities. 681 682 Linux 683 684 * Support for the [20]musl C library was added for the AArch64, ARM, 685 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 686 x86_64 targets. It can be selected using the new -mmusl option in 687 case musl is not the default libc. GCC defaults to musl libc if it 688 is built with a target triplet matching the *-linux-musl* pattern. 689 690 RTEMS 691 692 * The RTEMS thread model implementation changed. Mutexes now use 693 self-contained objects defined in Newlib <sys/lock.h> instead of 694 Classic API semaphores. The keys for thread specific data and the 695 once function are directly defined via <pthread.h>. Self-contained 696 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 697 thread model also supports C++11 threads. 698 * OpenMP support now uses self-contained objects provided by Newlib 699 <sys/lock.h> and offers a significantly better performance compared 700 to the POSIX configuration of libgomp. It is possible to configure 701 thread pools for each scheduler instance via the environment 702 variable GOMP_RTEMS_THREAD_POOLS. 703 704 Solaris 705 706 * Solaris 12 is now fully supported. Minimal support had already been 707 present in GCC 5.3. 708 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 709 crtn.o), which GCC now prefers over its own ones. 710 * Position independent executables (PIE) are now supported on Solaris 711 12. 712 * Constructor priority is now supported on Solaris 12 with the system 713 linker. 714 * libvtv has been ported to Solaris 11 and up. 715 716 Windows 717 718 * The option -mstackrealign is now automatically activated in 32-bit 719 mode whenever the use of SSE instructions is requested. 720 721 Other significant improvements 722 723 * The gcc and g++ driver programs will now provide suggestions for 724 misspelled command-line options. 725 726 $ gcc -static-libfortran test.f95 727 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 728 '-static-libgfortran'? 729 730 * The --enable-default-pie configure option enables generation of PIE 731 by default. 732 733 GCC 6.2 734 735 This is the [21]list of problem reports (PRs) from GCC's bug tracking 736 system that are known to be fixed in the 6.2 release. This list might 737 not be complete (that is, it is possible that some PRs that have been 738 fixed are not listed here). 739 740 Target Specific Changes 741 742 SPARC 743 744 * Support for --with-cpu-32 and --with-cpu-64 configure options has 745 been added on bi-architecture platforms. 746 * Support for the SPARC M7 (Niagara 7) processor has been added. 747 * Support for the VIS 4.0 instruction set has been added. 748 749 GCC 6.3 750 751 This is the [22]list of problem reports (PRs) from GCC's bug tracking 752 system that are known to be fixed in the 6.3 release. This list might 753 not be complete (that is, it is possible that some PRs that have been 754 fixed are not listed here). 755 756 Target Specific Changes 757 758 IA-32/x86-64 759 760 * Support for the [23]deprecated pcommit instruction has been 761 removed. 762 763 GCC 6.4 764 765 This is the [24]list of problem reports (PRs) from GCC's bug tracking 766 system that are known to be fixed in the 6.4 release. This list might 767 not be complete (that is, it is possible that some PRs that have been 768 fixed are not listed here). 769 770 Operating Systems 771 772 RTEMS 773 774 * The ABI changes on ARM so that no short enums are used by default. 775 776 777 For questions related to the use of GCC, please consult these web 778 pages and the [25]GCC manuals. If that fails, the 779 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 780 web pages and the development of GCC are welcome on our developer 781 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 782 archives. 783 784 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 785 distribution of this entire article is permitted in any medium, 786 provided this notice is preserved. 787 788 These pages are [30]maintained by the GCC team. Last modified 789 2017-07-04[31]. 790 791 References 792 793 1. http://gcc.gnu.org/gcc-6/porting_to.html 794 2. http://gcc.gnu.org/onlinedocs/index.html#current 795 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 796 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 797 5. https://www.openacc.org/ 798 6. https://gcc.gnu.org/wiki/OpenACC 799 7. https://gcc.gnu.org/wiki/Offloading 800 8. http://www.openmp.org/specifications/ 801 9. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 802 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 803 11. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 804 12. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 805 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 806 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 807 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 808 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 809 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 810 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 811 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 812 20. http://www.musl-libc.org/ 813 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 814 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 815 23. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 816 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 817 25. https://gcc.gnu.org/onlinedocs/ 818 26. mailto:gcc-help (a] gcc.gnu.org 819 27. mailto:gcc (a] gcc.gnu.org 820 28. https://gcc.gnu.org/lists.html 821 29. http://www.fsf.org/ 822 30. https://gcc.gnu.org/about.html 823 31. http://validator.w3.org/check/referer 824 ====================================================================== 825 http://gcc.gnu.org/gcc-5/index.html 826 827 GCC 5 Release Series 828 829 June 3, 2016 830 831 The [1]GNU project and the GCC developers are pleased to announce the 832 release of GCC 5.4. 833 834 This release is a bug-fix release, containing fixes for regressions in 835 GCC 5.3 relative to previous releases of GCC. 836 837 Release History 838 839 GCC 5.4 840 June 3, 2016 ([2]changes, [3]documentation) 841 842 GCC 5.3 843 December 4, 2015 ([4]changes, [5]documentation) 844 845 GCC 5.2 846 July 16, 2015 ([6]changes, [7]documentation) 847 848 GCC 5.1 849 April 22, 2015 ([8]changes, [9]documentation) 850 851 References and Acknowledgements 852 853 GCC used to stand for the GNU C Compiler, but since the compiler 854 supports several other languages aside from C, it now stands for the 855 GNU Compiler Collection. 856 857 A list of [10]successful builds is updated as new information becomes 858 available. 859 860 The GCC developers would like to thank the numerous people that have 861 contributed new features, improvements, bug fixes, and other changes as 862 well as test results to GCC. This [11]amazing group of volunteers is 863 what makes GCC successful. 864 865 For additional information about GCC please refer to the [12]GCC 866 project web site or contact the [13]GCC development mailing list. 867 868 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 869 870 871 For questions related to the use of GCC, please consult these web 872 pages and the [16]GCC manuals. If that fails, the 873 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 874 web pages and the development of GCC are welcome on our developer 875 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 876 archives. 877 878 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 879 distribution of this entire article is permitted in any medium, 880 provided this notice is preserved. 881 882 These pages are [21]maintained by the GCC team. Last modified 883 2016-06-03[22]. 884 885 References 886 887 1. http://www.gnu.org/ 888 2. http://gcc.gnu.org/gcc-5/changes.html 889 3. http://gcc.gnu.org/onlinedocs/5.4.0/ 890 4. http://gcc.gnu.org/gcc-5/changes.html 891 5. http://gcc.gnu.org/onlinedocs/5.3.0/ 892 6. http://gcc.gnu.org/gcc-5/changes.html 893 7. http://gcc.gnu.org/onlinedocs/5.2.0/ 894 8. http://gcc.gnu.org/gcc-5/changes.html 895 9. http://gcc.gnu.org/onlinedocs/5.1.0/ 896 10. http://gcc.gnu.org/gcc-5/buildstat.html 897 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 898 12. http://gcc.gnu.org/index.html 899 13. mailto:gcc (a] gcc.gnu.org 900 14. http://gcc.gnu.org/mirrors.html 901 15. http://gcc.gnu.org/svn.html 902 16. https://gcc.gnu.org/onlinedocs/ 903 17. mailto:gcc-help (a] gcc.gnu.org 904 18. mailto:gcc (a] gcc.gnu.org 905 19. https://gcc.gnu.org/lists.html 906 20. http://www.fsf.org/ 907 21. https://gcc.gnu.org/about.html 908 22. http://validator.w3.org/check/referer 909 ====================================================================== 910 http://gcc.gnu.org/gcc-5/changes.html 911 912 GCC 5 Release Series 913 Changes, New Features, and Fixes 914 915 Caveats 916 917 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 918 * The C++ runtime library (libstdc++) uses a new ABI by default (see 919 [1]below). 920 * The Graphite framework for loop optimizations no longer requires 921 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 922 The installation manual contains more information about 923 requirements to build GCC. 924 * The non-standard C++0x type traits has_trivial_default_constructor, 925 has_trivial_copy_constructor and has_trivial_copy_assign have been 926 deprecated and will be removed in a future version. The standard 927 C++11 traits is_trivially_default_constructible, 928 is_trivially_copy_constructible and is_trivially_copy_assignable 929 should be used instead. 930 * On AVR, support has been added for the devices 931 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 932 * The AVR port uses a new scheme to describe supported devices: For 933 each supported device the compiler provides a device-specific 934 [2]spec file. If the compiler is used together with AVR-LibC, this 935 requires at least GCC 5.2 and a version of AVR-LibC which 936 implements [3]feature #44574. 937 938 General Optimizer Improvements 939 940 * Inter-procedural optimization improvements: 941 + An Identical Code Folding (ICF) pass (controlled via 942 -fipa-icf) has been added. Compared to the identical code 943 folding performed by the Gold linker this pass does not 944 require function sections. It also performs merging before 945 inlining, so inter-procedural optimizations are aware of the 946 code re-use. On the other hand not all unifications performed 947 by a linker are doable by GCC which must honor aliasing 948 information. During link-time optimization of Firefox, this 949 pass unifies about 31000 functions, that is 14% overall. 950 + The devirtualization pass was significantly improved by adding 951 better support for speculative devirtualization and dynamic 952 type detection. About 50% of virtual calls in Firefox are now 953 speculatively devirtualized during link-time optimization. 954 + A new comdat localization pass allows the linker to eliminate 955 more dead code in presence of C++ inline functions. 956 + Virtual tables are now optimized. Local aliases are used to 957 reduce dynamic linking time of C++ virtual tables on ELF 958 targets and data alignment has been reduced to limit data 959 segment bloat. 960 + A new -fno-semantic-interposition option can be used to 961 improve code quality of shared libraries where interposition 962 of exported symbols is not allowed. 963 + Write-only variables are now detected and optimized out. 964 + With profile feedback the function inliner can now bypass 965 --param inline-insns-auto and --param inline-insns-single 966 limits for hot calls. 967 + The IPA reference pass was significantly sped up making it 968 feasible to enable -fipa-reference with -fprofile-generate. 969 This also solves a bottleneck seen when building Chromium with 970 link-time optimization. 971 + The symbol table and call-graph API was reworked to C++ and 972 simplified. 973 + The interprocedural propagation of constants now also 974 propagates alignments of pointer parameters. This for example 975 means that the vectorizer often does not need to generate loop 976 prologues and epilogues to make up for potential 977 misalignments. 978 * Link-time optimization improvements: 979 + One Definition Rule based merging of C++ types has been 980 implemented. Type merging enables better devirtualization and 981 alias analysis. Streaming extra information needed to merge 982 types adds about 2-6% of memory size and object size increase. 983 This can be controlled by -flto-odr-type-merging. 984 + Command-line optimization and target options are now streamed 985 on a per-function basis and honored by the link-time 986 optimizer. This change makes link-time optimization a more 987 transparent replacement of per-file optimizations. It is now 988 possible to build projects that require different optimization 989 settings for different translation units (such as -ffast-math, 990 -mavx, or -finline). Contrary to earlier GCC releases, the 991 optimization and target options passed on the link command 992 line are ignored. 993 Note that this applies only to those command-line options that 994 can be passed to optimize and target attributes. Command-line 995 options affecting global code generation (such as -fpic), 996 warnings (such as -Wodr), optimizations affecting the way 997 static variables are optimized (such as -fcommon), debug 998 output (such as -g), and --param parameters can be applied 999 only to the whole link-time optimization unit. In these cases, 1000 it is recommended to consistently use the same options at both 1001 compile time and link time. 1002 + GCC bootstrap now uses slim LTO object files. 1003 + Memory usage and link times were improved. Tree merging was 1004 sped up, memory usage of GIMPLE declarations and types was 1005 reduced, and, support for on-demand streaming of variable 1006 constructors was added. 1007 * Feedback directed optimization improvements: 1008 + A new auto-FDO mode uses profiles collected by low overhead 1009 profiling tools (perf) instead of more expensive program 1010 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 1011 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 1012 traditional feedback directed optimization. 1013 + Profile precision was improved in presence of C++ inline and 1014 extern inline functions. 1015 + The new gcov-tool utility allows manipulating profiles. 1016 + Profiles are now more tolerant to source file changes (this 1017 can be controlled by --param profile-func-internal-id). 1018 * Register allocation improvements: 1019 + A new local register allocator (LRA) sub-pass, controlled by 1020 -flra-remat, implements control-flow sensitive global register 1021 rematerialization. Instead of spilling and restoring a 1022 register value, it is recalculated if it is profitable. The 1023 sub-pass improved SPEC2000 generated code by 1% and 0.5% 1024 correspondingly on ARM and x86-64. 1025 + Reuse of the PIC hard register, instead of using a fixed 1026 register, was implemented on x86/x86-64 targets. This improves 1027 generated PIC code performance as more hard registers can be 1028 used. Shared libraries can significantly benefit from this 1029 optimization. Currently it is switched on only for x86/x86-64 1030 targets. As RA infrastructure is already implemented for PIC 1031 register reuse, other targets might follow this in the future. 1032 + A simple form of inter-procedural RA was implemented. When it 1033 is known that a called function does not use caller-saved 1034 registers, save/restore code is not generated around the call 1035 for such registers. This optimization can be controlled by 1036 -fipa-ra 1037 + LRA is now much more effective at generating spills of general 1038 registers into vector registers instead of memory on 1039 architectures (e.g., modern Intel processors) where this is 1040 profitable. 1041 * UndefinedBehaviorSanitizer gained a few new sanitization options: 1042 + -fsanitize=float-divide-by-zero: detect floating-point 1043 division by zero; 1044 + -fsanitize=float-cast-overflow: check that the result of 1045 floating-point type to integer conversions do not overflow; 1046 + -fsanitize=bounds: enable instrumentation of array bounds and 1047 detect out-of-bounds accesses; 1048 + -fsanitize=alignment: enable alignment checking, detect 1049 various misaligned objects; 1050 + -fsanitize=object-size: enable object size checking, detect 1051 various out-of-bounds accesses. 1052 + -fsanitize=vptr: enable checking of C++ member function calls, 1053 member accesses and some conversions between pointers to base 1054 and derived classes, detect if the referenced object does not 1055 have the correct dynamic type. 1056 * Pointer Bounds Checker, a bounds violation detector, has been added 1057 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 1058 instrumented with run-time checks of used pointers against their 1059 bounds to detect pointer bounds violations (overflows). The Pointer 1060 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 1061 new ISA extension Intel MPX support. See the Pointer Bounds Checker 1062 [4]Wiki page for more details. 1063 1064 New Languages and Language specific improvements 1065 1066 * [5]OpenMP 4.0 specification offloading features are now supported 1067 by the C, C++, and Fortran compilers. Generic changes: 1068 + Infrastructure (suitable for any vendor). 1069 + Testsuite which covers offloading from the [6]OpenMP 4.0 1070 Examples document. 1071 Specific for upcoming Intel Xeon Phi products: 1072 + Run-time library. 1073 + Card emulator. 1074 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 1075 specification. OpenACC is intended for programming accelerator 1076 devices such as GPUs. See [7]the OpenACC wiki page for more 1077 information. 1078 1079 C family 1080 1081 * The default setting of the -fdiagnostics-color= command-line option 1082 is now [8]configurable when building GCC using configuration option 1083 --with-diagnostics-color=. The possible values are: never, always, 1084 auto and auto-if-env. The new default auto uses color only when the 1085 standard error is a terminal. The default in GCC 4.9 was 1086 auto-if-env, which is equivalent to auto if there is a non-empty 1087 GCC_COLORS environment variable, and never otherwise. As in GCC 1088 4.9, an empty GCC_COLORS variable in the environment will always 1089 disable colors, no matter what the default is or what command-line 1090 options are used. 1091 * A new command-line option -Wswitch-bool has been added for the C 1092 and C++ compilers, which warns whenever a switch statement has an 1093 index of boolean type. 1094 * A new command-line option -Wlogical-not-parentheses has been added 1095 for the C and C++ compilers, which warns about "logical not" used 1096 on the left hand side operand of a comparison. 1097 * A new command-line option -Wsizeof-array-argument has been added 1098 for the C and C++ compilers, which warns when the sizeof operator 1099 is applied to a parameter that has been declared as an array in a 1100 function definition. 1101 * A new command-line option -Wbool-compare has been added for the C 1102 and C++ compilers, which warns about boolean expressions compared 1103 with an integer value different from true/false. 1104 * Full support for [9]Cilk Plus has been added to the GCC compiler. 1105 Cilk Plus is an extension to the C and C++ languages to support 1106 data and task parallelism. 1107 * A new attribute no_reorder prevents reordering of selected symbols 1108 against other such symbols or inline assembler. This enables to 1109 link-time optimize the Linux kernel without having to resort to 1110 -fno-toplevel-reorder that disables several optimizations. 1111 * New preprocessor constructs, __has_include and __has_include_next, 1112 to test the availability of headers have been added. 1113 This demonstrates a way to include the header <optional> only if it 1114 is available: 1115 1116 #ifdef __has_include 1117 # if __has_include(<optional>) 1118 # include <optional> 1119 # define have_optional 1 1120 # elif __has_include(<experimental/optional>) 1121 # include <experimental/optional> 1122 # define have_optional 1 1123 # define experimental_optional 1124 # else 1125 # define have_optional 0 1126 # endif 1127 #endif 1128 1129 The header search paths for __has_include and __has_include_next 1130 are equivalent to those of the standard directive #include and the 1131 extension #include_next respectively. 1132 * A new built-in function-like macro to determine the existence of an 1133 attribute, __has_attribute, has been added. The equivalent built-in 1134 macro __has_cpp_attribute was added to C++ to support 1135 [10]Feature-testing recommendations for C++. The macro 1136 __has_attribute is added to all C-like languages as an extension: 1137 1138 int 1139 #ifdef __has_attribute 1140 # if __has_attribute(__noinline__) 1141 __attribute__((__noinline__)) 1142 # endif 1143 #endif 1144 foo(int x); 1145 1146 If an attribute exists, a nonzero constant integer is returned. For 1147 standardized C++ attributes a date is returned, otherwise the 1148 constant returned is 1. Both __has_attribute and 1149 __has_cpp_attribute will add underscores to an attribute name if 1150 necessary to resolve the name. For C++11 and onwards the attribute 1151 may be scoped. 1152 * A new set of built-in functions for arithmetics with overflow 1153 checking has been added: __builtin_add_overflow, 1154 __builtin_sub_overflow and __builtin_mul_overflow and for 1155 compatibility with clang also other variants. These builtins have 1156 two integral arguments (which don't need to have the same type), 1157 the arguments are extended to infinite precision signed type, +, - 1158 or * is performed on those, and the result is stored in an integer 1159 variable pointed to by the last argument. If the stored value is 1160 equal to the infinite precision result, the built-in functions 1161 return false, otherwise true. The type of the integer variable that 1162 will hold the result can be different from the types of the first 1163 two arguments. The following snippet demonstrates how this can be 1164 used in computing the size for the calloc function: 1165 1166 void * 1167 calloc (size_t x, size_t y) 1168 { 1169 size_t sz; 1170 if (__builtin_mul_overflow (x, y, &sz)) 1171 return NULL; 1172 void *ret = malloc (sz); 1173 if (ret) memset (res, 0, sz); 1174 return ret; 1175 } 1176 1177 On e.g. i?86 or x86-64 the above will result in a mul instruction 1178 followed by a jump on overflow. 1179 * The option -fextended-identifiers is now enabled by default for 1180 C++, and for C99 and later C versions. Various bugs in the 1181 implementation of extended identifiers have been fixed. 1182 1183 C 1184 1185 * The default mode has been changed to -std=gnu11. 1186 * A new command-line option -Wc90-c99-compat has been added to warn 1187 about features not present in ISO C90, but present in ISO C99. 1188 * A new command-line option -Wc99-c11-compat has been added to warn 1189 about features not present in ISO C99, but present in ISO C11. 1190 * It is possible to disable warnings about conversions between 1191 pointers that have incompatible types via a new warning option 1192 -Wno-incompatible-pointer-types; warnings about implicit 1193 incompatible integer to pointer and pointer to integer conversions 1194 via a new warning option -Wno-int-conversion; and warnings about 1195 qualifiers on pointers being discarded via a new warning option 1196 -Wno-discarded-qualifiers. 1197 * To allow proper use of const qualifiers with multidimensional 1198 arrays, GCC will not warn about incompatible pointer types anymore 1199 for conversions between pointers to arrays with and without const 1200 qualifier (except when using -pedantic). Instead, a new warning is 1201 emitted only if the const qualifier is lost. This can be controlled 1202 with a new warning option -Wno-discarded-array-qualifiers. 1203 * The C front end now generates more precise caret diagnostics. 1204 * The -pg command-line option now only affects the current file in an 1205 LTO build. 1206 1207 C++ 1208 1209 * G++ now supports [11]C++14 variable templates. 1210 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 1211 * Excessive template instantiation depth is now a fatal error. This 1212 prevents excessive diagnostics that usually do not help to identify 1213 the problem. 1214 * G++ and libstdc++ now implement the feature-testing macros from 1215 [12]Feature-testing recommendations for C++. 1216 * G++ now allows typename in a template template parameter. 1217 1218 template<template<typename> typename X> struct D; // OK 1219 1220 * G++ now supports [13]C++14 aggregates with non-static data member 1221 initializers. 1222 1223 struct A { int i, j = i; }; 1224 A a = { 42 }; // a.j is also 42 1225 1226 * G++ now supports [14]C++14 extended constexpr. 1227 1228 constexpr int f (int i) 1229 { 1230 int j = 0; 1231 for (; i > 0; --i) 1232 ++j; 1233 return j; 1234 } 1235 1236 constexpr int i = f(42); // i is 42 1237 1238 * G++ now supports the [15]C++14 sized deallocation functions. 1239 1240 void operator delete (void *, std::size_t) noexcept; 1241 void operator delete[] (void *, std::size_t) noexcept; 1242 1243 * A new One Definition Rule violation warning (controlled by -Wodr) 1244 detects mismatches in type definitions and virtual table contents 1245 during link-time optimization. 1246 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 1247 developers to annotate programs with final specifiers (or anonymous 1248 namespaces) to improve code generation. These warnings can be used 1249 at compile time, but they are more useful in combination with 1250 link-time optimization. 1251 * G++ no longer supports [16]N3639 variable length arrays, as they 1252 were removed from the C++14 working paper prior to ratification. 1253 GNU VLAs are still supported, so VLA support is now the same in 1254 C++14 mode as in C++98 and C++11 modes. 1255 * G++ now allows passing a non-trivially-copyable class via C 1256 varargs, which is conditionally-supported with 1257 implementation-defined semantics in the standard. This uses the 1258 same calling convention as a normal value parameter. 1259 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 1260 various mangling bugs are fixed, but G++ will still emit aliases 1261 with the old, wrong mangling where feasible. -Wabi=2 will warn 1262 about differences between ABI version 2 and the current setting. 1263 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 1264 to be unaffected, but -Wabi=8 will warn about a non-static data 1265 member with type std::nullptr_t which changes position due to this 1266 change. 1267 1268 Runtime Library (libstdc++) 1269 1270 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 1271 default. The old ABI is still supported and can be used by defining 1272 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 1273 standard library headers. 1274 * A new implementation of std::string is enabled by default, using 1275 the small string optimization instead of copy-on-write reference 1276 counting. 1277 * A new implementation of std::list is enabled by default, with an 1278 O(1) size() function; 1279 * [18]Full support for C++11, including the following new features: 1280 + std::deque and std::vector<bool> meet the allocator-aware 1281 container requirements; 1282 + movable and swappable iostream classes; 1283 + support for std::align and std::aligned_union; 1284 + type traits std::is_trivially_copyable, 1285 std::is_trivially_constructible, std::is_trivially_assignable 1286 etc.; 1287 + I/O manipulators std::put_time, std::get_time, std::hexfloat 1288 and std::defaultfloat; 1289 + generic locale-aware std::isblank; 1290 + locale facets for Unicode conversion; 1291 + atomic operations for std::shared_ptr; 1292 + std::notify_all_at_thread_exit() and functions for making 1293 futures ready at thread exit. 1294 * Support for the C++11 hexfloat manipulator changes how the num_put 1295 facet formats floating point types when 1296 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 1297 This change affects all language modes, even though the C++98 1298 standard gave no special meaning to that combination of flags. To 1299 prevent the use of hexadecimal notation for floating point types 1300 use str.unsetf(std::ios_base::floatfield) to clear the relevant 1301 bits in str.flags(). 1302 * [19]Full experimental support for C++14, including the following 1303 new features: 1304 + std::is_final type trait; 1305 + heterogeneous comparison lookup in associative containers. 1306 + global functions cbegin, cend, rbegin, rend, crbegin, and 1307 crend for range access to containers, arrays and initializer 1308 lists. 1309 * [20]Improved experimental support for the Library Fundamentals TS, 1310 including: 1311 + class std::experimental::any; 1312 + function template std::experimental::apply; 1313 + function template std::experimental::sample; 1314 + function template std::experimental::search and related 1315 searcher types; 1316 + variable templates for type traits; 1317 + function template std::experimental::not_fn. 1318 * New random number distributions logistic_distribution and 1319 uniform_on_sphere_distribution as extensions. 1320 * [21]GDB Xmethods for containers and std::unique_ptr. 1321 1322 Fortran 1323 1324 * Compatibility notice: 1325 + The version of the module files (.mod) has been incremented. 1326 + For free-form source files [22]-Werror=line-truncation is now 1327 enabled by default. Note that comments exceeding the line 1328 length are not diagnosed. (For fixed-form source code, the 1329 same warning is available but turned off by default, such that 1330 excess characters are ignored. -ffree-line-length-n and 1331 -ffixed-line-length-n can be used to modify the default line 1332 lengths of 132 and 72 columns, respectively.) 1333 + The -Wtabs option is now more sensible: with -Wtabs the 1334 compiler warns if it encounters tabs and with -Wno-tabs this 1335 warning is turned off. Before, -Wno-tabs warned and -Wtabs 1336 disabled the warning. As before, this warning is also enabled 1337 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 1338 options of -std=. 1339 * Incomplete support for colorizing diagnostics emitted by gfortran 1340 has been added. The option [23]-fdiagnostics-color controls when 1341 color is used in diagnostics. The default value of this option can 1342 be [24]configured when building GCC. The GCC_COLORS environment 1343 variable can be used to customize the colors or disable coloring 1344 completely. Sample diagnostics output: 1345 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 1346 test.f90:6:1: 1347 1348 0 continue 1349 1 1350 Error: Zero is not a valid statement label at (1) 1351 test.f90:9:6: 1352 1353 USE foo 1354 1 1355 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 1356 1357 * The -Wuse-without-only option has been added to warn when a USE 1358 statement has no ONLY qualifier and thus implicitly imports all 1359 public entities of the used module. 1360 * Formatted READ and WRITE statements now work correctly in 1361 locale-aware programs. For more information and potential caveats, 1362 see [25]Section 5.3 Thread-safety of the runtime library in the 1363 manual. 1364 * [26]Fortran 2003: 1365 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 1366 IEEE_ARITHMETIC) are now supported. 1367 * [27]Fortran 2008: 1368 + [28]Coarrays: Full experimental support of Fortran 2008's 1369 coarrays with -fcoarray=lib except for allocatable/pointer 1370 components of derived-type coarrays. GCC currently only ships 1371 with a single-image library (libcaf_single), but multi-image 1372 support based on MPI and GASNet is provided by the libraries 1373 of the [29]OpenCoarrays project. 1374 * TS18508 Additional Parallel Features in Fortran: 1375 + Support for the collective intrinsic subroutines CO_MAX, 1376 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 1377 including -fcoarray=lib support. 1378 + Support for the new atomic intrinsics has been added, 1379 including -fcoarray=lib support. 1380 * Fortran 2015: 1381 + Support for IMPLICIT NONE (external, type). 1382 + ERROR STOP is now permitted in pure procedures. 1383 1384 Go 1385 1386 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 1387 * Building GCC 5 with Go enabled will install two new programs: 1388 [30]go and [31]gofmt. 1389 1390 libgccjit 1391 1392 New in GCC 5 is the ability to build GCC as a shared library for 1393 embedding in other processes (such as interpreters), suitable for 1394 Just-In-Time compilation to machine code. 1395 1396 The shared library has a [32]C API and a [33]C++ wrapper API providing 1397 some "syntactic sugar". There are also bindings available from 3rd 1398 parties for [34]Python and for [35]D. 1399 1400 For example, this library can be used by interpreters for [36]compiling 1401 functions from bytecode to machine code. 1402 1403 The library can also be used for ahead-of-time compilation, enabling 1404 GCC to be plugged into a pre-existing frontend. An example of using 1405 this to build a compiler for an esoteric language we'll refer to as 1406 "brainf" can be seen [37]here. 1407 1408 libgccjit is licensed under the GPLv3 (or at your option, any later 1409 version) 1410 1411 It should be regarded as experimental at this time. 1412 1413 New Targets and Target Specific Improvements 1414 1415 Reporting stack usage 1416 1417 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 1418 -fstack-usage option. 1419 1420 AArch64 1421 1422 * Code generation for the ARM Cortex-A57 processor has been improved. 1423 A more accurate instruction scheduling model for the processor is 1424 now used, and a number of compiler tuning parameters have been set 1425 to offer increased performance when compiling with -mcpu=cortex-a57 1426 or -mtune=cortex-a57. 1427 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 1428 and can be enabled by giving the -mfix-cortex-a53-835769 option. 1429 Alternatively it can be enabled by default by configuring GCC with 1430 the --enable-fix-cortex-a53-835769 option. 1431 * The optional cryptographic extensions to the ARMv8-A architecture 1432 are no longer enabled by default when specifying the 1433 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 1434 options. To enable these extensions add +crypto to the value of 1435 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 1436 * Support has been added for the following processors (GCC 1437 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 1438 initial support for its big.LITTLE combination with the ARM 1439 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 1440 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 1441 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1442 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 1443 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 1444 support for the Cortex-A72. 1445 * The transitional options -mlra and -mno-lra have been removed. The 1446 AArch64 backend now uses the local register allocator (LRA) only. 1447 1448 ARM 1449 1450 * Thumb-1 assembly code is now generated in unified syntax. The new 1451 option -masm-syntax-unified specifies whether inline assembly code 1452 is using unified syntax. By default the option is off which means 1453 non-unified syntax is used. However this is subject to change in 1454 future releases. Eventually the non-unified syntax will be 1455 deprecated. 1456 * It is now a configure-time error to use the --with-cpu configure 1457 option with either of --with-tune or --with-arch. 1458 * Code generation for the ARM Cortex-A57 processor has been improved. 1459 A more accurate instruction scheduling model for the processor is 1460 now used, and a number of compiler tuning parameters have been set 1461 to offer increased performance when compiling with -mcpu=cortex-a57 1462 or -mtune=cortex-a57. 1463 * Support has been added for the following processors (GCC 1464 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 1465 initial support for its big.LITTLE combination with the ARM 1466 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 1467 initial support for its big.LITTLE combination with the ARM 1468 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 1469 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 1470 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1471 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 1472 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 1473 support for the Cortex-A72. 1474 * The deprecated option -mwords-little-endian has been removed. 1475 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 1476 -mtpcs-leaf-frame which are only applicable to the old ABI have 1477 been deprecated. 1478 * The transitional options -mlra and -mno-lra have been removed. The 1479 ARM backend now uses the local register allocator (LRA) only. 1480 1481 AVR 1482 1483 * The compiler no more supports individual devices like ATmega8. 1484 Specifying, say, -mmcu=atmega8 triggers the usage of the 1485 device-specific [38]spec file specs-atmega8 which is part of the 1486 installation and describes options for the sub-processes like 1487 compiler proper, assembler and linker. You can add support for a 1488 new device -mmcu=mydevice as follows: 1489 1. In an empty directory /someplace, create a new directory 1490 device-specs. 1491 2. Copy a device spec file from the installed device-specs 1492 folder, follow the comments in that file and then save it as 1493 /someplace/device-specs/specs-mydevice. 1494 3. Add -B /someplace -mmcu=mydevice to the compiler's 1495 command-line options. Notice that /someplace must specify an 1496 absolute path and that mydevice must not start with "avr". 1497 4. Provided you have a device-specific library libmydevice.a 1498 available, you can put it at /someplace, dito for a 1499 device-specific startup file crtmydevice.o. 1500 The contents of the device spec files depend on the compiler's 1501 configuration, in particular on --with-avrlibc=no and whether or 1502 not it is configured for RTEMS. 1503 * A new command-line option -nodevicelib has been added. It prevents 1504 the compiler from linking against AVR-LibC's device-specific 1505 library libdevice.a. 1506 * The following three command-line options have been added: 1507 1508 -mrmw 1509 Set if the device supports the read-modify-write 1510 instructions LAC, LAS, LAT and XCH. 1511 1512 -mn-flash=size 1513 Specify the flash size of the device in units of 64 KiB, 1514 rounded up to the next integer as needed. This option 1515 affects the availability of the [39]AVR address-spaces. 1516 1517 -mskip-bug 1518 Set if the device is affected by the respective silicon 1519 bug. 1520 1521 In general, you don't need to set these options by hand. The new 1522 device-specific spec file will set them as needed. 1523 1524 IA-32/x86-64 1525 1526 * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 1527 Intel's CPU codenamed Skylake Server was added to GCC. That 1528 includes inline assembly support, new intrinsics, and basic 1529 autovectorization. These new AVX-512 extensions are available via 1530 the following GCC switches: AVX-512 Vector Length EVEX feature: 1531 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 1532 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 1533 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 1534 Instructions: -mavx512vbmi. 1535 * New ISA extensions support Intel MPX was added to GCC. This new 1536 extension is available via the -mmpx compiler switch. Intel MPX is 1537 a set of processor features which, with compiler, run-time library 1538 and OS support, brings increased robustness to software by run-time 1539 checking pointer references against their bounds. In GCC Intel MPX 1540 is supported by Pointer Bounds Checker and libmpx run-time 1541 libraries. 1542 * The new -mrecord-mcount option for -pg generates a Linux kernel 1543 style table of pointers to mcount or __fentry__ calls at the 1544 beginning of functions. The new -mnop-mcount option in addition 1545 also generates nops in place of the __fentry__ or mcount call, so 1546 that a call per function can be later patched in. This can be used 1547 for low overhead tracing or hot code patching. 1548 * The new -malign-data option controls how GCC aligns variables. 1549 -malign-data=compat uses increased alignment compatible with GCC 1550 4.8 and earlier, -malign-data=abi uses alignment as specified by 1551 the psABI, and -malign-data=cacheline uses increased alignment to 1552 match the cache line size. -malign-data=compat is the default. 1553 * The new -mskip-rax-setup option skips setting up the RAX register 1554 when SSE is disabled and there are no variable arguments passed in 1555 vector registers. This can be used to optimize the Linux kernel. 1556 1557 MIPS 1558 1559 * MIPS Releases 3 and 5 are now directly supported. Use the 1560 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 1561 to enable code-generation for these processors. 1562 * The Imagination P5600 processor is now supported using the 1563 -march=p5600 command-line option. 1564 * The Cavium Octeon3 processor is now supported using the 1565 -march=octeon3 command-line option. 1566 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 1567 command-line options. 1568 * The o32 ABI has been modified and extended. The o32 64-bit 1569 floating-point register support is now obsolete and has been 1570 removed. It has been replaced by three ABI extensions FPXX, FP64A, 1571 and FP64. The meaning of the -mfp64 command-line option has 1572 changed. It is now used to enable the FP64A and FP64 ABI 1573 extensions. 1574 + The FPXX extension requires that code generated to access 1575 double-precision values use even-numbered registers. Code that 1576 adheres to this extension is link-compatible with all other 1577 o32 double-precision ABI variants and will execute correctly 1578 in all hardware FPU modes. The command-line options -mabi=32 1579 -mfpxx can be used to enable this extension. MIPS II is the 1580 minimum processor required. 1581 + The o32 FP64A extension requires that floating-point registers 1582 be 64-bit and odd-numbered single-precision registers are not 1583 allowed. Code that adheres to the o32 FP64A variant is 1584 link-compatible with all other o32 double-precision ABI 1585 variants. The command-line options -mabi=32 -mfp64 1586 -mno-odd-spreg can be used to enable this extension. MIPS32R2 1587 is the minimum processor required. 1588 + The o32 FP64 extension also requires that floating-point 1589 registers be 64-bit, but permits the use of single-precision 1590 registers. Code that adheres to the o32 FP64 variant is 1591 link-compatible with o32 FPXX and o32 FP64A variants only, 1592 i.e. it is not compatible with the original o32 1593 double-precision ABI. The command-line options -mabi=32 -mfp64 1594 -modd-spreg can be used to enable this extension. MIPS32R2 is 1595 the minimum processor required. 1596 The new ABI variants can be enabled by default using the configure 1597 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 1598 It is strongly recommended that all vendors begin to set o32 FPXX 1599 as the default ABI. This will be required to run the generated code 1600 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 1601 MIPSR6 cores. 1602 * GCC will now pass all floating-point options to the assembler if 1603 GNU binutils 2.25 is used. As a result, any inline assembly code 1604 that uses hard-float instructions should be amended to include a 1605 .set directive to override the global assembler options when 1606 compiling for soft-float targets. 1607 1608 NDS32 1609 1610 * The variadic function ABI implementation is now compatible with 1611 past Andes toolchains where the caller uses registers to pass 1612 arguments and the callee is in charge of pushing them on stack. 1613 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 1614 been removed since they are not yet available in the nds32 port of 1615 GNU binutils. 1616 * A new option -mcmodel=[small|medium|large] supports varied code 1617 models on code generation. The -mgp-direct option became 1618 meaningless and can be discarded. 1619 1620 RX 1621 1622 * A new command line option -mno-allow-string-insns can be used to 1623 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 1624 SWHILE and RMPA instructions. An erratum released by Renesas shows 1625 that it is unsafe to use these instructions on addresses within the 1626 I/O space of the processor. The new option can be used when the 1627 programmer is concerned that the I/O space might be accessed. The 1628 default is still to enable these instructions. 1629 1630 SH 1631 1632 * The compiler will now pass the appropriate --isa= option to the 1633 assembler. 1634 * The default handling for the GBR has been changed from call 1635 clobbered to call preserved. The old behavior can be reinstated by 1636 specifying the option -fcall-used-gbr. 1637 * Support for the SH4A fpchg instruction has been added which will be 1638 utilized when switching between single and double precision FPU 1639 modes. 1640 * The compiler no longer uses the __fpscr_values array for switching 1641 between single and double FPU precision modes on non-SH4A targets. 1642 Instead mode switching will now be performed by storing, modifying 1643 and reloading the FPSCR, so that other FPSCR bits are preserved 1644 across mode switches. The __fpscr_values array that is defined in 1645 libgcc is still present for backwards compatibility, but it will 1646 not be referenced by compiler generated code anymore. 1647 * New builtin functions __builtin_sh_get_fpscr and 1648 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 1649 function will mask the specified bits in such a way that the SZ, PR 1650 and FR mode bits will be preserved, while changing the other bits. 1651 These new functions do not reference the __fpscr_values array. The 1652 old functions __set_fpscr and __get_fpscr in libgcc which access 1653 the __fpscr_values array are still present for backwards 1654 compatibility, but their usage is highly discouraged. 1655 * Some improvements to code generated for __atomic built-in 1656 functions. 1657 * When compiling for SH2E the compiler will no longer force the usage 1658 of delay slots for conditional branch instructions bt and bf. The 1659 old behavior can be reinstated (e.g. to work around a hardware bug 1660 in the original SH7055) by specifying the new option 1661 -mcbranch-force-delay-slot. 1662 1663 Operating Systems 1664 1665 AIX 1666 1667 * GCC now supports stabs debugging continuation lines to allow long 1668 stabs debug information without overflow that generates AIX linker 1669 errors. 1670 1671 DragonFly BSD 1672 1673 * GCC now supports the DragonFly BSD operating system. 1674 1675 FreeBSD 1676 1677 * GCC now supports the FreeBSD operating system for the arm port 1678 through the arm*-*-freebsd* target triplets. 1679 1680 VxWorks MILS 1681 1682 * GCC now supports the MILS (Multiple Independent Levels of Security) 1683 variant of WindRiver's VxWorks operating system for PowerPC 1684 targets. 1685 1686 Other significant improvements 1687 1688 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 1689 to set the compiler to use. 1690 1691 * When the new command-line option -freport-bug is used, GCC 1692 automatically generates a developer-friendly reproducer whenever an 1693 internal compiler error is encountered. 1694 1695 GCC 5.2 1696 1697 This is the [41]list of problem reports (PRs) from GCC's bug tracking 1698 system that are known to be fixed in the 5.2 release. This list might 1699 not be complete (that is, it is possible that some PRs that have been 1700 fixed are not listed here). 1701 1702 Target Specific Changes 1703 1704 IA-32/x86-64 1705 1706 * Support for new AMD instructions monitorx and mwaitx has been 1707 added. This includes new intrinsic and built-in support. It is 1708 enabled through option -mmwaitx. The instructions monitorx and 1709 mwaitx implement the same functionality as the old monitor and 1710 mwait instructions. In addition mwaitx adds a configurable timer. 1711 The timer value is received as third argument and stored in 1712 register %ebx. 1713 1714 S/390, System z, IBM z Systems 1715 1716 * Support for the IBM z13 processor has been added. When using the 1717 -march=z13 option, the compiler will generate code making use of 1718 the new instructions and registers introduced with the vector 1719 extension facility. The -mtune=z13 option enables z13 specific 1720 instruction scheduling without making use of new instructions. 1721 Compiling code with -march=z13 reduces the default alignment of 1722 vector types bigger than 8 bytes to 8. This is an ABI change and 1723 care must be taken when linking modules compiled with different 1724 arch levels which interchange variables containing vector type 1725 values. For newly compiled code the GNU linker will emit a warning. 1726 * The -mzvector option enables a C/C++ language extension. This 1727 extension provides a new keyword vector which can be used to define 1728 vector type variables. (Note: This is not available when enforcing 1729 strict standard compliance e.g. with -std=c99. Either enable GNU 1730 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 1731 Additionally a set of overloaded builtins is provided which is 1732 partially compatible to the PowerPC Altivec builtins. In order to 1733 make use of these builtins the vecintrin.h header file needs to be 1734 included. 1735 1736 GCC 5.3 1737 1738 This is the [42]list of problem reports (PRs) from GCC's bug tracking 1739 system that are known to be fixed in the 5.3 release. This list might 1740 not be complete (that is, it is possible that some PRs that have been 1741 fixed are not listed here). 1742 1743 Target Specific Changes 1744 1745 IA-32/x86-64 1746 1747 * GCC now supports the Intel CPU named Skylake with AVX-512 1748 extensions through -march=skylake-avx512. The switch enables the 1749 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 1750 AVX-512DQ. 1751 1752 S/390, System z, IBM z Systems 1753 1754 * With this version of GCC IBM z Systems support has been added to 1755 the GO runtime environment. GCC 5.3 has proven to be able to 1756 compile larger GO applications on IBM z Systems. 1757 1758 GCC 5.4 1759 1760 This is the [43]list of problem reports (PRs) from GCC's bug tracking 1761 system that are known to be fixed in the 5.4 release. This list might 1762 not be complete (that is, it is possible that some PRs that have been 1763 fixed are not listed here). 1764 1765 (Pending) GCC 5.5 1766 1767 This is the [44]list of problem reports (PRs) from GCC's bug tracking 1768 system that are known to be fixed in the 5.5 release. This list might 1769 not be complete (that is, it is possible that some PRs that have been 1770 fixed are not listed here). 1771 1772 Target Specific Changes 1773 1774 IA-32/x86-64 1775 1776 * Support for the [45]deprecated pcommit instruction has been 1777 removed. 1778 1779 1780 For questions related to the use of GCC, please consult these web 1781 pages and the [46]GCC manuals. If that fails, the 1782 [47]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1783 web pages and the development of GCC are welcome on our developer 1784 list at [48]gcc (a] gcc.gnu.org. All of [49]our lists have public 1785 archives. 1786 1787 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 1788 distribution of this entire article is permitted in any medium, 1789 provided this notice is preserved. 1790 1791 These pages are [51]maintained by the GCC team. Last modified 1792 2017-03-12[52]. 1793 1794 References 1795 1796 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 1797 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 1798 3. https://savannah.nongnu.org/bugs/?44574 1799 4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 1800 5. http://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 1801 6. http://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 1802 7. https://gcc.gnu.org/wiki/OpenACC 1803 8. https://gcc.gnu.org/install/configure.html 1804 9. https://www.cilkplus.org/ 1805 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 1806 11. http://gcc.gnu.org/projects/cxx1y.html 1807 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 1808 13. http://gcc.gnu.org/projects/cxx1y.html 1809 14. http://gcc.gnu.org/projects/cxx1y.html 1810 15. http://gcc.gnu.org/projects/cxx1y.html 1811 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 1812 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 1813 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 1814 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 1815 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 1816 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 1817 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 1818 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 1819 24. https://gcc.gnu.org/install/configure.html 1820 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 1821 26. https://gcc.gnu.org/wiki/Fortran2003Status 1822 27. https://gcc.gnu.org/wiki/Fortran2008Status 1823 28. https://gcc.gnu.org/wiki/Coarray 1824 29. http://www.opencoarrays.org/ 1825 30. https://golang.org/cmd/go/ 1826 31. https://golang.org/cmd/gofmt/ 1827 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 1828 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 1829 34. https://github.com/davidmalcolm/pygccjit 1830 35. https://github.com/ibuclaw/gccjitd 1831 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 1832 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 1833 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 1834 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 1835 40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 1836 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 1837 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 1838 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 1839 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 1840 45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 1841 46. https://gcc.gnu.org/onlinedocs/ 1842 47. mailto:gcc-help (a] gcc.gnu.org 1843 48. mailto:gcc (a] gcc.gnu.org 1844 49. https://gcc.gnu.org/lists.html 1845 50. http://www.fsf.org/ 1846 51. https://gcc.gnu.org/about.html 1847 52. http://validator.w3.org/check/referer 1848 ====================================================================== 1849 http://gcc.gnu.org/gcc-4.9/index.html 1850 1851 GCC 4.9 Release Series 1852 1853 Aug 3, 2016 1854 1855 The [1]GNU project and the GCC developers are pleased to announce the 1856 release of GCC 4.9.4. 1857 1858 This release is a bug-fix release, containing fixes for regressions in 1859 GCC 4.9.3 relative to previous releases of GCC. 1860 1861 This release series is no longer maintained. 1862 1863 Release History 1864 1865 GCC 4.9.4 1866 Aug 3, 2016 ([2]changes, [3]documentation) 1867 1868 GCC 4.9.3 1869 June 26, 2015 ([4]changes, [5]documentation) 1870 1871 GCC 4.9.2 1872 October 30, 2014 ([6]changes, [7]documentation) 1873 1874 GCC 4.9.1 1875 July 16, 2014 ([8]changes, [9]documentation) 1876 1877 GCC 4.9.0 1878 April 22, 2014 ([10]changes, [11]documentation) 1879 1880 References and Acknowledgements 1881 1882 GCC used to stand for the GNU C Compiler, but since the compiler 1883 supports several other languages aside from C, it now stands for the 1884 GNU Compiler Collection. 1885 1886 A list of [12]successful builds is updated as new information becomes 1887 available. 1888 1889 The GCC developers would like to thank the numerous people that have 1890 contributed new features, improvements, bug fixes, and other changes as 1891 well as test results to GCC. This [13]amazing group of volunteers is 1892 what makes GCC successful. 1893 1894 For additional information about GCC please refer to the [14]GCC 1895 project web site or contact the [15]GCC development mailing list. 1896 1897 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 1898 1899 1900 For questions related to the use of GCC, please consult these web 1901 pages and the [18]GCC manuals. If that fails, the 1902 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1903 web pages and the development of GCC are welcome on our developer 1904 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 1905 archives. 1906 1907 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 1908 distribution of this entire article is permitted in any medium, 1909 provided this notice is preserved. 1910 1911 These pages are [23]maintained by the GCC team. Last modified 1912 2016-09-30[24]. 1913 1914 References 1915 1916 1. http://www.gnu.org/ 1917 2. http://gcc.gnu.org/gcc-4.9/changes.html 1918 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 1919 4. http://gcc.gnu.org/gcc-4.9/changes.html 1920 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 1921 6. http://gcc.gnu.org/gcc-4.9/changes.html 1922 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 1923 8. http://gcc.gnu.org/gcc-4.9/changes.html 1924 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 1925 10. http://gcc.gnu.org/gcc-4.9/changes.html 1926 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 1927 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 1928 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1929 14. http://gcc.gnu.org/index.html 1930 15. mailto:gcc (a] gcc.gnu.org 1931 16. http://gcc.gnu.org/mirrors.html 1932 17. http://gcc.gnu.org/svn.html 1933 18. https://gcc.gnu.org/onlinedocs/ 1934 19. mailto:gcc-help (a] gcc.gnu.org 1935 20. mailto:gcc (a] gcc.gnu.org 1936 21. https://gcc.gnu.org/lists.html 1937 22. http://www.fsf.org/ 1938 23. https://gcc.gnu.org/about.html 1939 24. http://validator.w3.org/check/referer 1940 ====================================================================== 1941 http://gcc.gnu.org/gcc-4.9/changes.html 1942 1943 GCC 4.9 Release Series 1944 Changes, New Features, and Fixes 1945 1946 Caveats 1947 1948 * The mudflap run time checker has been removed. The mudflap options 1949 remain, but do nothing. 1950 * Support for a number of older systems and recently unmaintained or 1951 untested target ports of GCC has been declared obsolete in GCC 4.9. 1952 Unless there is activity to revive them, the next release of GCC 1953 will have their sources permanently removed. 1954 The following ports for individual systems on particular 1955 architectures have been obsoleted: 1956 + Solaris 9 (*-*-solaris2.9). Details can be found in the 1957 [1]announcement. 1958 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 1959 float64x1_t exported by arm_neon.h are defined to be the same as 1960 their base types. This results in incorrect application of 1961 parameter passing rules to arguments of types int64x1_t and 1962 uint64x1_t, with respect to the AAPCS64 ABI specification. In 1963 addition, names of C++ functions with parameters of these types 1964 (including float64x1_t) are not mangled correctly. The current 1965 typedef declarations also unintentionally allow implicit casting 1966 between singleton vector types and their base types. These issues 1967 will be resolved in a near future release. See [2]PR60825 for more 1968 information. 1969 1970 More information on porting to GCC 4.9 from previous versions of GCC 1971 can be found in the [3]porting guide for this release. 1972 1973 General Optimizer Improvements 1974 1975 * AddressSanitizer, a fast memory error detector, is now available on 1976 ARM. 1977 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 1978 detector, has been added and can be enabled via 1979 -fsanitize=undefined. Various computations will be instrumented to 1980 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 1981 currently available for the C and C++ languages. 1982 * Link-time optimization (LTO) improvements: 1983 + Type merging was rewritten. The new implementation is 1984 significantly faster and uses less memory. 1985 + Better partitioning algorithm resulting in less streaming 1986 during link time. 1987 + Early removal of virtual methods reduces the size of object 1988 files and improves link-time memory usage and compile time. 1989 + Function bodies are now loaded on-demand and released early 1990 improving overall memory usage at link time. 1991 + C++ hidden keyed methods can now be optimized out. 1992 + When using a linker plugin, compiling with the -flto option 1993 now generates slim object files (.o) which only contain 1994 intermediate language representation for LTO. Use 1995 -ffat-lto-objects to create files which contain additionally 1996 the object code. To generate static libraries suitable for LTO 1997 processing, use gcc-ar and gcc-ranlib; to list symbols from a 1998 slim object file use gcc-nm. (This requires that ar, ranlib 1999 and nm have been compiled with plugin support.) 2000 Memory usage building Firefox with debug enabled was reduced from 2001 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 2002 * Inter-procedural optimization improvements: 2003 + New type inheritance analysis module improving 2004 devirtualization. Devirtualization now takes into account 2005 anonymous name-spaces and the C++11 final keyword. 2006 + New speculative devirtualization pass (controlled by 2007 -fdevirtualize-speculatively. 2008 + Calls that were speculatively made direct are turned back to 2009 indirect where direct call is not cheaper. 2010 + Local aliases are introduced for symbols that are known to be 2011 semantically equivalent across shared libraries improving 2012 dynamic linking times. 2013 * Feedback directed optimization improvements: 2014 + Profiling of programs using C++ inline functions is now more 2015 reliable. 2016 + New time profiling determines typical order in which functions 2017 are executed. 2018 + A new function reordering pass (controlled by 2019 -freorder-functions) significantly reduces startup time of 2020 large applications. Until binutils support is completed, it is 2021 effective only with link-time optimization. 2022 + Feedback driven indirect call removal and devirtualization now 2023 handle cross-module calls when link-time optimization is 2024 enabled. 2025 2026 New Languages and Language specific improvements 2027 2028 * Version 4.0 of the [4]OpenMP specification is now supported in the 2029 C and C++ compilers and starting with the 4.9.1 release also in the 2030 Fortran compiler. The new -fopenmp-simd option can be used to 2031 enable OpenMP's SIMD directives while ignoring other OpenMP 2032 directives. The new [5]-fsimd-cost-model= option permits to tune 2033 the vectorization cost model for loops annotated with OpenMP and 2034 Cilk Plus simd directives. -Wopenmp-simd warns when the current 2035 cost model overrides simd directives set by the user. 2036 * The -Wdate-time option has been added for the C, C++ and Fortran 2037 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 2038 macros are used. Those macros might prevent bit-wise-identical 2039 reproducible compilations. 2040 2041 Ada 2042 2043 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 2044 2045 C family 2046 2047 * Support for colorizing diagnostics emitted by GCC has been added. 2048 The [6]-fdiagnostics-color=auto will enable it when outputting to 2049 terminals, -fdiagnostics-color=always unconditionally. The 2050 GCC_COLORS environment variable can be used to customize the colors 2051 or disable coloring. If GCC_COLORS variable is present in the 2052 environment, the default is -fdiagnostics-color=auto, otherwise 2053 -fdiagnostics-color=never. 2054 Sample diagnostics output: 2055 $ g++ -fdiagnostics-color=always -S -Wall test.C 2056 test.C: In function `int foo()': 2057 test.C:1:14: warning: no return statement in function returning non-void [-W 2058 return-type] 2059 int foo () { } 2060 ^ 2061 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 2062 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 2063 template <int N> struct X { static const int value = X<N-1>::value; }; temp 2064 late struct X<1000>; 2065 ^ 2066 test.C:2:46: recursively required from `const int X<999>::value' 2067 test.C:2:46: required from `const int X<1000>::value' 2068 test.C:2:88: required from here 2069 2070 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 2071 2072 * With the new [7]#pragma GCC ivdep, the user can assert that there 2073 are no loop-carried dependencies which would prevent concurrent 2074 execution of consecutive iterations using SIMD (single instruction 2075 multiple data) instructions. 2076 * Support for [8]Cilk Plus has been added and can be enabled with the 2077 -fcilkplus option. Cilk Plus is an extension to the C and C++ 2078 languages to support data and task parallelism. The present 2079 implementation follows ABI version 1.2; all features but _Cilk_for 2080 have been implemented. 2081 2082 C 2083 2084 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 2085 <stdatomic.h> header) are now supported. 2086 * ISO C11 generic selections (_Generic keyword) are now supported. 2087 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 2088 __thread) is now supported. 2089 * ISO C11 support is now at a similar level of completeness to ISO 2090 C99 support: substantially complete modulo bugs, extended 2091 identifiers (supported except for corner cases when 2092 -fextended-identifiers is used), floating-point issues (mainly but 2093 not entirely relating to optional C99 features from Annexes F and 2094 G) and the optional Annexes K (Bounds-checking interfaces) and L 2095 (Analyzability). 2096 * A new C extension __auto_type provides a subset of the 2097 functionality of C++11 auto in GNU C. 2098 2099 C++ 2100 2101 * The G++ implementation of [9]C++1y return type deduction for normal 2102 functions has been updated to conform to [10]N3638, the proposal 2103 accepted into the working paper. Most notably, it adds 2104 decltype(auto) for getting decltype semantics rather than the 2105 template argument deduction semantics of plain auto: 2106 2107 int& f(); 2108 auto i1 = f(); // int 2109 decltype(auto) i2 = f(); // int& 2110 2111 * G++ supports [11]C++1y lambda capture initializers: 2112 2113 [x = 42]{ ... }; 2114 2115 Actually, they have been accepted since GCC 4.5, but now the 2116 compiler doesn't warn about them with -std=c++1y, and supports 2117 parenthesized and brace-enclosed initializers as well. 2118 * G++ supports [12]C++1y variable length arrays. G++ has supported 2119 GNU/C99-style VLAs for a long time, but now additionally supports 2120 initializers and lambda capture by reference. In C++1y mode G++ 2121 will complain about VLA uses that are not permitted by the draft 2122 standard, such as forming a pointer to VLA type or applying sizeof 2123 to a VLA variable. Note that it now appears that VLAs will not be 2124 part of C++14, but will be part of a separate document and then 2125 perhaps C++17. 2126 2127 void f(int n) { 2128 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 2129 [&a]{ for (int i : a) { cout << i << endl; } }(); 2130 &a; // error, taking address of VLA 2131 } 2132 2133 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 2134 the underlying [[gnu::deprecated]] attribute. Classes and functions 2135 can be marked deprecated and a diagnostic message added: 2136 2137 class A; 2138 int bar(int n); 2139 #if __cplusplus > 201103 2140 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 2141 [[deprecated("bar is unsafe; use foo() instead")]] 2142 int bar(int n); 2143 2144 int foo(int n); 2145 class B; 2146 #endif 2147 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 2148 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 2149 o() instead 2150 2151 * G++ supports [14]C++1y digit separators. Long numeric literals can 2152 be subdivided with a single quote ' to enhance readability: 2153 2154 int i = 1048576; 2155 int j = 1'048'576; 2156 int k = 0x10'0000; 2157 int m = 0'004'000'000; 2158 int n = 0b0001'0000'0000'0000'0000'0000; 2159 2160 double x = 1.602'176'565e-19; 2161 double y = 1.602'176'565e-1'9; 2162 2163 * G++ supports [15]C++1y generic (polymorphic) lambdas. 2164 2165 // a functional object that will increment any type 2166 auto incr = [](auto x) { return x++; }; 2167 2168 * As a GNU extension, G++ supports explicit template parameter syntax 2169 for generic lambdas. This can be combined in the expected way with 2170 the standard auto syntax. 2171 2172 // a functional object that will add two like-type objects 2173 auto add = [] <typename T> (T a, T b) { return a + b; }; 2174 2175 * G++ supports unconstrained generic functions as specified by 2176 S:4.1.2 and S:5.1.1 of [16]N3889: Concepts Lite Specification. 2177 Briefly, auto may be used as a type-specifier in a parameter 2178 declaration of any function declarator in order to introduce an 2179 implicit function template parameter, akin to generic lambdas. 2180 2181 // the following two function declarations are equivalent 2182 auto incr(auto x) { return x++; } 2183 template <typename T> 2184 auto incr(T x) { return x++; } 2185 2186 Runtime Library (libstdc++) 2187 2188 * [17]Improved support for C++11, including: 2189 + support for <regex>; 2190 + The associative containers in <map> and <set> and the 2191 unordered associative containers in <unordered_map> and 2192 <unordered_set> meet the allocator-aware container 2193 requirements; 2194 * [18]Improved experimental support for the upcoming ISO C++ 2195 standard, C++14, including: 2196 + fixing constexpr member functions without const; 2197 + implementation of the std::exchange() utility function; 2198 + addressing tuples by type; 2199 + implemention of std::make_unique; 2200 + implemention of std::shared_lock; 2201 + making std::result_of SFINAE-friendly; 2202 + adding operator() to std::integral_constant; 2203 + adding user-defined literals for standard library types 2204 std::basic_string, std::chrono::duration, and std::complex; 2205 + adding two range overloads to non-modifying sequence oprations 2206 std::equal and std::mismatch; 2207 + adding IO manipulators for quoted strings; 2208 + adding constexpr members to <utility>, <complex>, <chrono>, 2209 and some containers; 2210 + adding compile-time std::integer_sequence; 2211 + adding cleaner transformation traits; 2212 + making <functional>s operator functors easier to use and more 2213 generic; 2214 * An implementation of std::experimental::optional. 2215 * An implementation of std::experimental::string_view. 2216 * The non-standard function std::copy_exception has been deprecated 2217 and will be removed in a future version. std::make_exception_ptr 2218 should be used instead. 2219 2220 Fortran 2221 2222 * Compatibility notice: 2223 + Module files: The version of the module files (.mod) has been 2224 incremented; additionally, module files are now compressed. 2225 Fortran MODULEs compiled by earlier GCC versions have to be 2226 recompiled, when they are USEd by files compiled with GCC 4.9. 2227 GCC 4.9 is not able to read .mod files of earlier GCC 2228 versions; attempting to do so gives an error message. Note: 2229 The ABI of the produced assembler data itself has not changed: 2230 object files and libraries are fully compatible with older 2231 versions (except as stated below). 2232 + ABI changes: 2233 o The [19]argument passing ABI has changed for scalar dummy 2234 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 2235 which have both the VALUE and the OPTIONAL attributes. 2236 o To support finalization the virtual table associated with 2237 polymorphic variables has changed. Code containing CLASS 2238 should be recompiled, including all files which define 2239 derived types involved in the type definition used by 2240 polymorphic variables. (Note: Due to the incremented 2241 module version, trying to mix old code with new code will 2242 usually give an error message.) 2243 + GNU Fortran no longer deallocates allocatable variables or 2244 allocatable components of variables declared in the main 2245 program. Since Fortran 2008, the standard explicitly states 2246 that variables declared in the Fortran main program 2247 automatically have the SAVE attribute. 2248 + When opening files, the close-on-exec flag is set if the 2249 system supports such a feature. This is generally considered 2250 good practice these days, but if there is a need to pass file 2251 descriptors to child processes the parent process must now 2252 remember to clear the close-on-exec flag by calling fcntl(), 2253 e.g. via ISO_C_BINDING, before executing the child process. 2254 * The deprecated command-line option -fno-whole-file has been 2255 removed. (-fwhole-file is the default since GCC 4.6.) 2256 -fwhole-file/-fno-whole-file continue to be accepted but do not 2257 influence the code generation. 2258 * The compiler no longer unconditionally warns about DO loops with 2259 zero iterations. This warning is now controlled by the -Wzerotrip 2260 option, which is implied by -Wall. 2261 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 2262 used to disable the type-kind-rank (TKR) argument check for a dummy 2263 argument. The feature is similar to ISO/IEC TS 29133:2012's 2264 TYPE(*), except that it additionally also disables the rank check. 2265 Variables with NO_ARG_CHECK have to be dummy arguments and may only 2266 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 2267 to another NO_ARG_CHECK dummy argument; also the other constraints 2268 of TYPE(*) apply. The dummy arguments should be declared as scalar 2269 or assumed-size variable of type type(*) (recommended) - or of type 2270 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 2271 the data without further type or shape information is passed, 2272 similar to C's void*. Note that also TS 29113's 2273 type(*),dimension(..) accepts arguments of any type and rank; 2274 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 2275 descriptor which contains the array shape and stride of the 2276 argument. 2277 * [21]Fortran 2003: 2278 + Finalization is now supported. It is currently only done for a 2279 subset of those situations in which it should occur. 2280 + Experimental support for scalar character components with 2281 deferred length (i.e. allocatable string length) in derived 2282 types has been added. (Deferred-length character variables are 2283 supported since GCC 4.6.) 2284 * [22]Fortran 2008: 2285 + When STOP or ERROR STOP are used to terminate the execution 2286 and any exception (but inexact) is signaling, a warning is 2287 printed to ERROR_UNIT, indicating which exceptions are 2288 signaling. The [23]-ffpe-summary= command-line option can be 2289 used to fine-tune for which exceptions the warning should be 2290 shown. 2291 + Rounding on input (READ) is now handled on systems where 2292 strtod honours the rounding mode. (For output, rounding is 2293 supported since GCC 4.5.) Note that for input, the compatible 2294 rounding mode is handled as nearest (i.e., rounding to an even 2295 least significant [cf. IEC 60559:1989] for a tie, while 2296 compatible rounds away from zero in that case). 2297 2298 Go 2299 2300 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 2301 2302 New Targets and Target Specific Improvements 2303 2304 AArch64 2305 2306 * The ARMv8-A crypto and CRC instructions are now supported through 2307 intrinsics. These are enabled when the architecture supports these 2308 and are available through the -march=armv8-a+crc and 2309 -march=armv8-a+crypto options. 2310 * Initial support for ILP32 has now been added to the compiler. This 2311 is now available through the command-line option -mabi=ilp32. 2312 Support for ILP32 is considered experimental as the ABI 2313 specification is still beta. 2314 * Coverage of more of the ISA including the SIMD extensions has been 2315 added. The Advanced SIMD intrinsics have also been improved. 2316 * The new local register allocator (LRA) is now on by default for the 2317 AArch64 backend. 2318 * The REE (Redundant extension elimination) pass has now been enabled 2319 by default for the AArch64 backend. 2320 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 2321 * Initial big.LITTLE tuning support for the combination of Cortex-A57 2322 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 2323 option. 2324 * A number of structural changes have been made to both the ARM and 2325 AArch64 backends to facilitate improved code-generation. 2326 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 2327 has been added and can be enabled by giving the 2328 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 2329 default by configuring GCC with the --enable-fix-cortex-a53-835769 2330 option. 2331 2332 ARC 2333 2334 * A port for Synopsys Designware ARC has been contributed by Embecosm 2335 and Synopsys Inc. 2336 2337 ARM 2338 2339 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 2340 disabled by default. This was found to generate better code in only 2341 a small number of cases. It can be turned back on with the 2342 -mneon-for-64bits option. 2343 * Further support for the ARMv8-A architecture, notably implementing 2344 the restriction around IT blocks in the Thumb32 instruction set has 2345 been added. The -mrestrict-it option can be used with 2346 -march=armv7-a or the -march=armv7ve options to make code 2347 generation fully compatible with the deprecated instructions in 2348 ARMv8-A. 2349 * Support has now been added for the ARMv7ve variant of the 2350 architecture. This can be used by the -march=armv7ve option. 2351 * The ARMv8-A crypto and CRC instructions are now supported through 2352 intrinsics and are available through the -march=armv8-a+crc and 2353 mfpu=crypto-neon-fp-armv8 options. 2354 * LRA is now on by default for the ARM target. This can be turned off 2355 using the -mno-lra option. This option is a purely transitionary 2356 command-line option and will be removed in a future release. We are 2357 interested in any bug reports regarding functional and performance 2358 regressions with LRA. 2359 * A new option -mslow-flash-data to improve performance of programs 2360 fetching data on slow flash memory has now been introduced for the 2361 ARMv7-M profile cores. 2362 * A new option -mpic-data-is-text-relative for targets that allows 2363 data segments to be relative to text segments has been added. This 2364 is on by default for all targets except VxWorks RTP. 2365 * A number of infrastructural changes have been made to both the ARM 2366 and AArch64 backends to facilitate improved code-generation. 2367 * GCC now supports Cortex-A12 and the Cortex-R7 through the 2368 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 2369 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 2370 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 2371 * Initial big.LITTLE tuning support for the combination of Cortex-A57 2372 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 2373 option. Similar support was added for the combination of Cortex-A15 2374 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 2375 * Further performance optimizations for the Cortex-A15 and the 2376 Cortex-M4 have been added. 2377 * A number of code generation improvements for Thumb2 to reduce code 2378 size when compiling for the M-profile processors. 2379 2380 AVR 2381 2382 * A new command-line option -mfract-convert-truncate has been added. 2383 It allows compiler to use truncation instead of rounding towards 2384 zero for fractional fixed-point types. 2385 2386 IA-32/x86-64 2387 2388 * -mfpmath=sse is now implied by -ffast-math on all targets where 2389 SSE2 is supported. 2390 * Intel AVX-512 support was added to GCC. That includes inline 2391 assembly support, new registers and extending existing ones, new 2392 intrinsics (covered by corresponding testsuite), and basic 2393 autovectorization. AVX-512 instructions are available via the 2394 following GCC switches: AVX-512 foundation instructions: -mavx512f, 2395 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 2396 reciprocal instructions: -mavx512er, AVX-512 conflict detection 2397 instructions: -mavx512cd. 2398 * It is now possible to call x86 intrinsics from select functions in 2399 a file that are tagged with the corresponding target attribute 2400 without having to compile the entire file with the -mxxx option. 2401 This improves the usability of x86 intrinsics and is particularly 2402 useful when doing [24]Function Multiversioning. 2403 * GCC now supports the new Intel microarchitecture named Silvermont 2404 through -march=silvermont. 2405 * GCC now supports the new Intel microarchitecture named Broadwell 2406 through -march=broadwell. 2407 * Optimizing for other Intel microarchitectures have been renamed to 2408 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 2409 * -march=generic has been retuned for better support of Intel core 2410 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 2411 Pentium-M, and Pentium4 based CPUs is no longer considered 2412 important for generic. 2413 * -mtune=intel can now be used to generate code running well on the 2414 most current Intel processors, which are Haswell and Silvermont for 2415 GCC 4.9. 2416 * Support to encode 32-bit assembly instructions in 16-bit format is 2417 now available through the -m16 command-line option. 2418 * Better inlining of memcpy and memset that is aware of value ranges 2419 and produces shorter alignment prologues. 2420 * -mno-accumulate-outgoing-args is now honored when unwind 2421 information is output. Argument accumulation is also now turned off 2422 for portions of programs optimized for size. 2423 * Support for new AMD family 15h processors (Excavator core) is now 2424 available through the -march=bdver4 and -mtune=bdver4 options. 2425 2426 MSP430 2427 2428 * A new command-line option -mcpu= has been added to the MSP430 2429 backend. This option is used to specify the ISA to be used. 2430 Accepted values are msp430 (the default), msp430x and msp430xv2. 2431 The ISA is no longer deduced from the -mmcu= option as there are 2432 far too many different MCU names. The -mmcu= option is still 2433 supported, and this is still used to select linker scripts and 2434 generate a C preprocessor symbol that will be recognised by the 2435 msp430.h header file. 2436 2437 NDS32 2438 2439 * A new nds32 port supports the 32-bit architecture from Andes 2440 Technology Corporation. 2441 * The port provides initial support for the V2, V3, V3m instruction 2442 set architectures. 2443 2444 Nios II 2445 2446 * A port for the Altera Nios II has been contributed by Mentor 2447 Graphics. 2448 2449 PowerPC / PowerPC64 / RS6000 2450 2451 * GCC now supports Power ISA 2.07, which includes support for 2452 Hardware Transactional Memory (HTM), Quadword atomics and several 2453 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 2454 integer and decimal integer operations. 2455 * Support for the POWER8 processor is now available through the 2456 -mcpu=power8 and -mtune=power8 options. 2457 * The libitm library has been modified to add a HTM fastpath that 2458 automatically uses POWER's HTM hardware instructions when it is 2459 executing on a HTM enabled processor. 2460 * Support for the new powerpc64le-linux platform has been added. It 2461 defaults to generating code that conforms to the ELFV2 ABI. 2462 2463 S/390, System z 2464 2465 * Support for the Transactional Execution Facility included with the 2466 IBM zEnterprise zEC12 processor has been added. A set of GCC style 2467 builtins as well as XLC style builtins are provided. The builtins 2468 are enabled by default when using the -march=zEC12 option but can 2469 explicitly be disabled with -mno-htm. Using the GCC builtins also 2470 libitm supports hardware transactions on S/390. 2471 * The hotpatch features allows to prepare functions for hotpatching. 2472 A certain amount of bytes is reserved before the function entry 2473 label plus a NOP is inserted at its very beginning to implement a 2474 backward jump when applying a patch. The feature can either be 2475 enabled per compilation unit via the command-line option -mhotpatch 2476 or per function using the hotpatch attribute. 2477 * The shrink wrap optimization is now supported on S/390 and enabled 2478 by default. 2479 * A major rework of the routines to determine which registers need to 2480 be saved and restored in function prologue/epilogue now allow to 2481 use floating point registers as save slots. This will happen for 2482 certain leaf function with -march=z10 or higher. 2483 * The LRA rtl pass replaces reload by default on S/390. 2484 2485 RX 2486 2487 * The port now allows to specify the RX100, RX200, and RX600 2488 processors with the command-line options -mcpu=rx100, -mcpu=rx200 2489 and -mcpu=rx600. 2490 2491 SH 2492 2493 * Minor improvements to code generated for integer arithmetic and 2494 code that involves the T bit. 2495 * Added support for the SH2A clips and clipu instructions. The 2496 compiler will now try to utilize them for min/max expressions such 2497 as max (-128, min (127, x)). 2498 * Added support for the cmp/str instruction through built-in 2499 functions such as __builtin_strlen. When not optimizing for size, 2500 the compiler will now expand calls to e.g. strlen as an inlined 2501 sequences which utilize the cmp/str instruction. 2502 * Improved code generated around volatile memory loads and stores. 2503 * The option -mcbranchdi has been deprecated. Specifying it will 2504 result in a warning and will not influence code generation. 2505 * The option -mcmpeqdi has been deprecated. Specifying it will result 2506 in a warning and will not influence code generation. 2507 2508 GCC 4.9.1 2509 2510 This is the [25]list of problem reports (PRs) from GCC's bug tracking 2511 system that are known to be fixed in the 4.9.1 release. This list might 2512 not be complete (that is, it is possible that some PRs that have been 2513 fixed are not listed here). 2514 2515 Version 4.0 of the OpenMP specification is supported even in Fortran, 2516 not just C and C++. 2517 2518 GCC 4.9.2 2519 2520 This is the [26]list of problem reports (PRs) from GCC's bug tracking 2521 system that are known to be fixed in the 4.9.2 release. This list might 2522 not be complete (that is, it is possible that some PRs that have been 2523 fixed are not listed here). 2524 2525 GCC 4.9.3 2526 2527 This is the [27]list of problem reports (PRs) from GCC's bug tracking 2528 system that are known to be fixed in the 4.9.3 release. This list might 2529 not be complete (that is, it is possible that some PRs that have been 2530 fixed are not listed here). 2531 2532 GCC 4.9.4 2533 2534 This is the [28]list of problem reports (PRs) from GCC's bug tracking 2535 system that are known to be fixed in the 4.9.4 release. This list might 2536 not be complete (that is, it is possible that some PRs that have been 2537 fixed are not listed here). 2538 2539 2540 For questions related to the use of GCC, please consult these web 2541 pages and the [29]GCC manuals. If that fails, the 2542 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2543 web pages and the development of GCC are welcome on our developer 2544 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 2545 archives. 2546 2547 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 2548 distribution of this entire article is permitted in any medium, 2549 provided this notice is preserved. 2550 2551 These pages are [34]maintained by the GCC team. Last modified 2552 2016-12-28[35]. 2553 2554 References 2555 2556 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 2557 2. https://gcc.gnu.org/PR60825 2558 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 2559 4. http://www.openmp.org/specifications/ 2560 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 2561 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 2562 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 2563 8. https://www.cilkplus.org/ 2564 9. http://gcc.gnu.org/projects/cxx1y.html 2565 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 2566 11. http://gcc.gnu.org/projects/cxx1y.html 2567 12. http://gcc.gnu.org/projects/cxx1y.html 2568 13. http://gcc.gnu.org/projects/cxx1y.html 2569 14. http://gcc.gnu.org/projects/cxx1y.html 2570 15. http://gcc.gnu.org/projects/cxx1y.html 2571 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 2572 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 2573 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 2574 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 2575 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 2576 21. https://gcc.gnu.org/wiki/Fortran2003Status 2577 22. https://gcc.gnu.org/wiki/Fortran2008Status 2578 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 2579 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 2580 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 2581 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 2582 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 2583 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 2584 29. https://gcc.gnu.org/onlinedocs/ 2585 30. mailto:gcc-help (a] gcc.gnu.org 2586 31. mailto:gcc (a] gcc.gnu.org 2587 32. https://gcc.gnu.org/lists.html 2588 33. http://www.fsf.org/ 2589 34. https://gcc.gnu.org/about.html 2590 35. http://validator.w3.org/check/referer 2591 ====================================================================== 2592 http://gcc.gnu.org/gcc-4.8/index.html 2593 2594 GCC 4.8 Release Series 2595 2596 June 23, 2015 2597 2598 The [1]GNU project and the GCC developers are pleased to announce the 2599 release of GCC 4.8.5. 2600 2601 This release is a bug-fix release, containing fixes for regressions in 2602 GCC 4.8.4 relative to previous releases of GCC. 2603 2604 This release series is no longer maintained. 2605 2606 Release History 2607 2608 GCC 4.8.5 2609 June 23, 2015 ([2]changes, [3]documentation) 2610 2611 GCC 4.8.4 2612 December 19, 2014 ([4]changes, [5]documentation) 2613 2614 GCC 4.8.3 2615 May 22, 2014 ([6]changes, [7]documentation) 2616 2617 GCC 4.8.2 2618 October 16, 2013 ([8]changes, [9]documentation) 2619 2620 GCC 4.8.1 2621 May 31, 2013 ([10]changes, [11]documentation) 2622 2623 GCC 4.8.0 2624 March 22, 2013 ([12]changes, [13]documentation) 2625 2626 References and Acknowledgements 2627 2628 GCC used to stand for the GNU C Compiler, but since the compiler 2629 supports several other languages aside from C, it now stands for the 2630 GNU Compiler Collection. 2631 2632 A list of [14]successful builds is updated as new information becomes 2633 available. 2634 2635 The GCC developers would like to thank the numerous people that have 2636 contributed new features, improvements, bug fixes, and other changes as 2637 well as test results to GCC. This [15]amazing group of volunteers is 2638 what makes GCC successful. 2639 2640 For additional information about GCC please refer to the [16]GCC 2641 project web site or contact the [17]GCC development mailing list. 2642 2643 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 2644 2645 2646 For questions related to the use of GCC, please consult these web 2647 pages and the [20]GCC manuals. If that fails, the 2648 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2649 web pages and the development of GCC are welcome on our developer 2650 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 2651 archives. 2652 2653 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 2654 distribution of this entire article is permitted in any medium, 2655 provided this notice is preserved. 2656 2657 These pages are [25]maintained by the GCC team. Last modified 2658 2016-09-30[26]. 2659 2660 References 2661 2662 1. http://www.gnu.org/ 2663 2. http://gcc.gnu.org/gcc-4.8/changes.html 2664 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 2665 4. http://gcc.gnu.org/gcc-4.8/changes.html 2666 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 2667 6. http://gcc.gnu.org/gcc-4.8/changes.html 2668 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 2669 8. http://gcc.gnu.org/gcc-4.8/changes.html 2670 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 2671 10. http://gcc.gnu.org/gcc-4.8/changes.html 2672 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 2673 12. http://gcc.gnu.org/gcc-4.8/changes.html 2674 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 2675 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 2676 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2677 16. http://gcc.gnu.org/index.html 2678 17. mailto:gcc (a] gcc.gnu.org 2679 18. http://gcc.gnu.org/mirrors.html 2680 19. http://gcc.gnu.org/svn.html 2681 20. https://gcc.gnu.org/onlinedocs/ 2682 21. mailto:gcc-help (a] gcc.gnu.org 2683 22. mailto:gcc (a] gcc.gnu.org 2684 23. https://gcc.gnu.org/lists.html 2685 24. http://www.fsf.org/ 2686 25. https://gcc.gnu.org/about.html 2687 26. http://validator.w3.org/check/referer 2688 ====================================================================== 2689 http://gcc.gnu.org/gcc-4.8/changes.html 2690 2691 GCC 4.8 Release Series 2692 Changes, New Features, and Fixes 2693 2694 Caveats 2695 2696 GCC now uses C++ as its implementation language. This means that to 2697 build GCC from sources, you will need a C++ compiler that understands 2698 C++ 2003. For more details on the rationale and specific changes, 2699 please refer to the [1]C++ conversion page. 2700 2701 To enable the Graphite framework for loop optimizations you now need 2702 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 2703 the [2]GCC infrastructure directory. The installation manual contains 2704 more information about requirements to build GCC. 2705 2706 GCC now uses a more aggressive analysis to derive an upper bound for 2707 the number of iterations of loops using constraints imposed by language 2708 standards. This may cause non-conforming programs to no longer work as 2709 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 2710 option, -fno-aggressive-loop-optimizations, was added to disable this 2711 aggressive analysis. In some loops that have known constant number of 2712 iterations, but undefined behavior is known to occur in the loop before 2713 reaching or during the last iteration, GCC will warn about the 2714 undefined behavior in the loop instead of deriving lower upper bound of 2715 the number of iterations for the loop. The warning can be disabled with 2716 -Wno-aggressive-loop-optimizations. 2717 2718 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 2719 for the layout of vectors that could lead to wrong code being 2720 generated. Vectors larger than 8 bytes in size are now by default 2721 aligned to an 8-byte boundary. This is an ABI change: code that makes 2722 explicit use of vector types may be incompatible with binary objects 2723 built with older versions of GCC. Auto-vectorized code is not affected 2724 by this change. 2725 2726 On AVR, support has been removed for the command-line option 2727 -mshort-calls deprecated in GCC 4.7. 2728 2729 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 2730 is turned on per default for all non-RTEMS configurations. This option 2731 arranges for a better integration of [3]AVR Libc with avr-gcc. For 2732 technical details, see [4]PR54461. To turn off the option in non-RTEMS 2733 configurations, use --with-avrlibc=no. If the compiler is configured 2734 for RTEMS, the option is always turned off. 2735 2736 More information on porting to GCC 4.8 from previous versions of GCC 2737 can be found in the [5]porting guide for this release. 2738 2739 General Optimizer Improvements (and Changes) 2740 2741 * DWARF4 is now the default when generating DWARF debug information. 2742 When -g is used on a platform that uses DWARF debugging 2743 information, GCC will now default to -gdwarf-4 2744 -fno-debug-types-section. 2745 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 2746 consumers support DWARF4 by default. Before GCC 4.8 the default 2747 version used was DWARF2. To make GCC 4.8 generate an older DWARF 2748 version use -g together with -gdwarf-2 or -gdwarf-3. The default 2749 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 2750 * A new general optimization level, -Og, has been introduced. It 2751 addresses the need for fast compilation and a superior debugging 2752 experience while providing a reasonable level of run-time 2753 performance. Overall experience for development should be better 2754 than the default optimization level -O0. 2755 * A new option -ftree-partial-pre was added to control the partial 2756 redundancy elimination (PRE) optimization. This option is enabled 2757 by default at the -O3 optimization level, and it makes PRE more 2758 aggressive. 2759 * The option -fconserve-space has been removed; it was no longer 2760 useful on most targets since GCC supports putting variables into 2761 BSS without making them common. 2762 * The struct reorg and matrix reorg optimizations (command-line 2763 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 2764 removed. They did not always work correctly, nor did they work with 2765 link-time optimization (LTO), hence were only applicable to 2766 programs consisting of a single translation unit. 2767 * Several scalability bottle-necks have been removed from GCC's 2768 optimization passes. Compilation of extremely large functions, e.g. 2769 due to the use of the flatten attribute in the "Eigen" C++ linear 2770 algebra templates library, is significantly faster than previous 2771 releases of GCC. 2772 * Link-time optimization (LTO) improvements: 2773 + LTO partitioning has been rewritten for better reliability and 2774 maintanibility. Several important bugs leading to link 2775 failures have been fixed. 2776 * Interprocedural optimization improvements: 2777 + A new symbol table has been implemented. It builds on existing 2778 callgraph and varpool modules and provide a new API. Unusual 2779 symbol visibilities and aliases are handled more consistently 2780 leading to, for example, more aggressive unreachable code 2781 removal with LTO. 2782 + The inline heuristic can now bypass limits on the size of of 2783 inlined functions when the inlining is particularly 2784 profitable. This happens, for example, when loop bounds or 2785 array strides get propagated. 2786 + Values passed through aggregates (either by value or 2787 reference) are now propagated at the inter-procedural level 2788 leading to better inlining decisions (for example in the case 2789 of Fortran array descriptors) and devirtualization. 2790 * [6]AddressSanitizer , a fast memory error detector, has been added 2791 and can be enabled via -fsanitize=address. Memory access 2792 instructions will be instrumented to detect heap-, stack-, and 2793 global-buffer overflow as well as use-after-free bugs. To get nicer 2794 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 2795 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 2796 x86-64 Darwin. 2797 * [7]ThreadSanitizer has been added and can be enabled via 2798 -fsanitize=thread. Instructions will be instrumented to detect data 2799 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 2800 * A new local register allocator (LRA) has been implemented, which 2801 replaces the 26 year old reload pass and improves generated code 2802 quality. For now it is active on the IA-32 and x86-64 targets. 2803 * Support for transactional memory has been implemented on the 2804 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 2805 Alpha. 2806 2807 New Languages and Language specific improvements 2808 2809 C family 2810 2811 * Each diagnostic emitted now includes the original source line and a 2812 caret '^' indicating the column. The option 2813 -fno-diagnostics-show-caret suppresses this information. 2814 * The option -ftrack-macro-expansion=2 is now enabled by default. 2815 This allows the compiler to display the macro expansion stack in 2816 diagnostics. Combined with the caret information, an example 2817 diagnostic showing these two features is: 2818 2819 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 2820 ') 2821 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 2822 _b = (B); __a < __b ? __b : __a; }) 2823 2824 ^ 2825 t.c:7:7: note: in expansion of macro 'MYMAX' 2826 X = MYMAX(P, F); 2827 ^ 2828 2829 * A new -Wsizeof-pointer-memaccess warning has been added (also 2830 enabled by -Wall) to warn about suspicious length parameters to 2831 certain string and memory built-in functions if the argument uses 2832 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 2833 (ptr)); if ptr is not an array, but a pointer, and suggests a 2834 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 2835 * The new option -Wpedantic is an alias for -pedantic, which is now 2836 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 2837 -Wno-error=pedantic work in the same way as for any other -W 2838 option. One caveat is that -Werror=pedantic is not equivalent to 2839 -pedantic-errors, since the latter makes into errors some warnings 2840 that are not controlled by -Wpedantic, and the former only affects 2841 diagnostics that are disabled when using -Wno-pedantic. 2842 * The option -Wshadow no longer warns if a declaration shadows a 2843 function declaration, unless the former declares a function or 2844 pointer to function, because this is [8]a common and valid case in 2845 real-world code. 2846 2847 C++ 2848 2849 * G++ now implements the [9]C++11 thread_local keyword; this differs 2850 from the GNU __thread keyword primarily in that it allows dynamic 2851 initialization and destruction semantics. Unfortunately, this 2852 support requires a run-time penalty for references to 2853 non-function-local thread_local variables defined in a different 2854 translation unit even if they don't need dynamic initialization, so 2855 users may want to continue to use __thread for TLS variables with 2856 static initialization semantics. 2857 If the programmer can be sure that no use of the variable in a 2858 non-defining TU needs to trigger dynamic initialization (either 2859 because the variable is statically initialized, or a use of the 2860 variable in the defining TU will be executed before any uses in 2861 another TU), they can avoid this overhead with the 2862 -fno-extern-tls-init option. 2863 OpenMP threadprivate variables now also support dynamic 2864 initialization and destruction by the same mechanism. 2865 * G++ now implements the [10]C++11 attribute syntax, e.g. 2866 2867 [[noreturn]] void f(); 2868 2869 and also the alignment specifier, e.g. 2870 2871 alignas(double) int i; 2872 2873 * G++ now implements [11]C++11 inheriting constructors, e.g. 2874 2875 struct A { A(int); }; 2876 struct B: A { using A::A; }; // defines B::B(int) 2877 B b(42); // OK 2878 2879 * As of GCC 4.8.1, G++ implements the change to decltype semantics 2880 from [12]N3276. 2881 2882 struct A f(); 2883 decltype(f()) g(); // OK, return type of f() is not required to be complete. 2884 2885 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 2886 2887 struct A { int f() &; }; 2888 int i = A().f(); // error, f() requires an lvalue object 2889 2890 * G++ now supports a -std=c++1y option for experimentation with 2891 features proposed for the next revision of the standard, expected 2892 around 2014. Currently the only difference from -std=c++11 is 2893 support for return type deduction in normal functions, as proposed 2894 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 2895 [15]here. 2896 * The G++ namespace association extension, __attribute ((strong)), 2897 has been deprecated. Inline namespaces should be used instead. 2898 * G++ now supports a -fext-numeric-literal option to control whether 2899 GNU numeric literal suffixes are accepted as extensions or 2900 processed as C++11 user-defined numeric literal suffixes. The flag 2901 is on (use suffixes for GNU literals) by default for -std=gnu++*, 2902 and -std=c++98. The flag is off (use suffixes for user-defined 2903 literals) by default for -std=c++11 and later. 2904 2905 Runtime Library (libstdc++) 2906 2907 * [16]Improved experimental support for the new ISO C++ standard, 2908 C++11, including: 2909 + forward_list meets the allocator-aware container requirements; 2910 + this_thread::sleep_for(), this_thread::sleep_until() and 2911 this_thread::yield() are defined without requiring the 2912 configure option --enable-libstdcxx-time; 2913 * Improvements to <random>: 2914 + SSE optimized normal_distribution. 2915 + Use of hardware RNG instruction for random_device on new x86 2916 processors (requires the assembler to support the 2917 instruction.) 2918 and <ext/random>: 2919 + New random number engine simd_fast_mersenne_twister_engine 2920 with an optimized SSE implementation. 2921 + New random number distributions beta_distribution, 2922 normal_mv_distribution, rice_distribution, 2923 nakagami_distribution, pareto_distribution, k_distribution, 2924 arcsine_distribution, hoyt_distribution. 2925 * Added --disable-libstdcxx-verbose configure option to disable 2926 diagnostic messages issued when a process terminates abnormally. 2927 This may be useful for embedded systems to reduce the size of 2928 executables that link statically to the library. 2929 2930 Fortran 2931 2932 * Compatibility notice: 2933 + Module files: The version of module files (.mod) has been 2934 incremented. Fortran MODULEs compiled by earlier GCC versions 2935 have to be recompiled, when they are USEd by files compiled 2936 with GCC 4.8. GCC 4.8 is not able to read .mod files created 2937 by earlier versions; attempting to do so gives an error 2938 message. 2939 Note: The ABI of the produced assembler data itself has not 2940 changed; object files and libraries are fully compatible with 2941 older versions except as noted below. 2942 + ABI: Some internal names (used in the assembler/object file) 2943 have changed for symbols declared in the specification part of 2944 a module. If an affected module - or a file using it via use 2945 association - is recompiled, the module and all files which 2946 directly use such symbols have to be recompiled as well. This 2947 change only affects the following kind of module symbols: 2948 o Procedure pointers. Note: C-interoperable function 2949 pointers (type(c_funptr)) are not affected nor are 2950 procedure-pointer components. 2951 o Deferred-length character strings. 2952 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 2953 backtrace at an arbitrary place in user code; program execution 2954 continues normally afterwards. 2955 * The [18]-Wc-binding-type warning option has been added (disabled by 2956 default). It warns if the a variable might not be C interoperable; 2957 in particular, if the variable has been declared using an intrinsic 2958 type with default kind instead of using a kind parameter defined 2959 for C interoperability in the intrinsic ISO_C_Binding module. 2960 Before, this warning was always printed. The -Wc-binding-type 2961 option is enabled by -Wall. 2962 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 2963 options have been added, which diagnose when code is inserted for 2964 automatic (re)allocation of a variable during assignment. This 2965 option can be used to decide whether it is safe to use 2966 [20]-fno-realloc-lhs. Additionally, it can be used to find 2967 automatic (re)allocation in hot loops. (For arrays, replacing 2968 "var=" by "var(:)=" disables the automatic reallocation.) 2969 * The [21]-Wcompare-reals command-line option has been added. When 2970 this is set, warnings are issued when comparing REAL or COMPLEX 2971 types for equality and inequality; consider replacing a == b by 2972 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by 2973 -Wextra. 2974 * The [22]-Wtarget-lifetime command-line option has been added 2975 (enabled with -Wall), which warns if the pointer in a pointer 2976 assignment might outlive its target. 2977 * Reading floating point numbers which use "q" for the exponential 2978 (such as 4.0q0) is now supported as vendor extension for better 2979 compatibility with old data files. It is strongly recommended to 2980 use for I/O the equivalent but standard conforming "e" (such as 2981 4.0e0). 2982 (For Fortran source code, consider replacing the "q" in 2983 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 2984 suitable qp). Note that - in Fortran source code - replacing "q" by 2985 a simple "e" is not equivalent.) 2986 * The GFORTRAN_TMPDIR environment variable for specifying a 2987 non-default directory for files opened with STATUS="SCRATCH", is 2988 not used anymore. Instead gfortran checks the POSIX/GNU standard 2989 TMPDIR environment variable. If TMPDIR is not defined, gfortran 2990 falls back to other methods to determine the directory for 2991 temporary files as documented in the [23]user manual. 2992 * [24]Fortran 2003: 2993 + Support for unlimited polymorphic variables (CLASS(*)) has 2994 been added. Nonconstant character lengths are not yet 2995 supported. 2996 * [25]TS 29113: 2997 + Assumed types (TYPE(*)) are now supported. 2998 + Experimental support for assumed-rank arrays (dimension(..)) 2999 has been added. Note that currently gfortran's own array 3000 descriptor is used, which is different from the one defined in 3001 TS29113, see [26]gfortran's header file or use the [27]Chasm 3002 Language Interoperability Tools. 3003 3004 Go 3005 3006 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 3007 release. 3008 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 3009 release. The library support is not quite complete. 3010 * Go has been tested on GNU/Linux and Solaris platforms for various 3011 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 3012 work on other platforms as well. 3013 3014 New Targets and Target Specific Improvements 3015 3016 AArch64 3017 3018 * A new port has been added to support AArch64, the new 64-bit 3019 architecture from ARM. Note that this is a separate port from the 3020 existing 32-bit ARM port. 3021 * The port provides initial support for the Cortex-A53 and the 3022 Cortex-A57 processors with the command line options 3023 -mcpu=cortex-a53 and -mcpu=cortex-a57. 3024 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 3025 has been added and can be enabled by giving the 3026 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 3027 default by configuring GCC with the --enable-fix-cortex-a53-835769 3028 option. 3029 3030 ARM 3031 3032 * Initial support has been added for the AArch32 extensions defined 3033 in the ARMv8 architecture. 3034 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 3035 * A new option, -mcpu=marvell-pj4, has been added to generate code 3036 for the Marvell PJ4 processor. 3037 * The compiler can now automatically generate the VFMA, VFMS, REVSH 3038 and REV16 instructions. 3039 * A new vectorizer cost model for Advanced SIMD configurations to 3040 improve the auto-vectorization strategies used. 3041 * The scheduler now takes into account the number of live registers 3042 to reduce the amount of spilling that can occur. This should 3043 improve code performance in large functions. The limit can be 3044 removed by using the option -fno-sched-pressure. 3045 * Improvements have been made to the Marvell iWMMX code generation 3046 and support for the iWMMX2 SIMD unit has been added. The option 3047 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 3048 * A number of code generation improvements for Thumb2 to reduce code 3049 size when compiling for the M-profile processors. 3050 * The RTEMS (arm-rtems) port has been updated to use the EABI. 3051 * Code generation support for the old FPA and Maverick floating-point 3052 architectures has been removed. Ports that previously relied on 3053 these features have also been removed. This includes the targets: 3054 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 3055 + arm*-*-elf (use arm*-*-eabi) 3056 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 3057 + arm*-*-ecos-elf (no alternative) 3058 + arm*-*-freebsd (no alternative) 3059 + arm*-wince-pe* (no alternative). 3060 3061 AVR 3062 3063 * Support for the "Embedded C" fixed-point has been added. For 3064 details, see the [28]GCC wiki and the [29]user manual. The support 3065 is not complete. 3066 * A new print modifier %r for register operands in inline assembler 3067 is supported. It will print the raw register number without the 3068 register prefix 'r': 3069 /* Return the most significant byte of 'val', a 64-bit value. */ 3070 3071 unsigned char msb (long long val) 3072 { 3073 unsigned char c; 3074 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 3075 return c; 3076 } 3077 The inline assembler in this example will generate code like 3078 mov r24, 8+7 3079 provided c is allocated to R24 and val is allocated to R8...R15. 3080 This works because the GNU assembler accepts plain register numbers 3081 without register prefix. 3082 * Static initializers with 3-byte symbols are supported now: 3083 extern const __memx char foo; 3084 const __memx void *pfoo = &foo; 3085 This requires at least Binutils 2.23. 3086 3087 IA-32/x86-64 3088 3089 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 3090 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 3091 stack alignment, this is ABI incompatible and intended to be used 3092 in controlled environments where stack space is an important 3093 limitation. This option will lead to wrong code when functions 3094 compiled with 16 byte stack alignment (such as functions from a 3095 standard library) are called with misaligned stack. In this case, 3096 SSE instructions may lead to misaligned memory access traps. In 3097 addition, variable arguments will be handled incorrectly for 16 3098 byte aligned objects (including x87 long double and __int128), 3099 leading to wrong results. You must build all modules with 3100 -mpreferred-stack-boundary=3, including any libraries. This 3101 includes the system libraries and startup modules. 3102 * Support for the new Intel processor codename Broadwell with RDSEED, 3103 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 3104 -mrdseed command-line options. 3105 * Support for the Intel RTM and HLE intrinsics, built-in functions 3106 and code generation is available via -mrtm and -mhle. 3107 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 3108 Intrinsics and built-in functions are available via -mfxsr, -mxsave 3109 and -mxsaveopt respectively. 3110 * New -maddress-mode=[short|long] options for x32. 3111 -maddress-mode=short overrides default 64-bit addresses to 32-bit 3112 by emitting the 0x67 address-size override prefix. This is the 3113 default address mode for x32. 3114 * New built-in functions to detect run-time CPU type and ISA: 3115 + A built-in function __builtin_cpu_is has been added to detect 3116 if the run-time CPU is of a particular type. It returns a 3117 positive integer on a match and zero otherwise. It accepts one 3118 string literal argument, the CPU name. For example, 3119 __builtin_cpu_is("westmere") returns a positive integer if the 3120 run-time CPU is an Intel Core i7 Westmere processor. Please 3121 refer to the [30]user manual for the list of valid CPU names 3122 recognized. 3123 + A built-in function __builtin_cpu_supports has been added to 3124 detect if the run-time CPU supports a particular ISA feature. 3125 It returns a positive integer on a match and zero otherwise. 3126 It accepts one string literal argument, the ISA feature. For 3127 example, __builtin_cpu_supports("ssse3") returns a positive 3128 integer if the run-time CPU supports SSSE3 instructions. 3129 Please refer to the [31]user manual for the list of valid ISA 3130 names recognized. 3131 Caveat: If these built-in functions are called before any static 3132 constructors are invoked, like during IFUNC initialization, then 3133 the CPU detection initialization must be explicitly run using this 3134 newly provided built-in function, __builtin_cpu_init. The 3135 initialization needs to be done only once. For example, this is how 3136 the invocation would look like inside an IFUNC initializer: 3137 static void (*some_ifunc_resolver(void))(void) 3138 { 3139 __builtin_cpu_init(); 3140 if (__builtin_cpu_is("amdfam10h") ... 3141 if (__builtin_cpu_supports("popcnt") ... 3142 } 3143 3144 * Function Multiversioning Support with G++: 3145 It is now possible to create multiple function versions each 3146 targeting a specific processor and/or ISA. Function versions have 3147 the same signature but different target attributes. For example, 3148 here is a program with function versions: 3149 __attribute__ ((target ("default"))) 3150 int foo(void) 3151 { 3152 return 1; 3153 } 3154 3155 __attribute__ ((target ("sse4.2"))) 3156 int foo(void) 3157 { 3158 return 2; 3159 } 3160 3161 int main (void) 3162 { 3163 int (*p) = &foo; 3164 assert ((*p)() == foo()); 3165 return 0; 3166 } 3167 3168 Please refer to this [32]wiki for more information. 3169 * The x86 back end has been improved to allow option -fschedule-insns 3170 to work reliably. This option can be used to schedule instructions 3171 better and leads to improved performace in certain cases. 3172 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 3173 from the Mingw-w64 trunk. 3174 * Support for new AMD family 15h processors (Steamroller core) is now 3175 available through the -march=bdver3 and -mtune=bdver3 options. 3176 * Support for new AMD family 16h processors (Jaguar core) is now 3177 available through the -march=btver2 and -mtune=btver2 options. 3178 3179 FRV 3180 3181 * This target now supports the -fstack-usage command-line option. 3182 3183 MIPS 3184 3185 * GCC can now generate code specifically for the R4700, Broadcom XLP 3186 and MIPS 34kn processors. The associated -march options are 3187 -march=r4700, -march=xlp and -march=34kn respectively. 3188 * GCC now generates better DSP code for MIPS 74k cores thanks to 3189 further scheduling optimizations. 3190 * The MIPS port now supports the -fstack-check option. 3191 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 3192 * Previous versions of GCC would silently accept -fpic and -fPIC for 3193 -mno-abicalls targets like mips*-elf. This combination was not 3194 intended or supported, and did not generate position-independent 3195 code. GCC 4.8 now reports an error when this combination is used. 3196 3197 PowerPC / PowerPC64 / RS6000 3198 3199 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 3200 restore or update the VRSAVE register by default. The respective 3201 operating systems manage the VRSAVE register directly. 3202 * Large TOC support has been added for AIX through the command line 3203 option -mcmodel=large. 3204 * Native Thread-Local Storage support has been added for AIX. 3205 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 3206 when targetting processors that support those hardware features on 3207 AIX 6.1 and above. 3208 3209 RX 3210 3211 * This target will now issue a warning message whenever multiple fast 3212 interrupt handlers are found in the same compilation unit. This 3213 feature can be turned off by the new 3214 -mno-warn-multiple-fast-interrupts command-line option. 3215 3216 S/390, System z 3217 3218 * Support for the IBM zEnterprise zEC12 processor has been added. 3219 When using the -march=zEC12 option, the compiler will generate code 3220 making use of the following new instructions: 3221 + load and trap instructions 3222 + 2 new compare and trap instructions 3223 + rotate and insert selected bits - without CC clobber 3224 The -mtune=zEC12 option enables zEC12 specific instruction 3225 scheduling without making use of new instructions. 3226 * Register pressure sensitive instruction scheduling is enabled by 3227 default. 3228 * The ifunc function attribute is enabled by default. 3229 * memcpy and memcmp invokations on big memory chunks or with run time 3230 lengths are not generated inline anymore when tuning for z10 or 3231 higher. The purpose is to make use of the IFUNC optimized versions 3232 in Glibc. 3233 3234 SH 3235 3236 * The default alignment settings have been reduced to be less 3237 aggressive. This results in more compact code for optimization 3238 levels other than -Os. 3239 * Improved support for the __atomic built-in functions: 3240 + A new option -matomic-model=model selects the model for the 3241 generated atomic sequences. The following models are 3242 supported: 3243 3244 soft-gusa 3245 Software gUSA sequences (SH3* and SH4* only). On 3246 SH4A targets this will now also partially utilize 3247 the movco.l and movli.l instructions. This is the 3248 default when the target is sh3*-*-linux* or 3249 sh4*-*-linux*. 3250 3251 hard-llcs 3252 Hardware movco.l / movli.l sequences (SH4A only). 3253 3254 soft-tcb 3255 Software thread control block sequences. 3256 3257 soft-imask 3258 Software interrupt flipping sequences (privileged 3259 mode only). This is the default when the target is 3260 sh1*-*-linux* or sh2*-*-linux*. 3261 3262 none 3263 Generates function calls to the respective __atomic 3264 built-in functions. This is the default for SH64 3265 targets or when the target is not sh*-*-linux*. 3266 3267 + The option -msoft-atomic has been deprecated. It is now an 3268 alias for -matomic-model=soft-gusa. 3269 + A new option -mtas makes the compiler generate the tas.b 3270 instruction for the __atomic_test_and_set built-in function 3271 regardless of the selected atomic model. 3272 + The __sync functions in libgcc now reflect the selected atomic 3273 model when building the toolchain. 3274 * Added support for the mov.b and mov.w instructions with 3275 displacement addressing. 3276 * Added support for the SH2A instructions movu.b and movu.w. 3277 * Various improvements to code generated for integer arithmetic. 3278 * Improvements to conditional branches and code that involves the T 3279 bit. A new option -mzdcbranch tells the compiler to favor 3280 zero-displacement branches. This is enabled by default for SH4* 3281 targets. 3282 * The pref instruction will now be emitted by the __builtin_prefetch 3283 built-in function for SH3* targets. 3284 * The fmac instruction will now be emitted by the fmaf standard 3285 function and the __builtin_fmaf built-in function. 3286 * The -mfused-madd option has been deprecated in favor of the 3287 machine-independent -ffp-contract option. Notice that the fmac 3288 instruction will now be generated by default for expressions like a 3289 * b + c. This is due to the compiler default setting 3290 -ffp-contract=fast. 3291 * Added new options -mfsrra and -mfsca to allow the compiler using 3292 the fsrra and fsca instructions on targets other than SH4A (where 3293 they are already enabled by default). 3294 * Added support for the __builtin_bswap32 built-in function. It is 3295 now expanded as a sequence of swap.b and swap.w instructions 3296 instead of a library function call. 3297 * The behavior of the -mieee option has been fixed and the negative 3298 form -mno-ieee has been added to control the IEEE conformance of 3299 floating point comparisons. By default -mieee is now enabled and 3300 the option -ffinite-math-only implicitly sets -mno-ieee. 3301 * Added support for the built-in functions __builtin_thread_pointer 3302 and __builtin_set_thread_pointer. This assumes that GBR is used to 3303 hold the thread pointer of the current thread. Memory loads and 3304 stores relative to the address returned by __builtin_thread_pointer 3305 will now also utilize GBR based displacement address modes. 3306 * The -mdiv= option for targets other than SHmedia has been fixed and 3307 documented. 3308 3309 SPARC 3310 3311 * Added optimized instruction scheduling for Niagara4. 3312 3313 TILE-Gx 3314 3315 * Added support for the -mcmodel=MODEL command-line option. The 3316 models supported are small and large. 3317 3318 V850 3319 3320 * This target now supports the E3V5 architecture via the use of the 3321 new -mv850e3v5 command-line option. It also has experimental 3322 support for the e3v5 LOOP instruction which can be enabled via the 3323 new -mloop command-line option. 3324 3325 XStormy16 3326 3327 * This target now supports the -fstack-usage command-line option. 3328 3329 Operating Systems 3330 3331 OpenBSD 3332 3333 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 3334 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 3335 3336 Windows (Cygwin) 3337 3338 * Executables are now linked against shared libgcc by default. The 3339 previous default was to link statically, which can still be done by 3340 explicitly specifying -static or static-libgcc on the command line. 3341 However it is strongly advised against, as it will cause problems 3342 for any application that makes use of DLLs compiled by GCC. It 3343 should be alright for a monolithic stand-alone application that 3344 only links against the Windows DLLs, but offers little or no 3345 benefit. 3346 3347 GCC 4.8.1 3348 3349 This is the [33]list of problem reports (PRs) from GCC's bug tracking 3350 system that are known to be fixed in the 4.8.1 release. This list might 3351 not be complete (that is, it is possible that some PRs that have been 3352 fixed are not listed here). 3353 3354 The C++11 <chrono> std::chrono::system_clock and 3355 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 3356 both are now separate (never typedefs of each other), both use 3357 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 3358 std::chrono::steady_clock is now finally monotonic, and both classes 3359 are mangled differently than in the previous GCC releases. 3360 std::chrono::system_clock::now() with std::chrono::microseconds resp. 3361 std::chrono::seconds resolution is still exported for backwards 3362 compatibility with default configured libstdc++. Note that libstdc++ 3363 configured with --enable-libstdcxx-time= used to be ABI incompatible 3364 with default configured libstdc++ for those two classes and no ABI 3365 compatibility can be offered for those configurations, so any C++11 3366 code that uses those classes and has been compiled and linked against 3367 libstdc++ configured with the non-default --enable-libstdcxx-time= 3368 configuration option needs to be recompiled. 3369 3370 GCC 4.8.2 3371 3372 This is the [34]list of problem reports (PRs) from GCC's bug tracking 3373 system that are known to be fixed in the 4.8.2 release. This list might 3374 not be complete (that is, it is possible that some PRs that have been 3375 fixed are not listed here). 3376 3377 GCC 4.8.3 3378 3379 This is the [35]list of problem reports (PRs) from GCC's bug tracking 3380 system that are known to be fixed in the 4.8.3 release. This list might 3381 not be complete (that is, it is possible that some PRs that have been 3382 fixed are not listed here). 3383 3384 Support for the new powerpc64le-linux platform has been added. It 3385 defaults to generating code that conforms to the ELFV2 ABI. 3386 3387 GCC 4.8.4 3388 3389 This is the [36]list of problem reports (PRs) from GCC's bug tracking 3390 system that are known to be fixed in the 4.8.4 release. This list might 3391 not be complete (that is, it is possible that some PRs that have been 3392 fixed are not listed here). 3393 3394 GCC 4.8.5 3395 3396 This is the [37]list of problem reports (PRs) from GCC's bug tracking 3397 system that are known to be fixed in the 4.8.5 release. This list might 3398 not be complete (that is, it is possible that some PRs that have been 3399 fixed are not listed here). 3400 3401 3402 For questions related to the use of GCC, please consult these web 3403 pages and the [38]GCC manuals. If that fails, the 3404 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3405 web pages and the development of GCC are welcome on our developer 3406 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 3407 archives. 3408 3409 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 3410 distribution of this entire article is permitted in any medium, 3411 provided this notice is preserved. 3412 3413 These pages are [43]maintained by the GCC team. Last modified 3414 2017-02-27[44]. 3415 3416 References 3417 3418 1. https://gcc.gnu.org/wiki/cxx-conversion 3419 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 3420 3. http://www.nongnu.org/avr-libc/ 3421 4. https://gcc.gnu.org/PR54461 3422 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 3423 6. https://github.com/google/sanitizers 3424 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 3425 8. https://lkml.org/lkml/2006/11/28/239 3426 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3427 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3428 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3429 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 3430 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3431 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 3432 15. http://gcc.gnu.org/projects/cxx1y.html 3433 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 3434 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 3435 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3436 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3437 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 3438 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3439 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3440 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 3441 24. https://gcc.gnu.org/wiki/Fortran2003Status 3442 25. https://gcc.gnu.org/wiki/TS29113Status 3443 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co 3444 27. http://chasm-interop.sourceforge.net/ 3445 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 3446 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 3447 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 3448 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 3449 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 3450 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 3451 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 3452 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 3453 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 3454 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 3455 38. https://gcc.gnu.org/onlinedocs/ 3456 39. mailto:gcc-help (a] gcc.gnu.org 3457 40. mailto:gcc (a] gcc.gnu.org 3458 41. https://gcc.gnu.org/lists.html 3459 42. http://www.fsf.org/ 3460 43. https://gcc.gnu.org/about.html 3461 44. http://validator.w3.org/check/referer 3462 ====================================================================== 3463 http://gcc.gnu.org/gcc-4.7/index.html 3464 3465 GCC 4.7 Release Series 3466 3467 June 12, 2014 3468 3469 The [1]GNU project and the GCC developers are pleased to announce the 3470 release of GCC 4.7.4. 3471 3472 This release is a bug-fix release, containing fixes for regressions in 3473 GCC 4.7.3 relative to previous releases of GCC. 3474 3475 This release series is no longer maintained. 3476 3477 Release History 3478 3479 GCC 4.7.4 3480 June 12, 2014 ([2]changes, [3]documentation) 3481 3482 GCC 4.7.3 3483 April 11, 2013 ([4]changes, [5]documentation) 3484 3485 GCC 4.7.2 3486 September 20, 2012 ([6]changes, [7]documentation) 3487 3488 GCC 4.7.1 3489 June 14, 2012 ([8]changes, [9]documentation) 3490 3491 GCC 4.7.0 3492 March 22, 2012 ([10]changes, [11]documentation) 3493 3494 References and Acknowledgements 3495 3496 GCC used to stand for the GNU C Compiler, but since the compiler 3497 supports several other languages aside from C, it now stands for the 3498 GNU Compiler Collection. 3499 3500 A list of [12]successful builds is updated as new information becomes 3501 available. 3502 3503 The GCC developers would like to thank the numerous people that have 3504 contributed new features, improvements, bug fixes, and other changes as 3505 well as test results to GCC. This [13]amazing group of volunteers is 3506 what makes GCC successful. 3507 3508 For additional information about GCC please refer to the [14]GCC 3509 project web site or contact the [15]GCC development mailing list. 3510 3511 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3512 3513 3514 For questions related to the use of GCC, please consult these web 3515 pages and the [18]GCC manuals. If that fails, the 3516 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3517 web pages and the development of GCC are welcome on our developer 3518 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3519 archives. 3520 3521 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3522 distribution of this entire article is permitted in any medium, 3523 provided this notice is preserved. 3524 3525 These pages are [23]maintained by the GCC team. Last modified 3526 2016-09-30[24]. 3527 3528 References 3529 3530 1. http://www.gnu.org/ 3531 2. http://gcc.gnu.org/gcc-4.7/changes.html 3532 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 3533 4. http://gcc.gnu.org/gcc-4.7/changes.html 3534 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 3535 6. http://gcc.gnu.org/gcc-4.7/changes.html 3536 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 3537 8. http://gcc.gnu.org/gcc-4.7/changes.html 3538 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 3539 10. http://gcc.gnu.org/gcc-4.7/changes.html 3540 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 3541 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 3542 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3543 14. http://gcc.gnu.org/index.html 3544 15. mailto:gcc (a] gcc.gnu.org 3545 16. http://gcc.gnu.org/mirrors.html 3546 17. http://gcc.gnu.org/svn.html 3547 18. https://gcc.gnu.org/onlinedocs/ 3548 19. mailto:gcc-help (a] gcc.gnu.org 3549 20. mailto:gcc (a] gcc.gnu.org 3550 21. https://gcc.gnu.org/lists.html 3551 22. http://www.fsf.org/ 3552 23. https://gcc.gnu.org/about.html 3553 24. http://validator.w3.org/check/referer 3554 ====================================================================== 3555 http://gcc.gnu.org/gcc-4.7/changes.html 3556 3557 GCC 4.7 Release Series 3558 Changes, New Features, and Fixes 3559 3560 Caveats 3561 3562 * The -fconserve-space flag has been deprecated. The flag had no 3563 effect for most targets: only targets without a global .bss section 3564 and without support for switchable sections. Furthermore, the flag 3565 only had an effect for G++, where it could result in wrong 3566 semantics (please refer to the GCC manual for further details). The 3567 flag will be removed in GCC 4.8 3568 * Support for a number of older systems and recently unmaintained or 3569 untested target ports of GCC has been declared obsolete in GCC 4.7. 3570 Unless there is activity to revive them, the next release of GCC 3571 will have their sources permanently removed. 3572 All GCC ports for the following processor architectures have been 3573 declared obsolete: 3574 + picoChip (picochip-*) 3575 The following ports for individual systems on particular 3576 architectures have been obsoleted: 3577 + IRIX 6.5 (mips-sgi-irix6.5) 3578 + MIPS OpenBSD (mips*-*-openbsd*) 3579 + Solaris 8 (*-*-solaris2.8). Details can be found in the 3580 [1]announcement. 3581 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 3582 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 3583 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 3584 default, which for some sources generates code that accesses memory 3585 on unaligned addresses. This requires the kernel of those systems 3586 to enable such accesses (controlled by CP15 register c1, refer to 3587 ARM documentation). Alternatively, or for compatibility with 3588 kernels where unaligned accesses are not supported, all code has to 3589 be compiled with -mno-unaligned-access. Upstream Linux kernel 3590 releases have automatically and unconditionally supported unaligned 3591 accesses as emitted by GCC due to this option being active since 3592 version 2.6.28. 3593 * Support on ARM for the legacy floating-point accelerator (FPA) and 3594 the mixed-endian floating-point format that it used has been 3595 obsoleted. The ports that still use this format have been obsoleted 3596 as well. Many legacy ARM ports already provide an alternative that 3597 uses the VFP floating-point format. The obsolete ports will be 3598 deleted in the next release. 3599 The obsolete ports with alternatives are: 3600 + arm*-*-rtems (use arm*-*-rtemseabi) 3601 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 3602 + arm*-*-elf (use arm*-*-eabi) 3603 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 3604 Note, however, that these alternatives are not binary compatible 3605 with their legacy counterparts (although some can support running 3606 legacy applications). 3607 The obsolete ports that currently lack a modern alternative are: 3608 + arm*-*-ecos-elf 3609 + arm*-*-freebsd 3610 + arm*-wince-pe* 3611 New ports that support more recent versions of the architecture are 3612 welcome. 3613 * Support for the Maverick co-processor on ARM has been obsoleted. 3614 Code to support it will be deleted in the next release. 3615 * Support has been removed for Unix International threads on Solaris 3616 2, so the --enable-threads=solaris configure option and the 3617 -threads compiler option don't work any longer. 3618 * Support has been removed for the Solaris BSD Compatibility Package, 3619 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 3620 from Solaris 11, and was only intended as a migration aid from 3621 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 3622 recognized any longer. 3623 * The AVR port's libgcc has been improved and its multilib structure 3624 has been enhanced. As a result, all objects contributing to an 3625 application must either be compiled with GCC versions up to 4.6.x 3626 or with GCC versions 4.7.1 or later. If the compiler is used with 3627 AVR Libc, you need a version that supports the new layout, i.e. 3628 implements [2]#35407. 3629 * The AVR port's -mshort-calls command-line option has been 3630 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 3631 for a replacement. 3632 * The AVR port only references startup code that clears .bss and the 3633 common section resp. initializes the .data and .rodata section 3634 provided respective sections (or subsections thereof) are not 3635 empty, see [3]PR18145. Applications that put all static storage 3636 objects into non-standard sections and / or define all static 3637 storage objects in assembler modules, must reference __do_clear_bss 3638 resp. __do_copy_data by hand or undefine the symbol(s) by means of 3639 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 3640 * The ARM port's -mwords-little-endian option has been deprecated. It 3641 will be removed in a future release. 3642 * Support has been removed for the NetWare x86 configuration 3643 obsoleted in GCC 4.6. 3644 * It is no longer possible to use the "l" constraint in MIPS16 asm 3645 statements. 3646 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 3647 library which affected the ABI in C++11 mode: a data member was 3648 added to std::list changing its size and altering the definitions 3649 of some member functions, and std::pair's move constructor was 3650 non-trivial which altered the calling convention for functions with 3651 std::pair arguments or return types. The ABI incompatibilities have 3652 been fixed for GCC version 4.7.2 but as a result C++11 code 3653 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 3654 code compiled with different GCC versions and with C++98/C++03 code 3655 compiled with any version. 3656 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 3657 rules for the layout of vectors that could lead to wrong code being 3658 generated. Vectors larger than 8 bytes in size are now by default 3659 aligned to an 8-byte boundary. This is an ABI change: code that 3660 makes explicit use of vector types may be incompatible with binary 3661 objects built with older versions of GCC. Auto-vectorized code is 3662 not affected by this change. (This change affects GCC versions 3663 4.7.2 and later.) 3664 * More information on porting to GCC 4.7 from previous versions of 3665 GCC can be found in the [4]porting guide for this release. 3666 3667 General Optimizer Improvements 3668 3669 * Support for a new parameter --param case-values-threshold=n was 3670 added to allow users to control the cutoff between doing switch 3671 statements as a series of if statements and using a jump table. 3672 * Link-time optimization (LTO) improvements: 3673 + Improved scalability and reduced memory usage. Link time 3674 optimization of Firefox now requires 3GB of RAM on a 64-bit 3675 system, while over 8GB was needed previously. Linking time has 3676 been improved, too. The serial stage of linking Firefox has 3677 been sped up by about a factor of 10. 3678 + Reduced size of object files and temporary storage used during 3679 linking. 3680 + Streaming performance (both outbound and inbound) has been 3681 improved. 3682 + ld -r is now supported with LTO. 3683 + Several bug fixes, especially in symbol table handling and 3684 merging. 3685 * Interprocedural optimization improvements: 3686 + Heuristics now take into account that after inlining code will 3687 be optimized out because of known values (or properties) of 3688 function parameters. For example: 3689 void foo(int a) 3690 { 3691 if (a > 10) 3692 ... huge code ... 3693 } 3694 void bar (void) 3695 { 3696 foo (0); 3697 } 3698 3699 The call of foo will be inlined into bar even when optimizing 3700 for code size. Constructs based on __builtin_constant_p are 3701 now understood by the inliner and code size estimates are 3702 evaluated a lot more realistically. 3703 + The representation of C++ virtual thunks and aliases (both 3704 implicit and defined via the alias attribute) has been 3705 re-engineered. Aliases no longer pose optimization barriers 3706 and calls to an alias can be inlined and otherwise optimized. 3707 + The inter-procedural constant propagation pass has been 3708 rewritten. It now performs generic function specialization. 3709 For example when compiling the following: 3710 void foo(bool flag) 3711 { 3712 if (flag) 3713 ... do something ... 3714 else 3715 ... do something else ... 3716 } 3717 void bar (void) 3718 { 3719 foo (false); 3720 foo (true); 3721 foo (false); 3722 foo (true); 3723 foo (false); 3724 foo (true); 3725 } 3726 3727 GCC will now produce two copies of foo. One with flag being 3728 true, while other with flag being false. This leads to 3729 performance improvements previously possible only by inlining 3730 all calls. Cloning causes a lot less code size growth. 3731 * A string length optimization pass has been added. It attempts to 3732 track string lengths and optimize various standard C string 3733 functions like strlen, strchr, strcpy, strcat, stpcpy and their 3734 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 3735 enabled by default at -O2 or above, unless optimizing for size, and 3736 can be disabled by the -fno-optimize-strlen option. The pass can 3737 e.g. optimize 3738 char *bar (const char *a) 3739 { 3740 size_t l = strlen (a) + 2; 3741 char *p = malloc (l); if (p == NULL) return p; 3742 strcpy (p, a); strcat (p, "/"); return p; 3743 } 3744 3745 into: 3746 char *bar (const char *a) 3747 { 3748 size_t tmp = strlen (a); 3749 char *p = malloc (tmp + 2); if (p == NULL) return p; 3750 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 3751 } 3752 3753 or for hosted compilations where stpcpy is available in the runtime 3754 and headers provide its prototype, e.g. 3755 void foo (char *a, const char *b, const char *c, const char *d) 3756 { 3757 strcpy (a, b); strcat (a, c); strcat (a, d); 3758 } 3759 3760 can be optimized into: 3761 void foo (char *a, const char *b, const char *c, const char *d) 3762 { 3763 strcpy (stpcpy (stpcpy (a, b), c), d); 3764 } 3765 3766 New Languages and Language specific improvements 3767 3768 * Version 3.1 of the [5]OpenMP specification is now supported for the 3769 C, C++, and Fortran compilers. 3770 3771 Ada 3772 3773 * The command-line option -feliminate-unused-debug-types has been 3774 re-enabled by default, as it is for the other languages, leading to 3775 a reduction in debug info size of 12.5% and more for relevant 3776 cases, as well as to a small compilation speedup. 3777 3778 C family 3779 3780 * A new built-in, __builtin_assume_aligned, has been added, through 3781 which the compiler can be hinted about pointer alignment and can 3782 use it to improve generated code. 3783 * A new warning option -Wunused-local-typedefs was added for C, C++, 3784 Objective-C and Objective-C++. This warning diagnoses typedefs 3785 locally defined in a function, and otherwise not used. 3786 * A new experimental command-line option -ftrack-macro-expansion was 3787 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 3788 the compiler to emit diagnostic about the current macro expansion 3789 stack when a compilation error occurs in a macro expansion. 3790 * Experimental support for transactional memory has been added. It 3791 includes support in the compiler, as well as a supporting runtime 3792 library called libitm. To compile code with transactional memory 3793 constructs, use the -fgnu-tm option. 3794 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 3795 and 32-bit/64-bit x86 platforms. 3796 For more details on transactional memory see [6]the GCC WiKi. 3797 * Support for atomic operations specifying the C++11/C11 memory model 3798 has been added. These new __atomic routines replace the existing 3799 __sync built-in routines. 3800 Atomic support is also available for memory blocks. Lock-free 3801 instructions will be used if a memory block is the same size and 3802 alignment as a supported integer type. Atomic operations which do 3803 not have lock-free support are left as function calls. A set of 3804 library functions is available on the GCC atomic wiki in the 3805 "External Atomics Library" section. 3806 For more details on the memory models and features, see the 3807 [7]atomic wiki. 3808 * When a binary operation is performed on vector types and one of the 3809 operands is a uniform vector, it is possible to replace the vector 3810 with the generating element. For example: 3811 typedef int v4si __attribute__ ((vector_size (16))); 3812 v4si res, a = {1,2,3,4}; 3813 int x; 3814 3815 res = 2 + a; /* means {2,2,2,2} + a */ 3816 res = a - x; /* means a - {x,x,x,x} */ 3817 3818 C 3819 3820 * There is support for some more features from the C11 revision of 3821 the ISO C standard. GCC now accepts the options -std=c11 and 3822 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 3823 + Unicode strings (previously supported only with options such 3824 as -std=gnu11, now supported with -std=c11), and the 3825 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 3826 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 3827 + Alignment support (_Alignas, _Alignof, max_align_t, 3828 <stdalign.h>). 3829 + A built-in function __builtin_complex is provided to support C 3830 library implementation of the CMPLX family of macros. 3831 3832 C++ 3833 3834 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 3835 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 3836 -Wc++0x-compat, respectively. 3837 * G++ now implements [8]C++11 extended friend syntax: 3838 3839 template<class W> 3840 class Q 3841 { 3842 static const int I = 2; 3843 public: 3844 friend W; 3845 }; 3846 3847 struct B 3848 { 3849 int ar[Q<B>::I]; 3850 }; 3851 3852 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit 3853 override control. 3854 3855 struct B { 3856 virtual void f() const final; 3857 virtual void f(int); 3858 }; 3859 3860 struct D : B { 3861 void f() const; // error: D::f attempts to override final B::f 3862 void f(long) override; // error: doesn't override anything 3863 void f(int) override; // ok 3864 }; 3865 3866 struct E final { }; 3867 struct F: E { }; // error: deriving from final class 3868 3869 * G++ now implements [10]C++11 non-static data member initializers. 3870 3871 struct A { 3872 int i = 42; 3873 } a; // initializes a.i to 42 3874 3875 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11 3876 user-defined literals. 3877 3878 // Not actually a good approximation. :) 3879 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 3880 long double pi = 180.0_degrees; 3881 3882 * G++ now implements [12]C++11 alias-declarations. 3883 3884 template <class T> using Ptr = T*; 3885 Ptr<int> ip; // decltype(ip) is int* 3886 3887 * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements 3888 [13]C++11 delegating constructors. 3889 3890 struct A { 3891 A(int); 3892 A(): A(42) { } // delegate to the A(int) constructor 3893 }; 3894 3895 * G++ now fully implements C++11 atomic classes rather than just 3896 integer derived classes. 3897 3898 class POD { 3899 int a; 3900 int b; 3901 }; 3902 std::atomic<POD> my_atomic_POD; 3903 3904 * G++ now sets the predefined macro __cplusplus to the correct value, 3905 199711L for C++98/03, and 201103L for C++11. 3906 * G++ now correctly implements the two-phase lookup rules such that 3907 an unqualified name used in a template must have an appropriate 3908 declaration found either in scope at the point of definition of the 3909 template or by argument-dependent lookup at the point of 3910 instantiation. As a result, code that relies on a second 3911 unqualified lookup at the point of instantiation to find functions 3912 declared after the template or in dependent bases will be rejected. 3913 The compiler will suggest ways to fix affected code, and using the 3914 -fpermissive compiler flag will allow the code to compile with a 3915 warning. 3916 3917 template <class T> 3918 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 3919 void g(int) { } // fix by moving this declaration before the declaration of f 3920 3921 template <class T> 3922 struct A: T { 3923 // error, B::g(B) not found by argument-dependent lookup 3924 void f() { g(T()); } // fix by using this->g or A::g 3925 }; 3926 3927 struct B { void g(B); }; 3928 3929 int main() 3930 { 3931 f<int>(); 3932 A<B>().f(); 3933 } 3934 3935 * G++ now properly re-uses stack space allocated for temporary 3936 objects when their lifetime ends, which can significantly lower 3937 stack consumption for some C++ functions. As a result of this, some 3938 code with undefined behavior will now break: 3939 3940 const int &f(const int &i) { return i; } 3941 .... 3942 const int &x = f(1); 3943 const int &y = f(2); 3944 3945 Here, x refers to the temporary allocated to hold the 1 argument, 3946 which only lives until the end of the initialization; it 3947 immediately becomes a dangling reference. So the next statement 3948 re-uses the stack slot to hold the 2 argument, and users of x get 3949 that value instead. 3950 Note that this should not cause any change of behavior for 3951 temporaries of types with non-trivial destructors, as they are 3952 already destroyed at end of full-expression; the change is that now 3953 the storage is released as well. 3954 * A new command-line option -Wdelete-non-virtual-dtor has been added 3955 to warn when delete is used to destroy an instance of a class which 3956 has virtual functions and non-virtual destructor. It is unsafe to 3957 delete an instance of a derived class through a pointer to a base 3958 class if the base class does not have a virtual destructor. This 3959 warning is enabled by -Wall. 3960 * A new command-line option -Wzero-as-null-pointer-constant has been 3961 added to warn when a literal '0' is used as null pointer constant. 3962 It can be useful to facilitate the conversion to nullptr in C++11. 3963 * As per C++98, access-declarations are now deprecated by G++. 3964 Using-declarations are to be used instead. Furthermore, some 3965 efforts have been made to improve the support of class scope 3966 using-declarations. In particular, using-declarations referring to 3967 a dependent type now work as expected ([14]bug c++/14258). 3968 * The ELF symbol visibility of a template instantiation is now 3969 properly constrained by the visibility of its template arguments 3970 ([15]bug c++/35688). 3971 3972 Runtime Library (libstdc++) 3973 3974 * [16]Improved experimental support for the new ISO C++ standard, 3975 C++11, including: 3976 + using noexcept in most of the library; 3977 + implementations of pointer_traits, allocator_traits and 3978 scoped_allocator_adaptor; 3979 + uses-allocator construction for tuple; 3980 + vector meets the allocator-aware container requirements; 3981 + replacing monotonic_clock with steady_clock; 3982 + enabling the thread support library on most POSIX targets; 3983 + many small improvements to conform to the FDIS. 3984 * Added --enable-clocale=newlib configure option. 3985 * Debug Mode iterators for unordered associative containers. 3986 * Avoid polluting the global namespace and do not include <unistd.h>. 3987 3988 Fortran 3989 3990 * The compile flag [17]-fstack-arrays has been added, which causes 3991 all local arrays to be put on stack memory. For some programs this 3992 will improve the performance significantly. If your program uses 3993 very large local arrays, it is possible that you will have to 3994 extend your runtime limits for stack memory. 3995 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 3996 [20]-fstack-arrays. 3997 * Front-end optimizations can now be selected by the 3998 [21]-ffrontend-optimize option and deselected by the 3999 -fno-frontend-optimize option. 4000 * When front-end optimization removes a function call, 4001 [22]-Wfunction-elimination warns about that. 4002 * When performing front-end-optimization, the 4003 [23]-faggressive-function-elimination option allows the removal of 4004 duplicate function calls even for impure functions. 4005 * The flag [24]-Wreal-q-constant has been added, which warns if 4006 floating-point literals have been specified using q (such as 4007 1.0q0); the q marker is now supported as a vendor extension to 4008 denote quad precision (REAL(16) or, if not available, REAL(10)). 4009 Consider using a kind parameter (such as in 1.0_qp) instead, which 4010 can be obtained via [25]SELECTED_REAL_KIND. 4011 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 4012 Fortran now always prints error messages to standard error. If you 4013 wish to redirect standard error, please consult the manual for your 4014 OS, shell, batch environment etc. as appropriate. 4015 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 4016 variable have been removed. When encountering a serious error, 4017 gfortran will now always abort the program. Whether a core dump is 4018 generated depends on the user environment settings; see the ulimit 4019 -c setting for POSIX shells, limit coredumpsize for C shells, and 4020 the [26]WER user-mode dumps settings on Windows. 4021 * The [27]-fbacktrace option is now enabled by default. When 4022 encountering a fatal error, gfortran will attempt to print a 4023 backtrace to standard error before aborting. It can be disabled 4024 with -fno-backtrace. Note: On POSIX targets with the addr2line 4025 utility from GNU binutils, GNU Fortran can print a backtrace with 4026 function name, file name, line number information in addition to 4027 the addresses; otherwise only the addresses are printed. 4028 * [28]Fortran 2003: 4029 + Generic interface names which have the same name as derived 4030 types are now supported, which allows to write constructor 4031 functions. Note that Fortran does not support static 4032 constructor functions; only default initialization or an 4033 explicit structure-constructor initialization are available. 4034 + [29]Polymorphic (class) arrays are now supported. 4035 * [30]Fortran 2008: 4036 + Support for the DO CONCURRENT construct has been added, which 4037 allows the user to specify that individual loop iterations 4038 have no interdependencies. 4039 + [31]Coarrays: Full single-image support except for polymorphic 4040 coarrays. Additionally, preliminary support for multiple 4041 images via an MPI-based [32]coarray communication library has 4042 been added. Note: The library version is not yet usable as 4043 remote coarray access is not yet possible. 4044 * [33]TS 29113: 4045 + New flag [34]-std=f2008ts permits programs that are expected 4046 to conform to the Fortran 2008 standard and the draft 4047 Technical Specification (TS) 29113 on Further Interoperability 4048 of Fortran with C. 4049 + The OPTIONAL attribute is now allowed for dummy arguments of 4050 BIND(C) procedures. 4051 + The RANK intrinsic has been added. 4052 + The implementation of the ASYNCHRONOUS attribute in GCC is 4053 compatible with the candidate draft of TS 29113 (since GCC 4054 4.6). 4055 4056 Go 4057 4058 * GCC 4.7 implements the [35]Go 1 language standard. The library 4059 support in 4.7.0 is not quite complete, due to release timing. 4060 Release 4.7.1 includes complete support for Go 1. The Go library is 4061 from the Go 1.0.1 release. 4062 * Go has been tested on GNU/Linux and Solaris platforms. It may work 4063 on other platforms as well. 4064 4065 New Targets and Target Specific Improvements 4066 4067 ARM 4068 4069 * GCC now supports the Cortex-A7 processor implementing the v7-a 4070 version of the architecture using the option -mcpu=cortex-a7. 4071 * The default vector size in auto-vectorization for NEON is now 128 4072 bits. If vectorization fails thusly, the vectorizer tries again 4073 with 64-bit vectors. 4074 * A new option -mvectorize-with-neon-double was added to allow users 4075 to change the vector size to 64 bits. 4076 4077 AVR 4078 4079 * GCC now supports the XMEGA architecture. This requires GNU binutils 4080 2.22 or later. 4081 * Support for the [36]named address spaces __flash, __flash1, ..., 4082 __flash5 and __memx has been added. These address spaces locate 4083 read-only data in flash memory and allow reading from flash memory 4084 by means of ordinary C code, i.e. without the need of (inline) 4085 assembler code: 4086 4087 const __flash int values[] = { 42, 31 }; 4088 4089 int add_values (const __flash int *p, int i) 4090 { 4091 return values[i] + *p; 4092 } 4093 4094 * Support has been added for the AVR-specific configure option 4095 --with-avrlibc=yes in order to arrange for better integration of 4096 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 4097 and newer and will only take effect in non-RTEMS configurations. If 4098 avr-gcc is configured for RTEMS, the option will be ignored which 4099 is the same as specifying --with-avrlibc=no. See [38]PR54461 for 4100 more technical details. 4101 * Support for AVR-specific [39]built-in functions has been added. 4102 * Support has been added for the signed and unsigned 24-bit scalar 4103 integer types __int24 and __uint24. 4104 * New command-line options -maccumulate-args, -mbranch-cost=cost and 4105 -mstrict-X were added to allow better fine-tuning of code 4106 optimization. 4107 * The command-line option -fdata-sections now also takes affect on 4108 the section names of variables with the progmem attribute. 4109 * A new inline assembler print modifier %i to print a RAM address as 4110 I/O address has been added: 4111 4112 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 4113 4114 void set_portb (uint8_t value) 4115 { 4116 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 4117 } 4118 4119 The offset between an I/O address and the RAM address for that I/O 4120 location is device-specific. This offset is taken into account when 4121 printing a RAM address with the %i modifier so that the address is 4122 suitable to be used as operand in an I/O command. The address must 4123 be a constant integer known at compile time. 4124 * The inline assembler constraint "R" to represent integers in the 4125 range -6 ... 5 has been removed without replacement. 4126 * Many optimizations to: 4127 + 64-bit integer arithmetic 4128 + Widening multiplication 4129 + Integer division by a constant 4130 + Avoid constant reloading in multi-byte instructions. 4131 + Micro-optimizations for special instruction sequences. 4132 + Generic built-in functions like __builtin_ffs*, 4133 __builtin_clz*, etc. 4134 + If-else decision trees generated by switch instructions 4135 + Merging of data located in flash memory 4136 + New libgcc variants for devices with 8-bit wide stack pointer 4137 + ... 4138 * Better documentation: 4139 + Handling of EIND and indirect jumps on devices with more than 4140 128 KiB of program memory. 4141 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 4142 registers. 4143 + Function attributes OS_main and OS_task. 4144 + AVR-specific built-in macros. 4145 4146 C6X 4147 4148 * Support has been added for the Texas Instruments C6X family of 4149 processors. 4150 4151 CR16 4152 4153 * Support has been added for National Semiconductor's CR16 4154 architecture. 4155 4156 Epiphany 4157 4158 * Support has been added for Adapteva's Epiphany architecture. 4159 4160 IA-32/x86-64 4161 4162 * Support for Intel AVX2 intrinsics, built-in functions and code 4163 generation is available via -mavx2. 4164 * Support for Intel BMI2 intrinsics, built-in functions and code 4165 generation is available via -mbmi2. 4166 * Implementation and automatic generation of __builtin_clz* using the 4167 lzcnt instruction is available via -mlzcnt. 4168 * Support for Intel FMA3 intrinsics and code generation is available 4169 via -mfma. 4170 * A new -mfsgsbase command-line option is available that makes GCC 4171 generate new segment register read/write instructions through 4172 dedicated built-ins. 4173 * Support for the new Intel rdrnd instruction is available via 4174 -mrdrnd. 4175 * Two additional AVX vector conversion instructions are available via 4176 -mf16c. 4177 * Support for new Intel processor codename IvyBridge with RDRND, 4178 FSGSBASE and F16C is available through -march=core-avx-i. 4179 * Support for the new Intel processor codename Haswell with AVX2, 4180 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 4181 * Support for new AMD family 15h processors (Piledriver core) is now 4182 available through -march=bdver2 and -mtune=bdver2 options. 4183 * Support for [40]the x32 psABI is now available through the -mx32 4184 option. 4185 * Windows mingw targets are using the -mms-bitfields option by 4186 default. 4187 * Windows x86 targets are using the __thiscall calling convention for 4188 C++ class-member functions. 4189 * Support for the configure option --with-threads=posix for Windows 4190 mingw targets. 4191 4192 MIPS 4193 4194 * GCC now supports thread-local storage (TLS) for MIPS16. This 4195 requires GNU binutils 2.22 or later. 4196 * GCC can now generate code specifically for the Cavium Octeon+ and 4197 Octeon2 processors. The associated command-line options are 4198 -march=octeon+ and -march=octeon2 respectively. Both options 4199 require GNU binutils 2.22 or later. 4200 * GCC can now work around certain 24k errata, under the control of 4201 the command-line option -mfix-24k. These workarounds require GNU 4202 binutils 2.20 or later. 4203 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 4204 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 4205 toolchain that generates 32-bit code by default. Use the 4206 configure-time option --enable-targets=all to select these extra 4207 multilibs. 4208 * Passing -fno-delayed-branch now also stops the assembler from 4209 automatically filling delay slots. 4210 4211 PowerPC/PowerPC64 4212 4213 * Vectors of type vector long long or vector long are passed and 4214 returned using the same method as other vectors with the VSX 4215 instruction set. Previously GCC did not adhere to the ABI for 4216 128-bit vectors with 64-bit integer base types (PR 48857). This 4217 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 4218 * A new option -mno-pointers-to-nested-functions was added to allow 4219 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 4220 that the compiler should not load up the chain register (r11) 4221 before calling a function through a pointer. If you use this 4222 option, you cannot call nested functions through a pointer, or call 4223 other languages that might use the static chain. 4224 * A new option msave-toc-indirect was added to allow AIX 4225 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 4226 save the TOC in the prologue for indirect calls or generate the 4227 save inline. This can speed up some programs that call through a 4228 function pointer a lot, but it can slow down other functions that 4229 only call through a function pointer in exceptional cases. 4230 * The PowerPC port will now enable machine-specific built-in 4231 functions when the user switches the target machine using the 4232 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 4233 sequences. In addition, the target macros are updated. However, due 4234 to the way the -save-temps switch is implemented, you won't see the 4235 effect of these additional macros being defined in preprocessor 4236 output. 4237 4238 SH 4239 4240 * A new option -msoft-atomic has been added. When it is specified, 4241 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 4242 the new __atomic routines. 4243 * Since it is neither supported by GAS nor officially documented, 4244 code generation for little endian SH2A has been disabled. 4245 Specifying -ml with -m2a* will now result in a compiler error. 4246 * The defunct -mbranch-cost option has been fixed. 4247 * Some improvements to the generated code of: 4248 + Utilization of the tst #imm,R0 instruction. 4249 + Dynamic shift instructions on SH2A. 4250 + Integer absolute value calculations. 4251 * The -mdiv= option for targets other than SHmedia has been fixed and 4252 documented. 4253 4254 SPARC 4255 4256 * The option -mflat has been reinstated. When it is specified, the 4257 compiler will generate code for a single register window model. 4258 This is essentially a new implementation and the corresponding 4259 debugger support has been added to GDB 7.4. 4260 * Support for the options -mtune=native and -mcpu=native has been 4261 added on selected native platforms (GNU/Linux and Solaris). 4262 * Support for the SPARC T3 (Niagara 3) processor has been added. 4263 * VIS: 4264 + An intrinsics header visintrin.h has been added. 4265 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 4266 compare instructions have been added. 4267 + The little-endian version of alignaddr is now supported. 4268 + When possible, VIS builtins are marked const, which should 4269 increase the compiler's ability to optimize VIS operations. 4270 + The compiler now properly tracks the %gsr register and how it 4271 behaves as an input for various VIS instructions. 4272 + Akin to fzero, the compiler can now generate fone instructions 4273 in order to set all of the bits of a floating-point register 4274 to 1. 4275 + The documentation for the VIS intrinsics in the GCC manual has 4276 been brought up to date and many inaccuracies were fixed. 4277 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 4278 non-condition-code setting edge instructions have been added. 4279 Their availability is controlled by the new -mvis2 and 4280 -mno-vis2 options. They are enabled by default on 4281 UltraSPARC-III and later CPUs. 4282 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 4283 has been added. These instructions are enabled by default on SPARC 4284 T3 (Niagara 3) and later CPUs. 4285 4286 TILE-Gx/TILEPro 4287 4288 * Support has been added for the Tilera TILE-Gx and TILEPro families 4289 of processors. 4290 4291 Other significant improvements 4292 4293 * A new option (-grecord-gcc-switches) was added that appends 4294 compiler command-line options that might affect code generation to 4295 the DW_AT_producer attribute string in the DWARF debugging 4296 information. 4297 * GCC now supports various new GNU extensions to the DWARF debugging 4298 information format, like [41]entry value and [42]call site 4299 information, [43]typed DWARF stack or [44]a more compact macro 4300 representation. Support for these extensions has been added to GDB 4301 7.4. They can be disabled through the -gstrict-dwarf command-line 4302 option. 4303 4304 GCC 4.7.1 4305 4306 This is the [45]list of problem reports (PRs) from GCC's bug tracking 4307 system that are known to be fixed in the 4.7.1 release. This list might 4308 not be complete (that is, it is possible that some PRs that have been 4309 fixed are not listed here). 4310 4311 The Go frontend in the 4.7.1 release fully supports the [46]Go 1 4312 language standard. 4313 4314 GCC 4.7.2 4315 4316 This is the [47]list of problem reports (PRs) from GCC's bug tracking 4317 system that are known to be fixed in the 4.7.2 release. This list might 4318 not be complete (that is, it is possible that some PRs that have been 4319 fixed are not listed here). 4320 4321 GCC 4.7.3 4322 4323 This is the [48]list of problem reports (PRs) from GCC's bug tracking 4324 system that are known to be fixed in the 4.7.3 release. This list might 4325 not be complete (that is, it is possible that some PRs that have been 4326 fixed are not listed here). 4327 4328 GCC 4.7.4 4329 4330 This is the [49]list of problem reports (PRs) from GCC's bug tracking 4331 system that are known to be fixed in the 4.7.4 release. This list might 4332 not be complete (that is, it is possible that some PRs that have been 4333 fixed are not listed here). 4334 4335 4336 For questions related to the use of GCC, please consult these web 4337 pages and the [50]GCC manuals. If that fails, the 4338 [51]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4339 web pages and the development of GCC are welcome on our developer 4340 list at [52]gcc (a] gcc.gnu.org. All of [53]our lists have public 4341 archives. 4342 4343 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 4344 distribution of this entire article is permitted in any medium, 4345 provided this notice is preserved. 4346 4347 These pages are [55]maintained by the GCC team. Last modified 4348 2017-03-03[56]. 4349 4350 References 4351 4352 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 4353 2. http://savannah.nongnu.org/bugs/?35407 4354 3. https://gcc.gnu.org/PR18145 4355 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 4356 5. http://www.openmp.org/specifications/ 4357 6. https://gcc.gnu.org/wiki/TransactionalMemory 4358 7. https://gcc.gnu.org/wiki/Atomic/GCCMM 4359 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4360 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4361 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4362 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4363 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4364 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4365 14. https://gcc.gnu.org/PR14258 4366 15. https://gcc.gnu.org/PR35688 4367 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 4368 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 4369 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 4370 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 4371 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 4372 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 4373 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 4374 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 4375 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 4376 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 4377 26. https://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx 4378 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 4379 28. https://gcc.gnu.org/wiki/Fortran2003Status 4380 29. https://gcc.gnu.org/wiki/OOP 4381 30. https://gcc.gnu.org/wiki/Fortran2008Status 4382 31. https://gcc.gnu.org/wiki/Coarray 4383 32. https://gcc.gnu.org/wiki/CoarrayLib 4384 33. https://gcc.gnu.org/wiki/TS29113Status 4385 34. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 4386 35. https://golang.org/doc/go1 4387 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 4388 37. http://nongnu.org/avr-libc/ 4389 38. https://gcc.gnu.org/PR54461 4390 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 4391 40. https://sites.google.com/site/x32abi/ 4392 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 4393 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 4394 43. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 4395 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 4396 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 4397 46. https://golang.org/doc/go1 4398 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 4399 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 4400 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 4401 50. https://gcc.gnu.org/onlinedocs/ 4402 51. mailto:gcc-help (a] gcc.gnu.org 4403 52. mailto:gcc (a] gcc.gnu.org 4404 53. https://gcc.gnu.org/lists.html 4405 54. http://www.fsf.org/ 4406 55. https://gcc.gnu.org/about.html 4407 56. http://validator.w3.org/check/referer 4408 ====================================================================== 4409 http://gcc.gnu.org/gcc-4.6/index.html 4410 4411 GCC 4.6 Release Series 4412 4413 April 12, 2013 4414 4415 The [1]GNU project and the GCC developers are pleased to announce the 4416 release of GCC 4.6.4. 4417 4418 This release is a bug-fix release, containing fixes for regressions in 4419 GCC 4.6.3 relative to previous releases of GCC. 4420 4421 This release series is no longer maintained. 4422 4423 Release History 4424 4425 GCC 4.6.4 4426 April 12, 2013 ([2]changes, [3]documentation) 4427 4428 GCC 4.6.3 4429 March 1, 2012 ([4]changes, [5]documentation) 4430 4431 GCC 4.6.2 4432 October 26, 2011 ([6]changes, [7]documentation) 4433 4434 GCC 4.6.1 4435 June 27, 2011 ([8]changes, [9]documentation) 4436 4437 GCC 4.6.0 4438 March 25, 2011 ([10]changes, [11]documentation) 4439 4440 References and Acknowledgements 4441 4442 GCC used to stand for the GNU C Compiler, but since the compiler 4443 supports several other languages aside from C, it now stands for the 4444 GNU Compiler Collection. 4445 4446 A list of [12]successful builds is updated as new information becomes 4447 available. 4448 4449 The GCC developers would like to thank the numerous people that have 4450 contributed new features, improvements, bug fixes, and other changes as 4451 well as test results to GCC. This [13]amazing group of volunteers is 4452 what makes GCC successful. 4453 4454 For additional information about GCC please refer to the [14]GCC 4455 project web site or contact the [15]GCC development mailing list. 4456 4457 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 4458 4459 4460 For questions related to the use of GCC, please consult these web 4461 pages and the [18]GCC manuals. If that fails, the 4462 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4463 web pages and the development of GCC are welcome on our developer 4464 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4465 archives. 4466 4467 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4468 distribution of this entire article is permitted in any medium, 4469 provided this notice is preserved. 4470 4471 These pages are [23]maintained by the GCC team. Last modified 4472 2016-09-30[24]. 4473 4474 References 4475 4476 1. http://www.gnu.org/ 4477 2. http://gcc.gnu.org/gcc-4.6/changes.html 4478 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 4479 4. http://gcc.gnu.org/gcc-4.6/changes.html 4480 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 4481 6. http://gcc.gnu.org/gcc-4.6/changes.html 4482 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 4483 8. http://gcc.gnu.org/gcc-4.6/changes.html 4484 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 4485 10. http://gcc.gnu.org/gcc-4.6/changes.html 4486 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 4487 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 4488 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4489 14. http://gcc.gnu.org/index.html 4490 15. mailto:gcc (a] gcc.gnu.org 4491 16. http://gcc.gnu.org/mirrors.html 4492 17. http://gcc.gnu.org/svn.html 4493 18. https://gcc.gnu.org/onlinedocs/ 4494 19. mailto:gcc-help (a] gcc.gnu.org 4495 20. mailto:gcc (a] gcc.gnu.org 4496 21. https://gcc.gnu.org/lists.html 4497 22. http://www.fsf.org/ 4498 23. https://gcc.gnu.org/about.html 4499 24. http://validator.w3.org/check/referer 4500 ====================================================================== 4501 http://gcc.gnu.org/gcc-4.6/changes.html 4502 4503 GCC 4.6 Release Series 4504 Changes, New Features, and Fixes 4505 4506 Caveats 4507 4508 * The options -b <machine> and -V <version> have been removed because 4509 they were unreliable. Instead, users should directly run 4510 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 4511 run a different version of gcc. 4512 * GCC now has stricter checks for invalid command-line options. In 4513 particular, when gcc was called to link object files rather than 4514 compile source code, it would previously accept and ignore all 4515 options starting with --, including linker options such as 4516 --as-needed and --export-dynamic, although such options would 4517 result in errors if any source code was compiled. Such options, if 4518 unknown to the compiler, are now rejected in all cases; if the 4519 intent was to pass them to the linker, options such as 4520 -Wl,--as-needed should be used. 4521 * Versions of the GNU C library up to and including 2.11.1 included 4522 an [1]incorrect implementation of the cproj function. GCC optimizes 4523 its builtin cproj according to the behavior specified and allowed 4524 by the ISO C99 standard. If you want to avoid discrepancies between 4525 the C library and GCC's builtin transformations when using cproj in 4526 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 4527 and actually rely on the incorrect behavior of cproj, then you can 4528 disable GCC's transformations using -fno-builtin-cproj. 4529 * The C-only intermodule optimization framework (IMA, enabled by 4530 -combine) has been removed in favor of the new generic link-time 4531 optimization framework (LTO) introduced in [2]GCC 4.5.0. 4532 * GCC now ships with the LGPL-licensed libquadmath library, which 4533 provides quad-precision mathematical functions for targets with a 4534 __float128 datatype. __float128 is available for targets on 32-bit 4535 x86, x86-64 and Itanium architectures. The libquadmath library is 4536 automatically built on such targets when building the Fortran 4537 compiler. 4538 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 4539 warnings were added for C, C++, Objective-C and Objective-C++. 4540 These warnings diagnose variables respective parameters which are 4541 only set in the code and never otherwise used. Usually such 4542 variables are useless and often even the value assigned to them is 4543 computed needlessly, sometimes expensively. The 4544 -Wunused-but-set-variable warning is enabled by default by -Wall 4545 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 4546 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 4547 rules for the layout of vectors that could lead to wrong code being 4548 generated. Vectors larger than 8 bytes in size are now by default 4549 aligned to an 8-byte boundary. This is an ABI change: code that 4550 makes explicit use of vector types may be incompatible with binary 4551 objects built with older versions of GCC. Auto-vectorized code is 4552 not affected by this change. (This change affects GCC versions 4553 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 4554 * On AVR, variables with the progmem attribute to locate data in 4555 flash memory must be qualified as const. 4556 * Support for a number of older systems and recently unmaintained or 4557 untested target ports of GCC has been declared obsolete in GCC 4.6. 4558 Unless there is activity to revive them, the next release of GCC 4559 will have their sources permanently removed. 4560 All GCC ports for the following processor architectures have been 4561 declared obsolete: 4562 + Argonaut ARC (arc-*) 4563 + National Semiconductor CRX (crx-*) 4564 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 4565 m68hc12-*-*, m6812-*-*) 4566 + Sunplus S+core (score-*) 4567 The following ports for individual systems on particular 4568 architectures have been obsoleted: 4569 + Interix (i[34567]86-*-interix3*) 4570 + NetWare x86 (i[3456x]86-*-netware*) 4571 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 4572 + MCore PE (mcore-*-pe*) 4573 + SH SymbianOS (sh*-*-symbianelf*) 4574 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 4575 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 4576 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 4577 vax-*-netbsd*, but not *-*-netbsdelf*) 4578 The i[34567]86-*-pe alias for Cygwin targets has also been 4579 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 4580 Certain configure options to control the set of libraries built 4581 with GCC on some targets have been obsoleted. On ARM targets, the 4582 options --disable-fpu, --disable-26bit, --disable-underscore, 4583 --disable-interwork, --disable-biendian and --disable-nofmult have 4584 been obsoleted. On MIPS targets, the options 4585 --disable-single-float, --disable-biendian and --disable-softfloat 4586 have been obsoleted. 4587 * Support has been removed for all the [3]configurations obsoleted in 4588 GCC 4.5. 4589 * More information on porting to GCC 4.6 from previous versions of 4590 GCC can be found in the [4]porting guide for this release. 4591 4592 General Optimizer Improvements 4593 4594 * A new general optimization level, -Ofast, has been introduced. It 4595 combines the existing optimization level -O3 with options that can 4596 affect standards compliance but result in better optimized code. 4597 For example, -Ofast enables -ffast-math. 4598 * Link-time optimization improvements: 4599 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 4600 stabilized to the point of being usable. It has become the 4601 default mode when using the LTO optimization model. Link time 4602 optimization can now split itself into multiple parallel 4603 compilations. Parallelism is controlled with -flto=n (where n 4604 specifies the number of compilations to execute in parallel). 4605 GCC can also cooperate with a GNU make job server by 4606 specifying the -flto=jobserver option and adding + to the 4607 beginning of the Makefile rule executing the linker. 4608 Classical LTO mode can be enforced by -flto-partition=none. 4609 This may result in small code quality improvements. 4610 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 4611 and other large applications can be built with LTO enabled. 4612 + The linker plugin support improvements 4613 o Linker plugin is now enabled by default when the linker 4614 is detected to have plugin support. This is the case for 4615 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 4616 the Gold linker on ELF targets. Plugin support of the 4617 Apple linker on Darwin is not compatible with GCC. The 4618 linker plugin can also be controlled by the 4619 -fuse-linker-plugin command-line option. 4620 o Resolution information from the linker plugin is used to 4621 drive whole program assumptions. Use of the linker plugin 4622 results in more aggressive optimization on binaries and 4623 on shared libraries that use the hidden visibility 4624 attribute. Consequently the use of -fwhole-program is not 4625 necessary in addition to LTO. 4626 + Hidden symbols used from non-LTO objects now have to be 4627 explicitly annotated with externally_visible when the linker 4628 plugin is not used. 4629 + C++ inline functions and virtual tables are now privatized 4630 more aggressively, leading to better inter-procedural 4631 optimization and faster dynamic linking. 4632 + Memory usage and intermediate language streaming performance 4633 have been improved. 4634 + Static constructors and destructors from individual units are 4635 inlined into a single function. This can significantly improve 4636 startup times of large C++ applications where static 4637 constructors are very common. For example, static constructors 4638 are used when including the iostream header. 4639 + Support for the Ada language has been added. 4640 * Interprocedural optimization improvements 4641 + The interprocedural framework was re-tuned for link time 4642 optimization. Several scalability issues were resolved. 4643 + Improved auto-detection of const and pure functions. Newly, 4644 noreturn functions are auto-detected. 4645 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 4646 available that informs users when adding attributes to headers 4647 might improve code generation. 4648 + A number of inlining heuristic improvements. In particular: 4649 o Partial inlining is now supported and enabled by default 4650 at -O2 and greater. The feature can be controlled via 4651 -fpartial-inlining. 4652 Partial inlining splits functions with short hot path to 4653 return. This allows more aggressive inlining of the hot 4654 path leading to better performance and often to code size 4655 reductions (because cold parts of functions are not 4656 duplicated). 4657 o Scalability for large compilation units was improved 4658 significantly. 4659 o Inlining of callbacks is now more aggressive. 4660 o Virtual methods are considered for inlining when the 4661 caller is inlined and devirtualization is then possible. 4662 o Inlining when optimizing for size (either in cold regions 4663 of a program or when compiling with -Os) was improved to 4664 better handle C++ programs with larger abstraction 4665 penalty, leading to smaller and faster code. 4666 + The IPA reference optimization pass detecting global variables 4667 used or modified by functions was strengthened and sped up. 4668 + Functions whose address was taken are now optimized out when 4669 all references to them are dead. 4670 + A new inter-procedural static profile estimation pass detects 4671 functions that are executed once or unlikely to be executed. 4672 Unlikely executed functions are optimized for size. Functions 4673 executed once are optimized for size except for the inner 4674 loops. 4675 + On most targets with named section support, functions used 4676 only at startup (static constructors and main), functions used 4677 only at exit and functions detected to be cold are placed into 4678 separate text segment subsections. This extends the 4679 -freorder-functions feature and is controlled by the same 4680 switch. The goal is to improve the startup time of large C++ 4681 programs. 4682 Proper function placement requires linker support. GNU ld 4683 2.21.51 on ELF targets was updated to place those functions 4684 together within the text section leading to better code 4685 locality and faster startup times of large C++ programs. The 4686 feature is also supported in the Apple linker. Support in the 4687 gold linker is planned. 4688 * A new switch -fstack-usage has been added. It makes the compiler 4689 output stack usage information for the program, on a per-function 4690 basis, in an auxiliary file. 4691 * A new switch -fcombine-stack-adjustments has been added. It can be 4692 used to enable or disable the compiler's stack-slot combining pass 4693 which before was enabled automatically at -O1 and above, but could 4694 not be controlled on its own. 4695 * A new switch -fstrict-volatile-bitfields has been added. Using it 4696 indicates that accesses to volatile bitfields should use a single 4697 access of the width of the field's type. This option can be useful 4698 for precisely defining and accessing memory-mapped peripheral 4699 registers from C or C++. 4700 4701 Compile time and memory usage improvements 4702 4703 * Datastructures used by the dataflow framework in GCC were 4704 reorganized for better memory usage and more cache locality. 4705 Compile time is improved especially on units with large functions 4706 (possibly resulting from a lot of inlining) not fitting into the 4707 processor cache. The compile time of the GCC C compiler binary with 4708 link-time optimization went down by over 10% (benchmarked on x86-64 4709 target). 4710 4711 New Languages and Language specific improvements 4712 4713 Ada 4714 4715 * Stack checking has been improved on selected architectures (Alpha, 4716 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 4717 overflows in all cases on these architectures. 4718 * Initial support for Ada 2012 has been added. 4719 4720 C family 4721 4722 * A new warning, enabled by -Wdouble-promotion, has been added that 4723 warns about cases where a value of type float is implicitly 4724 promoted to double. This is especially helpful for CPUs that handle 4725 the former in hardware, but emulate the latter in software. 4726 * A new function attribute leaf was introduced. This attribute allows 4727 better inter-procedural optimization across calls to functions that 4728 return to the current unit only via returning or exception 4729 handling. This is the case for most library functions that have no 4730 callbacks. 4731 * Support for a new data type __int128 for targets having wide enough 4732 machine-mode support. 4733 * The new function attribute callee_pop_aggregate allows to specify 4734 if the caller or callee is responsible for popping the aggregate 4735 return pointer value from the stack. 4736 * Support for selectively enabling and disabling warnings via #pragma 4737 GCC diagnostic has been added. For instance: 4738 #pragma GCC diagnostic error "-Wuninitialized" 4739 foo(a); /* error is given for this one */ 4740 #pragma GCC diagnostic push 4741 #pragma GCC diagnostic ignored "-Wuninitialized" 4742 foo(b); /* no diagnostic for this one */ 4743 #pragma GCC diagnostic pop 4744 foo(c); /* error is given for this one */ 4745 #pragma GCC diagnostic pop 4746 foo(d); /* depends on command-line options */ 4747 4748 * The -fmax-errors=N option is now supported. Using this option 4749 causes the compiler to exit after N errors have been issued. 4750 4751 C 4752 4753 * There is now experimental support for some features from the 4754 upcoming C1X revision of the ISO C standard. This support may be 4755 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 4756 Note that this support is experimental and may change incompatibly 4757 in future releases for consistency with changes to the C1X standard 4758 draft. The following features are newly supported as described in 4759 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 4760 meeting); some other features were already supported with no 4761 compiler changes being needed, or have some support but not in full 4762 accord with N1539 (as amended). 4763 + Static assertions (_Static_assert keyword) 4764 + Typedef redefinition 4765 + New macros in <float.h> 4766 + Anonymous structures and unions 4767 * The new -fplan9-extensions option directs the compiler to support 4768 some extensions for anonymous struct fields which are implemented 4769 by the Plan 9 compiler. A pointer to a struct may be automatically 4770 converted to a pointer to an anonymous field when calling a 4771 function, in order to make the types match. An anonymous struct 4772 field whose type is a typedef name may be referred to using the 4773 typedef name. 4774 4775 C++ 4776 4777 * Improved [7]experimental support for the upcoming C++0x ISO C++ 4778 standard, including support for constexpr (thanks to Gabriel Dos 4779 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 4780 noexcept, unrestricted unions, range-based for loops (thanks to 4781 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 4782 Rodrigo), implicitly deleted functions and implicit move 4783 constructors. 4784 * When an extern declaration within a function does not match a 4785 declaration in the enclosing context, G++ now properly declares the 4786 name within the namespace of the function rather than the namespace 4787 which was open just before the function definition ([8]c++/43145). 4788 * GCC now warns by default when casting integers to larger pointer 4789 types. These warnings can be disabled with the option 4790 -Wno-int-to-pointer-cast, which is now also available in C++. 4791 * G++ no longer optimizes using the assumption that a value of 4792 enumeration type will fall within the range specified by the 4793 standard, since that assumption is easily violated with a 4794 conversion from integer type ([9]c++/43680). The old behavior can 4795 be restored with -fstrict-enums. 4796 * The new -fnothrow-opt flag changes the semantics of a throw() 4797 exception specification to match the proposed semantics of the 4798 noexcept specification: just call terminate if an exception tries 4799 to propagate out of a function with such an exception 4800 specification. This dramatically reduces or eliminates the code 4801 size overhead from adding the exception specification. 4802 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 4803 a function that the compiler can tell doesn't throw if it would 4804 change the value of a noexcept expression. 4805 * The -Wshadow option now warns if a local variable or type 4806 declaration shadows another type in C++. Note that the compiler 4807 will not warn if a local variable shadows a struct/class/enum, but 4808 will warn if it shadows an explicit typedef. 4809 * When an identifier is not found in the current scope, G++ now 4810 offers suggestions about which identifier might have been intended. 4811 * G++ now issues clearer diagnostics for missing semicolons after 4812 class, struct, and union definitions. 4813 * G++ now issues clearer diagnostics for missing semicolons after 4814 class member declarations. 4815 * G++ now issues clearer diagnostics when a colon is used in a place 4816 where a double-colon was intended. 4817 * G++ no longer accepts mutable on reference members ([10]c++/33558). 4818 Use -fpermissive to allow the old, non-conforming behaviour. 4819 * A few mangling fixes have been made, to attribute const/volatile on 4820 function pointer types, decltype of a plain decl, and use of a 4821 function parameter in the declaration of another parameter. By 4822 default the compiler still uses the old mangling, but emits aliases 4823 with the new mangling on targets that support strong aliases. Users 4824 can switch over entirely to the new mangling with -fabi-version=5 4825 or -fabi-version=0. -Wabi will now warn about code that uses the 4826 old mangling. 4827 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 4828 type to be default initialized unless the type has a user-declared 4829 default constructor. In 4.6.2 G++ implements the proposed 4830 resolution of [11]DR 253, so default initialization is allowed if 4831 it initializes all subobjects. Code that fails to compile can be 4832 fixed by providing an initializer e.g. 4833 struct A { A(); }; 4834 struct B : A { int i; }; 4835 const B b = B(); 4836 Use -fpermissive to allow the old, non-conforming behaviour. 4837 4838 Runtime Library (libstdc++) 4839 4840 * [12]Improved experimental support for the upcoming ISO C++ 4841 standard, C++0x, including using constexpr and nullptr. 4842 * Performance improvements to the [13]Debug Mode, thanks to Franc,ois 4843 Dumont. 4844 * Atomic operations used for reference-counting are annotated so that 4845 they can be understood by race detectors such as Helgrind, see 4846 [14]Data Race Hunting. 4847 * Most libstdc++ standard headers have been changed to no longer 4848 include the cstddef header as an implementation detail. Code that 4849 relied on that header being included as side-effect of including 4850 other standard headers will need to include cstddef explicitly. 4851 4852 Fortran 4853 4854 * On systems supporting the libquadmath library, GNU Fortran now also 4855 supports a quad-precision, kind=16 floating-point data type 4856 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 4857 hardware, calculations might be one to two orders of magnitude 4858 slower than with the 4, 8 or 10 bytes floating-point data types. 4859 This change does not affect systems which support REAL(16) in 4860 hardware nor those which do not support libquadmath. 4861 * Much improved compile time for large array constructors. 4862 * In order to reduce execution time and memory consumption, use of 4863 temporary arrays in assignment expressions is avoided for many 4864 cases. The compiler now reverses loops in order to avoid generating 4865 a temporary array where possible. 4866 * Improved diagnostics, especially with -fwhole-file. 4867 * The -fwhole-file flag is now enabled by default. This improves code 4868 generation and diagnostics. It can be disabled using the deprecated 4869 -fno-whole-file flag. 4870 * Support the generation of Makefile dependencies via the [15]-M... 4871 flags of GCC; you may need to specify the -cpp option in addition. 4872 The dependencies take modules, Fortran's include, and CPP's 4873 #include into account. Note: Using -M for the module path is no 4874 longer supported, use -J instead. 4875 * The flag -Wconversion has been modified to only issue warnings 4876 where a conversion leads to information loss. This drastically 4877 reduces the number of warnings; -Wconversion is thus now enabled 4878 with -Wall. The flag -Wconversion-extra has been added and also 4879 warns about other conversions; -Wconversion-extra typically issues 4880 a huge number of warnings, most of which can be ignored. 4881 * A new command-line option -Wunused-dummy-argument warns about 4882 unused dummy arguments and is included in -Wall. Before, 4883 -Wunused-variable also warned about unused dummy arguments. 4884 * Fortran 2003 support has been extended: 4885 + Improved support for polymorphism between libraries and 4886 programs and for complicated inheritance patterns (cf. 4887 [16]object-oriented programming). 4888 + Experimental support of the ASSOCIATE construct. 4889 + In pointer assignments it is now possible to specify the lower 4890 bounds of the pointer and, for a rank-1 or a simply contiguous 4891 data-target, to remap the bounds. 4892 + Automatic (re)allocation: In intrinsic assignments to 4893 allocatable variables the left-hand side will be automatically 4894 allocated (if unallocated) or reallocated (if the shape or 4895 type parameter is different). To avoid the small performance 4896 penalty, you can use a(:) = ... instead of a = ... for arrays 4897 and character strings - or disable the feature using -std=f95 4898 or -fno-realloc-lhs. 4899 + Deferred type parameter: For scalar allocatable and pointer 4900 variables the character length can be deferred. 4901 + Namelist variables with allocatable and pointer attribute and 4902 nonconstant length type parameter are supported. 4903 * Fortran 2008 support has been extended: 4904 + Experimental [17]coarray support (for one image only, i.e. 4905 num_images() == 1); use the [18]-fcoarray=single flag to 4906 enable it. 4907 + The STOP and the new ERROR STOP statements now support all 4908 constant expressions. 4909 + Support for the CONTIGUOUS attribute. 4910 + Support for ALLOCATE with MOLD. 4911 + Support for the STORAGE_SIZE intrinsic inquiry function. 4912 + Support of the NORM2 and PARITY intrinsic functions. 4913 + The following bit intrinsics were added: POPCNT and POPPAR for 4914 counting the number of 1 bits and returning the parity; BGE, 4915 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 4916 for combined left and right shifts, MASKL and MASKR for simple 4917 left and right justified masks, MERGE_BITS for a bitwise merge 4918 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 4919 and the transformational bit intrinsics IALL, IANY and 4920 IPARITY. 4921 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 4922 + Support for the IMPURE attribute for procedures, which allows 4923 for ELEMENTAL procedures without the restrictions of PURE. 4924 + Null pointers (including NULL()) and not allocated variables 4925 can be used as actual argument to optional non-pointer, 4926 non-allocatable dummy arguments, denoting an absent argument. 4927 + Non-pointer variables with TARGET attribute can be used as 4928 actual argument to POINTER dummies with INTENT(IN) 4929 + Pointers including procedure pointers and those in a derived 4930 type (pointer components) can now be initialized by a target 4931 instead of only by NULL. 4932 + The EXIT statement (with construct-name) can now be used to 4933 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 4934 SELECT CASE and SELECT TYPE constructs. 4935 + Internal procedures can now be used as actual argument. 4936 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 4937 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 4938 have been added; these arrays contain the supported kind 4939 values for the respective types. 4940 + The module procedures C_SIZEOF of the intrinsic module 4941 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 4942 ISO_FORTRAN_ENV have been implemented. 4943 + Minor changes: obsolescence diagnostics for ENTRY was added 4944 for -std=f2008; a line may start with a semicolon; for 4945 internal and module procedures END can be used instead of END 4946 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 4947 a RADIX argument; intrinsic types are supported for 4948 TYPE(intrinsic-type-spec); multiple type-bound procedures can 4949 be declared in a single PROCEDURE statement; implied-shape 4950 arrays are supported for named constants (PARAMETER). The 4951 transformational, three argument versions of BESSEL_JN and 4952 BESSEL_YN were added - the elemental, two-argument version had 4953 been added in GCC 4.4; note that the transformational 4954 functions use a recurrence algorithm. 4955 4956 Go 4957 4958 Support for the [19]Go programming language has been added to GCC. It 4959 is not enabled by default when you build GCC; use the 4960 --enable-languages configure option to build it. The driver program for 4961 compiling Go code is gccgo. 4962 4963 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 4964 is in progress. It may or may not work on other platforms. 4965 4966 Objective-C and Objective-C++ 4967 4968 * The -fobjc-exceptions flag is now required to enable Objective-C 4969 exception and synchronization syntax (introduced by the keywords 4970 @try, @catch, @finally and @synchronized). 4971 * A number of Objective-C 2.0 features and extensions are now 4972 supported by GCC. These features are enabled by default; you can 4973 disable them by using the new -fobjc-std=objc1 command-line option. 4974 * The Objective-C 2.0 dot-syntax is now supported. It is an 4975 alternative syntax for using getters and setters; object.count is 4976 automatically converted into [object count] or [object setCount: 4977 ...] depending on context; for example if (object.count > 0) is 4978 automatically compiled into the equivalent of if ([object count] > 4979 0) while object.count = 0; is automatically compiled into the 4980 equivalent ot [object setCount: 0];. The dot-syntax can be used 4981 with instance and class objects and with any setters or getters, no 4982 matter if they are part of a declared property or not. 4983 * Objective-C 2.0 declared properties are now supported. They are 4984 declared using the new @property keyword, and are most commonly 4985 used in conjunction with the new Objective-C 2.0 dot-syntax. The 4986 nonatomic, readonly, readwrite, assign, retain, copy, setter and 4987 getter attributes are all supported. Marking declared properties 4988 with __attribute__ ((deprecated)) is supported too. 4989 * The Objective-C 2.0 @synthesize and @dynamic keywords are 4990 supported. @synthesize causes the compiler to automatically 4991 synthesize a declared property, while @dynamic is used to disable 4992 all warnings for a declared property for which no implementation is 4993 provided at compile time. Synthesizing declared properties requires 4994 runtime support in most useful cases; to be able to use it with the 4995 GNU runtime, appropriate helper functions have been added to the 4996 GNU Objective-C runtime ABI, and are implemented by the GNU 4997 Objective-C runtime library shipped with GCC. 4998 * The Objective-C 2.0 fast enumeration syntax is supported in 4999 Objective-C. This is currently not yet available in Objective-C++. 5000 Fast enumeration requires support in the runtime, and such support 5001 has been added to the GNU Objective-C runtime library (shipped with 5002 GCC). 5003 * The Objective-C 2.0 @optional keyword is supported. It allows you 5004 to mark methods or properties in a protocol as optional as opposed 5005 to required. 5006 * The Objective-C 2.0 @package keyword is supported. It has currently 5007 the same effect as the @public keyword. 5008 * Objective-C 2.0 method attributes are supported. Currently the 5009 supported attributes are deprecated, sentinel, noreturn and format. 5010 * Objective-C 2.0 method argument attributes are supported. The most 5011 widely used attribute is unused, to mark an argument as unused in 5012 the implementation. 5013 * Objective-C 2.0 class and protocol attributes are supported. 5014 Currently the only supported attribute is deprecated. 5015 * Objective-C 2.0 class extensions are supported. A class extension 5016 has the same syntax as a category declaration with no category 5017 name, and the methods and properties declared in it are added 5018 directly to the main class. It is mostly used as an alternative to 5019 a category to add methods to a class without advertising them in 5020 the public headers, with the advantage that for class extensions 5021 the compiler checks that all the privately declared methods are 5022 actually implemented. 5023 * As a result of these enhancements, GCC can now be used to build 5024 Objective-C and Objective-C++ software that uses Foundation and 5025 other important system frameworks with the NeXT runtime on Darwin 9 5026 and Darwin 10 (OSX 10.5 and 10.6). 5027 * Many bugs in the compiler have been fixed in this release; in 5028 particular, LTO can now be used when compiling Objective-C and 5029 Objective-C++ and the parser is much more robust in dealing with 5030 invalid code. 5031 5032 Runtime Library (libobjc) 5033 5034 * The GNU Objective-C runtime library now defines the macro 5035 __GNU_LIBOBJC__ (with a value that is increased at every release 5036 where there is any change to the API) in objc/objc.h, making it 5037 easy to determine if the GNU Objective-C runtime library is being 5038 used, and if so, which version. Previous versions of the GNU 5039 Objective-C runtime library (and other Objective-C runtime 5040 libraries such as the Apple one) do not define this macro. 5041 * A new Objective-C 2.0 API, almost identical to the one implemented 5042 by the Apple Objective-C runtime, has been implemented in the GNU 5043 Objective-C runtime library. The new API hides the internals of 5044 most runtime structures but provides a more extensive set of 5045 functions to operate on them. It is much easier, for example, to 5046 create or modify classes at runtime. The new API also makes it 5047 easier to port software from Apple to GNU as almost no changes 5048 should be required. The old API is still supported for backwards 5049 compatibility; including the old objc/objc-api.h header file 5050 automatically selects the old API, while including the new 5051 objc/runtime.h header file automatically selects the new API. 5052 Support for the old API is being phased out and upgrading the 5053 software to use the new API is strongly recommended. To check for 5054 the availability of the new API, the __GNU_LIBOBJC__ macro can be 5055 used as older versions of the GNU Objective-C runtime library, 5056 which do not support the new API, do not define such a macro. 5057 * Runtime support for @synchronized has been added. 5058 * Runtime support for Objective-C 2.0 synthesized property accessors 5059 has been added. 5060 * Runtime support for Objective-C 2.0 fast enumeration has been 5061 added. 5062 5063 New Targets and Target Specific Improvements 5064 5065 ARM 5066 5067 * GCC now supports the Cortex-M4 processor implementing the v7-em 5068 version of the architecture using the option -mcpu=cortex-m4. 5069 * Scheduling descriptions for the Cortex-M4, the Neon and the 5070 floating point units of the Cortex-A9 and a pipeline description 5071 for the Cortex-A5 have been added. 5072 * Synchronization primitives such as __sync_fetch_and_add and friends 5073 are now inlined for supported architectures rather than calling 5074 into a kernel helper function. 5075 * SSA loop prefetching is enabled by default for the Cortex-A9 at 5076 -O3. 5077 * Several improvements were committed to improve code generation for 5078 the ARM architecture including a rewritten implementation for load 5079 and store multiples. 5080 * Several enhancements were committed to improve SIMD code generation 5081 for NEON by adding support for widening instructions, misaligned 5082 loads and stores, vector conditionals and support for 64 bit 5083 arithmetic. 5084 * Support was added for the Faraday cores fa526, fa606te, fa626te, 5085 fmp626te, fmp626 and fa726te and can be used with the respective 5086 names as parameters to the -mcpu= option. 5087 * Basic support was added for Cortex-A15 and is available through 5088 -mcpu=cortex-a15. 5089 * GCC for AAPCS configurations now more closely adheres to the AAPCS 5090 specification by enabling -fstrict-volatile-bitfields by default. 5091 5092 IA-32/x86-64 5093 5094 * The new -fsplit-stack option permits programs to use a 5095 discontiguous stack. This is useful for threaded programs, in that 5096 it is no longer necessary to specify the maximum stack size when 5097 creating a thread. This feature is currently only implemented for 5098 32-bit and 64-bit x86 GNU/Linux targets. 5099 * Support for emitting profiler counter calls before function 5100 prologues. This is enabled via a new command-line option -mfentry. 5101 * Optimization for the Intel Core 2 processors is now available 5102 through the -march=core2 and -mtune=core2 options. 5103 * Support for Intel Core i3/i5/i7 processors is now available through 5104 the -march=corei7 and -mtune=corei7 options. 5105 * Support for Intel Core i3/i5/i7 processors with AVX is now 5106 available through the -march=corei7-avx and -mtune=corei7-avx 5107 options. 5108 * Support for AMD Bobcat (family 14) processors is now available 5109 through the -march=btver1 and -mtune=btver1 options. 5110 * Support for AMD Bulldozer (family 15) processors is now available 5111 through the -march=bdver1 and -mtune=bdver1 options. 5112 * The default setting (when not optimizing for size) for 32-bit 5113 GNU/Linux and Darwin x86 targets has been changed to 5114 -fomit-frame-pointer. The default can be reverted to 5115 -fno-omit-frame-pointer by configuring GCC with the 5116 --enable-frame-pointer configure option. 5117 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 5118 __float128 on 32-bit and 64-bit x86 targets. 5119 * AVX floating-point arithmetic can now be enabled by default at 5120 configure time with the new --with-fpmath=avx option. 5121 * The SSA loop prefetching pass is enabled when using -O3 when 5122 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 5123 than K6). 5124 * Support for TBM (Trailing Bit Manipulation) built-in functions and 5125 code generation is available via -mtbm. 5126 * Support for AMD's BMI (Bit Manipulation) built-in functions and 5127 code generation is available via -mbmi. 5128 5129 MicroBlaze 5130 5131 * Support has been added for the Xilinx MicroBlaze softcore processor 5132 (microblaze-elf) embedded target. This configurable processor is 5133 supported on several Xilinx Spartan and Virtex FPGAs. 5134 5135 MIPS 5136 5137 * GCC now supports the Loongson 3A processor. Its canonical -march= 5138 and -mtune= name is loongson3a. 5139 5140 MN10300 / AM33 5141 5142 * The inline assembly register constraint "A" has been renamed "c". 5143 This constraint is used to select a floating-point register that 5144 can be used as the destination of a multiply-accumulate 5145 instruction. 5146 * New inline assembly register constraints "A" and "D" have been 5147 added. These constraint letters resolve to all general registers 5148 when compiling for AM33, and resolve to address registers only or 5149 data registers only when compiling for MN10300. 5150 * The MDR register is represented in the compiler. One can access the 5151 register via the "z" constraint in inline assembly. It can be 5152 marked as clobbered or used as a local register variable via the 5153 "mdr" name. The compiler uses the RETF instruction if the function 5154 does not modify the MDR register, so it is important that inline 5155 assembly properly annotate any usage of the register. 5156 5157 PowerPC/PowerPC64 5158 5159 * GCC now supports the Applied Micro Titan processor with 5160 -mcpu=titan. 5161 * The -mrecip option has been added, which indicates whether the 5162 reciprocal and reciprocal square root instructions should be used. 5163 * The -mveclibabi=mass option can be used to enable the compiler to 5164 autovectorize mathematical functions using the Mathematical 5165 Acceleration Subsystem library. 5166 * The -msingle-pic-base option has been added, which instructs the 5167 compiler to avoid loading the PIC base register in function 5168 prologues. The PIC base register must be initialized by the runtime 5169 system. 5170 * The -mblock-move-inline-limit option has been added, which enables 5171 the user to control the maximum size of inlined memcpy calls and 5172 similar. 5173 * PowerPC64 GNU/Linux support for applications requiring a large TOC 5174 section has been improved. A new command-line option, 5175 -mcmodel=MODEL, controls this feature; valid values for MODEL are 5176 small, medium, or large. 5177 * The Altivec builtin functions vec_ld and vec_st have been modified 5178 to generate the Altivec memory instructions LVX and STVX, even if 5179 the -mvsx option is used. In the initial GCC 4.5 release, these 5180 builtin functions were changed to generate VSX memory reference 5181 instructions instead of Altivec memory instructions, but there are 5182 differences between the two instructions. If the VSX instruction 5183 set is available, you can now use the new builtin functions 5184 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 5185 instructions. 5186 * The GCC compiler on AIX now defaults to a process layout with a 5187 larger data space allowing larger programs to be compiled. 5188 * The GCC long double type on AIX 6.1 and above has reverted to 64 5189 bit double precision, matching the AIX XL compiler default, because 5190 of missing C99 symbols required by the GCC runtime. 5191 * The default processor scheduling model and tuning for PowerPC64 5192 GNU/Linux and for AIX 6.1 and above now is POWER7. 5193 * Starting with GCC 4.6.1, vectors of type vector long long or vector 5194 long are passed and returned in the same method as other vectors 5195 with the VSX instruction set. Previously the GCC compiler did not 5196 adhere to the ABI for 128-bit vectors with 64-bit integer base 5197 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 5198 5199 S/390, zSeries and System z9/z10, IBM zEnterprise z196 5200 5201 * Support for the zEnterprise z196 processor has been added. When 5202 using the -march=z196 option, the compiler will generate code 5203 making use of the following instruction facilities: 5204 + Conditional load/store 5205 + Distinct-operands 5206 + Floating-point-extension 5207 + Interlocked-access 5208 + Population-count 5209 The -mtune=z196 option avoids the compare and branch instructions 5210 as well as the load address instruction with an index register as 5211 much as possible and performs instruction scheduling appropriate 5212 for the new out-of-order pipeline architecture. 5213 * When using the -m31 -mzarch options the generated code still 5214 conforms to the 32-bit ABI but uses the general purpose registers 5215 as 64-bit registers internally. This requires a Linux kernel saving 5216 the whole 64-bit registers when doing a context switch. Kernels 5217 providing that feature indicate that by the 'highgprs' string in 5218 /proc/cpuinfo. 5219 * The SSA loop prefetching pass is enabled when using -O3. 5220 5221 SPARC 5222 5223 * GCC now supports the LEON series of SPARC V8 processors. The code 5224 generated by the compiler can either be tuned to it by means of the 5225 --with-tune=leon configure option and -mtune=leon compilation 5226 option, or the compiler can be built for the sparc-leon-{elf,linux} 5227 and sparc-leon3-{elf,linux} targets directly. 5228 * GCC has stopped sign/zero-extending parameter registers in the 5229 callee for functions taking parameters with sub-word size in 32-bit 5230 mode, since this is redundant with the specification of the ABI. 5231 GCC has never done so in 64-bit mode since this is also redundant. 5232 * The command-line option -mfix-at697f has been added to enable the 5233 documented workaround for the single erratum of the Atmel AT697F 5234 processor. 5235 5236 Operating Systems 5237 5238 Android 5239 5240 * GCC now supports the Bionic C library and provides a convenient way 5241 of building native libraries and applications for the Android 5242 platform. Refer to the documentation of the -mandroid and -mbionic 5243 options for details on building native code. At the moment, Android 5244 support is enabled only for ARM. 5245 5246 Darwin/Mac OS X 5247 5248 * General 5249 + Initial support for CFString types has been added. 5250 This allows GCC to build projects including the system Core 5251 Foundation frameworks. The GCC Objective-C family supports 5252 CFString "toll-free bridged" as per the Mac OS X system tools. 5253 CFString is also recognized in the context of format 5254 attributes and arguments (see the documentation for format 5255 attributes for limitations). At present, 8-bit character types 5256 are supported. 5257 + Object file size reduction. 5258 The Darwin zeroed memory allocators have been re-written to 5259 make more use of .zerofill sections. For non-debug code, this 5260 can reduce object file size significantly. 5261 + Objective-C family 64-bit support (NeXT ABI 2). 5262 Initial support has been added to support 64-bit Objective-C 5263 code using the Darwin/OS X native (NeXT) runtime. ABI version 5264 2 will be selected automatically when 64-bit code is built. 5265 + Objective-C family 32-bit ABI 1. 5266 For 32-bit code ABI 1 is also now also allowed. At present it 5267 must be selected manually using -fobjc-abi-version=1 where 5268 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 5269 * x86 Architecture 5270 + The -mdynamic-no-pic option has been enabled. 5271 Code supporting -mdynamic-no-pic optimization has been added 5272 and is applicable to -m32 builds. The compiler bootstrap uses 5273 the option where appropriate. 5274 + The default value for -mtune= has been changed. 5275 Since Darwin systems are primarily Xeon, Core-2 or similar the 5276 default tuning has been changed to -mtune=core2. 5277 + Enable 128-bit long double (__float128) support on Darwin. 5278 * PPC Architecture 5279 + Darwin64 ABI. 5280 Several significant bugs have been fixed, such that GCC now 5281 produces code compatible with the Darwin64 PowerPC ABI. 5282 + libffi and boehm-gc. 5283 The Darwin ports of the libffi and boehm-gc libraries have 5284 been upgraded to include a Darwin64 implementation. This means 5285 that powerpc*-*-darwin9 platforms may now, for example, build 5286 Java applications with -m64 enabled. 5287 + Plug-in support has been enabled. 5288 + The -fsection-anchors option is now available although, 5289 presently, not heavily tested. 5290 5291 Solaris 2 5292 5293 New Features 5294 5295 * Support symbol versioning with the Sun linker. 5296 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 5297 * Support thread-local storage (TLS) with the Sun assembler on 5298 Solaris 2/x86. 5299 * Support TLS on Solaris 8/9 if prerequisites are met. 5300 * Support COMDAT group with the GNU assembler and recent Sun linker. 5301 * Support the Sun assembler visibility syntax. 5302 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 5303 -march=pentiumpro (Solaris 8/9). 5304 * Don't use SSE on Solaris 8/9 x86 by default. 5305 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 5306 5307 ABI Change 5308 5309 * Change the ABI for returning 8-byte vectors like __m64 in MMX 5310 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 5311 compilers. This is an incompatible change. If you use such types, 5312 you must either recompile all your code with the new compiler or 5313 use the new -mvect8-ret-in-mem option to remain compatible with 5314 previous versions of GCC and Sun Studio. 5315 5316 Windows x86/x86_64 5317 5318 * Initial support for decimal floating point. 5319 * Support for the __thiscall calling-convention. 5320 * Support for hot-patchable function prologues via the 5321 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 5322 * Improvements of stack-probing and stack-allocation mechanisms. 5323 * Support of push/pop-macro pragma as preprocessor command. 5324 With #pragma push_macro("macro-name") the current definition of 5325 macro-name is saved and can be restored with #pragma 5326 pop_macro("macro-name") to its saved definition. 5327 * Enable 128-bit long double (__float128) support on MinGW and 5328 Cygwin. 5329 5330 Other significant improvements 5331 5332 Installation changes 5333 5334 * An install-strip make target is provided that installs stripped 5335 executables, and may install libraries with unneeded or debugging 5336 sections stripped. 5337 * On Power7 systems, there is a potential problem if you build the 5338 GCC compiler with a host compiler using options that enable the VSX 5339 instruction set generation. If the host compiler has been patched 5340 so that the vec_ld and vec_st builtin functions generate Altivec 5341 memory instructions instead of VSX memory instructions, then you 5342 should be able to build the compiler with VSX instruction 5343 generation. 5344 5345 Changes for GCC Developers 5346 5347 Note: these changes concern developers that develop GCC itself or 5348 software that integrates with GCC, such as plugins, and not the general 5349 GCC users. 5350 * The gengtype utility, which previously was internal to the GCC 5351 build process, has been enchanced to provide GC root information 5352 for plugins as necessary. 5353 * The old GC allocation interface of ggc_alloc and friends was 5354 replaced with a type-safe alternative. 5355 5356 GCC 4.6.1 5357 5358 This is the [20]list of problem reports (PRs) from GCC's bug tracking 5359 system that are known to be fixed in the 4.6.1 release. This list might 5360 not be complete (that is, it is possible that some PRs that have been 5361 fixed are not listed here). 5362 5363 GCC 4.6.2 5364 5365 This is the [21]list of problem reports (PRs) from GCC's bug tracking 5366 system that are known to be fixed in the 4.6.2 release. This list might 5367 not be complete (that is, it is possible that some PRs that have been 5368 fixed are not listed here). 5369 5370 GCC 4.6.3 5371 5372 This is the [22]list of problem reports (PRs) from GCC's bug tracking 5373 system that are known to be fixed in the 4.6.3 release. This list might 5374 not be complete (that is, it is possible that some PRs that have been 5375 fixed are not listed here). 5376 5377 GCC 4.6.4 5378 5379 This is the [23]list of problem reports (PRs) from GCC's bug tracking 5380 system that are known to be fixed in the 4.6.4 release. This list might 5381 not be complete (that is, it is possible that some PRs that have been 5382 fixed are not listed here). 5383 5384 5385 For questions related to the use of GCC, please consult these web 5386 pages and the [24]GCC manuals. If that fails, the 5387 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5388 web pages and the development of GCC are welcome on our developer 5389 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 5390 archives. 5391 5392 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 5393 distribution of this entire article is permitted in any medium, 5394 provided this notice is preserved. 5395 5396 These pages are [29]maintained by the GCC team. Last modified 5397 2016-11-27[30]. 5398 5399 References 5400 5401 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 5402 2. http://gcc.gnu.org/gcc-4.5/changes.html 5403 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 5404 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 5405 5. http://gcc.gnu.org/projects/lto/whopr.pdf 5406 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 5407 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 5408 8. https://gcc.gnu.org/PR43145 5409 9. https://gcc.gnu.org/PR43680 5410 10. https://gcc.gnu.org/PR33558 5411 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 5412 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 5413 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 5414 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 5415 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 5416 16. https://gcc.gnu.org/wiki/OOP 5417 17. https://gcc.gnu.org/wiki/Coarray 5418 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 5419 19. https://golang.org/ 5420 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 5421 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 5422 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 5423 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 5424 24. https://gcc.gnu.org/onlinedocs/ 5425 25. mailto:gcc-help (a] gcc.gnu.org 5426 26. mailto:gcc (a] gcc.gnu.org 5427 27. https://gcc.gnu.org/lists.html 5428 28. http://www.fsf.org/ 5429 29. https://gcc.gnu.org/about.html 5430 30. http://validator.w3.org/check/referer 5431 ====================================================================== 5432 http://gcc.gnu.org/gcc-4.5/index.html 5433 5434 GCC 4.5 Release Series 5435 5436 Jul 2, 2012 5437 5438 The [1]GNU project and the GCC developers are pleased to announce the 5439 release of GCC 4.5.4. 5440 5441 This release is a bug-fix release, containing fixes for regressions in 5442 GCC 4.5.3 relative to previous releases of GCC. 5443 5444 This release series is no longer maintained. 5445 5446 Release History 5447 5448 GCC 4.5.4 5449 Jul 2, 2012 ([2]changes) 5450 5451 GCC 4.5.3 5452 Apr 28, 2011 ([3]changes) 5453 5454 GCC 4.5.2 5455 Dec 16, 2010 ([4]changes) 5456 5457 GCC 4.5.1 5458 Jul 31, 2010 ([5]changes) 5459 5460 GCC 4.5.0 5461 April 14, 2010 ([6]changes) 5462 5463 References and Acknowledgements 5464 5465 GCC used to stand for the GNU C Compiler, but since the compiler 5466 supports several other languages aside from C, it now stands for the 5467 GNU Compiler Collection. 5468 5469 A list of [7]successful builds is updated as new information becomes 5470 available. 5471 5472 The GCC developers would like to thank the numerous people that have 5473 contributed new features, improvements, bug fixes, and other changes as 5474 well as test results to GCC. This [8]amazing group of volunteers is 5475 what makes GCC successful. 5476 5477 For additional information about GCC please refer to the [9]GCC project 5478 web site or contact the [10]GCC development mailing list. 5479 5480 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 5481 5482 5483 For questions related to the use of GCC, please consult these web 5484 pages and the [13]GCC manuals. If that fails, the 5485 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5486 web pages and the development of GCC are welcome on our developer 5487 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 5488 archives. 5489 5490 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 5491 distribution of this entire article is permitted in any medium, 5492 provided this notice is preserved. 5493 5494 These pages are [18]maintained by the GCC team. Last modified 5495 2016-09-30[19]. 5496 5497 References 5498 5499 1. http://www.gnu.org/ 5500 2. http://gcc.gnu.org/gcc-4.5/changes.html 5501 3. http://gcc.gnu.org/gcc-4.5/changes.html 5502 4. http://gcc.gnu.org/gcc-4.5/changes.html 5503 5. http://gcc.gnu.org/gcc-4.5/changes.html 5504 6. http://gcc.gnu.org/gcc-4.5/changes.html 5505 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 5506 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5507 9. http://gcc.gnu.org/index.html 5508 10. mailto:gcc (a] gcc.gnu.org 5509 11. http://gcc.gnu.org/mirrors.html 5510 12. http://gcc.gnu.org/svn.html 5511 13. https://gcc.gnu.org/onlinedocs/ 5512 14. mailto:gcc-help (a] gcc.gnu.org 5513 15. mailto:gcc (a] gcc.gnu.org 5514 16. https://gcc.gnu.org/lists.html 5515 17. http://www.fsf.org/ 5516 18. https://gcc.gnu.org/about.html 5517 19. http://validator.w3.org/check/referer 5518 ====================================================================== 5519 http://gcc.gnu.org/gcc-4.5/changes.html 5520 5521 GCC 4.5 Release Series 5522 Changes, New Features, and Fixes 5523 5524 Caveats 5525 5526 * GCC now requires the [1]MPC library in order to build. See the 5527 [2]prerequisites page for version requirements. 5528 * Support for a number of older systems and recently unmaintained or 5529 untested target ports of GCC has been declared obsolete in GCC 4.5. 5530 Unless there is activity to revive them, the next release of GCC 5531 will have their sources permanently removed. 5532 The following ports for individual systems on particular 5533 architectures have been obsoleted: 5534 + IRIX releases before 6.5 (mips-sgi-irix5*, 5535 mips-sgi-irix6.[0-4]) 5536 + Solaris 7 (*-*-solaris2.7) 5537 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 5538 alpha-dec-osf5.0*) 5539 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 5540 can be found in the [3]announcement. 5541 Support for the classic POWER architecture implemented in the 5542 original RIOS and RIOS2 processors of the old IBM RS/6000 product 5543 line has been obsoleted in the rs6000 port. This does not affect 5544 the new generation Power and PowerPC architectures. 5545 * Support has been removed for all the [4]configurations obsoleted in 5546 GCC 4.4. 5547 * Support has been removed for the protoize and unprotoize utilities, 5548 obsoleted in GCC 4.4. 5549 * Support has been removed for tuning for Itanium1 (Merced) variants. 5550 Note that code tuned for Itanium2 should also run correctly on 5551 Itanium1. 5552 * GCC now generates unwind info also for epilogues. DWARF debuginfo 5553 generated by GCC now uses more features of DWARF3 than before, and 5554 also some DWARF4 features. GDB older than 7.0 is not able to handle 5555 either of these, so to debug GCC 4.5 generated binaries or 5556 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 5557 features with the -gdwarf-3 -gstrict-dwarf options, or use 5558 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 5559 epilogue unwind info is emitted unconditionally whenever unwind 5560 info is emitted. 5561 * On x86 targets, code containing floating-point calculations may run 5562 significantly more slowly when compiled with GCC 4.5 in strict C99 5563 conformance mode than they did with earlier GCC versions. This is 5564 due to stricter standard conformance of the compiler and can be 5565 avoided by using the option -fexcess-precision=fast; also see 5566 [5]below. 5567 * The function attribute noinline no longer prevents GCC from cloning 5568 the function. A new attribute noclone has been introduced for this 5569 purpose. Cloning a function means that it is duplicated and the new 5570 copy is specialized for certain contexts (for example when a 5571 parameter is a known constant). 5572 5573 General Optimizer Improvements 5574 5575 * The -save-temps now takes an optional argument. The -save-temps and 5576 -save-temps=cwd switches write the temporary files in the current 5577 working directory based on the original source file. The 5578 -save-temps=obj switch will write files into the directory 5579 specified with the -o option, and the intermediate filenames are 5580 based on the output file. This will allow the user to get the 5581 compiler intermediate files when doing parallel builds without two 5582 builds of the same filename located in different directories from 5583 interfering with each other. 5584 * Debugging dumps are now created in the same directory as the object 5585 file rather than in the current working directory. This allows the 5586 user to get debugging dumps when doing parallel builds without two 5587 builds of the same filename interfering with each other. 5588 * GCC has been integrated with the [6]MPC library. This allows GCC to 5589 evaluate complex arithmetic at compile time [7]more accurately. It 5590 also allows GCC to evaluate calls to complex built-in math 5591 functions having constant arguments and replace them at compile 5592 time with their mathematically equivalent results. In doing so, GCC 5593 can generate correct results regardless of the math library 5594 implementation or floating point precision of the host platform. 5595 This also allows GCC to generate identical results regardless of 5596 whether one compiles in native or cross-compile configurations to a 5597 particular target. The following built-in functions take advantage 5598 of this new capability: cacos, cacosh, casin, casinh, catan, 5599 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 5600 and ctanh. The float and long double variants of these functions 5601 (e.g. csinf and csinl) are also handled. 5602 * A new link-time optimizer has been added ([8]-flto). When this 5603 option is used, GCC generates a bytecode representation of each 5604 input file and writes it to specially-named sections in each object 5605 file. When the object files are linked together, all the function 5606 bodies are read from these named sections and instantiated as if 5607 they had been part of the same translation unit. This enables 5608 interprocedural optimizations to work across different files (and 5609 even different languages), potentially improving the performance of 5610 the generated code. To use the link-timer optimizer, -flto needs to 5611 be specified at compile time and during the final link. If the 5612 program does not require any symbols to be exported, it is possible 5613 to combine -flto and the experimental [9]-fwhopr with 5614 [10]-fwhole-program to allow the interprocedural optimizers to use 5615 more aggressive assumptions. 5616 * The automatic parallelization pass was enhanced to support 5617 parallelization of outer loops. 5618 * Automatic parallelization can be enabled as part of Graphite. In 5619 addition to -ftree-parallelize-loops=, specify 5620 -floop-parallelize-all to enable the Graphite-based optimization. 5621 * The infrastructure for optimizing based on [11]restrict qualified 5622 pointers has been rewritten and should result in code generation 5623 improvements. Optimizations based on restrict qualified pointers 5624 are now also available when using -fno-strict-aliasing. 5625 * There is a new optimization pass that attempts to change prototype 5626 of functions to avoid unused parameters, pass only relevant parts 5627 of structures and turn arguments passed by reference to arguments 5628 passed by value when possible. It is enabled by -O2 and above as 5629 well as -Os and can be manually invoked using the new command-line 5630 switch -fipa-sra. 5631 * GCC now optimize exception handling code. In particular cleanup 5632 regions that are proved to not have any effect are optimized out. 5633 5634 New Languages and Language specific improvements 5635 5636 All languages 5637 5638 * The -fshow-column option is now on by default. This means error 5639 messages now have a column associated with them. 5640 5641 Ada 5642 5643 * Compilation of programs heavily using discriminated record types 5644 with variant parts has been sped up and generates more compact 5645 code. 5646 * Stack checking now works reasonably well on most plaforms. In some 5647 specific cases, stack overflows may still fail to be detected, but 5648 a compile-time warning will be issued for these cases. 5649 5650 C family 5651 5652 * If a header named in a #include directive is not found, the 5653 compiler exits immediately. This avoids a cascade of errors arising 5654 from declarations expected to be found in that header being 5655 missing. 5656 * A new built-in function __builtin_unreachable() has been added that 5657 tells the compiler that control will never reach that point. It may 5658 be used after asm statements that terminate by transferring control 5659 elsewhere, and in other places that are known to be unreachable. 5660 * The -Wlogical-op option now warns for logical expressions such as 5661 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 5662 mistakes. This option is disabled by default. 5663 * An asm goto feature has been added to allow asm statements that 5664 jump to C labels. 5665 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 5666 * The deprecated attribute now takes an optional string argument, for 5667 example, __attribute__((deprecated("text string"))), that will be 5668 printed together with the deprecation warning. 5669 5670 C 5671 5672 * The -Wenum-compare option, which warns when comparing values of 5673 different enum types, now works for C. It formerly only worked for 5674 C++. This warning is enabled by -Wall. It may be avoided by using a 5675 type cast. 5676 * The -Wcast-qual option now warns about casts which are unsafe in 5677 that they permit const-correctness to be violated without further 5678 warnings. Specifically, it warns about cases where a qualifier is 5679 added when all the lower types are not const. For example, it warns 5680 about a cast from char ** to const char **. 5681 * The -Wc++-compat option is significantly improved. It issues new 5682 warnings for: 5683 + Using C++ reserved operator names as identifiers. 5684 + Conversions to enum types without explicit casts. 5685 + Using va_arg with an enum type. 5686 + Using different enum types in the two branches of ?:. 5687 + Using ++ or -- on a variable of enum type. 5688 + Using the same name as both a struct, union or enum tag and a 5689 typedef, unless the typedef refers to the tagged type itself. 5690 + Using a struct, union, or enum which is defined within another 5691 struct or union. 5692 + A struct field defined using a typedef if there is a field in 5693 the struct, or an enclosing struct, whose name is the typedef 5694 name. 5695 + Duplicate definitions at file scope. 5696 + Uninitialized const variables. 5697 + A global variable with an anonymous struct, union, or enum 5698 type. 5699 + Using a string constant to initialize a char array whose size 5700 is the length of the string. 5701 * The new -Wjump-misses-init option warns about cases where a goto or 5702 switch skips the initialization of a variable. This sort of branch 5703 is an error in C++ but not in C. This warning is enabled by 5704 -Wc++-compat. 5705 * GCC now ensures that a C99-conforming <stdint.h> is present on most 5706 targets, and uses information about the types in this header to 5707 implement the Fortran bindings to those types. GCC does not ensure 5708 the presence of such a header, and does not implement the Fortran 5709 bindings, on the following targets: NetBSD, VxWorks, VMS, 5710 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 5711 * GCC now implements C90- and C99-conforming rules for constant 5712 expressions. This may cause warnings or errors for some code using 5713 expressions that can be folded to a constant but are not constant 5714 expressions as defined by ISO C. 5715 * All known target-independent C90 and C90 Amendment 1 conformance 5716 bugs, and all known target-independent C99 conformance bugs not 5717 related to floating point or extended identifiers, have been fixed. 5718 * The C decimal floating point support now includes support for the 5719 FLOAT_CONST_DECIMAL64 pragma. 5720 * The named address space feature from ISO/IEC TR 18037 is now 5721 supported. This is currently only implemented for the SPU 5722 processor. 5723 5724 C++ 5725 5726 * Improved [12]experimental support for the upcoming C++0x ISO C++ 5727 standard, including support for raw strings, lambda expressions and 5728 explicit type conversion operators. 5729 * When printing the name of a class template specialization, G++ will 5730 now omit any template arguments which come from default template 5731 arguments. This behavior (and the pretty-printing of function 5732 template specializations as template signature and arguments) can 5733 be disabled with the -fno-pretty-templates option. 5734 * Access control is now applied to typedef names used in a template, 5735 which may cause G++ to reject some ill-formed code that was 5736 accepted by earlier releases. The -fno-access-control option can be 5737 used as a temporary workaround until the code is corrected. 5738 * Compilation time for code that uses templates should now scale 5739 linearly with the number of instantiations rather than 5740 quadratically, as template instantiations are now looked up using 5741 hash tables. 5742 * Declarations of functions that look like builtin declarations of 5743 library functions are only considered to be redeclarations if they 5744 are declared with extern "C". This may cause problems with code 5745 that omits extern "C" on hand-written declarations of C library 5746 functions such as abort or memcpy. Such code is ill-formed, but was 5747 accepted by earlier releases. 5748 * Diagnostics that used to complain about passing non-POD types to 5749 ... or jumping past the declaration of a non-POD variable now check 5750 for triviality rather than PODness, as per C++0x. 5751 * In C++0x mode local and anonymous classes are now allowed as 5752 template arguments, and in declarations of variables and functions 5753 with linkage, so long as any such declaration that is used is also 5754 defined ([13]DR 757). 5755 * Labels may now have attributes, as has been permitted for a while 5756 in C. This is only permitted when the label definition and the 5757 attribute specifier is followed by a semicolon--i.e., the label 5758 applies to an empty statement. The only useful attribute for a 5759 label is unused. 5760 * G++ now implements [14]DR 176. Previously G++ did not support using 5761 the injected-class-name of a template base class as a type name, 5762 and lookup of the name found the declaration of the template in the 5763 enclosing scope. Now lookup of the name finds the 5764 injected-class-name, which can be used either as a type or as a 5765 template, depending on whether or not the name is followed by a 5766 template argument list. As a result of this change, some code that 5767 was previously accepted may be ill-formed because 5768 1. The injected-class-name is not accessible because it's from a 5769 private base, or 5770 2. The injected-class-name cannot be used as an argument for a 5771 template template parameter. 5772 In either of these cases, the code can be fixed by adding a 5773 nested-name-specifier to explicitly name the template. The first 5774 can be worked around with -fno-access-control; the second is only 5775 rejected with -pedantic. 5776 * A new standard mangling for SIMD vector types has been added, to 5777 avoid name clashes on systems with vectors of varying length. By 5778 default the compiler still uses the old mangling, but emits aliases 5779 with the new mangling on targets that support strong aliases. Users 5780 can switch over entirely to the new mangling with -fabi-version=4 5781 or -fabi-version=0. -Wabi will now warn about code that uses the 5782 old mangling. 5783 * The command-line option -ftemplate-depth-N is now written as 5784 -ftemplate-depth=N and the old form is deprecated. 5785 * Conversions between NULL and non-pointer types are now warned by 5786 default. The new option -Wno-conversion-null disables these 5787 warnings. Previously these warnings were only available when using 5788 -Wconversion explicitly. 5789 5790 Runtime Library (libstdc++) 5791 5792 * Improved experimental support for the upcoming ISO C++ standard, 5793 C++0x, including: 5794 + Support for <future>, <functional>, and <random>. 5795 + Existing facilities now exploit explicit operators and the 5796 newly implemented core C++0x features. 5797 + The header <cstdatomic> has been renamed to <atomic>. 5798 * An experimental [15]profile mode has been added. This is an 5799 implementation of many C++ standard library constructs with an 5800 additional analysis layer that gives performance improvement advice 5801 based on recognition of suboptimal usage patterns. For example, 5802 #include <vector> 5803 int main() 5804 { 5805 std::vector<int> v; 5806 for (int k = 0; k < 1024; ++k) 5807 v.insert(v.begin(), k); 5808 } 5809 5810 When instrumented via the profile mode, can return suggestions 5811 about the initial size and choice of the container used as follows: 5812 vector-to-list: improvement = 5: call stack = 0x804842c ... 5813 : advice = change std::vector to std::list 5814 vector-size: improvement = 3: call stack = 0x804842c ... 5815 : advice = change initial container size from 0 to 1024 5816 5817 These constructs can be substituted for the normal libstdc++ 5818 constructs on a piecemeal basis, or all existing components can be 5819 transformed via the -D_GLIBCXX_PROFILE macro. 5820 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR 5821 24733) has been added. This support is in header file 5822 <decimal/decimal>, uses namespace std::decimal, and includes 5823 classes decimal32, decimal64, and decimal128. 5824 * Sources have been audited for application of function attributes 5825 nothrow, const, pure, and noreturn. 5826 * Python pretty-printers have been added for many standard library 5827 components that simplify the internal representation and present a 5828 more intuitive view of components when used with 5829 appropriately-advanced versions of GDB. For more information, 5830 please consult the more [17]detailed description. 5831 * The default behavior for comparing typeinfo names has changed, so 5832 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 5833 * The new -static-libstdc++ option directs g++ to link the C++ 5834 library statically, even if the default would normally be to link 5835 it dynamically. 5836 5837 Fortran 5838 5839 * The COMMON default padding has been changed - instead of adding the 5840 padding before a variable it is now added afterwards, which 5841 increases the compatibility with other vendors and helps to obtain 5842 the correct output in some cases. Cf. also the -falign-commons 5843 option ([18]added in 4.4). 5844 * The -finit-real= option now also supports the value snan for 5845 signaling not-a-number; to be effective, one additionally needs to 5846 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 5847 optimizations can turn a signaling NaN into a quiet one. 5848 * The new option -fcheck= has been added with the options bounds, 5849 array-temps, do, pointer, and recursive. The bounds and array-temps 5850 options are equivalent to -fbounds-check and 5851 -fcheck-array-temporaries. The do option checks for invalid 5852 modification of loop iteration variables, and the recursive option 5853 tests for recursive calls to subroutines/functions which are not 5854 marked as recursive. With pointer pointer association checks in 5855 calls are performed; however, neither undefined pointers nor 5856 pointers in expressions are handled. Using -fcheck=all enables all 5857 these run-time checks. 5858 * The run-time checking -fcheck=bounds now warns about invalid string 5859 lengths of character dummy arguments. Additionally, more 5860 compile-time checks have been added. 5861 * The new option [19]-fno-protect-parens has been added; if set, the 5862 compiler may reorder REAL and COMPLEX expressions without regard to 5863 parentheses. 5864 * GNU Fortran no longer links against libgfortranbegin. As before, 5865 MAIN__ (assembler symbol name) is the actual Fortran main program, 5866 which is invoked by the main function. However, main is now 5867 generated and put in the same object file as MAIN__. For the time 5868 being, libgfortranbegin still exists for backward compatibility. 5869 For details see the new [20]Mixed-Language Programming chapter in 5870 the manual. 5871 * The I/O library was restructured for performance and cleaner code. 5872 * Array assignments and WHERE are now run in parallel when OpenMP's 5873 WORKSHARE is used. 5874 * The experimental option -fwhole-file was added. The option allows 5875 whole-file checking of procedure arguments and allows for better 5876 optimizations. It can also be used with -fwhole-program, which is 5877 now also supported in gfortran. 5878 * More Fortran 2003 and Fortran 2008 mathematical functions can now 5879 be used as initialization expressions. 5880 * Some extended attributes such as STDCALL are now supported via the 5881 [21]GCC$ compiler directive. 5882 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 5883 intrinsic behaves now as if zero were always positive. 5884 * For legacy compatibiliy: On Cygwin and MinGW, the special files 5885 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 5886 supported. 5887 * Fortran 2003 support has been extended: 5888 + Procedure-pointer function results and procedure-pointer 5889 components (including PASS), 5890 + allocatable scalars (experimental), 5891 + DEFERRED type-bound procedures, 5892 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 5893 have been implemented. 5894 + The ALLOCATE statement supports type-specs and the SOURCE= 5895 argument. 5896 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 5897 type-bound procedure (i.e. as type-bound operators). 5898 + Rounding (ROUND=, RZ, ...) for output is now supported. 5899 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 5900 intrinsic module ISO_C_BINDING are now supported, except for 5901 the targets listed above as ones where GCC does not have 5902 <stdint.h> type information. 5903 + Extensible derived types with type-bound procedure or 5904 procedure pointer with PASS attribute now have to use CLASS in 5905 line with the Fortran 2003 standard; the workaround to use 5906 TYPE is no longer supported. 5907 + [22]Experimental, incomplete support for polymorphism, 5908 including CLASS, SELECT TYPE and dynamic dispatch of 5909 type-bound procedure calls. Some features do not work yet such 5910 as unlimited polymorphism (CLASS(*)). 5911 * Fortran 2008 support has been extended: 5912 + The OPEN statement now supports the NEWUNIT= option, which 5913 returns a unique file unit, thus preventing inadvertent use of 5914 the same unit in different parts of the program. 5915 + Support for unlimited format items has been added. 5916 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 5917 the intrinsic module ISO_FORTRAN_ENV are now supported. 5918 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 5919 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 5920 and ATANH have been added (for real and complex arguments) and 5921 ATAN(Y,X) is now an alias for ATAN2(Y,X). 5922 + The BLOCK construct has been implemented. 5923 5924 New Targets and Target Specific Improvements 5925 5926 AIX 5927 5928 * Full cross-toolchain support now available with GNU Binutils 5929 5930 ARM 5931 5932 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 5933 * GCC now supports the ARM v7E-M architecture. 5934 * GCC now supports VFPv4-based FPUs and FPUs with 5935 single-precision-only VFP. 5936 * GCC has many improvements to optimization for other ARM processors, 5937 including scheduling support for the integer pipeline on Cortex-A9. 5938 * GCC now supports the IEEE 754-2008 half-precision floating-point 5939 type, and a variant ARM-specific half-precision type. This type is 5940 specified using __fp16, with the layout determined by 5941 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 5942 VFPv4 half-precision instructions will be used. 5943 * GCC now supports the variant of AAPCS that uses VFP registers for 5944 parameter passing and return values. 5945 5946 AVR 5947 5948 * The -mno-tablejump option has been removed because it has the same 5949 effect as the -fno-jump-tables option. 5950 * Added support for these new AVR devices: 5951 + ATmega8U2 5952 + ATmega16U2 5953 + ATmega32U2 5954 5955 IA-32/x86-64 5956 5957 * GCC now will set the default for -march= based on the configure 5958 target. 5959 * GCC now supports handling floating-point excess precision arising 5960 from use of the x87 floating-point unit in a way that conforms to 5961 ISO C99. This is enabled with -fexcess-precision=standard and with 5962 standards conformance options such as -std=c99, and may be disabled 5963 using -fexcess-precision=fast. 5964 * Support for the Intel Atom processor is now available through the 5965 -march=atom and -mtune=atom options. 5966 * A new -mcrc32 option is now available to enable crc32 intrinsics. 5967 * A new -mmovbe option is now available to enable GCC to use the 5968 movbe instruction to implement __builtin_bswap32 and 5969 __builtin_bswap64. 5970 * SSE math now can be enabled by default at configure time with the 5971 new --with-fpmath=sse option. 5972 * There is a new intrinsic header file, <x86intrin.h>. It should be 5973 included before using any IA-32/x86-64 intrinsics. 5974 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 5975 Orochi processors are now available with the -mxop, -mfma4, and 5976 -mlwp options. 5977 * The -mabm option enables GCC to use the popcnt and lzcnt 5978 instructions on AMD processors. 5979 * The -mpopcnt option enables GCC to use the popcnt instructions on 5980 both AMD and Intel processors. 5981 5982 M68K/ColdFire 5983 5984 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 5985 and 5441x devices. 5986 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 5987 processors. 5988 5989 MeP 5990 5991 Support has been added for the Toshiba Media embedded Processor (MeP, 5992 or mep-elf) embedded target. 5993 5994 MIPS 5995 5996 * GCC now supports MIPS 1004K processors. 5997 * GCC can now be configured with options --with-arch-32, 5998 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 5999 default optimization separately for 32-bit and 64-bit modes. 6000 * MIPS targets now support an alternative _mcount interface, in which 6001 register $12 points to the function's save slot for register $31. 6002 This interface is selected by the -mcount-ra-address option; see 6003 the documentation for more details. 6004 * GNU/Linux targets can now generate read-only .eh_frame sections. 6005 This optimization requires GNU binutils 2.20 or above, and is only 6006 available if GCC is configured with a suitable version of binutils. 6007 * GNU/Linux targets can now attach special relocations to indirect 6008 calls, so that the linker can turn them into direct jumps or 6009 branches. This optimization requires GNU binutils 2.20 or later, 6010 and is automatically selected if GCC is configured with an 6011 appropriate version of binutils. It can be explicitly enabled or 6012 disabled using the -mrelax-pic-calls command-line option. 6013 * GCC now generates more heavily-optimized atomic operations on 6014 Octeon processors. 6015 * MIPS targets now support the -fstack-protector option. 6016 * GCC now supports an -msynci option, which specifies that synci is 6017 enough to flush the instruction cache, without help from the 6018 operating system. GCC uses this information to optimize 6019 automatically-generated cache flush operations, such as those used 6020 for nested functions in C. There is also a --with-synci 6021 configure-time option, which makes -msynci the default. 6022 * GCC supports four new function attributes for interrupt handlers: 6023 interrupt, use_shadow_register_set, keep_interrupts_masked and 6024 use_debug_exception_return. See the documentation for more details 6025 about these attributes. 6026 6027 RS/6000 (POWER/PowerPC) 6028 6029 * GCC now supports the Power ISA 2.06, which includes the VSX 6030 instructions that add vector 64-bit floating point support, new 6031 population count instructions, and conversions between floating 6032 point and unsigned types. 6033 * Support for the power7 processor is now available through the 6034 -mcpu=power7 and -mtune=power7. 6035 * GCC will now vectorize loops that contain simple math functions 6036 like copysign when generating code for altivec or VSX targets. 6037 * Support for the A2 processor is now available through the -mcpu=a2 6038 and -mtune=a2 options. 6039 * Support for the 476 processor is now available through the 6040 -mcpu={476,476fp} and -mtune={476,476fp} options. 6041 * Support for the e500mc64 processor is now available through the 6042 -mcpu=e500mc64 and -mtune=e500mc64 options. 6043 * GCC can now be configured with options --with-cpu-32, 6044 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 6045 default optimization separately for 32-bit and 64-bit modes. 6046 * Starting with GCC 4.5.4, vectors of type vector long long or vector 6047 long are passed and returned in the same method as other vectors 6048 with the VSX instruction set. Previously the GCC compiler did not 6049 adhere to the ABI for 128-bit vectors with 64-bit integer base 6050 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 6051 6052 RX 6053 6054 Support has been added for the Renesas RX Processor (rx-elf) target. 6055 6056 Operating Systems 6057 6058 Windows (Cygwin and MinGW) 6059 6060 * GCC now installs all the major language runtime libraries as DLLs 6061 when configured with the --enable-shared option. 6062 * GCC now makes use of the new support for aligned common variables 6063 in versions of binutils >= 2.20 to fix bugs in the support for SSE 6064 data types. 6065 * Improvements to the libffi support library increase the reliability 6066 of code generated by GCJ on all Windows platforms. Libgcj is 6067 enabled by default for the first time. 6068 * Libtool improvements simplify installation by placing the generated 6069 DLLs in the correct binaries directory. 6070 * Numerous other minor bugfixes and improvements, and substantial 6071 enhancements to the Fortran language support library. 6072 6073 > 6074 6075 Other significant improvements 6076 6077 Plugins 6078 6079 * It is now possible to extend the compiler without having to modify 6080 its source code. A new option -fplugin=file.so tells GCC to load 6081 the shared object file.so and execute it as part of the compiler. 6082 The internal documentation describes the details on how plugins can 6083 interact with the compiler. 6084 6085 Installation changes 6086 6087 * The move to newer autotools changed default installation 6088 directories and switches to control them: The --with-datarootdir, 6089 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 6090 used any more. Instead, you can now use --datarootdir, --docdir, 6091 --htmldir, and --pdfdir. The default installation directories have 6092 changed as follows according to the GNU Coding Standards: 6093 6094 datarootdir read-only architecture-independent data root [PREFIX/share] 6095 localedir locale-specific message catalogs [DATAROOTDIR/locale] 6096 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 6097 htmldir html documentation [DOCDIR] 6098 dvidir dvi documentation [DOCDIR] 6099 pdfdir pdf documentation [DOCDIR] 6100 psdir ps documentation [DOCDIR] 6101 The following variables have new default values: 6102 6103 datadir read-only architecture-independent data [DATAROOTDIR] 6104 infodir info documentation [DATAROOTDIR/info] 6105 mandir man documentation [DATAROOTDIR/man] 6106 6107 GCC 4.5.1 6108 6109 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6110 system that are known to be fixed in the 4.5.1 release. This list might 6111 not be complete (that is, it is possible that some PRs that have been 6112 fixed are not listed here). 6113 6114 All languages 6115 6116 * GCC's new link-time optimizer ([24]-flto) now also works on a few 6117 non-ELF targets: 6118 + Cygwin (*-cygwin*) 6119 + MinGW (*-mingw*) 6120 + Darwin on x86-64 (x86_64-apple-darwin*) 6121 LTO is not enabled by default for these targets. To enable LTO, you 6122 should configure with the --enable-lto option. 6123 6124 GCC 4.5.2 6125 6126 This is the [25]list of problem reports (PRs) from GCC's bug tracking 6127 system that are known to be fixed in the 4.5.2 release. This list might 6128 not be complete (that is, it is possible that some PRs that have been 6129 fixed are not listed here). 6130 6131 GCC 4.5.3 6132 6133 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6134 system that are known to be fixed in the 4.5.3 release. This list might 6135 not be complete (that is, it is possible that some PRs that have been 6136 fixed are not listed here). 6137 6138 On the PowerPC compiler, the Altivec builtin functions vec_ld and 6139 vec_st have been modified to generate the Altivec memory instructions 6140 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 6141 release, these builtin functions were changed to generate VSX memory 6142 reference instructions instead of Altivec memory instructions, but 6143 there are differences between the two instructions. If the VSX 6144 instruction set is available, you can now use the new builtin functions 6145 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 6146 instructions. 6147 6148 GCC 4.5.4 6149 6150 This is the [27]list of problem reports (PRs) from GCC's bug tracking 6151 system that are known to be fixed in the 4.5.4 release. This list might 6152 not be complete (that is, it is possible that some PRs that have been 6153 fixed are not listed here). 6154 6155 6156 For questions related to the use of GCC, please consult these web 6157 pages and the [28]GCC manuals. If that fails, the 6158 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6159 web pages and the development of GCC are welcome on our developer 6160 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 6161 archives. 6162 6163 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 6164 distribution of this entire article is permitted in any medium, 6165 provided this notice is preserved. 6166 6167 These pages are [33]maintained by the GCC team. Last modified 6168 2017-02-28[34]. 6169 6170 References 6171 6172 1. http://www.multiprecision.org/ 6173 2. https://gcc.gnu.org/install/prerequisites.html 6174 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 6175 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 6176 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 6177 6. http://www.multiprecision.org/ 6178 7. https://gcc.gnu.org/PR30789 6179 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 6180 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 6181 10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 6182 11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 6183 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 6184 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 6185 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 6186 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 6187 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 6188 17. https://sourceware.org/gdb/wiki/STLSupport 6189 18. http://gcc.gnu.org/gcc-4.4/changes.html 6190 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 6191 20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 6192 21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 6193 22. https://gcc.gnu.org/wiki/OOP 6194 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 6195 24. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 6196 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 6197 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 6198 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 6199 28. https://gcc.gnu.org/onlinedocs/ 6200 29. mailto:gcc-help (a] gcc.gnu.org 6201 30. mailto:gcc (a] gcc.gnu.org 6202 31. https://gcc.gnu.org/lists.html 6203 32. http://www.fsf.org/ 6204 33. https://gcc.gnu.org/about.html 6205 34. http://validator.w3.org/check/referer 6206 ====================================================================== 6207 http://gcc.gnu.org/gcc-4.4/index.html 6208 6209 GCC 4.4 Release Series 6210 6211 March 13, 2012 6212 6213 The [1]GNU project and the GCC developers are pleased to announce the 6214 release of GCC 4.4.7. 6215 6216 This release is a bug-fix release, containing fixes for regressions in 6217 GCC 4.4.6 relative to previous releases of GCC. 6218 6219 This release series is no longer maintained. 6220 6221 Release History 6222 6223 GCC 4.4.7 6224 March 13, 2012 ([2]changes) 6225 6226 GCC 4.4.6 6227 April 16, 2011 ([3]changes) 6228 6229 GCC 4.4.5 6230 October 1, 2010 ([4]changes) 6231 6232 GCC 4.4.4 6233 April 29, 2010 ([5]changes) 6234 6235 GCC 4.4.3 6236 January 21, 2010 ([6]changes) 6237 6238 GCC 4.4.2 6239 October 15, 2009 ([7]changes) 6240 6241 GCC 4.4.1 6242 July 22, 2009 ([8]changes) 6243 6244 GCC 4.4.0 6245 April 21, 2009 ([9]changes) 6246 6247 References and Acknowledgements 6248 6249 GCC used to stand for the GNU C Compiler, but since the compiler 6250 supports several other languages aside from C, it now stands for the 6251 GNU Compiler Collection. 6252 6253 A list of [10]successful builds is updated as new information becomes 6254 available. 6255 6256 The GCC developers would like to thank the numerous people that have 6257 contributed new features, improvements, bug fixes, and other changes as 6258 well as test results to GCC. This [11]amazing group of volunteers is 6259 what makes GCC successful. 6260 6261 For additional information about GCC please refer to the [12]GCC 6262 project web site or contact the [13]GCC development mailing list. 6263 6264 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 6265 6266 6267 For questions related to the use of GCC, please consult these web 6268 pages and the [16]GCC manuals. If that fails, the 6269 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6270 web pages and the development of GCC are welcome on our developer 6271 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 6272 archives. 6273 6274 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 6275 distribution of this entire article is permitted in any medium, 6276 provided this notice is preserved. 6277 6278 These pages are [21]maintained by the GCC team. Last modified 6279 2016-09-30[22]. 6280 6281 References 6282 6283 1. http://www.gnu.org/ 6284 2. http://gcc.gnu.org/gcc-4.4/changes.html 6285 3. http://gcc.gnu.org/gcc-4.4/changes.html 6286 4. http://gcc.gnu.org/gcc-4.4/changes.html 6287 5. http://gcc.gnu.org/gcc-4.4/changes.html 6288 6. http://gcc.gnu.org/gcc-4.4/changes.html 6289 7. http://gcc.gnu.org/gcc-4.4/changes.html 6290 8. http://gcc.gnu.org/gcc-4.4/changes.html 6291 9. http://gcc.gnu.org/gcc-4.4/changes.html 6292 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 6293 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6294 12. http://gcc.gnu.org/index.html 6295 13. mailto:gcc (a] gcc.gnu.org 6296 14. http://gcc.gnu.org/mirrors.html 6297 15. http://gcc.gnu.org/svn.html 6298 16. https://gcc.gnu.org/onlinedocs/ 6299 17. mailto:gcc-help (a] gcc.gnu.org 6300 18. mailto:gcc (a] gcc.gnu.org 6301 19. https://gcc.gnu.org/lists.html 6302 20. http://www.fsf.org/ 6303 21. https://gcc.gnu.org/about.html 6304 22. http://validator.w3.org/check/referer 6305 ====================================================================== 6306 http://gcc.gnu.org/gcc-4.4/changes.html 6307 6308 GCC 4.4 Release Series 6309 Changes, New Features, and Fixes 6310 6311 The latest release in the 4.4 release series is [1]GCC 4.4.7. 6312 6313 Caveats 6314 6315 * __builtin_stdarg_start has been completely removed from GCC. 6316 Support for <varargs.h> had been deprecated since GCC 4.0. Use 6317 __builtin_va_start as a replacement. 6318 * Some of the errors issued by the C++ front end that could be 6319 downgraded to warnings in previous releases by using -fpermissive 6320 are now warnings by default. They can be converted into errors by 6321 using -pedantic-errors. 6322 * Use of the cpp assertion extension will now emit a warning when 6323 -Wdeprecated or -pedantic is used. This extension has been 6324 deprecated for many years, but never warned about. 6325 * Packed bit-fields of type char were not properly bit-packed on many 6326 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 6327 causes an ABI change. For example there is no longer a 4-bit 6328 padding between field a and b in this structure: 6329 struct foo 6330 { 6331 char a:4; 6332 char b:8; 6333 } __attribute__ ((packed)); 6334 There is a new warning to help identify fields that are affected: 6335 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 6336 The warning can be disabled with -Wno-packed-bitfield-compat. 6337 * On ARM EABI targets, the C++ mangling of the va_list type has been 6338 changed to conform to the current revision of the EABI. This does 6339 not affect the libstdc++ library included with GCC. 6340 * The SCOUNT and POS bits of the MIPS DSP control register are now 6341 treated as global. Previous versions of GCC treated these fields as 6342 call-clobbered instead. 6343 * The MIPS port no longer recognizes the h asm constraint. It was 6344 necessary to remove this constraint in order to avoid generating 6345 unpredictable code sequences. 6346 One of the main uses of the h constraint was to extract the high 6347 part of a multiplication on 64-bit targets. For example: 6348 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 6349 You can now achieve the same effect using 128-bit types: 6350 typedef unsigned int uint128_t __attribute__((mode(TI))); 6351 result = ((uint128_t) x * y) >> 64; 6352 The second sequence is better in many ways. For example, if x and y 6353 are constants, the compiler can perform the multiplication at 6354 compile time. If x and y are not constants, the compiler can 6355 schedule the runtime multiplication better than it can schedule an 6356 asm statement. 6357 * Support for a number of older systems and recently unmaintained or 6358 untested target ports of GCC has been declared obsolete in GCC 4.4. 6359 Unless there is activity to revive them, the next release of GCC 6360 will have their sources permanently removed. 6361 The following ports for individual systems on particular 6362 architectures have been obsoleted: 6363 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 6364 m68k-*-aout*) 6365 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 6366 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 6367 sh-*-*). This does not affect other more specific targets 6368 using the COFF object format on those architectures, or the 6369 more specific H8300 and SH targets (h8300-*-rtems*, 6370 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 6371 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 6372 + 2BSD on PDP-11 (pdp11-*-bsd) 6373 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 6374 powerpc-ibm-aix4.[12]*) 6375 + Tuning support for Itanium1 (Merced) variants. Note that code 6376 tuned for Itanium2 should also run correctly on Itanium1. 6377 * The protoize and unprotoize utilities have been obsoleted and will 6378 be removed in GCC 4.5. These utilities have not been installed by 6379 default since GCC 3.0. 6380 * Support has been removed for all the [2]configurations obsoleted in 6381 GCC 4.3. 6382 * Unknown -Wno-* options are now silently ignored by GCC if no other 6383 diagnostics are issued. If other diagnostics are issued, then GCC 6384 warns about the unknown options. 6385 * More information on porting to GCC 4.4 from previous versions of 6386 GCC can be found in the [3]porting guide for this release. 6387 6388 General Optimizer Improvements 6389 6390 * A new command-line switch -findirect-inlining has been added. When 6391 turned on it allows the inliner to also inline indirect calls that 6392 are discovered to have known targets at compile time thanks to 6393 previous inlining. 6394 * A new command-line switch -ftree-switch-conversion has been added. 6395 This new pass turns simple initializations of scalar variables in 6396 switch statements into initializations from a static array, given 6397 that all the values are known at compile time and the ratio between 6398 the new array size and the original switch branches does not exceed 6399 the parameter --param switch-conversion-max-branch-ratio (default 6400 is eight). 6401 * A new command-line switch -ftree-builtin-call-dce has been added. 6402 This optimization eliminates unnecessary calls to certain builtin 6403 functions when the return value is not used, in cases where the 6404 calls can not be eliminated entirely because the function may set 6405 errno. This optimization is on by default at -O2 and above. 6406 * A new command-line switch -fconserve-stack directs the compiler to 6407 minimize stack usage even if it makes the generated code slower. 6408 This affects inlining decisions. 6409 * When the assembler supports it, the compiler will now emit unwind 6410 information using assembler .cfi directives. This makes it possible 6411 to use such directives in inline assembler code. The new option 6412 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 6413 directives. 6414 * The [4]Graphite branch has been merged. This merge has brought in a 6415 new framework for loop optimizations based on a polyhedral 6416 intermediate representation. These optimizations apply to all the 6417 languages supported by GCC. The following new code transformations 6418 are available in GCC 4.4: 6419 + -floop-interchange performs loop interchange transformations 6420 on loops. Interchanging two nested loops switches the inner 6421 and outer loops. For example, given a loop like: 6422 DO J = 1, M 6423 DO I = 1, N 6424 A(J, I) = A(J, I) * C 6425 ENDDO 6426 ENDDO 6427 6428 loop interchange will transform the loop as if the user had 6429 written: 6430 DO I = 1, N 6431 DO J = 1, M 6432 A(J, I) = A(J, I) * C 6433 ENDDO 6434 ENDDO 6435 6436 which can be beneficial when N is larger than the caches, 6437 because in Fortran, the elements of an array are stored in 6438 memory contiguously by column, and the original loop iterates 6439 over rows, potentially creating at each access a cache miss. 6440 + -floop-strip-mine performs loop strip mining transformations 6441 on loops. Strip mining splits a loop into two nested loops. 6442 The outer loop has strides equal to the strip size and the 6443 inner loop has strides of the original loop within a strip. 6444 For example, given a loop like: 6445 DO I = 1, N 6446 A(I) = A(I) + C 6447 ENDDO 6448 6449 loop strip mining will transform the loop as if the user had 6450 written: 6451 DO II = 1, N, 4 6452 DO I = II, min (II + 3, N) 6453 A(I) = A(I) + C 6454 ENDDO 6455 ENDDO 6456 6457 + -floop-block performs loop blocking transformations on loops. 6458 Blocking strip mines each loop in the loop nest such that the 6459 memory accesses of the element loops fit inside caches. For 6460 example, given a loop like: 6461 DO I = 1, N 6462 DO J = 1, M 6463 A(J, I) = B(I) + C(J) 6464 ENDDO 6465 ENDDO 6466 6467 loop blocking will transform the loop as if the user had 6468 written: 6469 DO II = 1, N, 64 6470 DO JJ = 1, M, 64 6471 DO I = II, min (II + 63, N) 6472 DO J = JJ, min (JJ + 63, M) 6473 A(J, I) = B(I) + C(J) 6474 ENDDO 6475 ENDDO 6476 ENDDO 6477 ENDDO 6478 6479 which can be beneficial when M is larger than the caches, 6480 because the innermost loop will iterate over a smaller amount 6481 of data that can be kept in the caches. 6482 * A new register allocator has replaced the old one. It is called 6483 integrated register allocator (IRA) because coalescing, register 6484 live range splitting, and hard register preferencing are done 6485 on-the-fly during coloring. It also has better integration with the 6486 reload pass. IRA is a regional register allocator which uses modern 6487 Chaitin-Briggs coloring instead of Chow's priority coloring used in 6488 the old register allocator. More info about IRA internals and 6489 options can be found in the GCC manuals. 6490 * A new instruction scheduler and software pipeliner, based on the 6491 selective scheduling approach, has been added. The new pass 6492 performs instruction unification, register renaming, substitution 6493 through register copies, and speculation during scheduling. The 6494 software pipeliner is able to pipeline non-countable loops. The new 6495 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 6496 it is available for the Intel Itanium platform working by default 6497 as the second scheduling pass (after register allocation) at the 6498 -O3 optimization level. 6499 * When using -fprofile-generate with a multi-threaded program, the 6500 profile counts may be slightly wrong due to race conditions. The 6501 new -fprofile-correction option directs the compiler to apply 6502 heuristics to smooth out the inconsistencies. By default the 6503 compiler will give an error message when it finds an inconsistent 6504 profile. 6505 * The new -fprofile-dir=PATH option permits setting the directory 6506 where profile data files are stored when using -fprofile-generate 6507 and friends, and the directory used when reading profile data files 6508 using -fprofile-use and friends. 6509 6510 New warning options 6511 6512 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 6513 warning if any stack frame is larger than NUMBER bytes. This may be 6514 used to help ensure that code fits within a limited amount of stack 6515 space. 6516 * The command-line option -Wlarger-than-N is now written as 6517 -Wlarger-than=N and the old form is deprecated. 6518 * The new -Wno-mudflap option disables warnings about constructs 6519 which can not be instrumented when using -fmudflap. 6520 6521 New Languages and Language specific improvements 6522 6523 * Version 3.0 of the OpenMP specification is now supported for the C, 6524 C++, and Fortran compilers. 6525 * New character data types, per [5]TR 19769: New character types in 6526 C, are now supported for the C compiler in -std=gnu99 mode, as 6527 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 6528 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 6529 6530 C family 6531 6532 * A new optimize attribute was added to allow programmers to change 6533 the optimization level and particular optimization options for an 6534 individual function. You can also change the optimization options 6535 via the GCC optimize pragma for functions defined after the pragma. 6536 The GCC push_options pragma and the GCC pop_options pragma allow 6537 you temporarily save and restore the options used. The GCC 6538 reset_options pragma restores the options to what was specified on 6539 the command line. 6540 * Uninitialized warnings do not require enabling optimization 6541 anymore, that is, -Wuninitialized can be used together with -O0. 6542 Nonetheless, the warnings given by -Wuninitialized will probably be 6543 more accurate if optimization is enabled. 6544 * -Wparentheses now warns about expressions such as (!x | y) and (!x 6545 & y). Using explicit parentheses, such as in ((!x) | y), silences 6546 this warning. 6547 * -Wsequence-point now warns within if, while,do while and for 6548 conditions, and within for begin/end expressions. 6549 * A new option -dU is available to dump definitions of preprocessor 6550 macros that are tested or expanded. 6551 6552 C++ 6553 6554 * [6]Improved experimental support for the upcoming ISO C++ standard, 6555 C++0x. Including support for auto, inline namespaces, generalized 6556 initializer lists, defaulted and deleted functions, new character 6557 types, and scoped enums. 6558 * Those errors that may be downgraded to warnings to build legacy 6559 code now mention -fpermissive when -fdiagnostics-show-option is 6560 enabled. 6561 * -Wconversion now warns if the result of a static_cast to enumeral 6562 type is unspecified because the value is outside the range of the 6563 enumeral type. 6564 * -Wuninitialized now warns if a non-static reference or non-static 6565 const member appears in a class without constructors. 6566 * G++ now properly implements value-initialization, so objects with 6567 an initializer of () and an implicitly defined default constructor 6568 will be zero-initialized before the default constructor is called. 6569 6570 Runtime Library (libstdc++) 6571 6572 * Improved experimental support for the upcoming ISO C++ standard, 6573 C++0x, including: 6574 + Support for <chrono>, <condition_variable>, <cstdatomic>, 6575 <forward_list>, <initializer_list>, <mutex>, <ratio>, 6576 <system_error>, and <thread>. 6577 + unique_ptr, <algorithm> additions, exception propagation, and 6578 support for the new character types in <string> and <limits>. 6579 + Existing facilities now exploit initializer lists, defaulted 6580 and deleted functions, and the newly implemented core C++0x 6581 features. 6582 + Some standard containers are more efficient together with 6583 stateful allocators, i.e., no allocator is constructed on the 6584 fly at element construction time. 6585 * Experimental support for non-standard pointer types in containers. 6586 * The long standing libstdc++/30928 has been fixed for targets 6587 running glibc 2.10 or later. 6588 * As usual, many small and larger bug fixes, in particular quite a 6589 few corner cases in <locale>. 6590 6591 Fortran 6592 6593 * GNU Fortran now employs libcpp directly instead of using cc1 as an 6594 external preprocessor. The [7]-cpp option was added to allow manual 6595 invocation of the preprocessor without relying on filename 6596 extensions. 6597 * The [8]-Warray-temporaries option warns about array temporaries 6598 generated by the compiler, as an aid to optimization. 6599 * The [9]-fcheck-array-temporaries option has been added, printing a 6600 notification at run time, when an array temporary had to be created 6601 for an function argument. Contrary to -Warray-temporaries the 6602 warning is only printed if the array is noncontiguous. 6603 * Improved generation of DWARF debugging symbols 6604 * If using an intrinsic not part of the selected standard (via -std= 6605 and -fall-intrinsics) gfortran will now treat it as if this 6606 procedure were declared EXTERNAL and try to link to a user-supplied 6607 procedure. -Wintrinsics-std will warn whenever this happens. The 6608 now-useless option -Wnonstd-intrinsic was removed. 6609 * The flag -falign-commons has been added to control the alignment of 6610 variables in COMMON blocks, which is enabled by default in line 6611 with previous GCC version. Using -fno-align-commons one can force 6612 commons to be contiguous in memory as required by the Fortran 6613 standard, however, this slows down the memory access. The option 6614 -Walign-commons, which is enabled by default, warns when padding 6615 bytes were added for alignment. The proper solution is to sort the 6616 common objects by decreasing storage size, which avoids the 6617 alignment problems. 6618 * Fortran 2003 support has been extended: 6619 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 6620 now supported (except internal reads from/writes to wide 6621 strings). [10]-fbackslash now supports also \unnnn and 6622 \Unnnnnnnn to enter Unicode characters. 6623 + Asynchronous I/O (implemented as synchronous I/O) and the 6624 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 6625 are now supported in I/O statements. 6626 + Support for Fortran 2003 structure constructors and for array 6627 constructor with typespec has been added. 6628 + Procedure Pointers (but not yet as component in derived types 6629 and as function results) are now supported. 6630 + Abstract types, type extension, and type-bound procedures 6631 (both PROCEDURE and GENERIC but not as operators). Note: As 6632 CLASS/polymorphyic types are not implemented, type-bound 6633 procedures with PASS accept as non-standard extension TYPE 6634 arguments. 6635 * Fortran 2008 support has been added: 6636 + The -std=f2008 option and support for the file extensions 6637 .f2008 and .F2008 has been added. 6638 + The g0 format descriptor is now supported. 6639 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 6640 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 6641 are now available (some of them existed as GNU extension 6642 before). Note: The hyperbolic functions are not yet supporting 6643 complex arguments and the three- argument version of BESSEL_*N 6644 is not available. 6645 + The bit intrinsics LEADZ and TRAILZ have been added. 6646 6647 Java (GCJ) 6648 6649 Ada 6650 6651 * The Ada runtime now supports multilibs on many platforms including 6652 x86_64, SPARC and PowerPC. Their build is enabled by default. 6653 6654 New Targets and Target Specific Improvements 6655 6656 ARM 6657 6658 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 6659 Cortex-R4F processors and has many other improvements to 6660 optimization for ARM processors. 6661 * GCC now supports the VFPv3 variant with 16 double-precision 6662 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 6663 renamed to -mfpu=vfpv3. 6664 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 6665 erratum on Cortex-M3 processors. 6666 * GCC now supports the __sync_* atomic operations for ARM EABI 6667 GNU/Linux. 6668 * The section anchors optimization is now enabled by default when 6669 optimizing for ARM. 6670 * GCC now uses a new EABI-compatible profiling interface for EABI 6671 targets. This requires a function __gnu_mcount_nc, which is 6672 provided by GNU libc versions 2.8 and later. 6673 6674 AVR 6675 6676 * The -mno-tablejump option has been deprecated because it has the 6677 same effect as the -fno-jump-tables option. 6678 * Added support for these new AVR devices: 6679 + ATA6289 6680 + ATtiny13A 6681 + ATtiny87 6682 + ATtiny167 6683 + ATtiny327 6684 + ATmega8C1 6685 + ATmega16C1 6686 + ATmega32C1 6687 + ATmega8M1 6688 + ATmega16M1 6689 + ATmega32M1 6690 + ATmega32U4 6691 + ATmega16HVB 6692 + ATmega4HVD 6693 + ATmega8HVD 6694 + ATmega64C1 6695 + ATmega64M1 6696 + ATmega16U4 6697 + ATmega32U6 6698 + ATmega128RFA1 6699 + AT90PWM81 6700 + AT90SCR100 6701 + M3000F 6702 + M3000S 6703 + M3001B 6704 6705 IA-32/x86-64 6706 6707 * Support for Intel AES built-in functions and code generation is 6708 available via -maes. 6709 * Support for Intel PCLMUL built-in function and code generation is 6710 available via -mpclmul. 6711 * Support for Intel AVX built-in functions and code generation is 6712 available via -mavx. 6713 * Automatically align the stack for local variables with alignment 6714 requirement. 6715 * GCC can now utilize the SVML library for vectorizing calls to a set 6716 of C99 functions if -mveclibabi=svml is specified and you link to 6717 an SVML ABI compatible library. 6718 * On x86-64, the ABI has been changed in the following cases to 6719 conform to the x86-64 ABI: 6720 + Passing/returning structures with flexible array member: 6721 struct foo 6722 { 6723 int i; 6724 int flex[]; 6725 }; 6726 + Passing/returning structures with complex float member: 6727 struct foo 6728 { 6729 int i; 6730 __complex__ float f; 6731 }; 6732 + Passing/returning unions with long double member: 6733 union foo 6734 { 6735 int x; 6736 long double ld; 6737 }; 6738 Code built with previous versions of GCC that uses any of these is 6739 not compatible with code built with GCC 4.4.0 or later. 6740 * A new target attribute was added to allow programmers to change the 6741 target options like -msse2 or -march=k8 for an individual function. 6742 You can also change the target options via the GCC target pragma 6743 for functions defined after the pragma. 6744 * GCC can now be configured with options --with-arch-32, 6745 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 6746 --with-tune-64 to control the default optimization separately for 6747 32-bit and 64-bit modes. 6748 6749 IA-32/IA64 6750 6751 * Support for __float128 (TFmode) IEEE quad type and corresponding 6752 TCmode IEEE complex quad type is available via the soft-fp library 6753 on IA-32/IA64 targets. This includes basic arithmetic operations 6754 (addition, subtraction, negation, multiplication and division) on 6755 __float128 real and TCmode complex values, the full set of IEEE 6756 comparisons between __float128 values, conversions to and from 6757 float, double and long double floating point types, as well as 6758 conversions to and from signed or unsigned integer, signed or 6759 unsigned long integer and signed or unsigned quad (TImode, IA64 6760 only) integer types. Additionally, all operations generate the full 6761 set of IEEE exceptions and support the full set of IEEE rounding 6762 modes. 6763 6764 M68K/ColdFire 6765 6766 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 6767 processors. (Scheduling support for ColdFire V2 processors was 6768 added in GCC 4.3.) 6769 * GCC now supports the -mxgot option to support programs requiring 6770 many GOT entries on ColdFire. 6771 * The m68k-*-linux-gnu target now builds multilibs by default. 6772 6773 MIPS 6774 6775 * MIPS Technologies have extended the original MIPS SVR4 ABI to 6776 include support for procedure linkage tables (PLTs) and copy 6777 relocations. These extensions allow GNU/Linux executables to use a 6778 significantly more efficient code model than the one defined by the 6779 original ABI. 6780 GCC support for this code model is available via a new command-line 6781 option, -mplt. There is also a new configure-time option, 6782 --with-mips-plt, to make -mplt the default. 6783 The new code model requires support from the assembler, the linker, 6784 and the runtime C library. This support is available in binutils 6785 2.19 and GLIBC 2.9. 6786 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 6787 and 32-bit GNU/Linux shared libraries. This feature requires GNU 6788 binutils 2.19 or above. 6789 * Support for RMI's XLR processor is now available through the 6790 -march=xlr and -mtune=xlr options. 6791 * 64-bit targets can now perform 128-bit multiplications inline, 6792 instead of relying on a libgcc function. 6793 * Native GNU/Linux toolchains now support -march=native and 6794 -mtune=native, which select the host processor. 6795 * GCC now supports the R10K, R12K, R14K and R16K processors. The 6796 canonical -march= and -mtune= names for these processors are 6797 r10000, r12000, r14000 and r16000 respectively. 6798 * GCC can now work around the side effects of speculative execution 6799 on R10K processors. Please see the documentation of the 6800 -mr10k-cache-barrier option for details. 6801 * Support for the MIPS64 Release 2 instruction set has been added. 6802 The option -march=mips64r2 enables generation of these 6803 instructions. 6804 * GCC now supports Cavium Networks' Octeon processor. This support is 6805 available through the -march=octeon and -mtune=octeon options. 6806 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 6807 canonical -march= and -mtune= names for these processors are 6808 loongson2e and loongson2f. 6809 6810 picochip 6811 6812 Picochip is a 16-bit processor. A typical picoChip contains over 250 6813 small cores, each with small amounts of memory. There are three 6814 processor variants (STAN, MEM and CTRL) with different instruction sets 6815 and memory configurations and they can be chosen using the -mae option. 6816 6817 This port is intended to be a "C" only port. 6818 6819 Power Architecture and PowerPC 6820 6821 * GCC now supports the e300c2, e300c3 and e500mc processors. 6822 * GCC now supports Xilinx processors with a single-precision FPU. 6823 * Decimal floating point is now supported for e500 processors. 6824 6825 S/390, zSeries and System z9/z10 6826 6827 * Support for the IBM System z10 EC/BC processor has been added. When 6828 using the -march=z10 option, the compiler will generate code making 6829 use of instructions provided by the General-Instruction-Extension 6830 Facility and the Execute-Extension Facility. 6831 6832 VxWorks 6833 6834 * GCC now supports the thread-local storage mechanism used on 6835 VxWorks. 6836 6837 Xtensa 6838 6839 * GCC now supports thread-local storage (TLS) for Xtensa processor 6840 configurations that include the Thread Pointer option. TLS also 6841 requires support from the assembler and linker; this support is 6842 provided in the GNU binutils beginning with version 2.19. 6843 6844 Documentation improvements 6845 6846 Other significant improvements 6847 6848 GCC 4.4.1 6849 6850 This is the [11]list of problem reports (PRs) from GCC's bug tracking 6851 system that are known to be fixed in the 4.4.1 release. This list might 6852 not be complete (that is, it is possible that some PRs that have been 6853 fixed are not listed here). 6854 6855 GCC 4.4.2 6856 6857 This is the [12]list of problem reports (PRs) from GCC's bug tracking 6858 system that are known to be fixed in the 4.4.2 release. This list might 6859 not be complete (that is, it is possible that some PRs that have been 6860 fixed are not listed here). 6861 6862 GCC 4.4.3 6863 6864 This is the [13]list of problem reports (PRs) from GCC's bug tracking 6865 system that are known to be fixed in the 4.4.3 release. This list might 6866 not be complete (that is, it is possible that some PRs that have been 6867 fixed are not listed here). 6868 6869 GCC 4.4.4 6870 6871 This is the [14]list of problem reports (PRs) from GCC's bug tracking 6872 system that are known to be fixed in the 4.4.4 release. This list might 6873 not be complete (that is, it is possible that some PRs that have been 6874 fixed are not listed here). 6875 6876 GCC 4.4.5 6877 6878 This is the [15]list of problem reports (PRs) from GCC's bug tracking 6879 system that are known to be fixed in the 4.4.5 release. This list might 6880 not be complete (that is, it is possible that some PRs that have been 6881 fixed are not listed here). 6882 6883 GCC 4.4.6 6884 6885 This is the [16]list of problem reports (PRs) from GCC's bug tracking 6886 system that are known to be fixed in the 4.4.6 release. This list might 6887 not be complete (that is, it is possible that some PRs that have been 6888 fixed are not listed here). 6889 6890 GCC 4.4.7 6891 6892 This is the [17]list of problem reports (PRs) from GCC's bug tracking 6893 system that are known to be fixed in the 4.4.7 release. This list might 6894 not be complete (that is, it is possible that some PRs that have been 6895 fixed are not listed here). 6896 6897 6898 For questions related to the use of GCC, please consult these web 6899 pages and the [18]GCC manuals. If that fails, the 6900 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6901 web pages and the development of GCC are welcome on our developer 6902 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6903 archives. 6904 6905 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6906 distribution of this entire article is permitted in any medium, 6907 provided this notice is preserved. 6908 6909 These pages are [23]maintained by the GCC team. Last modified 6910 2016-12-28[24]. 6911 6912 References 6913 6914 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 6915 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 6916 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 6917 4. https://gcc.gnu.org/wiki/Graphite 6918 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 6919 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 6920 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 6921 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 6922 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 6923 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 6924 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 6925 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 6926 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 6927 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 6928 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 6929 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 6930 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 6931 18. https://gcc.gnu.org/onlinedocs/ 6932 19. mailto:gcc-help (a] gcc.gnu.org 6933 20. mailto:gcc (a] gcc.gnu.org 6934 21. https://gcc.gnu.org/lists.html 6935 22. http://www.fsf.org/ 6936 23. https://gcc.gnu.org/about.html 6937 24. http://validator.w3.org/check/referer 6938 ====================================================================== 6939 http://gcc.gnu.org/gcc-4.3/index.html 6940 6941 GCC 4.3 Release Series 6942 6943 Jun 27, 2011 6944 6945 The [1]GNU project and the GCC developers are pleased to announce the 6946 release of GCC 4.3.6. 6947 6948 This release is a bug-fix release, containing fixes for regressions in 6949 GCC 4.3.5 relative to previous releases of GCC. 6950 6951 This release series is no longer maintained. 6952 6953 Release History 6954 6955 GCC 4.3.6 6956 Jun 27, 2011 ([2]changes) 6957 6958 GCC 4.3.5 6959 May 22, 2010 ([3]changes) 6960 6961 GCC 4.3.4 6962 August 4, 2009 ([4]changes) 6963 6964 GCC 4.3.3 6965 January 24, 2009 ([5]changes) 6966 6967 GCC 4.3.2 6968 August 27, 2008 ([6]changes) 6969 6970 GCC 4.3.1 6971 June 6, 2008 ([7]changes) 6972 6973 GCC 4.3.0 6974 March 5, 2008 ([8]changes) 6975 6976 References and Acknowledgements 6977 6978 GCC used to stand for the GNU C Compiler, but since the compiler 6979 supports several other languages aside from C, it now stands for the 6980 GNU Compiler Collection. 6981 6982 A list of [9]successful builds is updated as new information becomes 6983 available. 6984 6985 The GCC developers would like to thank the numerous people that have 6986 contributed new features, improvements, bug fixes, and other changes as 6987 well as test results to GCC. This [10]amazing group of volunteers is 6988 what makes GCC successful. 6989 6990 For additional information about GCC please refer to the [11]GCC 6991 project web site or contact the [12]GCC development mailing list. 6992 6993 To obtain GCC please use [13]our mirror sites or [14]our SVN server. 6994 6995 6996 For questions related to the use of GCC, please consult these web 6997 pages and the [15]GCC manuals. If that fails, the 6998 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6999 web pages and the development of GCC are welcome on our developer 7000 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 7001 archives. 7002 7003 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 7004 distribution of this entire article is permitted in any medium, 7005 provided this notice is preserved. 7006 7007 These pages are [20]maintained by the GCC team. Last modified 7008 2016-09-30[21]. 7009 7010 References 7011 7012 1. http://www.gnu.org/ 7013 2. http://gcc.gnu.org/gcc-4.3/changes.html 7014 3. http://gcc.gnu.org/gcc-4.3/changes.html 7015 4. http://gcc.gnu.org/gcc-4.3/changes.html 7016 5. http://gcc.gnu.org/gcc-4.3/changes.html 7017 6. http://gcc.gnu.org/gcc-4.3/changes.html 7018 7. http://gcc.gnu.org/gcc-4.3/changes.html 7019 8. http://gcc.gnu.org/gcc-4.3/changes.html 7020 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 7021 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7022 11. http://gcc.gnu.org/index.html 7023 12. mailto:gcc (a] gcc.gnu.org 7024 13. http://gcc.gnu.org/mirrors.html 7025 14. http://gcc.gnu.org/svn.html 7026 15. https://gcc.gnu.org/onlinedocs/ 7027 16. mailto:gcc-help (a] gcc.gnu.org 7028 17. mailto:gcc (a] gcc.gnu.org 7029 18. https://gcc.gnu.org/lists.html 7030 19. http://www.fsf.org/ 7031 20. https://gcc.gnu.org/about.html 7032 21. http://validator.w3.org/check/referer 7033 ====================================================================== 7034 http://gcc.gnu.org/gcc-4.3/changes.html 7035 7036 GCC 4.3 Release Series 7037 Changes, New Features, and Fixes 7038 7039 The latest release in the 4.3 release series is [1]GCC 4.3.5. 7040 7041 Caveats 7042 7043 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 7044 various front-end languages it supports. See the [4]prerequisites 7045 page for version requirements. 7046 * ColdFire targets now treat long double as having the same format as 7047 double. In earlier versions of GCC, they used the 68881 long double 7048 format instead. 7049 * The m68k-uclinux target now uses the same calling conventions as 7050 m68k-linux-gnu. You can select the original calling conventions by 7051 configuring for m68k-uclinuxoldabi instead. Note that 7052 m68k-uclinuxoldabi also retains the original 80-bit long double on 7053 ColdFire targets. 7054 * The -fforce-mem option has been removed because it has had no 7055 effect in the last few GCC releases. 7056 * The i386 -msvr3-shlib option has been removed since it is no longer 7057 used. 7058 * Fastcall for i386 has been changed not to pass aggregate arguments 7059 in registers, following Microsoft compilers. 7060 * Support for the AOF assembler has been removed from the ARM back 7061 end; this affects only the targets arm-semi-aof and armel-semi-aof, 7062 which are no longer recognized. We removed these targets without a 7063 deprecation period because we discovered that they have been 7064 unusable since GCC 4.0.0. 7065 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 7066 has been removed. This support had been deprecated since GCC 4.0.0. 7067 * Support for a number of older systems and recently unmaintained or 7068 untested target ports of GCC has been declared obsolete in GCC 4.3. 7069 Unless there is activity to revive them, the next release of GCC 7070 will have their sources permanently removed. 7071 All GCC ports for the following processor architectures have been 7072 declared obsolete: 7073 + Morpho MT (mt-*) 7074 The following aliases for processor architectures have been 7075 declared obsolete. Users should use the indicated generic target 7076 names instead, with compile-time options such as -mcpu or 7077 configure-time options such as --with-cpu to control the 7078 configuration more precisely. 7079 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 7080 instead). 7081 + parisc*-*-* (use hppa*-*-* instead). 7082 + m680[012]0-*-* (use m68k-*-* instead). 7083 All GCC ports for the following operating systems have been 7084 declared obsolete: 7085 + BeOS (*-*-beos*) 7086 + kaOS (*-*-kaos*) 7087 + GNU/Linux using the a.out object format (*-*-linux*aout*) 7088 + GNU/Linux using version 1 of the GNU C Library 7089 (*-*-linux*libc1*) 7090 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 7091 *-*-solaris2.[0-6].*) 7092 + Miscellaneous System V (*-*-sysv*) 7093 + WindISS (*-*-windiss*) 7094 Also, those for some individual systems on particular architectures 7095 have been obsoleted: 7096 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 7097 + CRIS with a.out object format (cris-*-aout) 7098 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 7099 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 7100 + PRO on PA-RISC (hppa1.1-*-pro*) 7101 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 7102 i[34567]86-sequent-sysv4*) 7103 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 7104 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 7105 was previously [5]removed in 2001, leaving only the support 7106 for UWIN as a target now being deprecated) 7107 + ChorusOS on PowerPC (powerpc-*-chorusos*) 7108 + All VAX configurations apart from NetBSD and OpenBSD 7109 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 7110 * The [6]-Wconversion option has been modified. Its purpose now is to 7111 warn for implicit conversions that may alter a value. This new 7112 behavior is available for both C and C++. Warnings about 7113 conversions between signed and unsigned integers can be disabled by 7114 using -Wno-sign-conversion. In C++, they are disabled by default 7115 unless -Wsign-conversion is explicitly requested. The old behavior 7116 of -Wconversion, that is, warn for prototypes causing a type 7117 conversion that is different from what would happen to the same 7118 argument in the absence of a prototype, has been moved to a new 7119 option -Wtraditional-conversion, which is only available for C. 7120 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 7121 been removed because they were deprecated for more than 3 GCC major 7122 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 7123 -mtune=pentiumpro as a replacement. 7124 * The -funsafe-math-optimizations option now automatically turns on 7125 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 7126 reassociation and thus may introduce or remove traps. 7127 * The -ftree-vectorize option is now on by default under -O3. In 7128 order to generate code for a SIMD extension, it has to be enabled 7129 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 7130 i?86 and x86_64. 7131 * More information on porting to GCC 4.3 from previous versions of 7132 GCC can be found in the [7]porting guide for this release. 7133 7134 General Optimizer Improvements 7135 7136 * The GCC middle-end has been integrated with the [8]MPFR library. 7137 This allows GCC to evaluate and replace at compile-time calls to 7138 built-in math functions having constant arguments with their 7139 mathematically equivalent results. In making use of [9]MPFR, GCC 7140 can generate correct results regardless of the math library 7141 implementation or floating point precision of the host platform. 7142 This also allows GCC to generate identical results regardless of 7143 whether one compiles in native or cross-compile configurations to a 7144 particular target. The following built-in functions take advantage 7145 of this new capability: acos, acosh, asin, asinh, atan2, atan, 7146 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 7147 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 7148 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 7149 tan, tanh, tgamma, y0, y1 and yn. The float and long double 7150 variants of these functions (e.g. sinf and sinl) are also handled. 7151 The sqrt and cabs functions with constant arguments were already 7152 optimized in prior GCC releases. Now they also use [10]MPFR. 7153 * A new forward propagation pass on RTL was added. The new pass 7154 replaces several slower transformations, resulting in compile-time 7155 improvements as well as better code generation in some cases. 7156 * A new command-line switch -frecord-gcc-switches has been added to 7157 GCC, although it is only enabled for some targets. The switch 7158 causes the command line that was used to invoke the compiler to be 7159 recorded into the object file that is being created. The exact 7160 format of this recording is target and binary file format 7161 dependent, but it usually takes the form of a note section 7162 containing ASCII text. The switch is related to the -fverbose-asm 7163 switch, but that one only records the information in the assembler 7164 output file as comments, so the information never reaches the 7165 object file. 7166 * The inliner heuristic is now aware of stack frame consumption. New 7167 command-line parameters --param large-stack-frame and --param 7168 large-stack-frame-growth can be used to limit stack frame size 7169 growth caused by inlining. 7170 * During feedback directed optimizations, the expected block size the 7171 memcpy, memset and bzero functions operate on is discovered and for 7172 cases of commonly used small sizes, specialized inline code is 7173 generated. 7174 * __builtin_expect no longer requires its argument to be a compile 7175 time constant. 7176 * Interprocedural optimization was reorganized to work on functions 7177 in SSA form. This enables more precise and cheaper dataflow 7178 analysis and makes writing interprocedural optimizations easier. 7179 The following improvements have been implemented on top of this 7180 framework: 7181 + Pre-inline optimization: Selected local optimization passes 7182 are run before the inliner (and other interprocedural passes) 7183 are executed. This significantly improves the accuracy of code 7184 growth estimates used by the inliner and reduces the overall 7185 memory footprint for large compilation units. 7186 + Early inlining (a simple bottom-up inliner pass inlining only 7187 functions whose body is smaller than the expected call 7188 overhead) is now executed with the early optimization passes, 7189 thus inlining already optimized function bodies into an 7190 unoptimized function that is subsequently optimized by early 7191 optimizers. This enables the compiler to quickly eliminate 7192 abstraction penalty in C++ programs. 7193 + Interprocedural constant propagation now operate on SSA form 7194 increasing accuracy of the analysis. 7195 * A new internal representation for GIMPLE statements has been 7196 contributed, resulting in compile-time memory savings. 7197 * The vectorizer was enhanced to support vectorization of outer 7198 loops, intra-iteration parallelism (loop-aware SLP), vectorization 7199 of strided accesses and loops with multiple data-types. Run-time 7200 dependency testing using loop versioning was added. The cost model, 7201 turned on by -fvect-cost-model, was developed. 7202 7203 New Languages and Language specific improvements 7204 7205 * We have added new command-line options 7206 -finstrument-functions-exclude-function-list and 7207 -finstrument-functions-exclude-file-list. They provide more control 7208 over which functions are annotated by the -finstrument-functions 7209 option. 7210 7211 C family 7212 7213 * Implicit conversions between generic vector types are now only 7214 permitted when the two vectors in question have the same number of 7215 elements and compatible element types. (Note that the restriction 7216 involves compatible element types, not implicitly-convertible 7217 element types: thus, a vector type with element type int may not be 7218 implicitly converted to a vector type with element type unsigned 7219 int.) This restriction, which is in line with specifications for 7220 SIMD architectures such as AltiVec, may be relaxed using the flag 7221 -flax-vector-conversions. This flag is intended only as a 7222 compatibility measure and should not be used for new code. 7223 * -Warray-bounds has been added and is now enabled by default for 7224 -Wall . It produces warnings for array subscripts that can be 7225 determined at compile time to be always out of bounds. 7226 -Wno-array-bounds will disable the warning. 7227 * The constructor and destructor function attributes now accept 7228 optional priority arguments which control the order in which the 7229 constructor and destructor functions are run. 7230 * New [11]command-line options -Wtype-limits, 7231 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body, 7232 -Wclobbered and -Wignored-qualifiers have been added for finer 7233 control of the diverse warnings enabled by -Wextra. 7234 * A new function attribute alloc_size has been added to mark up 7235 malloc style functions. For constant sized allocations this can be 7236 used to find out the size of the returned pointer using the 7237 __builtin_object_size() function for buffer overflow checking and 7238 similar. This supplements the already built-in malloc and calloc 7239 constant size handling. 7240 * Integer constants written in binary are now supported as a GCC 7241 extension. They consist of a prefix 0b or 0B, followed by a 7242 sequence of 0 and 1 digits. 7243 * A new predefined macro __COUNTER__ has been added. It expands to 7244 sequential integral values starting from 0. In conjunction with the 7245 ## operator, this provides a convenient means to generate unique 7246 identifiers. 7247 * A new command-line option -fdirectives-only has been added. It 7248 enables a special preprocessing mode which improves the performance 7249 of applications like distcc and ccache. 7250 * Fixed-point data types and operators have been added. They are 7251 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 7252 Currently, only MIPS targets are supported. 7253 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 7254 N1241, is now supported as a GCC extension to C for targets 7255 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 7256 and x86_64-*-linux-gnu. The feature introduces new data types 7257 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 7258 DD, and DL. 7259 7260 C++ 7261 7262 * [12]Experimental support for the upcoming ISO C++ standard, C++0x. 7263 * -Wc++0x-compat has been added and is now enabled by default for 7264 -Wall. It produces warnings for constructs whose meaning differs 7265 between ISO C++ 1998 and C++0x. 7266 * The -Wparentheses option now works for C++ as it does for C. It 7267 warns if parentheses are omitted when operators with confusing 7268 precedence are nested. It also warns about ambiguous else 7269 statements. Since -Wparentheses is enabled by -Wall, this may cause 7270 additional warnings with existing C++ code which uses -Wall. These 7271 new warnings may be disabled by using -Wall -Wno-parentheses. 7272 * The -Wmissing-declarations now works for C++ as it does for C. 7273 * The -fvisibility-ms-compat flag was added, to make it easier to 7274 port larger projects using shared libraries from Microsoft's Visual 7275 Studio to ELF and Mach-O systems. 7276 * C++ attribute handling has been overhauled for template arguments 7277 (ie dependent types). In particular, __attribute__((aligned(T))); 7278 works for C++ types. 7279 7280 Runtime Library (libstdc++) 7281 7282 * [13]Experimental support for the upcoming ISO C++ standard, C++0x. 7283 * Support for TR1 mathematical special functions and regular 7284 expressions. 7285 * Default what implementations give more elaborate exception strings 7286 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 7287 * Header dependencies have been streamlined, reducing unnecessary 7288 includes and pre-processed bloat. 7289 * Variadic template implementations of items in <tuple> and 7290 <functional>. 7291 * An experimental [14]parallel mode has been added. This is a 7292 parallel implementation of many C++ Standard library algorithms, 7293 like std::accumulate, std::for_each, std::transform, or std::sort, 7294 to give but four examples. These algorithms can be substituted for 7295 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 7296 or all existing algorithms can be transformed via the 7297 -D_GLIBCXX_PARALLEL macro. 7298 * Debug mode versions of classes in <unordered_set> and 7299 <unordered_map>. 7300 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 7301 now <backward/hash_set> and <backward/hash_map>. This code: 7302 #include <ext/hash_set> 7303 __gnu_cxx::hash_set<int> s; 7304 7305 Can be transformed (in order of preference) to: 7306 #include <tr1/unordered_set> 7307 std::tr1::unordered_set<int> s; 7308 7309 or 7310 #include <backward/hash_set> 7311 __gnu_cxx::hash_set<int> s; 7312 7313 Similar transformations apply to __gnu_cxx::hash_map, 7314 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 7315 __gnu_cxx::hash_multiset. 7316 7317 Fortran 7318 7319 * Due to the fact that the GMP and MPFR libraries are required for 7320 all languages, Fortran is no longer special in this regard and is 7321 available by default. 7322 * The [15]-fexternal-blas option has been added, which generates 7323 calls to BLAS routines for intrinsic matrix operations such as 7324 matmul rather than using the built-in algorithms. 7325 * Support to give a backtrace (compiler flag -fbacktrace or 7326 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 7327 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 7328 run-time error occured. 7329 * GNU Fortran now defines __GFORTRAN__ when it runs the C 7330 preprocessor (CPP). 7331 * The [16]-finit-local-zero, -finit-real, -finit-integer, 7332 -finit-character, and -finit-logical options have been added, which 7333 can be used to initialize local variables. 7334 * The intrinsic procedures [17]GAMMA and [18]LGAMMA have been added, 7335 which calculate the Gamma function and its logarithm. Use EXTERNAL 7336 gamma if you want to use your own gamma function. 7337 * GNU Fortran now regards the backslash character as literal (as 7338 required by the Fortran 2003 standard); using [19]-fbackslash GNU 7339 Fortran interprets backslashes as C-style escape characters. 7340 * The [20]interpretation of binary, octal and hexadecimal (BOZ) 7341 literal constants has been changed. Before they were always 7342 interpreted as integer; now they are bit-wise transferred as 7343 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 7344 2003 standard, and for real and complex variables in DATA 7345 statements or when directly assigned to real and complex variables. 7346 Everywhere else and especially in expressions they are still 7347 regarded as integer constants. 7348 * Fortran 2003 support has been extended: 7349 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 7350 + Pointer intent 7351 + Intrinsic module ISO_ENV_FORTRAN 7352 + Interoperability with C (ISO C Bindings) 7353 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 7354 attribute) 7355 + Fortran 2003 BOZ 7356 7357 Java (GCJ) 7358 7359 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 7360 This enables the use of all 1.5 language features, and fixes most 7361 existing front end bugs. 7362 * libgcj now supports all 1.5 language features which require runtime 7363 support: foreach, enum, annotations, generics, and auto-boxing. 7364 * We've made many changes to the tools shipped with gcj. 7365 + The old jv-scan tool has been removed. This tool never really 7366 worked properly. There is no replacement. 7367 + gcjh has been rewritten. Some of its more obscure options no 7368 longer work, but are still recognized in an attempt at 7369 compatibility. gjavah is a new program with similar 7370 functionality but different command-line options. 7371 + grmic and grmiregistry have been rewritten. grmid has been 7372 added. 7373 + gjar replaces the old fastjar. 7374 + gjarsigner (used for signing jars), gkeytool (used for key 7375 management), gorbd (for CORBA), gserialver (computes 7376 serialization UIDs), and gtnameserv (also for CORBA) are now 7377 installed. 7378 * The ability to dump the contents of the java run time heap to a 7379 file for off-line analysis has been added. The heap dumps may be 7380 analyzed with the new gc-analyze tool. They may be generated on 7381 out-of-memory conditions or on demand and are controlled by the new 7382 run time class gnu.gcj.util.GCInfo. 7383 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 7384 provide correct, updated, timezone information. This means that 7385 packagers no longer have to update libgcj when a time zone change 7386 is published. 7387 7388 New Targets and Target Specific Improvements 7389 7390 IA-32/x86-64 7391 7392 * Tuning for Intel Core 2 processors is available via -mtune=core2 7393 and -march=core2. 7394 * Tuning for AMD Geode processors is available via -mtune=geode and 7395 -march=geode. 7396 * Code generation of block move (memcpy) and block set (memset) was 7397 rewritten. GCC can now pick the best algorithm (loop, unrolled 7398 loop, instruction with rep prefix or a library call) based on the 7399 size of the block being copied and the CPU being optimized for. A 7400 new option -minline-stringops-dynamically has been added. With this 7401 option string operations of unknown size are expanded such that 7402 small blocks are copied by in-line code, while for large blocks a 7403 library call is used. This results in faster code than 7404 -minline-all-stringops when the library implementation is capable 7405 of using cache hierarchy hints. The heuristic choosing the 7406 particular algorithm can be overwritten via -mstringop-strategy. 7407 Newly also memset of values different from 0 is inlined. 7408 * GCC no longer places the cld instruction before string operations. 7409 Both i386 and x86-64 ABI documents mandate the direction flag to be 7410 clear at the entry of a function. It is now invalid to set the flag 7411 in asm statement without reseting it afterward. 7412 * Support for SSSE3 built-in functions and code generation are 7413 available via -mssse3. 7414 * Support for SSE4.1 built-in functions and code generation are 7415 available via -msse4.1. 7416 * Support for SSE4.2 built-in functions and code generation are 7417 available via -msse4.2. 7418 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 7419 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 7420 allow explicit control of x87 floating point precision. 7421 * Support for __float128 (TFmode) IEEE quad type and corresponding 7422 TCmode IEEE complex quad type is available via the soft-fp library 7423 on x86_64 targets. This includes basic arithmetic operations 7424 (addition, subtraction, negation, multiplication and division) on 7425 __float128 real and TCmode complex values, the full set of IEEE 7426 comparisons between __float128 values, conversions to and from 7427 float, double and long double floating point types, as well as 7428 conversions to and from signed or unsigned integer, signed or 7429 unsigned long integer and signed or unsigned quad (TImode) integer 7430 types. Additionally, all operations generate the full set of IEEE 7431 exceptions and support the full set of IEEE rounding modes. 7432 * GCC can now utilize the ACML library for vectorizing calls to a set 7433 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 7434 link to an ACML ABI compatible library. 7435 7436 ARM 7437 7438 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 7439 has been added. 7440 7441 CRIS 7442 7443 New features 7444 7445 * Compiler and Library support for the CRIS v32 architecture, as 7446 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 7447 added. 7448 7449 Configuration changes 7450 7451 * The cris-*-elf target now includes support for CRIS v32, including 7452 libraries, through the -march=v32 option. 7453 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 7454 * A new crisv32-*-linux* target defaults to generate code for CRIS 7455 v32. 7456 * The cris-*-aout target has been obsoleted. 7457 7458 Improved support for built-in functions 7459 7460 * GCC can now use the lz and swapwbr instructions to implement the 7461 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 7462 * __builtin_bswap32 is now implemented using the swapwb instruction, 7463 when available. 7464 7465 m68k and ColdFire 7466 7467 New features 7468 7469 * Support for several new ColdFire processors has been added. You can 7470 generate code for them using the new -mcpu option. 7471 * All targets now support ColdFire processors. 7472 * m68k-uclinux targets have improved support for C++ constructors and 7473 destructors, and for shared libraries. 7474 * It is now possible to set breakpoints on the first or last line of 7475 a function, even if there are no statements on that line. 7476 7477 Optimizations 7478 7479 * Support for sibling calls has been added. 7480 * More use is now made of the ColdFire mov3q instruction. 7481 * __builtin_clz is now implemented using the ff1 ColdFire 7482 instruction, when available. 7483 * GCC now honors the -m68010 option. 68010 code now uses clr rather 7484 than move to zero volatile memory. 7485 * 68020 targets and above can now use symbol(index.size*scale) 7486 addresses for indexed array accesses. Earlier compilers would 7487 always load the symbol into a base register first. 7488 7489 Configuration changes 7490 7491 * All m68k and ColdFire targets now allow the default processor to be 7492 set at configure time using --with-cpu. 7493 * A --with-arch configuration option has been added. This option 7494 allows you to restrict a target to ColdFire or non-ColdFire 7495 processors. 7496 7497 Preprocessor macros 7498 7499 * An __mcfv*__ macro is now defined for all ColdFire targets. 7500 (Earlier versions of GCC only defined __mcfv4e__.) 7501 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 7502 * All targets now define __mc68010 and __mc68010__ when generating 7503 68010 code. 7504 7505 Command-line changes 7506 7507 * New command-line options -march, -mcpu, -mtune and -mhard-float 7508 have been added. These options apply to both m68k and ColdFire 7509 targets. 7510 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 7511 versions of -mshort, etc. 7512 * -fforce-addr has been removed. It is now ignored by the compiler. 7513 7514 Other improvements 7515 7516 * ColdFire targets now try to maintain a 4-byte-aligned stack where 7517 possible. 7518 * m68k-uclinux targets now try to avoid situations that lead to the 7519 load-time error: BINFMT_FLAT: reloc outside program. 7520 7521 MIPS 7522 7523 Changes to existing configurations 7524 7525 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 7526 and n64. Every GNU/Linux configuration now builds these libraries 7527 by default. 7528 * GNU/Linux configurations now generate -mno-shared code unless 7529 overridden by -fpic, -fPIC, -fpie or -fPIE. 7530 * mipsisa32*-linux-gnu configurations now generate hard-float code by 7531 default, just like other mipsisa32* and mips*-linux-gnu 7532 configurations. You can build a soft-float version of any 7533 mips*-linux-gnu configuration by passing --with-float=soft to 7534 configure. 7535 * mips-wrs-vxworks now supports run-time processes (RTPs). 7536 7537 Changes to existing command-line options 7538 7539 * The -march and -mtune options no longer accept 24k as a processor 7540 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 7541 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 7542 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 7543 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 7544 24kx, 24kex and 34kx. 7545 7546 New configurations 7547 7548 GCC now supports the following configurations: 7549 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 7550 default. Earlier releases also recognized this configuration, but 7551 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 7552 you can customize any mips*-linux-gnu* configuration to a 7553 particular ISA or processor by passing an appropriate --with-arch 7554 option to configure. 7555 * mipsisa*-sde-elf*, which provides compatibility with MIPS 7556 Technologies' SDE toolchains. The configuration uses the SDE 7557 libraries by default, but you can use it like other newlib-based 7558 ELF configurations by passing --with-newlib to configure. It is the 7559 only configuration besides mips64vr*-elf* to build MIPS16 as well 7560 as non-MIPS16 libraries. 7561 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 7562 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 7563 and 64-bit forms of the EABI. 7564 7565 New processors and application-specific extensions 7566 7567 * Support for the SmartMIPS ASE is available through the new 7568 -msmartmips option. 7569 * Support for revision 2 of the DSP ASE is available through the new 7570 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 7571 indicates the revision of the ASE in use. 7572 * Support for the 4KS and 74K families of processors is available 7573 through the -march and -mtune options. 7574 7575 Improved support for built-in functions 7576 7577 * GCC can now use load-linked, store-conditional and sync 7578 instructions to implement atomic built-in functions such as 7579 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 7580 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 7581 * GCC can now use the clz and dclz instructions to implement the 7582 __builtin_ctz and __builtin_ffs families of functions. 7583 * There is a new __builtin___clear_cache function for flushing the 7584 instruction cache. GCC expands this function inline on MIPS32 7585 revision 2 targets, otherwise it calls the function specified by 7586 -mcache-flush-func. 7587 7588 MIPS16 improvements 7589 7590 * GCC can now compile objects that contain a mixture of MIPS16 and 7591 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 7592 for specifying which mode a function should use. 7593 * A new option called -minterlink-mips16 makes non-MIPS16 code 7594 link-compatible with MIPS16 code. 7595 * After many bug fixes, the long-standing MIPS16 -mhard-float support 7596 should now work fairly reliably. 7597 * GCC can now use the MIPS16e save and restore instructions. 7598 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 7599 with -G0 -fsection-anchors is often smaller than code compiled with 7600 -G8. However, please note that you must usually compile all objects 7601 in your application with the same -G option; see the documentation 7602 of -G for details. 7603 * A new option called-mcode-readable specifies which instructions are 7604 allowed to load from the code segment. -mcode-readable=yes is the 7605 default and says that any instruction may load from the code 7606 segment. The other alternatives are -mcode-readable=pcrel, which 7607 says that only PC-relative MIPS16 instructions may load from the 7608 code segment, and -mcode-readable=no, which says that no 7609 instruction may do so. Please see the documentation for more 7610 details, including example uses. 7611 7612 Small-data improvements 7613 7614 There are three new options for controlling small data: 7615 * -mno-extern-sdata, which disables small-data accesses for 7616 externally-defined variables. Code compiled with -Gn 7617 -mno-extern-sdata will be link-compatible with any -G setting 7618 between -G0 and -Gn inclusive. 7619 * -mno-local-sdata, which disables the use of small-data sections for 7620 data that is not externally visible. This option can be a useful 7621 way of reducing small-data usage in less performance-critical parts 7622 of an application. 7623 * -mno-gpopt, which disables the use of the $gp register while still 7624 honoring the -G limit when placing externally-visible data. This 7625 option implies -mno-extern-sdata and -mno-local-sdata and it can be 7626 useful in situations where $gp does not necessarily hold the 7627 expected value. 7628 7629 Miscellaneous improvements 7630 7631 * There is a new option called -mbranch-cost for tweaking the 7632 perceived cost of branches. 7633 * If GCC is configured to use a version of GAS that supports the 7634 .gnu_attribute directive, it will use that directive to record 7635 certain properties of the output code. .gnu_attribute is new to GAS 7636 2.18. 7637 * There are two new function attributes, near and far, for overriding 7638 the command-line setting of -mlong-calls on a function-by-function 7639 basis. 7640 * -mfp64, which previously required a 64-bit target, now works with 7641 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 7642 mipsisa*-sde-elf* configurations provide suitable library support. 7643 * GCC now recognizes the -mdmx and -mmt options and passes them down 7644 to the assembler. It does nothing else with the options at present. 7645 7646 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 7647 (BEA) 7648 7649 * Support has been added for this new architecture. 7650 7651 RS6000 (POWER/PowerPC) 7652 7653 * Support for the PowerPC 750CL paired-single instructions has been 7654 added with a new powerpc-*-linux*paired* target configuration. It 7655 is enabled by an associated -mpaired option and can be accessed 7656 using new built-in functions. 7657 * Support for auto-detecting architecture and system configuration to 7658 auto-select processor optimization tuning. 7659 * Support for VMX on AIX 5.3 has been added. 7660 * Support for AIX Version 6.1 has been added. 7661 7662 S/390, zSeries and System z9 7663 7664 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 7665 added. When using the -march=z9-ec option, the compiler will 7666 generate code making use of instructions provided by the decimal 7667 floating point facility and the floating point conversion facility 7668 (pfpo). Besides the instructions used to implement decimal floating 7669 point operations these facilities also contain instructions to move 7670 between general purpose and floating point registers and to modify 7671 and copy the sign-bit of floating point values. 7672 * When the -march=z9-ec option is used the new 7673 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 7674 decimal floating point hardware instructions will be used or not. 7675 If none of them is given the hardware support is enabled by 7676 default. 7677 * The -mstack-guard option can now be omitted when using stack 7678 checking via -mstack-size in order to let GCC choose a sensible 7679 stack guard value according to the frame size of each function. 7680 * Various changes to improve performance of generated code have been 7681 implemented, including: 7682 + The condition code set by an add logical with carry 7683 instruction is now available for overflow checks like: a + b + 7684 carry < b. 7685 + The test data class instruction is now used to implement 7686 sign-bit and infinity checks of binary and decimal floating 7687 point numbers. 7688 7689 SPARC 7690 7691 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 7692 added. 7693 7694 Xtensa 7695 7696 * Stack unwinding for exception handling now uses by default a 7697 specialized version of DWARF unwinding. This is not 7698 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 7699 Xtensa with previous versions of GCC. 7700 * For Xtensa processors that include the Conditional Store option, 7701 the built-in functions for atomic memory access are now implemented 7702 using S32C1I instructions. 7703 * If the Xtensa NSA option is available, GCC will use it to implement 7704 the __builtin_ctz and __builtin_clz functions. 7705 7706 Documentation improvements 7707 7708 * Existing libstdc++ documentation has been edited and restructured 7709 into a single DocBook XML manual. The results can be viewed online 7710 [21]here. 7711 7712 Other significant improvements 7713 7714 * The compiler's --help command-line option has been extended so that 7715 it now takes an optional set of arguments. These arguments restrict 7716 the information displayed to specific classes of command-line 7717 options, and possibly only a subset of those options. It is also 7718 now possible to replace the descriptive text associated with each 7719 displayed option with an indication of its current value, or for 7720 binary options, whether it has been enabled or disabled. 7721 Here are some examples. The following will display all the options 7722 controlling warning messages: 7723 --help=warnings 7724 7725 Whereas this will display all the undocumented, target specific 7726 options: 7727 --help=target,undocumented 7728 7729 This sequence of commands will display the binary optimizations 7730 that are enabled by -O3: 7731 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 7732 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 7733 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 7734 7735 * The configure options --with-pkgversion and --with-bugurl have been 7736 added. These allow distributors of GCC to include a 7737 distributor-specific string in manuals and --version output and to 7738 specify the URL for reporting bugs in their versions of GCC. 7739 7740 GCC 4.3.1 7741 7742 This is the [22]list of problem reports (PRs) from GCC's bug tracking 7743 system that are known to be fixed in the 4.3.1 release. This list might 7744 not be complete (that is, it is possible that some PRs that have been 7745 fixed are not listed here). 7746 7747 Target Specific Changes 7748 7749 IA-32/x86-64 7750 7751 ABI changes 7752 7753 * Starting with GCC 4.3.1, decimal floating point variables are 7754 aligned to their natural boundaries when they are passed on the 7755 stack for i386. 7756 7757 Command-line changes 7758 7759 * Starting with GCC 4.3.1, the -mcld option has been added to 7760 automatically generate a cld instruction in the prologue of 7761 functions that use string instructions. This option is used for 7762 backward compatibility on some operating systems and can be enabled 7763 by default for 32-bit x86 targets by configuring GCC with the 7764 --enable-cld configure option. 7765 7766 GCC 4.3.2 7767 7768 This is the [23]list of problem reports (PRs) from GCC's bug tracking 7769 system that are known to be fixed in the 4.3.2 release. This list might 7770 not be complete (that is, it is possible that some PRs that have been 7771 fixed are not listed here). 7772 7773 GCC 4.3.3 7774 7775 This is the [24]list of problem reports (PRs) from GCC's bug tracking 7776 system that are known to be fixed in the 4.3.3 release. This list might 7777 not be complete (that is, it is possible that some PRs that have been 7778 fixed are not listed here). 7779 7780 GCC 4.3.4 7781 7782 This is the [25]list of problem reports (PRs) from GCC's bug tracking 7783 system that are known to be fixed in the 4.3.4 release. This list might 7784 not be complete (that is, it is possible that some PRs that have been 7785 fixed are not listed here). 7786 7787 GCC 4.3.5 7788 7789 This is the [26]list of problem reports (PRs) from GCC's bug tracking 7790 system that are known to be fixed in the 4.3.5 release. This list might 7791 not be complete (that is, it is possible that some PRs that have been 7792 fixed are not listed here). 7793 7794 GCC 4.3.6 7795 7796 This is the [27]list of problem reports (PRs) from GCC's bug tracking 7797 system that are known to be fixed in the 4.3.6 release. This list might 7798 not be complete (that is, it is possible that some PRs that have been 7799 fixed are not listed here). 7800 7801 7802 For questions related to the use of GCC, please consult these web 7803 pages and the [28]GCC manuals. If that fails, the 7804 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7805 web pages and the development of GCC are welcome on our developer 7806 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 7807 archives. 7808 7809 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 7810 distribution of this entire article is permitted in any medium, 7811 provided this notice is preserved. 7812 7813 These pages are [33]maintained by the GCC team. Last modified 7814 2017-02-07[34]. 7815 7816 References 7817 7818 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 7819 2. https://gmplib.org/ 7820 3. http://www.mpfr.org/ 7821 4. https://gcc.gnu.org/install/prerequisites.html 7822 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 7823 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 7824 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 7825 8. http://www.mpfr.org/ 7826 9. http://www.mpfr.org/ 7827 10. http://www.mpfr.org/ 7828 11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 7829 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 7830 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 7831 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 7832 15. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 7833 16. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 7834 17. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 7835 18. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 7836 19. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 7837 20. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 7838 21. https://gcc.gnu.org/onlinedocs/libstdc++/ 7839 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 7840 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 7841 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 7842 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 7843 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 7844 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 7845 28. https://gcc.gnu.org/onlinedocs/ 7846 29. mailto:gcc-help (a] gcc.gnu.org 7847 30. mailto:gcc (a] gcc.gnu.org 7848 31. https://gcc.gnu.org/lists.html 7849 32. http://www.fsf.org/ 7850 33. https://gcc.gnu.org/about.html 7851 34. http://validator.w3.org/check/referer 7852 ====================================================================== 7853 http://gcc.gnu.org/gcc-4.2/index.html 7854 7855 GCC 4.2 Release Series 7856 7857 May 19, 2008 7858 7859 The [1]GNU project and the GCC developers are pleased to announce the 7860 release of GCC 4.2.4. 7861 7862 This release is a bug-fix release, containing fixes for regressions in 7863 GCC 4.2.3 relative to previous releases of GCC. 7864 7865 This release series is no longer maintained. 7866 7867 Release History 7868 7869 GCC 4.2.4 7870 May 19, 2008 ([2]changes) 7871 7872 GCC 4.2.3 7873 February 1, 2008 ([3]changes) 7874 7875 GCC 4.2.2 7876 October 7, 2007 ([4]changes) 7877 7878 GCC 4.2.1 7879 July 18, 2007 ([5]changes) 7880 7881 GCC 4.2.0 7882 May 13, 2007 ([6]changes) 7883 7884 References and Acknowledgements 7885 7886 GCC used to stand for the GNU C Compiler, but since the compiler 7887 supports several other languages aside from C, it now stands for the 7888 GNU Compiler Collection. 7889 7890 A list of [7]successful builds is updated as new information becomes 7891 available. 7892 7893 The GCC developers would like to thank the numerous people that have 7894 contributed new features, improvements, bug fixes, and other changes as 7895 well as test results to GCC. This [8]amazing group of volunteers is 7896 what makes GCC successful. 7897 7898 For additional information about GCC please refer to the [9]GCC project 7899 web site or contact the [10]GCC development mailing list. 7900 7901 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 7902 7903 7904 For questions related to the use of GCC, please consult these web 7905 pages and the [13]GCC manuals. If that fails, the 7906 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7907 web pages and the development of GCC are welcome on our developer 7908 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 7909 archives. 7910 7911 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 7912 distribution of this entire article is permitted in any medium, 7913 provided this notice is preserved. 7914 7915 These pages are [18]maintained by the GCC team. Last modified 7916 2016-09-30[19]. 7917 7918 References 7919 7920 1. http://www.gnu.org/ 7921 2. http://gcc.gnu.org/gcc-4.2/changes.html 7922 3. http://gcc.gnu.org/gcc-4.2/changes.html 7923 4. http://gcc.gnu.org/gcc-4.2/changes.html 7924 5. http://gcc.gnu.org/gcc-4.2/changes.html 7925 6. http://gcc.gnu.org/gcc-4.2/changes.html 7926 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 7927 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7928 9. http://gcc.gnu.org/index.html 7929 10. mailto:gcc (a] gcc.gnu.org 7930 11. http://gcc.gnu.org/mirrors.html 7931 12. http://gcc.gnu.org/svn.html 7932 13. https://gcc.gnu.org/onlinedocs/ 7933 14. mailto:gcc-help (a] gcc.gnu.org 7934 15. mailto:gcc (a] gcc.gnu.org 7935 16. https://gcc.gnu.org/lists.html 7936 17. http://www.fsf.org/ 7937 18. https://gcc.gnu.org/about.html 7938 19. http://validator.w3.org/check/referer 7939 ====================================================================== 7940 http://gcc.gnu.org/gcc-4.2/changes.html 7941 7942 GCC 4.2 Release Series 7943 Changes, New Features, and Fixes 7944 7945 Caveats 7946 7947 * GCC no longer accepts the -fshared-data option. This option has had 7948 no effect in any GCC 4 release; the targets to which the option 7949 used to apply had been removed before GCC 4.0. 7950 7951 General Optimizer Improvements 7952 7953 * New command-line options specify the possible relationships among 7954 parameters and between parameters and global data. For example, 7955 -fargument-noalias-anything specifies that arguments do not alias 7956 any other storage. 7957 Each language will automatically use whatever option is required by 7958 the language standard. You should not need to use these options 7959 yourself. 7960 7961 New Languages and Language specific improvements 7962 7963 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 7964 * New command-line options -fstrict-overflow and -Wstrict-overflow 7965 have been added. -fstrict-overflow tells the compiler that it may 7966 assume that the program follows the strict signed overflow 7967 semantics permitted for the language: for C and C++ this means that 7968 the compiler may assume that signed overflow does not occur. For 7969 example, a loop like 7970 for (i = 1; i > 0; i *= 2) 7971 7972 is presumably intended to continue looping until i overflows. With 7973 -fstrict-overflow, the compiler may assume that signed overflow 7974 will not occur, and transform this into an infinite loop. 7975 -fstrict-overflow is turned on by default at -O2, and may be 7976 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 7977 be used to warn about cases where the compiler assumes that signed 7978 overflow will not occur. It takes five different levels: 7979 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 7980 -Wstrict-overflow=1 is enabled by -Wall. 7981 * The new command-line option -fno-toplevel-reorder directs GCC to 7982 emit top-level functions, variables, and asm statements in the same 7983 order that they appear in the input file. This is intended to 7984 support existing code which relies on a particular ordering (for 7985 example, code which uses top-level asm statements to switch 7986 sections). For new code, it is generally better to use function and 7987 variable attributes. The -fno-toplevel-reorder option may be used 7988 for most cases which currently use -fno-unit-at-a-time. The 7989 -fno-unit-at-a-time option will be removed in some future version 7990 of GCC. If you know of a case which requires -fno-unit-at-a-time 7991 which is not fixed by -fno-toplevel-reorder, please open a bug 7992 report. 7993 7994 C family 7995 7996 * The pragma redefine_extname will now macro expand its tokens for 7997 compatibility with SunPRO. 7998 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 7999 GCC to handle inline functions as specified in the C99 standard. In 8000 preparation for this, GCC 4.2 will warn about any use of non-static 8001 inline functions in gnu99 or c99 mode. This new warning may be 8002 disabled with the new gnu_inline function attribute or the new 8003 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 8004 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 8005 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 8006 in the current compilation. 8007 * A new command-line option -Waddress has been added to warn about 8008 suspicious uses of memory addresses as, for example, using the 8009 address of a function in a conditional expression, and comparisons 8010 against the memory address of a string literal. This warning is 8011 enabled by -Wall. 8012 8013 C++ 8014 8015 * C++ visibility handling has been overhauled. 8016 Restricted visiblity is propagated from classes to members, from 8017 functions to local statics, and from templates and template 8018 arguments to instantiations, unless the latter has explicitly 8019 declared visibility. 8020 The visibility attribute for a class must come between the 8021 class-key and the name, not after the closing brace. 8022 Attributes are now allowed for enums and elaborated-type-specifiers 8023 that only declare a type. 8024 Members of the anonymous namespace are now local to a particular 8025 translation unit, along with any other declarations which use them, 8026 though they are still treated as having external linkage for 8027 language semantics. 8028 * The (undocumented) extension which permitted templates with default 8029 arguments to be bound to template template parameters with fewer 8030 parameters has been removed. For example: 8031 template <template <typename> class C> 8032 void f(C<double>) {} 8033 8034 template <typename T, typename U = int> 8035 struct S {}; 8036 8037 template void f(S<double>); 8038 8039 is no longer accepted by G++. The reason this code is not accepted 8040 is that S is a template with two parameters; therefore, it cannot 8041 be bound to C which has only one parameter. 8042 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 8043 releases, have been removed. 8044 * The command-line option -fconst-strings, deprecated in previous GCC 8045 releases, has been removed. 8046 * The configure variable enable-__cxa_atexit is now enabled by 8047 default for more targets. Enabling this variable is necessary in 8048 order for static destructors to be executed in the correct order, 8049 but it depends upon the presence of a non-standard C library in the 8050 target library in order to work. The variable is now enabled for 8051 more targets which are known to have suitable C libraries. 8052 * -Wextra will produce warnings for if statements with a semicolon as 8053 the only body, to catch code like: 8054 if (a); 8055 return 1; 8056 return 0; 8057 8058 To suppress the warning in valid cases, use { } instead. 8059 * The C++ frontend now also produces strict aliasing warnings when 8060 -fstrict-aliasing -Wstrict-aliasing is in effect. 8061 8062 Runtime Library (libstdc++) 8063 8064 * Added support for TR1 <random>, <complex>, and C compatibility 8065 headers. In addition, a lock-free version of shared_ptr was 8066 contributed as part of Phillip Jordan's Google Summer of Code 8067 project on lock-free containers. 8068 * In association with the Summer of Code work on lock-free 8069 containers, the interface for atomic builtins was adjusted, 8070 creating simpler alternatives for non-threaded code paths. Also, 8071 usage was consolidated and all elements were moved from namespace 8072 std to namespace__gnu_cxx. Affected interfaces are the functions 8073 __exchange_and_add, __atomic_add, and the objects __mutex, 8074 __recursive_mutex, and __scoped_lock. 8075 * Support for versioning weak symbol names via namespace association 8076 was added. However, as this changes the names of exported symbols, 8077 this is turned off by default in the current ABI. Intrepid users 8078 can enable this feature by using 8079 --enable-symvers=gnu-versioned-namespace during configuration. 8080 * Revised, simplified, and expanded policy-based associative 8081 containers, including data types for tree and trie forms 8082 (basic_tree, tree, trie), lists (list_update), and both 8083 collision-chaining and probing hash-based containers 8084 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 8085 the [3]documentation. 8086 * The implementation of the debug mode was modified, whereby the 8087 debug namespaces were nested inside of namespace std and namespace 8088 __gnu_cxx in order to resolve some long standing corner cases 8089 involving name lookup. Debug functionality from the policy-based 8090 data structures was consolidated and enabled with the single macro, 8091 _GLIBCXX_DEBUG. See PR 26142 for more information. 8092 * Added extensions for type traits: __conditional_type, 8093 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 8094 * Added a typelist implementation for compile-time meta-programming. 8095 Elements for typelist construction and operation can be found 8096 within namespace __gnu_cxx::typelist. 8097 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 8098 exception-safety. 8099 * Enabled library-wide visibility control, allowing -fvisibility to 8100 be used. 8101 * Consolidated all nested namespaces and the conversion of 8102 __gnu_internal implementation-private details to anonymous 8103 namespaces whenever possible. 8104 * Implemented LWG resolutions DR 431 and DR 538. 8105 8106 Fortran 8107 8108 * Support for allocatable components has been added (TR 15581 and 8109 Fortran 2003). 8110 * Support for the Fortran 2003 streaming IO extension has been added. 8111 * The GNU Fortran compiler now uses 4-byte record markers by default 8112 for unformatted files to be compatible with g77 and most other 8113 compilers. The implementation allows for records greater than 2 GB 8114 and is compatible with several other compilers. Older versions of 8115 gfortran used 8-byte record markers by default (on most systems). 8116 In order to change the length of the record markers, e.g. to read 8117 unformatted files created by older gfortran versions, the 8118 [4]-frecord-marker=8 option can be used. 8119 8120 Java (GCJ) 8121 8122 * A new command-line option -static-libgcj has been added for targets 8123 that use a linker compatible with GNU Binutils. As its name 8124 implies, this causes libgcj to be linked statically. In some cases 8125 this causes the resulting executable to start faster and use less 8126 memory than if the shared version of libgcj were used. However 8127 caution should be used as it can also cause essential parts of the 8128 library to be omitted. Some of these issues are discussed in: 8129 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 8130 * fastjar is no longer bundled with GCC. To build libgcj, you will 8131 need either InfoZIP (both zip and unzip) or an external jar 8132 program. In the former case, the GCC build will install a jar shell 8133 script that is based on InfoZIP and provides the same functionality 8134 as fastjar. 8135 8136 New Targets and Target Specific Improvements 8137 8138 IA-32/x86-64 8139 8140 * -mtune=generic can now be used to generate code running well on 8141 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 8142 Pentium-M, Intel Pentium 4 and Intel Core 2. 8143 * -mtune=native and -march=native will produce code optimized for the 8144 host architecture as detected using the cpuid instruction. 8145 * Added a new command-line option -fstackrealign and and 8146 __attribute__ ((force_align_arg_pointer)) to realign the stack at 8147 runtime. This allows functions compiled with a vector-aligned stack 8148 to be invoked from legacy objects that keep only word-alignment. 8149 8150 SPARC 8151 8152 * The default CPU setting has been changed from V7 to V9 in 32-bit 8153 mode on Solaris 7 and above. This is already the case in 64-bit 8154 mode. It can be overridden by specifying --with-cpu at configure 8155 time. 8156 * Back-end support of built-in functions for atomic memory access has 8157 been implemented. 8158 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 8159 added. 8160 8161 M32C 8162 8163 * Various bug fixes have made some functions (notably, functions 8164 returning structures) incompatible with previous releases. 8165 Recompiling all libraries is recommended. Note that code quality 8166 has considerably improved since 4.1, making a recompile even more 8167 beneficial. 8168 8169 MIPS 8170 8171 * Added support for the Broadcom SB-1A core. 8172 8173 IA-64 8174 8175 * Added support for IA-64 data and control speculation. By default 8176 speculation is enabled only during second scheduler pass. A number 8177 of machine flags was introduced to control the usage of speculation 8178 for both scheduler passes. 8179 8180 HPPA 8181 8182 * Added Java language support (libffi and libjava) for 32-bit HP-UX 8183 11 target. 8184 8185 Obsolete Systems 8186 8187 Documentation improvements 8188 8189 PDF Documentation 8190 8191 * A make pdf target has been added to the top-level makefile, 8192 enabling automated production of PDF documentation files. 8193 (Front-ends external to GCC should modify their Make-lang.in file 8194 to add a lang.pdf: target.) 8195 8196 Other significant improvements 8197 8198 Build system improvements 8199 8200 * All the components of the compiler are now bootstrapped by default. 8201 This improves the resilience to bugs in the system compiler or 8202 binary compatibility problems, as well as providing better testing 8203 of GCC 4.2 itself. In addition, if you build the compiler from a 8204 combined tree, the assembler, linker, etc. will also be 8205 bootstrapped (i.e. built with themselves). 8206 You can disable this behavior, and go back to the pre-GCC 4.2 set 8207 up, by configuring GCC with --disable-bootstrap. 8208 * The rules that configure follows to find target tools resemble more 8209 closely the locations that the built compiler will search. In 8210 addition, you can use the new configure option --with-target-tools 8211 to specify where to find the target tools used during the build, 8212 without affecting what the built compiler will use. 8213 This can be especially useful when building packages of GCC. For 8214 example, you may want to build GCC with GNU as or ld, even if the 8215 resulting compiler to work with the native assembler and linker. To 8216 do so, you can use --with-target-tools to point to the native 8217 tools. 8218 8219 Incompatible changes to the build system 8220 8221 * Front-ends external to GCC should modify their Make-lang.in file to 8222 replace double-colon rules (e.g. dvi::) with normal rules (like 8223 lang.dvi:). Front-end makefile hooks do not use double-colon rules 8224 anymore. 8225 * Up to GCC 4.1, a popular way to specify the target tools used 8226 during the build was to create directories named gas, binutils, 8227 etc. in the build tree, and create links to the tools from there. 8228 This does not work any more when the compiler is bootstrapped. The 8229 new configure option --with-target-tools provides a better way to 8230 achieve the same effect, and works for all native and cross 8231 settings. 8232 8233 8234 For questions related to the use of GCC, please consult these web 8235 pages and the [6]GCC manuals. If that fails, the 8236 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8237 web pages and the development of GCC are welcome on our developer 8238 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 8239 8240 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 8241 distribution of this entire article is permitted in any medium, 8242 provided this notice is preserved. 8243 8244 These pages are [11]maintained by the GCC team. Last modified 8245 2017-02-07[12]. 8246 8247 References 8248 8249 1. http://gcc.gnu.org/projects/gomp/ 8250 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 8251 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 8252 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 8253 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 8254 6. https://gcc.gnu.org/onlinedocs/ 8255 7. mailto:gcc-help (a] gcc.gnu.org 8256 8. mailto:gcc (a] gcc.gnu.org 8257 9. https://gcc.gnu.org/lists.html 8258 10. http://www.fsf.org/ 8259 11. https://gcc.gnu.org/about.html 8260 12. http://validator.w3.org/check/referer 8261 ====================================================================== 8262 http://gcc.gnu.org/gcc-4.1/index.html 8263 8264 GCC 4.1 Release Series 8265 8266 February 13, 2007 8267 8268 The [1]GNU project and the GCC developers are pleased to announce the 8269 release of GCC 4.1.2. 8270 8271 This release is a bug-fix release, containing fixes for regressions in 8272 GCC 4.1.1 relative to previous releases of GCC. 8273 8274 This release series is no longer maintained. 8275 8276 Release History 8277 8278 GCC 4.1.2 8279 February 13, 2007 ([2]changes) 8280 8281 GCC 4.1.1 8282 May 24, 2006 ([3]changes) 8283 8284 GCC 4.1.0 8285 February 28, 2006 ([4]changes) 8286 8287 References and Acknowledgements 8288 8289 GCC used to stand for the GNU C Compiler, but since the compiler 8290 supports several other languages aside from C, it now stands for the 8291 GNU Compiler Collection. 8292 8293 A list of [5]successful builds is updated as new information becomes 8294 available. 8295 8296 The GCC developers would like to thank the numerous people that have 8297 contributed new features, improvements, bug fixes, and other changes as 8298 well as test results to GCC. This [6]amazing group of volunteers is 8299 what makes GCC successful. 8300 8301 For additional information about GCC please refer to the [7]GCC project 8302 web site or contact the [8]GCC development mailing list. 8303 8304 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 8305 8306 8307 For questions related to the use of GCC, please consult these web 8308 pages and the [11]GCC manuals. If that fails, the 8309 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8310 web pages and the development of GCC are welcome on our developer 8311 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 8312 archives. 8313 8314 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 8315 distribution of this entire article is permitted in any medium, 8316 provided this notice is preserved. 8317 8318 These pages are [16]maintained by the GCC team. Last modified 8319 2016-09-30[17]. 8320 8321 References 8322 8323 1. http://www.gnu.org/ 8324 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 8325 3. http://gcc.gnu.org/gcc-4.1/changes.html 8326 4. http://gcc.gnu.org/gcc-4.1/changes.html 8327 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 8328 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8329 7. http://gcc.gnu.org/index.html 8330 8. mailto:gcc (a] gcc.gnu.org 8331 9. http://gcc.gnu.org/mirrors.html 8332 10. http://gcc.gnu.org/svn.html 8333 11. https://gcc.gnu.org/onlinedocs/ 8334 12. mailto:gcc-help (a] gcc.gnu.org 8335 13. mailto:gcc (a] gcc.gnu.org 8336 14. https://gcc.gnu.org/lists.html 8337 15. http://www.fsf.org/ 8338 16. https://gcc.gnu.org/about.html 8339 17. http://validator.w3.org/check/referer 8340 ====================================================================== 8341 http://gcc.gnu.org/gcc-4.1/changes.html 8342 8343 GCC 4.1 Release Series 8344 Changes, New Features, and Fixes 8345 8346 The latest release in the 4.1 release series is [1]GCC 4.1.2. 8347 8348 Caveats 8349 8350 General Optimizer Improvements 8351 8352 * GCC now has infrastructure for inter-procedural optimizations and 8353 the following inter-procedural optimizations are implemented: 8354 + Profile guided inlining. When doing profile feedback guided 8355 optimization, GCC can now use the profile to make better 8356 informed decisions on whether inlining of a function is 8357 profitable or not. This means that GCC will no longer inline 8358 functions at call sites that are not executed very often, and 8359 that functions at hot call sites are more likely to be 8360 inlined. 8361 A new parameter min-inline-recursive-probability is also now 8362 available to throttle recursive inlining of functions with 8363 small average recursive depths. 8364 + Discovery of pure and const functions, a form of side-effects 8365 analysis. While older GCC releases could also discover such 8366 special functions, the new IPA-based pass runs earlier so that 8367 the results are available to more optimizers. The pass is also 8368 simply more powerful than the old one. 8369 + Analysis of references to static variables and type escape 8370 analysis, also forms of side-effects analysis. The results of 8371 these passes allow the compiler to be less conservative about 8372 call-clobbered variables and references. This results in more 8373 redundant loads being eliminated and in making static 8374 variables candidates for register promotion. 8375 + Improvement of RTL-based alias analysis. The results of type 8376 escape analysis are fed to the RTL type-based alias analyzer, 8377 allowing it to disambiguate more memory references. 8378 + Interprocedural constant propagation and function versioning. 8379 This pass looks for functions that are always called with the 8380 same constant value for one or more of the function arguments, 8381 and propagates those constants into those functions. 8382 + GCC will now eliminate static variables whose usage was 8383 optimized out. 8384 + -fwhole-program --combine can now be used to make all 8385 functions in program static allowing whole program 8386 optimization. As an exception, the main function and all 8387 functions marked with the new externally_visible attribute are 8388 kept global so that programs can link with runtime libraries. 8389 * GCC can now do a form of partial dead code elimination (PDCE) that 8390 allows code motion of expressions to the paths where the result of 8391 the expression is actually needed. This is not always a win, so the 8392 pass has been limited to only consider profitable cases. Here is an 8393 example: 8394 int foo (int *, int *); 8395 int 8396 bar (int d) 8397 { 8398 int a, b, c; 8399 b = d + 1; 8400 c = d + 2; 8401 a = b + c; 8402 if (d) 8403 { 8404 foo (&b, &c); 8405 a = b + c; 8406 } 8407 printf ("%d\n", a); 8408 } 8409 8410 The a = b + c can be sunk to right before the printf. Normal code 8411 sinking will not do this, it will sink the first one above into the 8412 else-branch of the conditional jump, which still gives you two 8413 copies of the code. 8414 * GCC now has a value range propagation pass. This allows the 8415 compiler to eliminate bounds checks and branches. The results of 8416 the pass can also be used to accurately compute branch 8417 probabilities. 8418 * The pass to convert PHI nodes to straight-line code (a form of 8419 if-conversion for GIMPLE) has been improved significantly. The two 8420 most significant improvements are an improved algorithm to 8421 determine the order in which the PHI nodes are considered, and an 8422 improvement that allow the pass to consider if-conversions of basic 8423 blocks with more than two predecessors. 8424 * Alias analysis improvements. GCC can now differentiate between 8425 different fields of structures in Tree-SSA's virtual operands form. 8426 This lets stores/loads from non-overlapping structure fields not 8427 conflict. A new algorithm to compute points-to sets was contributed 8428 that can allows GCC to see now that p->a and p->b, where p is a 8429 pointer to a structure, can never point to the same field. 8430 * Various enhancements to auto-vectorization: 8431 + Incrementally preserve SSA form when vectorizing. 8432 + Incrementally preserve loop-closed form when vectorizing. 8433 + Improvements to peeling for alignment: generate better code 8434 when the misalignment of an access is known at compile time, 8435 or when different accesses are known to have the same 8436 misalignment, even if the misalignment amount itself is 8437 unknown. 8438 + Consider dependence distance in the vectorizer. 8439 + Externalize generic parts of data reference analysis to make 8440 this analysis available to other passes. 8441 + Vectorization of conditional code. 8442 + Reduction support. 8443 * GCC can now partition functions in sections of hot and cold code. 8444 This can significantly improve performance due to better 8445 instruction cache locality. This feature works best together with 8446 profile feedback driven optimization. 8447 * A new pass to avoid saving of unneeded arguments to the stack in 8448 vararg functions if the compiler can prove that they will not be 8449 needed. 8450 * Transition of basic block profiling to tree level implementation 8451 has been completed. The new implementation should be considerably 8452 more reliable (hopefully avoiding profile mismatch errors when 8453 using -fprofile-use or -fbranch-probabilities) and can be used to 8454 drive higher level optimizations, such as inlining. 8455 The -ftree-based-profiling command-line option was removed and 8456 -fprofile-use now implies disabling old RTL level loop optimizer 8457 (-fno-loop-optimize). Speculative prefetching optimization 8458 (originally enabled by -fspeculative-prefetching) was removed. 8459 8460 New Languages and Language specific improvements 8461 8462 C and Objective-C 8463 8464 * The old Bison-based C and Objective-C parser has been replaced by a 8465 new, faster hand-written recursive-descent parser. 8466 8467 Ada 8468 8469 * The build infrastructure for the Ada runtime library and tools has 8470 been changed to be better integrated with the rest of the build 8471 infrastructure of GCC. This should make doing cross builds of Ada a 8472 bit easier. 8473 8474 C++ 8475 8476 * ARM-style name-injection of friend declarations is no longer the 8477 default. For example: 8478 struct S { 8479 friend void f(); 8480 }; 8481 8482 void g() { f(); } 8483 will not be accepted; instead a declaration of f will need to be 8484 present outside of the scope of S. The new -ffriend-injection 8485 option will enable the old behavior. 8486 * The (undocumented) extension which permitted templates with default 8487 arguments to be bound to template template parameters with fewer 8488 parameters has been deprecated, and will be removed in the next 8489 major release of G++. For example: 8490 template <template <typename> class C> 8491 void f(C<double>) {} 8492 8493 template <typename T, typename U = int> 8494 struct S {}; 8495 8496 template void f(S<double>); 8497 8498 makes use of the deprecated extension. The reason this code is not 8499 valid ISO C++ is that S is a template with two parameters; 8500 therefore, it cannot be bound to C which has only one parameter. 8501 8502 Runtime Library (libstdc++) 8503 8504 * Optimization work: 8505 + A new implementation of std::search_n is provided, better 8506 performing in case of random access iterators. 8507 + Added further efficient specializations of istream functions, 8508 i.e., character array and string extractors. 8509 + Other smaller improvements throughout. 8510 * Policy-based associative containers, designed for high-performance, 8511 flexibility and semantic safety are delivered in ext/pb_assoc. 8512 * A versatile string class, __gnu_cxx::__versa_string, providing 8513 facilities conforming to the standard requirements for 8514 basic_string, is delivered in <ext/vstring.h>. In particular: 8515 + Two base classes are provided: the default one avoids 8516 reference counting and is optimized for short strings; the 8517 alternate one, still uses it while improving in a few low 8518 level areas (e.g., alignment). See vstring_fwd.h for some 8519 useful typedefs. 8520 + Various algorithms have been rewritten (e.g., replace), the 8521 code streamlined and simple optimizations added. 8522 + Option 3 of DR 431 is implemented for both available bases, 8523 thus improving the support for stateful allocators. 8524 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 8525 libstdc++/23953) and LWG resolutions put into effect for the first 8526 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 8527 Issue 6.19). The implementation status of TR1 is now tracked in the 8528 docs in tr1.html. 8529 8530 Objective-C++ 8531 8532 * A new language front end for Objective-C++ has been added. This 8533 language allows users to mix the object oriented features of 8534 Objective-C with those of C++. 8535 8536 Java (GCJ) 8537 8538 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 8539 features (plus some 0.20 bug-fixes) 8540 + Networking 8541 o The java.net.HttpURLConnection implementation no longer 8542 buffers the entire response body in memory. This means 8543 that response bodies larger than available memory can now 8544 be handled. 8545 + (N)IO 8546 o NIO FileChannel.map implementation, fast bulk put 8547 implementation for DirectByteBuffer (speeds up this 8548 method 10x). 8549 o FileChannel.lock() and FileChannel.force() implemented. 8550 + XML 8551 o gnu.xml fix for nodes created outside a namespace 8552 context. 8553 o Add support for output indenting and 8554 cdata-section-elements output instruction in 8555 xml.transform. 8556 o xml.xpath corrections for cases where elements/attributes 8557 might have been created in non-namespace-aware mode. 8558 Corrections to handling of XSL variables and minor 8559 conformance updates. 8560 + AWT 8561 o GNU JAWT implementation, the AWT Native Interface, which 8562 allows direct access to native screen resources from 8563 within a Canvas's paint method. GNU Classpath Examples 8564 comes with a Demo, see libjava/classpath/examples/README. 8565 o awt.datatransfer updated to 1.5 with support for 8566 FlavorEvents. The gtk+ awt peers now allow copy/paste of 8567 text, images, URIs/files and serialized objects with 8568 other applications and tracking clipboard change events 8569 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 8570 objects are supported). A GNU Classpath Examples 8571 datatransfer Demo was added to show the new 8572 functionality. 8573 o Split gtk+ awt peers event handling in two threads and 8574 improve gdk lock handling (solves several awt lock ups). 8575 o Speed up awt Image loading. 8576 o Better gtk+ scrollbar peer implementation when using gtk+ 8577 >= 2.6. 8578 o Handle image loading errors correctly for gdkpixbuf and 8579 MediaTracker. 8580 o Better handle GDK lock. Properly prefix gtkpeer native 8581 functions (cp_gtk). 8582 o GdkGraphics2D has been updated to use Cairo 0.5.x or 8583 higher. 8584 o BufferedImage and GtkImage rewrites. All image drawing 8585 operations should now work correctly (flipping requires 8586 gtk+ >= 2.6) 8587 o Future Graphics2D, image and text work is documented at: 8588 [2]http://developer.classpath.org/mediation/ClasspathGrap 8589 hicsImagesText 8590 o When gtk+ 2.6 or higher is installed the default log 8591 handler will produce stack traces whenever a WARNING, 8592 CRITICAL or ERROR message is produced. 8593 + Free Swing 8594 o The RepaintManager has been reworked for more efficient 8595 painting, especially for large GUIs. 8596 o The layout manager OverlayLayout has been implemented, 8597 the BoxLayout has been rewritten to make use of the 8598 SizeRequirements utility class and caching for more 8599 efficient layout. 8600 o Improved accessibility support. 8601 o Significant progress has been made in the implementation 8602 of the javax.swing.plaf.metal package, with most UI 8603 delegates in a working state now. Please test this with 8604 your own applications and provide feedback that will help 8605 us to improve this package. 8606 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 8607 extended to highlight various features in our Free Swing 8608 implementation. And it includes a look and feel switcher 8609 for Metal (default), Ocean and GNU themes. 8610 o The javax.swing.plaf.multi package is now implemented. 8611 o Editing and several key actions for JTree and JTable were 8612 implemented. 8613 o Lots of icons and look and feel improvements for Free 8614 Swing basic and metal themes were added. Try running the 8615 GNU Classpath Swing Demo in examples 8616 (gnu.classpath.examples.swing.Demo) with: 8617 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 8618 l or 8619 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 8620 l 8621 o Start of styled text capabilites for java.swing.text. 8622 o DefaultMutableTreeNode pre-order, post-order, depth-first 8623 and breadth-first traversal enumerations implemented. 8624 o JInternalFrame colors and titlebar draw properly. 8625 o JTree is working up to par (icons, selection and keyboard 8626 traversal). 8627 o JMenus were made more compatible in visual and 8628 programmatic behavior. 8629 o JTable changeSelection and multiple selections 8630 implemented. 8631 o JButton and JToggleButton change states work properly 8632 now. 8633 o JFileChooser fixes. 8634 o revalidate() and repaint() fixes which make Free Swing 8635 much more responsive. 8636 o MetalIconFactory implemented. 8637 o Free Swing Top-Level Compatibility. JFrame, JDialog, 8638 JApplet, JInternalFrame, and JWindow are now 1.5 8639 compatible in the sense that you can call add() and 8640 setLayout() directly on them, which will have the same 8641 effect as calling getContentPane().add() and 8642 getContentPane().setLayout(). 8643 o The JTree interface has been completed. JTrees now 8644 recognizes mouse clicks and selections work. 8645 o BoxLayout works properly now. 8646 o Fixed GrayFilter to actually work. 8647 o Metal SplitPane implemented. 8648 o Lots of Free Swing text and editor stuff work now. 8649 + Free RMI and Corba 8650 o Andrew Watson, Vice President and Technical Director of 8651 the Object Management Group, has officially assigned us 8652 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 8653 mark remote classpath-specific system exceptions. 8654 Obtaining the VMCID means that GNU Classpath now is a 8655 recogniseable type of node in a highly interoperable 8656 CORBA world. 8657 o GNU Classpath now includes the first working draft to 8658 support the RMI over IIOP protocol. The current 8659 implementation is capable of remote invocations, 8660 transferring various Serializables and Externalizables 8661 via RMI-IIOP protocol. It can flatten graphs and, at 8662 least for the simple cases, is interoperable with 1.5 8663 JDKs. 8664 o org.omg.PortableInterceptor and related functionality in 8665 other packages is now implemented: 8666 # The sever and client interceptors work as required 8667 since 1.4. 8668 # The IOR interceptor works as needed for 1.5. 8669 o The org.omg.DynamicAny package is completed and passes 8670 the prepared tests. 8671 o The Portable Object Adapter should now support the output 8672 of the recent IDL to java compilers. These compilers now 8673 generate servants and not CORBA objects as before, making 8674 the output depend on the existing POA implementation. 8675 Completing POA means that such code can already be tried 8676 to run on Classpath. Our POA is tested for the following 8677 usager scenarios: 8678 # POA converts servant to the CORBA object. 8679 # Servant provides to the CORBA object. 8680 # POA activates new CORBA object with the given Object 8681 Id (byte array) that is later accessible for the 8682 servant. 8683 # During the first call, the ServantActivator provides 8684 servant for this and all subsequent calls on the 8685 current object. 8686 # During each call, the ServantLocator provides 8687 servant for this call only. 8688 # ServantLocator or ServantActivator forwards call to 8689 another server. 8690 # POA has a single servant, responsible for all 8691 objects. 8692 # POA has a default servant, but some objects are 8693 explicitly connected to they specific servants. 8694 The POA is verified using tests from the former 8695 cost.omg.org. 8696 o The CORBA implementation is now a working prototype that 8697 should support features up to 1.3 inclusive. We invite 8698 groups writing CORBA dependent applications to try 8699 Classpath implementation, reporting any possible bugs. 8700 The CORBA prototype is interoperable with Sun's 8701 implementation v 1.4, transferring object references, 8702 primitive types, narrow and wide strings, arrays, 8703 structures, trees, abstract interfaces and value types 8704 (feature of CORBA 2.3) between these two platforms. 8705 Remote exceptions are transferred and handled correctly. 8706 The stringified object references (IORs) from various 8707 sources are parsed as required. The transient (for 8708 current session) and permanent (till jre restart) 8709 redirections work. Both Little and Big Endian encoded 8710 messages are accepted. The implementation is verified 8711 using tests from the former cost.omg.org. The current 8712 release includes working examples (see the examples 8713 directory), demonstrating the client-server 8714 communication, using either CORBA Request or IDL-based 8715 stub (usually generated by a IDL to java compiler). These 8716 examples also show how to use the Classpath CORBA naming 8717 service. The IDL to java compiler is not yet written, but 8718 as our library must be compatible, it naturally accepts 8719 the output of other idlj implementations. 8720 + Misc 8721 o Updated TimeZone data against Olson tzdata2005l. 8722 o Make zip and jar packages UTF-8 clean. 8723 o "native" code builds and compiles (warning free) on 8724 Darwin and Solaris. 8725 o java.util.logging.FileHandler now rotates files. 8726 o Start of a generic JDWP framework in gnu/classpath/jdwp. 8727 This is unfinished, but feedback (at classpath (a] gnu.org) 8728 from runtime hackers is greatly appreciated. Although 8729 most of the work is currently being done around gcj/gij 8730 we want this framework to be as VM neutral as possible. 8731 Early design is described in: 8732 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 8733 o QT4 AWT peers, enable by giving configure 8734 --enable-qt-peer. Included, but not ready for production 8735 yet. They are explicitly disabled and not supported. But 8736 if you want to help with the development of these new 8737 features we are interested in feedback. You will have to 8738 explicitly enable them to try them out (and they will 8739 most likely contain bugs). 8740 o Documentation fixes all over the place. See 8741 [4]http://developer.classpath.org/doc/ 8742 8743 New Targets and Target Specific Improvements 8744 8745 IA-32/x86-64 8746 8747 * The x86-64 medium model (that allows building applications whose 8748 data segment exceeds 4GB) was redesigned to match latest ABI draft. 8749 New implementation split large datastructures into separate segment 8750 improving performance of accesses to small datastructures and also 8751 allows linking of small model libraries into medium model programs 8752 as long as the libraries are not accessing the large datastructures 8753 directly. Medium model is also supported in position independent 8754 code now. 8755 The ABI change results in partial incompatibility among medium 8756 model objects. Linking medium model libraries (or objects) compiled 8757 with new compiler into medium model program compiled with older 8758 will likely result in exceeding ranges of relocations. 8759 Binutils 2.16.91 or newer are required for compiling medium model 8760 now. 8761 8762 RS6000 (POWER/PowerPC) 8763 8764 * The AltiVec vector primitives in <altivec.h> are now implemented in 8765 a way that puts a smaller burden on the preprocessor, instead 8766 processing the "overloading" in the front ends. This should benefit 8767 compilation speed on AltiVec vector code. 8768 * AltiVec initializers now are generated more efficiently. 8769 * The popcountb instruction available on POWER5 now is generated. 8770 * The floating point round to integer instructions available on 8771 POWER5+ now is generated. 8772 * Floating point divides can be synthesized using the floating point 8773 reciprocal estimate instructions. 8774 * Double precision floating point constants are initialized as single 8775 precision values if they can be represented exactly. 8776 8777 S/390, zSeries and System z9 8778 8779 * Support for the IBM System z9 109 processor has been added. When 8780 using the -march=z9-109 option, the compiler will generate code 8781 making use of instructions provided by the extended immediate 8782 facility. 8783 * Support for 128-bit IEEE floating point has been added. When using 8784 the -mlong-double-128 option, the compiler will map the long double 8785 data type to 128-bit IEEE floating point. Using this option 8786 constitutes an ABI change, and requires glibc support. 8787 * Various changes to improve performance of generated code have been 8788 implemented, including: 8789 + In functions that do not require a literal pool, register %r13 8790 (which is traditionally reserved as literal pool pointer), can 8791 now be freely used for other purposes by the compiler. 8792 + More precise tracking of register use allows the compiler to 8793 generate more efficient function prolog and epilog code in 8794 certain cases. 8795 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 8796 instructions are now used to implement C string functions. 8797 + The MOVE CHARACTER instruction with single byte overlap is now 8798 used to implement the memset function with non-zero fill byte. 8799 + The LOAD ZERO instructions are now used where appropriate. 8800 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 8801 and INSERT IMMEDIATE instructions are now used more frequently 8802 to optimize bitfield operations. 8803 + The BRANCH ON COUNT instruction is now used more frequently. 8804 In particular, the fact that a loop contains a subroutine call 8805 no longer prevents the compiler from using this instruction. 8806 + The compiler is now aware that all shift and rotate 8807 instructions implicitly truncate the shift count to six bits. 8808 * Back-end support for the following generic features has been 8809 implemented: 8810 + The full set of [5]built-in functions for atomic memory 8811 access. 8812 + The -fstack-protector feature. 8813 + The optimization pass avoiding unnecessary stores of incoming 8814 argument registers in functions with variable argument list. 8815 8816 SPARC 8817 8818 * The default code model in 64-bit mode has been changed from 8819 Medium/Anywhere to Medium/Middle on Solaris. 8820 * TLS support is disabled by default on Solaris prior to release 10. 8821 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 8822 and later) by specifying --enable-tls at configure time. 8823 8824 MorphoSys 8825 8826 * Support has been added for this new architecture. 8827 8828 Obsolete Systems 8829 8830 Documentation improvements 8831 8832 Other significant improvements 8833 8834 * GCC can now emit code for protecting applications from 8835 stack-smashing attacks. The protection is realized by buffer 8836 overflow detection and reordering of stack variables to avoid 8837 pointer corruption. 8838 * Some built-in functions have been fortified to protect them against 8839 various buffer overflow (and format string) vulnerabilities. 8840 Compared to the mudflap bounds checking feature, the safe builtins 8841 have far smaller overhead. This means that programs built using 8842 safe builtins should not experience any measurable slowdown. 8843 8844 GCC 4.1.2 8845 8846 This is the [6]list of problem reports (PRs) from GCC's bug tracking 8847 system that are known to be fixed in the 4.1.2 release. This list might 8848 not be complete (that is, it is possible that some PRs that have been 8849 fixed are not listed here). 8850 8851 When generating code for a shared library, GCC now recognizes that 8852 global functions may be replaced when the program runs. Therefore, it 8853 is now more conservative in deducing information from the bodies of 8854 functions. For example, in this example: 8855 void f() {} 8856 void g() { 8857 try { f(); } 8858 catch (...) { 8859 cout << "Exception"; 8860 } 8861 } 8862 8863 G++ would previously have optimized away the catch clause, since it 8864 would have concluded that f cannot throw exceptions. Because users may 8865 replace f with another function in the main body of the program, this 8866 optimization is unsafe, and is no longer performed. If you wish G++ to 8867 continue to optimize as before, you must add a throw() clause to the 8868 declaration of f to make clear that it does not throw exceptions. 8869 8870 8871 For questions related to the use of GCC, please consult these web 8872 pages and the [7]GCC manuals. If that fails, the 8873 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8874 web pages and the development of GCC are welcome on our developer 8875 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 8876 archives. 8877 8878 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 8879 distribution of this entire article is permitted in any medium, 8880 provided this notice is preserved. 8881 8882 These pages are [12]maintained by the GCC team. Last modified 8883 2016-01-30[13]. 8884 8885 References 8886 8887 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 8888 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 8889 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 8890 4. http://developer.classpath.org/doc/ 8891 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 8892 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 8893 7. https://gcc.gnu.org/onlinedocs/ 8894 8. mailto:gcc-help (a] gcc.gnu.org 8895 9. mailto:gcc (a] gcc.gnu.org 8896 10. https://gcc.gnu.org/lists.html 8897 11. http://www.fsf.org/ 8898 12. https://gcc.gnu.org/about.html 8899 13. http://validator.w3.org/check/referer 8900 ====================================================================== 8901 http://gcc.gnu.org/gcc-4.0/index.html 8902 8903 GCC 4.0 Release Series 8904 8905 January 31, 2007 8906 8907 The [1]GNU project and the GCC developers are pleased to announce the 8908 release of GCC 4.0.4. 8909 8910 This release is a bug-fix release, containing fixes for regressions in 8911 GCC 4.0.3 relative to previous releases of GCC. 8912 8913 This release series is no longer maintained. 8914 8915 Release History 8916 8917 GCC 4.0.4 8918 January 31, 2007 ([2]changes) 8919 8920 GCC 4.0.3 8921 March 10, 2006 ([3]changes) 8922 8923 GCC 4.0.2 8924 September 28, 2005 ([4]changes) 8925 8926 GCC 4.0.1 8927 July 7, 2005 ([5]changes) 8928 8929 GCC 4.0.0 8930 April 20, 2005 ([6]changes) 8931 8932 References and Acknowledgements 8933 8934 GCC used to stand for the GNU C Compiler, but since the compiler 8935 supports several other languages aside from C, it now stands for the 8936 GNU Compiler Collection. 8937 8938 A list of [7]successful builds is updated as new information becomes 8939 available. 8940 8941 The GCC developers would like to thank the numerous people that have 8942 contributed new features, improvements, bug fixes, and other changes as 8943 well as test results to GCC. This [8]amazing group of volunteers is 8944 what makes GCC successful. 8945 8946 For additional information about GCC please refer to the [9]GCC project 8947 web site or contact the [10]GCC development mailing list. 8948 8949 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 8950 8951 8952 For questions related to the use of GCC, please consult these web 8953 pages and the [13]GCC manuals. If that fails, the 8954 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8955 web pages and the development of GCC are welcome on our developer 8956 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 8957 archives. 8958 8959 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 8960 distribution of this entire article is permitted in any medium, 8961 provided this notice is preserved. 8962 8963 These pages are [18]maintained by the GCC team. Last modified 8964 2016-09-30[19]. 8965 8966 References 8967 8968 1. http://www.gnu.org/ 8969 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8970 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 8971 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 8972 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 8973 6. http://gcc.gnu.org/gcc-4.0/changes.html 8974 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 8975 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8976 9. http://gcc.gnu.org/index.html 8977 10. mailto:gcc (a] gcc.gnu.org 8978 11. http://gcc.gnu.org/mirrors.html 8979 12. http://gcc.gnu.org/svn.html 8980 13. https://gcc.gnu.org/onlinedocs/ 8981 14. mailto:gcc-help (a] gcc.gnu.org 8982 15. mailto:gcc (a] gcc.gnu.org 8983 16. https://gcc.gnu.org/lists.html 8984 17. http://www.fsf.org/ 8985 18. https://gcc.gnu.org/about.html 8986 19. http://validator.w3.org/check/referer 8987 ====================================================================== 8988 http://gcc.gnu.org/gcc-4.0/changes.html 8989 8990 GCC 4.0 Release Series 8991 Changes, New Features, and Fixes 8992 8993 The latest release in the 4.0 release series is [1]GCC 4.0.4. 8994 8995 Caveats 8996 8997 * GCC now generates location lists by default when compiling with 8998 debug info and optimization. 8999 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 9000 or later is needed to debug binaries containing location 9001 lists. 9002 + When you are trying to view a value of a variable in a part of 9003 a function where it has no location (for example when the 9004 variable is no longer used and thus its location was used for 9005 something else) GDB will say that it is not available. 9006 You can disable generating location lists by -fno-var-tracking. 9007 * GCC no longer accepts the -fwritable-strings option. Use named 9008 character arrays when you need a writable string. 9009 * The options -freduce-all-givs and -fmove-all-movables have been 9010 discontinued. They were used to circumvent a shortcoming in the 9011 heuristics of the old loop optimization code with respect to common 9012 Fortran constructs. The new (tree) loop optimizer works differently 9013 and doesn't need those work-arounds. 9014 * The graph-coloring register allocator, formerly enabled by the 9015 option -fnew-ra, has been discontinued. 9016 * -I- has been deprecated. -iquote is meant to replace the need for 9017 this option. 9018 * The MIPS -membedded-pic and -mrnames options have been removed. 9019 * All MIPS targets now require the GNU assembler. In particular, IRIX 9020 configurations can no longer use the MIPSpro assemblers, although 9021 they do still support the MIPSpro linkers. 9022 * The SPARC option -mflat has been removed. 9023 * English-language diagnostic messages will now use Unicode quotation 9024 marks in UTF-8 locales. (Non-English messages already used the 9025 quotes appropriate for the language in previous releases.) If your 9026 terminal does not support UTF-8 but you are using a UTF-8 locale 9027 (such locales are the default on many GNU/Linux systems) then you 9028 should set LC_CTYPE=C in the environment to disable that locale. 9029 Programs that parse diagnostics and expect plain ASCII 9030 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 9031 explanation of Unicode quotation marks for more information. 9032 * The specs file is no longer installed on most platforms. Most users 9033 will be totally unaffected. However, if you are accustomed to 9034 editing the specs file yourself, you will now have to use the 9035 -dumpspecs option to generate the specs file, and then edit the 9036 resulting file. 9037 9038 General Optimizer Improvements 9039 9040 * The [3]tree ssa branch has been merged. This merge has brought in a 9041 completely new optimization framework based on a higher level 9042 intermediate representation than the existing RTL representation. 9043 Numerous new code transformations based on the new framework are 9044 available in GCC 4.0, including: 9045 + Scalar replacement of aggregates 9046 + Constant propagation 9047 + Value range propagation 9048 + Partial redundancy elimination 9049 + Load and store motion 9050 + Strength reduction 9051 + Dead store elimination 9052 + Dead and unreachable code elimination 9053 + [4]Autovectorization 9054 + Loop interchange 9055 + Tail recursion by accumulation 9056 Many of these passes outperform their counterparts from previous 9057 GCC releases. 9058 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 9059 scheduling optimization intended for loops that perform heavy 9060 computations. 9061 9062 New Languages and Language specific improvements 9063 9064 C family 9065 9066 * The sentinel attribute has been added to GCC. This function 9067 attribute allows GCC to warn when variadic functions such as execl 9068 are not NULL terminated. See the GCC manual for a complete 9069 description of its behavior. 9070 * Given __attribute__((alias("target"))) it is now an error if target 9071 is not a symbol, defined in the same translation unit. This also 9072 applies to aliases created by #pragma weak alias=target. This is 9073 because it's meaningless to define an alias to an undefined symbol. 9074 On Solaris, the native assembler would have caught this error, but 9075 GNU as does not. 9076 9077 C and Objective-C 9078 9079 * The -Wstrict-aliasing=2 option has been added. This warning catches 9080 all unsafe cases, but it may also give a warning for some cases 9081 that are safe. 9082 * The cast-as-lvalue, conditional-expression-as-lvalue and 9083 compound-expression-as-lvalue extensions, which were deprecated in 9084 3.3.4 and 3.4, have been removed. 9085 * The -fwritable-strings option, which was deprecated in 3.4, has 9086 been removed. 9087 * #pragma pack() semantics have been brought closer to those used by 9088 other compilers. This also applies to C++. 9089 * Taking the address of a variable with register storage is invalid 9090 in C. GCC now issues an error instead of a warning. 9091 * Arrays of incomplete element type are invalid in C. GCC now issues 9092 an error for such arrays. Declarations such as extern struct s x[]; 9093 (where struct s has not been defined) can be moved after the 9094 definition of struct s. Function parameters declared as arrays of 9095 incomplete type can instead be declared as pointers. 9096 9097 C++ 9098 9099 * When compiling without optimizations (-O0), the C++ frontend is 9100 much faster than in any previous versions of GCC. Independent 9101 testers have measured speed-ups up to 25% in real-world production 9102 code, compared to the 3.4 family (which was already the fastest 9103 version to date). Upgrading from older versions might show even 9104 bigger improvements. 9105 * ELF visibility attributes can now be applied to a class type, so 9106 that it affects every member function of a class at once, without 9107 having to specify each individually: 9108 class __attribute__ ((visibility("hidden"))) Foo 9109 { 9110 int foo1(); 9111 void foo2(); 9112 }; 9113 The syntax is deliberately similar to the __declspec() system used 9114 by Microsoft Windows based compilers, allowing cross-platform 9115 projects to easily reuse their existing macro system for denoting 9116 exports and imports. By explicitly marking internal classes never 9117 used outside a binary as hidden, one can completely avoid PLT 9118 indirection overheads during their usage by the compiler. You can 9119 find out more about the advantages of this at 9120 [6]https://www.akkadia.org/drepper/dsohowto.pdf 9121 * The -fvisibility-inlines-hidden option has been added which marks 9122 all inlineable functions as having hidden ELF visibility, thus 9123 removing their symbol and typeinfo from the exported symbol table 9124 of the output ELF binary. Using this option can reduce the exported 9125 symbol count of template-heavy code by up to 40% with no code 9126 change at all, thus notably improving link and load times for the 9127 binary as well as a reduction in size of up to 10%. Also, check the 9128 new [7]-fvisibility option. 9129 * The compiler now uses the library interface specified by the [8]C++ 9130 ABI for thread-safe initialization of function-scope static 9131 variables. Most users should leave this alone, but embedded 9132 programmers may want to disable this by specifying 9133 -fno-threadsafe-statics for a small savings in code size. 9134 * Taking the address of an explicit register variable is no longer 9135 supported. Note that C++ allows taking the address of variables 9136 with register storage so this will continue to compile with a 9137 warning. For example, assuming that r0 is a machine register: 9138 register int foo asm ("r0"); 9139 register int bar; 9140 &foo; // error, no longer accepted 9141 &bar; // OK, with a warning 9142 * G++ has an undocumented extension to virtual function covariancy 9143 rules that allowed the overrider to return a type that was 9144 implicitly convertable to the overridden function's return type. 9145 For instance a function returning void * could be overridden by a 9146 function returning T *. This is now deprecated and will be removed 9147 in a future release. 9148 * The G++ minimum and maximum operators (<? and >?) and their 9149 compound forms (<?=) and >?=) have been deprecated and will be 9150 removed in a future version. Code using these operators should be 9151 modified to use std::min and std::max instead. 9152 * Declaration of nested classes of class templates as friends are 9153 supported: 9154 template <typename T> struct A { 9155 class B {}; 9156 }; 9157 class C { 9158 template <typename T> friend class A<T>::B; 9159 }; 9160 This complements the feature member functions of class templates as 9161 friends introduced in GCC 3.4.0. 9162 * When declaring a friend class using an unqualified name, classes 9163 outside the innermost non-class scope are not searched: 9164 class A; 9165 namespace N { 9166 class B { 9167 friend class A; // Refer to N::A which has not been declared yet 9168 // because name outside namespace N are not searched 9169 friend class ::A; // Refer to ::A 9170 }; 9171 } 9172 Hiding the friend name until declaration is still not implemented. 9173 * Friends of classes defined outside their namespace are correctly 9174 handled: 9175 namespace N { 9176 class A; 9177 } 9178 class N::A { 9179 friend class B; // Refer to N::B in GCC 4.0.0 9180 // but ::B in earlier versions of GCC 9181 }; 9182 9183 Runtime Library (libstdc++) 9184 9185 * Optimization work: 9186 + Added efficient specializations of istream functions for char 9187 and wchar_t. 9188 + Further performance tuning of strings, in particular wrt 9189 single-char append and getline. 9190 + iter_swap - and therefore most of the mutating algorithms - 9191 now makes an unqualified call to swap when the value_type of 9192 the two iterators is the same. 9193 * A large subset of the features in Technical Report 1 (TR1 for 9194 short) is experimentally delivered (i.e., no guarantees about the 9195 implementation are provided. In particular it is not promised that 9196 the library will remain link-compatible when code using TR1 is 9197 used): 9198 + General utilities such as reference_wrapper and shared_ptr. 9199 + Function objects, i.e., result_of, mem_fn, bind, function. 9200 + Support for metaprogramming. 9201 + New containers such as tuple, array, unordered_set, 9202 unordered_map, unordered_multiset, unordered_multimap. 9203 * As usual, many bugs have been fixed and LWG resolutions implemented 9204 for the first time (e.g., DR 409). 9205 9206 Java 9207 9208 * In order to prevent naming conflicts with other implementations of 9209 these tools, some GCJ binaries have been renamed: 9210 + rmic is now grmic, 9211 + rmiregistry is now grmiregistry, and 9212 + jar is now fastjar. 9213 In particular, these names were problematic for the jpackage.org 9214 packaging conventions which install symlinks in /usr/bin that point 9215 to the preferred versions of these tools. 9216 * The -findirect-dispatch argument to the compiler now works and 9217 generates code following a new "binary compatibility" ABI. Code 9218 compiled this way follows the binary compatibility rules of the 9219 Java Language Specification. 9220 * libgcj now has support for using GCJ as a JIT, using the 9221 gnu.gcj.jit family of system properties. 9222 * libgcj can now find a shared library corresponding to the bytecode 9223 representation of a class. See the documentation for the new 9224 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 9225 property. 9226 * There have been many improvements to the class library. Here are 9227 some highlights: 9228 + Much more of AWT and Swing exist. 9229 + Many new packages and classes were added, including 9230 java.util.regex, java.net.URI, javax.crypto, 9231 javax.crypto.interfaces, javax.crypto.spec, javax.net, 9232 javax.net.ssl, javax.security.auth, 9233 javax.security.auth.callback, javax.security.auth.login, 9234 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 9235 javax.imageio, javax.imageio.event, javax.imageio.spi, 9236 javax.print, javax.print.attribute, 9237 javax.print.attribute.standard, javax.print.event, and 9238 javax.xml 9239 + Updated SAX and DOM, and imported GNU JAXP 9240 9241 Fortran 9242 9243 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 9244 front end. The new front end supports Fortran 90 and Fortran 95. It 9245 may not yet be as stable as the old Fortran front end. 9246 9247 Ada 9248 9249 * Ada (with tasking and Zero Cost Exceptions) is now available on 9250 many more targets, including but not limited to: alpha-linux, 9251 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 9252 s390x-linux, sparc-linux. 9253 * Some of the new Ada 2005 features are now implemented like 9254 Wide_Wide_Character and Ada.Containers. 9255 * Many bugs have been fixed, tools and documentation improved. 9256 * To compile Ada from the sources, install an older working Ada 9257 compiler and then use --enable-languages=ada at configuration time, 9258 since the Ada frontend is not currently activated by default. See 9259 the [10]Installing GCC for details. 9260 9261 New Targets and Target Specific Improvements 9262 9263 H8/300 9264 9265 * The frame layout has changed. In the new layout, the prologue of a 9266 function first saves registers and then allocate space for locals, 9267 resulting in an 1% improvement on code size. 9268 9269 IA-32/x86-64 (AMD64) 9270 9271 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 9272 log1p, log2, logb and tan mathematical builtins (and their float 9273 and long double variants) are now implemented as inline x87 9274 intrinsics when using -ffast-math. 9275 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 9276 (and their float and long double variants) are now implemented as 9277 inline x87 intrinsics when using -ffast-math. 9278 * The x87's fsincos instruction is now used automatically with 9279 -ffast-math when calculating both the sin and cos of the same 9280 argument. 9281 * Instruction selection for multiplication and division by constants 9282 has been improved. 9283 9284 IA-64 9285 9286 * Floating point division, integer division and sqrt are now inlined, 9287 resulting in significant performance improvements on some codes. 9288 9289 MIPS 9290 9291 * Division by zero checks now use conditional traps if the target 9292 processor supports them. This decreases code size by one word per 9293 division operation. The old behavior (branch and break) can be 9294 obtained either at configure time by passing --with-divide=breaks 9295 to configure or at runtime by passing -mdivide-breaks to GCC. 9296 * Support for MIPS64 paired-single instructions has been added. It is 9297 enabled by -mpaired-single and can be accessed using both the 9298 target-independent vector extensions and new MIPS-specific built-in 9299 functions. 9300 * Support for the MIPS-3D ASE has been added. It is enabled by 9301 -mips3d and provides new MIPS-3D-specific built-in functions. 9302 * The -mexplicit-relocs option now supports static n64 code (as is 9303 used, for example, in 64-bit linux kernels). -mexplicit-relocs 9304 should now be feature-complete and is enabled by default when GCC 9305 is configured to use a compatible assembler. 9306 * Support for the NEC VR4130 series has been added. This support 9307 includes the use of VR-specific instructions and a new VR4130 9308 scheduler. Full VR4130 support can be selected with -march=vr4130 9309 while code for any ISA can be tuned for the VR4130 using 9310 -mtune=vr4130. There is also a new -mvr4130-align option that 9311 produces better schedules at the cost of increased code size. 9312 * Support for the Broadcom SB-1 has been extended. There is now an 9313 SB-1 scheduler as well as support for the SB-1-specific 9314 paired-single instructions. Full SB-1 support can be selected with 9315 -march=sb1 while code for any ISA can be optimized for the SB-1 9316 using -mtune=sb1. 9317 * The compiler can now work around errata in R4000, R4400, VR4120 and 9318 VR4130 processors. These workarounds are enabled by -mfix-r4000, 9319 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 9320 and VR4130 workarounds need binutils 2.16 or above. 9321 * IRIX shared libraries are now installed into the standard library 9322 directories: o32 libraries go into lib/, n32 libraries go into 9323 lib32/ and n64 libraries go into lib64/. 9324 * The compiler supports a new -msym32 option. It can be used to 9325 optimize n64 code in which all symbols are known to have 32-bit 9326 values. 9327 9328 S/390 and zSeries 9329 9330 * New command-line options help to generate code intended to run in 9331 an environment where stack space is restricted, e.g. Linux kernel 9332 code: 9333 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 9334 warnings for single functions that require large or dynamic 9335 stack frames. 9336 + -mstack-size and -mstack-guard generate code that checks for 9337 stack overflow at run time. 9338 + -mpacked-stack generates code that reduces the stack frame 9339 size of many functions by reusing unneeded parts of the stack 9340 bias area. 9341 * The -msoft-float option now ensures that generated code never 9342 accesses floating point registers. 9343 * The s390x-ibm-tpf target now fully supports C++, including 9344 exceptions and threads. 9345 * Various changes to improve performance of the generated code have 9346 been implemented, including: 9347 + GCC now uses sibling calls where possible. 9348 + Condition code handling has been optimized, allowing GCC to 9349 omit redundant comparisons in certain cases. 9350 + The cost function guiding many optimizations has been refined 9351 to more accurately represent the z900 and z990 processors. 9352 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 9353 instructions are now used to avoid conditional branches in 9354 certain cases. 9355 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 9356 optimize address arithmetic required to access large stack 9357 frames. 9358 + GCC now makes more efficient use of memory-to-memory type 9359 instructions (MVC, CLC, ...). 9360 + More precise tracking of special register use allows better 9361 instruction scheduling, in particular of the function prologue 9362 and epilogue sequences. 9363 + The Java front end now generates inline code to implement 9364 integer division, instead of calling library routines. 9365 9366 SPARC 9367 9368 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 9369 -mf934 have been removed. They have been replaced with -mcpu=xxx. 9370 * The internal model used to estimate the relative cost of each 9371 instruction has been updated. It is expected to give better results 9372 on recent UltraSPARC processors. 9373 * Code generation for function prologues and epilogues has been 9374 improved, resulting in better scheduling and allowing multiple exit 9375 points in functions. 9376 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 9377 It is enabled by -mvis and provides new built-in functions for VIS 9378 instructions on UltraSPARC processors. 9379 * The option -mapp-regs has been turned on by default on Solaris too. 9380 9381 NetWare 9382 9383 * Novell NetWare (on ix86, no other hardware platform was ever really 9384 supported by this OS) has been re-enabled and the ABI supported by 9385 GCC has been brought into sync with that of MetroWerks CodeWarrior 9386 (the ABI previously supported was that of some Unix systems, which 9387 NetWare never tried to support). 9388 9389 Obsolete Systems 9390 9391 Support for a number of older systems has been declared obsolete in GCC 9392 4.0. Unless there is activity to revive them, the next release of GCC 9393 will have their sources permanently removed. 9394 9395 All GCC ports for the following processor architectures have been 9396 declared obsolete: 9397 * Intel i860 9398 * Ubicom IP2022 9399 * National Semiconductor NS32K (ns32k) 9400 * Texas Instruments TMS320C[34]x 9401 9402 Also, those for some individual systems have been obsoleted: 9403 * SPARC family 9404 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 9405 sparc86x-*-elf) 9406 + OpenBSD 32-bit (sparc-*-openbsd*) 9407 9408 Documentation improvements 9409 9410 Other significant improvements 9411 9412 * Location lists are now generated by default when compiling with 9413 debug info and optimization. Location lists provide more accurate 9414 debug info about locations of variables and they allow debugging 9415 code compiled with -fomit-frame-pointer. 9416 * The -fvisibility option has been added which allows the default ELF 9417 visibility of all symbols to be set per compilation and the new 9418 #pragma GCC visibility preprocessor command allows the setting of 9419 default ELF visibility for a region of code. Using 9420 -fvisibility=hidden especially in combination with the new 9421 -fvisibility-inlines-hidden can yield substantial improvements in 9422 output binary quality including avoiding PLT indirection overheads, 9423 reduction of the exported symbol count by up to 60% (with resultant 9424 improvements to link and load times), better scope for the 9425 optimizer to improve code and up to a 20% reduction in binary size. 9426 Using these options correctly yields a binary with a similar symbol 9427 count to a Windows DLL. 9428 Perhaps more importantly, this new feature finally allows (with 9429 careful planning) complete avoidance of symbol clashes when 9430 manually loading shared objects with RTLD_GLOBAL, thus finally 9431 solving problems many projects such as python were forced to use 9432 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 9433 can find more information about using these options at 9434 [11]https://gcc.gnu.org/wiki/Visibility. 9435 __________________________________________________________________ 9436 9437 GCC 4.0.1 9438 9439 This is the [12]list of problem reports (PRs) from GCC's bug tracking 9440 system that are known to be fixed in the 4.0.1 release. This list might 9441 not be complete (that is, it is possible that some PRs that have been 9442 fixed are not listed here). 9443 9444 GCC 4.0.2 9445 9446 This is the [13]list of problem reports (PRs) from GCC's bug tracking 9447 system that are known to be fixed in the 4.0.2 release. This list might 9448 not be complete (that is, it is possible that some PRs that have been 9449 fixed are not listed here). 9450 9451 Unfortunately, due to a release engineering failure, this release has a 9452 regression on Solaris that will affect some C++ programs. We suggest 9453 that Solaris users apply a [14]patch that corrects the problem. Users 9454 who do not wish to apply the patch should explicitly link C++ programs 9455 with the -pthreads option, even if they do not use threads. This 9456 problem has been corrected in the current 4.0 branch sources and will 9457 not be present in GCC 4.0.3. 9458 9459 GCC 4.0.3 9460 9461 Starting with this release, the function getcontext is recognized by 9462 the compiler as having the same semantics as the setjmp function. In 9463 particular, the compiler will ensure that all registers are dead before 9464 calling such a function and will emit a warning about the variables 9465 that may be clobbered after the second return from the function. 9466 9467 GCC 4.0.4 9468 9469 This is the [15]list of problem reports (PRs) from GCC's bug tracking 9470 system that are known to be fixed in the 4.0.4 release. This list might 9471 not be complete (that is, it is possible that some PRs that have been 9472 fixed are not listed here). 9473 9474 The 4.0.4 release is provided for those that require a high degree of 9475 binary compatibility with previous 4.0.x releases. For most users, the 9476 GCC team recommends that version 4.1.1 or later be used instead." 9477 9478 9479 For questions related to the use of GCC, please consult these web 9480 pages and the [16]GCC manuals. If that fails, the 9481 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9482 web pages and the development of GCC are welcome on our developer 9483 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 9484 archives. 9485 9486 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 9487 distribution of this entire article is permitted in any medium, 9488 provided this notice is preserved. 9489 9490 These pages are [21]maintained by the GCC team. Last modified 9491 2017-03-29[22]. 9492 9493 References 9494 9495 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 9496 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 9497 3. http://gcc.gnu.org/projects/tree-ssa/ 9498 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 9499 5. http://gcc.gnu.org/news/sms.html 9500 6. https://www.akkadia.org/drepper/dsohowto.pdf 9501 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 9502 8. https://itanium-cxx-abi.github.io/cxx-abi/ 9503 9. http://gcc.gnu.org/fortran/ 9504 10. https://gcc.gnu.org/install/ 9505 11. https://gcc.gnu.org/wiki/Visibility 9506 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 9507 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 9508 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 9509 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 9510 16. https://gcc.gnu.org/onlinedocs/ 9511 17. mailto:gcc-help (a] gcc.gnu.org 9512 18. mailto:gcc (a] gcc.gnu.org 9513 19. https://gcc.gnu.org/lists.html 9514 20. http://www.fsf.org/ 9515 21. https://gcc.gnu.org/about.html 9516 22. http://validator.w3.org/check/referer 9517 ====================================================================== 9518 http://gcc.gnu.org/gcc-3.4/index.html 9519 9520 GCC 3.4 Release Series 9521 9522 May 26, 2006 9523 9524 The [1]GNU project and the GCC developers are pleased to announce the 9525 release of GCC 3.4.6. 9526 9527 This release is a bug-fix release, containing fixes for regressions in 9528 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 9529 3.4.x series. 9530 9531 The GCC 3.4 release series includes numerous [2]new features, 9532 improvements, bug fixes, and other changes, thanks to an [3]amazing 9533 group of volunteers. 9534 9535 This release series is no longer maintained. 9536 9537 Release History 9538 9539 GCC 3.4.6 9540 March 6, 2006 ([4]changes) 9541 9542 GCC 3.4.5 9543 November 30, 2005 ([5]changes) 9544 9545 GCC 3.4.4 9546 May 18, 2005 ([6]changes) 9547 9548 GCC 3.4.3 9549 November 4, 2004 ([7]changes) 9550 9551 GCC 3.4.2 9552 September 6, 2004 ([8]changes) 9553 9554 GCC 3.4.1 9555 July 1, 2004 ([9]changes) 9556 9557 GCC 3.4.0 9558 April 18, 2004 ([10]changes) 9559 9560 References and Acknowledgements 9561 9562 GCC used to stand for the GNU C Compiler, but since the compiler 9563 supports several other languages aside from C, it now stands for the 9564 GNU Compiler Collection. 9565 9566 A list of [11]successful builds is updated as new information becomes 9567 available. 9568 9569 The GCC developers would like to thank the numerous people that have 9570 contributed new features, improvements, bug fixes, and other changes as 9571 well as test results to GCC. This [12]amazing group of volunteers is 9572 what makes GCC successful. 9573 9574 For additional information about GCC please refer to the [13]GCC 9575 project web site or contact the [14]GCC development mailing list. 9576 9577 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 9578 9579 9580 For questions related to the use of GCC, please consult these web 9581 pages and the [17]GCC manuals. If that fails, the 9582 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9583 web pages and the development of GCC are welcome on our developer 9584 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 9585 archives. 9586 9587 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 9588 distribution of this entire article is permitted in any medium, 9589 provided this notice is preserved. 9590 9591 These pages are [22]maintained by the GCC team. Last modified 9592 2016-09-30[23]. 9593 9594 References 9595 9596 1. http://www.gnu.org/ 9597 2. http://gcc.gnu.org/gcc-3.4/changes.html 9598 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9599 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 9600 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 9601 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 9602 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 9603 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 9604 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 9605 10. http://gcc.gnu.org/gcc-3.4/changes.html 9606 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 9607 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9608 13. http://gcc.gnu.org/index.html 9609 14. mailto:gcc (a] gcc.gnu.org 9610 15. http://gcc.gnu.org/mirrors.html 9611 16. http://gcc.gnu.org/svn.html 9612 17. https://gcc.gnu.org/onlinedocs/ 9613 18. mailto:gcc-help (a] gcc.gnu.org 9614 19. mailto:gcc (a] gcc.gnu.org 9615 20. https://gcc.gnu.org/lists.html 9616 21. http://www.fsf.org/ 9617 22. https://gcc.gnu.org/about.html 9618 23. http://validator.w3.org/check/referer 9619 ====================================================================== 9620 http://gcc.gnu.org/gcc-3.4/changes.html 9621 9622 GCC 3.4 Release Series 9623 Changes, New Features, and Fixes 9624 9625 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 9626 is now closed. 9627 9628 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting 9629 a bug, please make sure it's really GCC, and not your code, that is 9630 broken. 9631 9632 Caveats 9633 9634 * GNU Make is now required to build GCC. 9635 * With -nostdinc the preprocessor used to ignore both standard 9636 include paths and include paths contained in environment variables. 9637 It was neither documented nor intended that environment variable 9638 paths be ignored, so this has been corrected. 9639 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 9640 -fvolatile-static. It is unlikely that they worked correctly in any 9641 3.x release. 9642 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 9643 * Support for all the systems [3]obsoleted in GCC 3.3 has been 9644 removed from GCC 3.4. See below for a [4]list of systems which are 9645 obsoleted in this release. 9646 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 9647 compilers will not work. 9648 * The implementation of the [5]MIPS ABIs has changed. As a result, 9649 the code generated for certain MIPS targets will not be binary 9650 compatible with earlier releases. 9651 * In previous releases, the MIPS port had a fake "hilo" register with 9652 the user-visible name accum. This register has been removed. 9653 * The implementation of the [6]SPARC ABIs has changed. As a result, 9654 the code generated will not be binary compatible with earlier 9655 releases in certain cases. 9656 * The configure option --enable-threads=pthreads has been removed; 9657 use --enable-threads=posix instead, which should have the same 9658 effect. 9659 * Code size estimates used by inlining heuristics for C, Objective-C, 9660 C++ and Java have been redesigned significantly. As a result the 9661 parameters of -finline-insns, --param max-inline-insns-single and 9662 --param max-inline-insns-auto need to be reconsidered. 9663 * --param max-inline-slope and --param min-inline-insns have been 9664 removed; they are not needed for the new bottom-up inlining 9665 heuristics. 9666 * The new unit-at-a-time compilation scheme has several compatibility 9667 issues: 9668 + The order in which functions, variables, and top-level asm 9669 statements are emitted may have changed. Code relying on some 9670 particular ordering needs to be updated. The majority of such 9671 top-level asm statements can be replaced by section 9672 attributes. 9673 + Unreferenced static variables and functions are removed. This 9674 may result in undefined references when an asm statement 9675 refers to the variable/function directly. In that case either 9676 the variable/function shall be listed in asm statement operand 9677 or in the case of top-level asm statements the attribute used 9678 shall be used to force function/variable to be always output 9679 and considered as a possibly used by unknown code. 9680 For variables the attribute is accepted only by GCC 3.4 and 9681 newer, while for earlier versions it is sufficient to use 9682 unused to silence warnings about the variables not being 9683 referenced. To keep code portable across different GCC 9684 versions, you can use appropriate preprocessor conditionals. 9685 + Static functions now can use non-standard passing conventions 9686 that may break asm statements calling functions directly. 9687 Again the attribute used shall be used to prevent this 9688 behavior. 9689 As a temporary workaround, -fno-unit-at-a-time can be used, but 9690 this scheme may not be supported by future releases of GCC. 9691 * GCC 3.4 automatically places zero-initialized variables in the .bss 9692 section on some operating systems. Versions of GNU Emacs up to (and 9693 including) 21.3 will not work correctly when using this 9694 optimization; you can use -fno-zero-initialized-in-bss to disable 9695 it. 9696 * If GCC 3.4 is configured with --enable-threads=posix (the default 9697 on most targets that support pthreads) then _REENTRANT will be 9698 defined unconditionally by some libstdc++ headers. C++ code which 9699 relies on that macro to detect whether multi-threaded code is being 9700 compiled might change in meaning, possibly resulting in linker 9701 errors for single-threaded programs. Affected users of [7]Boost 9702 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 9703 See Bugzilla for [8]more information. 9704 9705 General Optimizer Improvements 9706 9707 * Usability of the profile feedback and coverage testing has been 9708 improved. 9709 + Performance of profiled programs has been improved by faster 9710 profile merging code. 9711 + Better use of the profile feedback for optimization (loop 9712 unrolling and loop peeling). 9713 + File locking support allowing fork() calls and parallel runs 9714 of profiled programs. 9715 + Coverage file format has been redesigned. 9716 + gcov coverage tool has been improved. 9717 + make profiledbootstrap available to build a faster compiler. 9718 Experiments made on i386 hardware showed an 11% speedup on -O0 9719 and a 7.5% speedup on -O2 compilation of a [9]large C++ 9720 testcase. 9721 + New value profiling pass enabled via -fprofile-values 9722 + New value profile transformations pass enabled via -fvpt aims 9723 to optimize some code sequences by exploiting knowledge about 9724 value ranges or other properties of the operands. At the 9725 moment a conversion of expensive divisions into cheaper 9726 operations has been implemented. 9727 + New -fprofile-generate and -fprofile-use command-line options 9728 to simplify the use of profile feedback. 9729 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 9730 Java which is enabled via -funit-at-a-time (and implied by -O2). In 9731 this scheme a whole file is parsed first and optimized later. The 9732 following basic inter-procedural optimizations are implemented: 9733 + Removal of unreachable functions and variables 9734 + Discovery of local functions (functions with static linkage 9735 whose address is never taken) 9736 + On i386, these local functions use register parameter passing 9737 conventions. 9738 + Reordering of functions in topological order of the call graph 9739 to enable better propagation of optimizing hints (such as the 9740 stack alignments needed by functions) in the back end. 9741 + Call graph based out-of-order inlining heuristics which allows 9742 to limit overall compilation unit growth (--param 9743 inline-unit-growth). 9744 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 9745 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 9746 CPU). 9747 * More realistic code size estimates used by inlining for C, 9748 Objective-C, C++ and Java. The growth of large functions can now be 9749 limited via --param large-function-insns and --param 9750 large-function-growth. 9751 * A new cfg-level loop optimizer pass replaces the old loop unrolling 9752 pass and adds two other loop transformations -- loop peeling and 9753 loop unswitching -- and also uses the profile feedback to limit 9754 code growth. (The three optimizations are enabled by 9755 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 9756 respectively). 9757 The old loop unroller still can be enabled by -fold-unroll-loops 9758 and may produce better code in some cases, especially when the 9759 webizer optimization pass is not run. 9760 * A new web construction pass enabled via -fweb (and implied by -O3) 9761 improves the quality of register allocation, CSE, first scheduling 9762 pass and some other optimization passes by avoiding re-use of 9763 pseudo registers with non-overlapping live ranges. The pass almost 9764 always improves code quality but does make debugging difficult and 9765 thus is not enabled by default by -O2 9766 The pass is especially effective as cleanup after code duplication 9767 passes, such as the loop unroller or the tracer. 9768 * Experimental implementations of superblock or trace scheduling in 9769 the second scheduling pass can be enabled via 9770 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 9771 9772 New Languages and Language specific improvements 9773 9774 Ada 9775 9776 * The Ada front end has been updated to include numerous bug fixes 9777 and enhancements. These include: 9778 + Improved project file support 9779 + Additional set of warnings about potential wrong code 9780 + Improved error messages 9781 + Improved code generation 9782 + Improved cross reference information 9783 + Improved inlining 9784 + Better run-time check elimination 9785 + Better error recovery 9786 + More efficient implementation of unbounded strings 9787 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 9788 ... 9789 + New GNAT.xxxx packages (e.g. GNAT.Strings, 9790 GNAT.Exception_Action) 9791 + New pragmas 9792 + New -gnatS switch replacing gnatpsta 9793 + Implementation of new Ada features (in particular limited 9794 with, limited aggregates) 9795 9796 C/Objective-C/C++ 9797 9798 * Precompiled headers are now supported. Precompiled headers can 9799 dramatically speed up compilation of some projects. There are some 9800 known defects in the current precompiled header implementation that 9801 will result in compiler crashes in relatively rare situations. 9802 Therefore, precompiled headers should be considered a "technology 9803 preview" in this release. Read the manual for details about how to 9804 use precompiled headers. 9805 * File handling in the preprocessor has been rewritten. GCC no longer 9806 gets confused by symlinks and hardlinks, and now has a correct 9807 implementation of #import and #pragma once. These two directives 9808 have therefore been un-deprecated. 9809 * The undocumented extension that allowed C programs to have a label 9810 at the end of a compound statement, which has been deprecated since 9811 GCC 3.0, has been removed. 9812 * The cast-as-lvalue extension has been removed for C++ and 9813 deprecated for C and Objective-C. In particular, code like this: 9814 int i; 9815 (char) i = 5; 9816 9817 or this: 9818 char *p; 9819 ((int *) p)++; 9820 9821 is no longer accepted for C++ and will not be accepted for C and 9822 Objective-C in a future version. 9823 * The conditional-expression-as-lvalue extension has been deprecated 9824 for C and Objective-C. In particular, code like this: 9825 int a, b, c; 9826 (a ? b : c) = 2; 9827 9828 will not be accepted for C and Objective-C in a future version. 9829 * The compound-expression-as-lvalue extension has been deprecated for 9830 C and Objective-C. In particular, code like this: 9831 int a, b; 9832 (a, b) = 2; 9833 9834 will not be accepted for C and Objective-C in a future version. A 9835 possible non-intrusive workaround is the following: 9836 (*(a, &b)) = 2; 9837 9838 * Several [10]built-in functions such as __builtin_popcount for 9839 counting bits, finding the highest and lowest bit in a word, and 9840 parity have been added. 9841 * The -fwritable-strings option has been deprecated and will be 9842 removed. 9843 * Many C math library functions are now recognized as built-ins and 9844 optimized. 9845 * The C, C++, and Objective-C compilers can now handle source files 9846 written in any character encoding supported by the host C library. 9847 The default input character set is taken from the current locale, 9848 and may be overridden with the -finput-charset command line option. 9849 In the future we will add support for inline encoding markers. 9850 9851 C++ 9852 9853 * G++ is now much closer to full conformance to the ISO/ANSI C++ 9854 standard. This means, among other things, that a lot of invalid 9855 constructs which used to be accepted in previous versions will now 9856 be rejected. It is very likely that existing C++ code will need to 9857 be fixed. This document lists some of the most common issues. 9858 * A hand-written recursive-descent C++ parser has replaced the 9859 YACC-derived C++ parser from previous GCC releases. The new parser 9860 contains much improved infrastructure needed for better parsing of 9861 C++ source codes, handling of extensions, and clean separation 9862 (where possible) between proper semantics analysis and parsing. The 9863 new parser fixes many bugs that were found in the old parser. 9864 * You must now use the typename and template keywords to disambiguate 9865 dependent names, as required by the C++ standard. 9866 struct K { 9867 typedef int mytype_t; 9868 }; 9869 9870 template <class T1> struct A { 9871 template <class T2> struct B { 9872 void callme(void); 9873 }; 9874 9875 template <int N> void bar(void) 9876 { 9877 // Use 'typename' to tell the parser that T1::mytype_t names 9878 // a type. This is needed because the name is dependent (in 9879 // this case, on template parameter T1). 9880 typename T1::mytype_t x; 9881 x = 0; 9882 } 9883 }; 9884 9885 template <class T> void template_func(void) 9886 { 9887 // Use 'template' to prefix member templates within 9888 // dependent types (a has type A<T>, which depends on 9889 // the template parameter T). 9890 A<T> a; 9891 a.template bar<0>(); 9892 9893 // Use 'template' to tell the parser that B is a nested 9894 // template class (dependent on template parameter T), and 9895 // 'typename' because the whole A<T>::B<int> is 9896 // the name of a type (again, dependent). 9897 typename A<T>::template B<int> b; 9898 b.callme(); 9899 } 9900 9901 void non_template_func(void) 9902 { 9903 // Outside of any template class or function, no names can be 9904 // dependent, so the use of the keyword 'typename' and 'template' 9905 // is not needed (and actually forbidden). 9906 A<K> a; 9907 a.bar<0>(); 9908 A<K>::B<float> b; 9909 b.callme(); 9910 } 9911 * In a template definition, unqualified names will no longer find 9912 members of a dependent base (as specified by [temp.dep]/3 in the 9913 C++ standard). For example, 9914 template <typename T> struct B { 9915 int m; 9916 int n; 9917 int f (); 9918 int g (); 9919 }; 9920 int n; 9921 int g (); 9922 template <typename T> struct C : B<T> { 9923 void h () 9924 { 9925 m = 0; // error 9926 f (); // error 9927 n = 0; // ::n is modified 9928 g (); // ::g is called 9929 } 9930 }; 9931 You must make the names dependent, e.g. by prefixing them with 9932 this->. Here is the corrected definition of C<T>::h, 9933 template <typename T> void C<T>::h () 9934 { 9935 this->m = 0; 9936 this->f (); 9937 this->n = 0 9938 this->g (); 9939 } 9940 As an alternative solution (unfortunately not backwards compatible 9941 with GCC 3.3), you may use using declarations instead of this->: 9942 template <typename T> struct C : B<T> { 9943 using B<T>::m; 9944 using B<T>::f; 9945 using B<T>::n; 9946 using B<T>::g; 9947 void h () 9948 { 9949 m = 0; 9950 f (); 9951 n = 0; 9952 g (); 9953 } 9954 }; 9955 * In templates, all non-dependent names are now looked up and bound 9956 at definition time (while parsing the code), instead of later when 9957 the template is instantiated. For instance: 9958 void foo(int); 9959 9960 template <int> struct A { 9961 static void bar(void){ 9962 foo('a'); 9963 } 9964 }; 9965 9966 void foo(char); 9967 9968 int main() 9969 { 9970 A<0>::bar(); // Calls foo(int), used to call foo(char). 9971 } 9972 9973 * In an explicit instantiation of a class template, you must use 9974 class or struct before the template-id: 9975 template <int N> 9976 class A {}; 9977 9978 template A<0>; // error, not accepted anymore 9979 template class A<0>; // OK 9980 * The "named return value" and "implicit typename" extensions have 9981 been removed. 9982 * Default arguments in function types have been deprecated and will 9983 be removed. 9984 * ARM-style name-injection of friend declarations has been deprecated 9985 and will be removed. For example: struct S { friend void f(); }; 9986 void g() { f(); } will not be accepted by future versions of G++; 9987 instead a declaration of "f" will need to be present outside of the 9988 scope of "S". 9989 * Covariant returns are implemented for all but varadic functions 9990 that require an adjustment. 9991 * When -pedantic is used, G++ now issues errors about spurious 9992 semicolons. For example, 9993 namespace N {}; // Invalid semicolon. 9994 void f() {}; // Invalid semicolon. 9995 * G++ no longer accepts attributes for a declarator after the 9996 initializer associated with that declarator. For example, 9997 X x(1) __attribute__((...)); 9998 is no longer accepted. Instead, use: 9999 X x __attribute__((...)) (1); 10000 * Inside the scope of a template class, the name of the class itself 10001 can be treated as either a class or a template. So GCC used to 10002 accept the class name as argument of type template, and template 10003 template parameter. However this is not C++ standard compliant. Now 10004 the name is not treated as a valid template template argument 10005 unless you qualify the name by its scope. For example, the code 10006 below no longer compiles. 10007 template <template <class> class TT> class X {}; 10008 template <class T> class Y { 10009 X<Y> x; // Invalid, Y is always a type template parameter. 10010 }; 10011 The valid code for the above example is 10012 X< ::Y> x; // Valid. 10013 (Notice the space between < and : to prevent GCC to interpret this 10014 as a digraph for [.) 10015 * Friend declarations that refer to template specializations are 10016 rejected if the template has not already been declared. For 10017 example, 10018 template <typename T> 10019 class C { 10020 friend void f<> (C&); 10021 }; 10022 is rejected. You must first declare f as a template, 10023 template <typename T> 10024 void f(T); 10025 * In case of friend declarations, every name used in the friend 10026 declaration must be accessible at the point of that declaration. 10027 Previous versions of G++ used to be less strict about this and 10028 allowed friend declarations for private class members, for example. 10029 See the ISO C++ Standard Committee's [11]defect report #209 for 10030 details. 10031 * Declaration of member functions of class templates as friends are 10032 supported. For example, 10033 template <typename T> struct A { 10034 void f(); 10035 }; 10036 class C { 10037 template <typename T> friend void A<T>::f(); 10038 }; 10039 * You must use template <> to introduce template specializations, as 10040 required by the standard. For example, 10041 template <typename T> 10042 struct S; 10043 10044 struct S<int> { }; 10045 is rejected. You must write, 10046 template <> struct S<int> {}; 10047 * G++ used to accept code like this, 10048 struct S { 10049 int h(); 10050 void f(int i = g()); 10051 int g(int i = h()); 10052 }; 10053 This behavior is not mandated by the standard. Now G++ issues an 10054 error about this code. To avoid the error, you must move the 10055 declaration of g before the declaration of f. The default arguments 10056 for g must be visible at the point where it is called. 10057 * The C++ ABI Section 3.3.3 specifications for the array construction 10058 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 10059 NULL when the allocator argument returns NULL. These changes are 10060 incorporated into the libstdc++ runtime library. 10061 * Using a name introduced by a typedef in a friend declaration or in 10062 an explicit instantiation is now rejected, as specified by the ISO 10063 C++ standard. 10064 class A; 10065 typedef A B; 10066 class C { 10067 friend class B; // error, no typedef name here 10068 friend B; // error, friend always needs class/struct/enum 10069 friend class A; // OK 10070 }; 10071 10072 template <int> class Q {}; 10073 typedef Q<0> R; 10074 template class R; // error, no typedef name here 10075 template class Q<0>; // OK 10076 * When allocating an array with a new expression, GCC used to allow 10077 parentheses around the type name. This is actually ill-formed and 10078 it is now rejected: 10079 int* a = new (int)[10]; // error, not accepted anymore 10080 int* a = new int[10]; // OK 10081 * When binding an rvalue of class type to a reference, the copy 10082 constructor of the class must be accessible. For instance, consider 10083 the following code: 10084 class A 10085 { 10086 public: 10087 A(); 10088 10089 private: 10090 A(const A&); // private copy ctor 10091 }; 10092 10093 A makeA(void); 10094 void foo(const A&); 10095 10096 void bar(void) 10097 { 10098 foo(A()); // error, copy ctor is not accessible 10099 foo(makeA()); // error, copy ctor is not accessible 10100 10101 A a1; 10102 foo(a1); // OK, a1 is a lvalue 10103 } 10104 This might be surprising at first sight, especially since most 10105 popular compilers do not correctly implement this rule ([12]further 10106 details). 10107 * When forming a pointer to member or a pointer to member function, 10108 access checks for class visibility (public, protected, private) are 10109 now performed using the qualifying scope of the name itself. This 10110 is better explained with an example: 10111 class A 10112 { 10113 public: 10114 void pub_func(); 10115 protected: 10116 void prot_func(); 10117 private: 10118 void priv_func(); 10119 }; 10120 10121 class B : public A 10122 { 10123 public: 10124 void foo() 10125 { 10126 &A::pub_func; // OK, pub_func is accessible through A 10127 &A::prot_func; // error, cannot access prot_func through A 10128 &A::priv_func; // error, cannot access priv_func through A 10129 10130 &B::pub_func; // OK, pub_func is accessible through B 10131 &B::prot_func; // OK, can access prot_func through B (within B) 10132 &B::priv_func; // error, cannot access priv_func through B 10133 } 10134 }; 10135 10136 Runtime Library (libstdc++) 10137 10138 * Optimization work: 10139 + Streamlined streambuf, filebuf, separate synched with C 10140 Standard I/O streambuf. 10141 + All formatted I/O now uses cached locale information. 10142 + STL optimizations (memory/speed for list, red-black trees as 10143 used by sets and maps). 10144 + More use of GCC builtins. 10145 + String optimizations (avoid contention on 10146 increment/decrement-and-test of the reference count in the 10147 empty-string object, constructor from input_iterators 10148 speedup). 10149 * Static linkage size reductions. 10150 * Large File Support (files larger than 2 GB on 32-bit systems). 10151 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 10152 * Generic character traits. 10153 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 10154 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 10155 * The allocator class is now standard-conformant, and two additional 10156 extension allocators have been added, mt_alloc and 10157 bitmap_allocator. 10158 * PCH support: -include bits/stdc++.h (2x compile speedup). 10159 * Rewrote __cxa_demangle with support for C++ style allocators. 10160 * New debug modes for STL containers and iterators. 10161 * Testsuite rewrite: five times as many tests, plus increasingly 10162 sophisticated tests, including I/O, MT, multi-locale, wide and 10163 narrow characters. 10164 * Use current versions of GNU "autotools" for build/configuration. 10165 10166 Objective-C 10167 10168 * The Objective-C front end has been updated to include the numerous 10169 bug fixes and enhancements previously available only in Apple's 10170 version of GCC. These include: 10171 + Structured exception (@try... @catch... @finally, @throw) and 10172 synchronization (@synchronized) support. These are accessible 10173 via the -fobjc-exceptions switch; as of this writing, they may 10174 only be used in conjunction with -fnext-runtime on Mac OS X 10175 10.3 and later. See [13]Options Controlling Objective-C 10176 Dialect for more information. 10177 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 10178 may now be encoded as 'B'. In addition, the back-end/codegen 10179 dependencies have been removed. 10180 + An overhaul of message dispatch construction, ensuring that 10181 the various receiver types (and casts thereof) are handled 10182 properly, and that correct diagnostics are issued. 10183 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 10184 (-freplace-objc-classes) debugging modes, currently available 10185 on Mac OS X 10.3 and later. See [14]Options Controlling 10186 Objective-C Dialect for more information. 10187 + Access to optimized runtime entry points (-fno-nil-receivers ) 10188 on the assumption that message receivers are never nil. This 10189 is currently available on Mac OS X 10.3 and later. See 10190 [15]Options Controlling Objective-C Dialect for more 10191 information. 10192 10193 Java 10194 10195 * Compiling a .jar file will now cause non-.class entries to be 10196 automatically compiled as resources. 10197 * libgcj has been ported to Darwin. 10198 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 10199 to gcj. 10200 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 10201 code from shared libraries. 10202 * libgcj has been much more completely merged with [16]GNU Classpath. 10203 * Class loading is now much more correct; in particular the caller's 10204 class loader is now used when that is required. 10205 * [17]Eclipse 2.x will run out of the box using gij. 10206 * Parts of java.nio have been implemented. Direct and indirect 10207 buffers work, as do fundamental file and socket operations. 10208 * java.awt has been improved, though it is still not ready for 10209 general use. 10210 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 10211 method. 10212 * The MinGW port has matured. Enhancements include socket timeout 10213 support, thread interruption, improved Runtime.exec() handling and 10214 support for accented characters in filenames. 10215 10216 Fortran 10217 10218 * Fortran improvements are listed in the [18]Fortran documentation. 10219 10220 New Targets and Target Specific Improvements 10221 10222 Alpha 10223 10224 * Several [19]built-in functions have been added such as 10225 __builtin_alpha_zap to allow utilizing the more obscure 10226 instructions of the CPU. 10227 * Parameter passing of complex arguments has changed to match the 10228 ABI. This change is incompatible with previous GCC versions, but 10229 does fix compatibility with the Tru64 compiler and several corner 10230 cases where GCC was incompatible with itself. 10231 10232 ARM 10233 10234 * Nicolas Pitre has contributed his hand-coded floating-point support 10235 code for ARM. It is both significantly smaller and faster than the 10236 existing C-based implementation, even when building applications 10237 for Thumb. The arm-elf configuration has been converted to use the 10238 new code. 10239 * Support for the Intel's iWMMXt architecture, a second generation 10240 XScale processor, has been added. Enabled at run time with the 10241 -mcpu=iwmmxt command line switch. 10242 * A new ARM target has been added: arm-wince-pe. This is similar to 10243 the arm-pe target, but it defaults to using the APCS32 ABI. 10244 * The existing ARM pipeline description has been converted to the use 10245 the [20]DFA processor pipeline model. There is not much change in 10246 code performance, but the description is now [21]easier to 10247 understand. 10248 * Support for the Cirrus EP9312 Maverick floating point co-processor 10249 added. Enabled at run time with the -mcpu=ep9312 command line 10250 switch. Note however that the multilibs to support this chip are 10251 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 10252 enable their production you will have to uncomment the entries in 10253 that file. 10254 10255 H8/300 10256 10257 * Support for long long has been added. 10258 * Support for saveall attribute has been added. 10259 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 10260 for H8/300H and H8S, which is much faster than the previous 10261 implementation. 10262 * A lot of small performance improvements. 10263 10264 IA-32/AMD64 (x86-64) 10265 10266 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 10267 -march=k8 and -mcpu=k8. 10268 * Scalar SSE code generation carefully avoids reformatting penalties, 10269 hidden dependencies and minimizes the number of uops generated on 10270 both Intel and AMD CPUs. 10271 * Vector MMX and SSE operands are now passed in registers to improve 10272 performance and match the argument passing convention used by the 10273 Intel C++ Compiler. As a result it is not possible to call 10274 functions accepting vector arguments compiled by older GCC version. 10275 * Conditional jump elimination is now more aggressive on modern CPUs. 10276 * The Athlon ports has been converted to use the DFA processor 10277 pipeline description. 10278 * Optimization of indirect tail calls is now possible in a similar 10279 fashion as direct sibcall optimization. 10280 * Further small performance improvements. 10281 * -m128bit-long-double is now less buggy. 10282 * __float128 support in 64-bit compilation. 10283 * Support for data structures exceeding 2GB in 64-bit mode. 10284 * -mcpu has been renamed to -mtune. 10285 10286 IA-64 10287 10288 * Tuning code for the Itanium 2 processor has been added. The 10289 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 10290 enabled by default now. To generate code tuned for Itanium 1 the 10291 option -mtune=itanium1 should be used. 10292 * [22]DFA processor pipeline descriptions for the IA-64 processors 10293 have been added. This resulted in about 3% improvement on the 10294 SPECInt2000 benchmark for Itanium 2. 10295 * Instruction bundling for the IA-64 processors has been rewritten 10296 using the DFA pipeline hazard recognizer. It resulted in about 60% 10297 compiler speedup on the SPECInt2000 C programs. 10298 10299 M32R 10300 10301 * Support for the M32R/2 processor has been added by Renesas. 10302 * Support for an M32R GNU/Linux target and PIC code generation has 10303 been added by Renesas. 10304 10305 M68000 10306 10307 * Bernardo Innocenti (Develer S.r.l.) has contributed the 10308 m68k-uclinux target, based on former work done by Paul Dale 10309 (SnapGear Inc.). Code generation for the ColdFire processors family 10310 has been enhanced and extended to support the MCF 53xx and MCF 54xx 10311 cores, integrating former work done by Peter Barada (Motorola). 10312 10313 MIPS 10314 10315 Processor-specific changes 10316 10317 * Support for the RM7000 and RM9000 processors has been added. It can 10318 be selected using the -march compiler option and should work with 10319 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 10320 * Support for revision 2 of the MIPS32 ISA has been added. It can be 10321 selected with the command-line option -march=mips32r2. 10322 * There is a new option, -mfix-sb1, to work around certain SB-1 10323 errata. 10324 10325 Configuration 10326 10327 * It is possible to customize GCC using the following configure-time 10328 options: 10329 + --with-arch, which specifies the default value of the -march 10330 option. 10331 + --with-tune, which specifies the default value of the -mtune 10332 option. 10333 + --with-abi, which specifies the default ABI. 10334 + --with-float=soft, which tells GCC to use software floating 10335 point by default. 10336 + --with-float=hard, which tells GCC to use hardware floating 10337 point by default. 10338 * A 64-bit GNU/Linux port has been added. The associated 10339 configurations are mips64-linux-gnu and mips64el-linux-gnu. 10340 * The 32-bit GNU/Linux port now supports Java. 10341 * The IRIX 6 configuration now supports the o32 ABI and will build 10342 o32 multilibs by default. This support is compatible with both 10343 binutils and the SGI tools, but note that several features, 10344 including debugging information and DWARF2 exception handling, are 10345 only available when using the GNU assembler. Use of the GNU 10346 assembler and linker (version 2.15 or above) is strongly 10347 recommended. 10348 * The IRIX 6 configuration now supports 128-bit long doubles. 10349 * There are two new RTEMS-specific configurations, mips-rtems and 10350 mipsel-rtems. 10351 * There are two new *-elf configurations, mipsisa32r2-elf and 10352 mipsisa32r2el-elf. 10353 10354 General 10355 10356 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 10357 will break binary compatibility with earlier releases. 10358 * GCC can now use explicit relocation operators when generating 10359 -mabicalls code. This behavior is controlled by -mexplicit-relocs 10360 and can have several performance benefits. For example: 10361 + It allows for more optimization of GOT accesses, including 10362 better scheduling and redundancy elimination. 10363 + It allows sibling calls to be implemented as jumps. 10364 + n32 and n64 leaf functions can use a call-clobbered global 10365 pointer instead of $28. 10366 + The code to set up $gp can be removed from functions that 10367 don't need it. 10368 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 10369 option is equivalent to the assembler's -xgot option and should be 10370 used instead of -Wa,-xgot. 10371 * Frame pointer elimination is now supported when generating 64-bit 10372 MIPS16 code. 10373 * Inline block moves have been optimized to take more account of 10374 alignment information. 10375 * Many internal changes have been made to the MIPS port, mostly aimed 10376 at reducing the reliance on assembler macros. 10377 10378 PowerPC 10379 10380 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 10381 [24]ABI incompatibilities regarding the way parameters are passed 10382 during functions calls. These changes may result in incompatibility 10383 between code compiled with GCC 3.3 and GCC 3.4. 10384 10385 PowerPC Darwin 10386 10387 * Support for shared/dylib gcc libraries has been added. It is 10388 enabled by default on powerpc-apple-darwin7.0.0 and up. 10389 * Libgcj is enabled by default. On systems older than 10390 powerpc-apple-darwin7.0.0 you need to install dlcompat. 10391 * 128-bit IBM extended precision format support added for long 10392 double. 10393 10394 PowerPC64 GNU/Linux 10395 10396 * By default, PowerPC64 GNU/Linux now uses natural alignment of 10397 structure elements. The old four byte alignment for double, with 10398 special rules for a struct starting with a double, can be chosen 10399 with -malign-power. This change may result in incompatibility 10400 between code compiled with GCC 3.3 and GCC 3.4. 10401 * -mabi=altivec is now the default rather than -mabi=no-altivec. 10402 * 128-bit IBM extended precision format support added for long 10403 double. 10404 10405 S/390 and zSeries 10406 10407 * New command-line options allow to specify the intended execution 10408 environment for generated code: 10409 + -mesa/-mzarch allows to specify whether to generate code 10410 running in ESA/390 mode or in z/Architecture mode (this is 10411 applicable to 31-bit code only). 10412 + -march allows to specify a minimum processor architecture 10413 level (g5, g6, z900, or z990). 10414 + -mtune allows to specify which processor to tune for. 10415 * It is possible to customize GCC using the following configure-time 10416 options: 10417 + --with-mode, which specifies whether to default to assuming 10418 ESA/390 or z/Architecture mode. 10419 + --with-arch, which specifies the default value of the -march 10420 option. 10421 + --with-tune, which specifies the default value of the -mtune 10422 option. 10423 * Support for the z990 processor has been added, and can be selected 10424 using -march=z990 or -mtune=z990. This includes instruction 10425 scheduling tuned for the superscalar instruction pipeline of the 10426 z990 processor as well as support for all new instructions provided 10427 by the long-displacement facility. 10428 * Support to generate 31-bit code optimized for zSeries processors 10429 (running in ESA/390 or in z/Architecture mode) has been added. This 10430 can be selected using -march=z900 and -mzarch respectively. 10431 * Instruction scheduling for the z900 and z990 processors now uses 10432 the DFA pipeline hazard recognizer. 10433 * GCC no longer generates code to maintain a stack backchain, 10434 previously used to generate stack backtraces for debugging 10435 purposes. As replacement that does not incur runtime overhead, 10436 DWARF-2 call frame information is provided by GCC; this is 10437 supported by GDB 6.1. The old behavior can be restored using the 10438 -mbackchain option. 10439 * The stack frame size of functions may now exceed 2 GB in 64-bit 10440 code. 10441 * A port for the 64-bit IBM TPF operating system has been added; the 10442 configuration is s390x-ibm-tpf. This configuration is supported as 10443 cross-compilation target only. 10444 * Various changes to improve the generated code have been 10445 implemented, including: 10446 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 10447 instructions to significantly speed up many floating-point 10448 applications. 10449 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 10450 WITH BORROW instructions to speed up long long arithmetic. 10451 + GCC now uses the SEARCH STRING instruction to implement 10452 strlen(). 10453 + In many cases, function call overhead for 31-bit code has been 10454 reduced by placing the literal pool after the function code 10455 instead of after the function prolog. 10456 + Register 14 is no longer reserved in 64-bit code. 10457 + Handling of global register variables has been improved. 10458 10459 SPARC 10460 10461 * The option -mflat is deprecated. 10462 * Support for large (> 2GB) frames has been added to the 64-bit port. 10463 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 10464 will break binary compatibility with earlier releases. 10465 * The default debugging format has been switched from STABS to 10466 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 10467 the default debugging format for 64-bit code on Solaris. 10468 10469 SuperH 10470 10471 * Support for the SH2E processor has been added. Enabled at run time 10472 with the -m2e command line switch, or at configure time by 10473 specifying sh2e as the machine part of the target triple. 10474 10475 V850 10476 10477 * Support for the Mitsubishi V850E1 processor has been added. This is 10478 a variant of the V850E processor with some additional debugging 10479 instructions. 10480 10481 Xtensa 10482 10483 * Several ABI bugs have been fixed. Unfortunately, these changes 10484 break binary compatibility with earlier releases. 10485 + For big-endian processors, the padding of aggregate return 10486 values larger than a word has changed. If the size of an 10487 aggregate return value is not a multiple of 32 bits, previous 10488 versions of GCC inserted padding in the most-significant bytes 10489 of the first return value register. Aggregates larger than a 10490 word are now padded in the least-significant bytes of the last 10491 return value register used. Aggregates smaller than a word are 10492 still padded in the most-significant bytes. The return value 10493 padding has not changed for little-endian processors. 10494 + Function arguments with 16-byte alignment are now properly 10495 aligned. 10496 + The implementation of the va_list type has changed. A va_list 10497 value created by va_start from a previous release cannot be 10498 used with va_arg from this release, or vice versa. 10499 * More processor configuration options for Xtensa processors are 10500 supported: 10501 + the ABS instruction is now optional; 10502 + the ADDX* and SUBX* instructions are now optional; 10503 + an experimental CONST16 instruction can be used to synthesize 10504 constants instead of loading them from constant pools. 10505 These and other Xtensa processor configuration options can no 10506 longer be enabled or disabled by command-line options; the 10507 processor configuration must be specified by the xtensa-config.h 10508 header file when building GCC. Additionally, the 10509 -mno-serialize-volatile option is no longer supported. 10510 10511 Obsolete Systems 10512 10513 Support for a number of older systems has been declared obsolete in GCC 10514 3.4. Unless there is activity to revive them, the next release of GCC 10515 will have their sources permanently removed. 10516 10517 All configurations of the following processor architectures have been 10518 declared obsolete: 10519 * Mitsubishi D30V, d30v-* 10520 * AT&T DSP1600 and DSP1610, dsp16xx-* 10521 * Intel 80960, i960 10522 10523 Also, some individual systems have been obsoleted: 10524 * ARM Family 10525 + Support for generating code for operation in APCS/26 mode 10526 (-mapcs-26). 10527 * IBM ESA/390 10528 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 10529 maintained and supported.) 10530 * Intel 386 family 10531 + MOSS, i?86-moss-msdos and i?86-*-moss* 10532 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 10533 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 10534 i?86-*-freebsd2* 10535 + GNU/Linux with a.out object format, i?86-linux*aout* 10536 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 10537 + Interix versions before Interix 3, i?86-*-interix 10538 + Mach microkernel, i?86-mach* 10539 + SCO UnixWare with UDK, i?86-*-udk* 10540 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 10541 + VSTa microkernel, i386-*-vsta 10542 * Motorola M68000 family 10543 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 10544 + NetBSD with a.out object format (before NetBSD 1.4), 10545 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 10546 + Generic System V r.4, m68k-*-sysv4* 10547 * VAX 10548 + Generic VAX, vax-*-* (This is generic VAX only; we have not 10549 obsoleted any VAX triples for specific operating systems.) 10550 10551 Documentation improvements 10552 10553 Other significant improvements 10554 10555 * The build system has undergone several significant cleanups. 10556 Subdirectories will only be configured if they are being built, and 10557 all subdirectory configures are run from the make command. The top 10558 level has been autoconfiscated. 10559 * Building GCC no longer writes to its source directory. This should 10560 help those wishing to share a read-only source directory over NFS 10561 or build from a CD. The exceptions to this feature are if you 10562 configure with either --enable-maintainer-mode or 10563 --enable-generated-files-in-srcdir. 10564 * The -W warning option has been renamed to -Wextra, which is more 10565 easily understood. The older spelling will be retained for 10566 backwards compatibility. 10567 * Substantial improvements in compile time have been made, 10568 particularly for non-optimizing compilations. 10569 __________________________________________________________________ 10570 10571 GCC 3.4.0 10572 10573 Bug Fixes 10574 10575 A vast number of bugs have been fixed in 3.4.0, too many to publish a 10576 complete list here. [26]Follow this link to query the Bugzilla database 10577 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 10578 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 10579 regressions. 10580 __________________________________________________________________ 10581 10582 GCC 3.4.1 10583 10584 Bug Fixes 10585 10586 This section lists the problem reports (PRs) from GCC's bug tracking 10587 system that are known to be fixed in the 3.4.1 release. This list might 10588 not be complete (that is, it is possible that some PRs that have been 10589 fixed are not listed here). 10590 10591 Bootstrap failures 10592 10593 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 10594 emitted - PIC related 10595 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 10596 * [29]14760 A bug in configure.in prevents using both 10597 --program-suffix and --program-prefix 10598 * [30]14671 [hppa64] bootstrap fails: ICE in 10599 save_call_clobbered_regs, in caller_save.c 10600 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 10601 Alpha 10602 * [32]15178 Solaris 9/x86 fails linking after stage 3 10603 10604 Multi-platform internal compiler errors (ICEs) 10605 10606 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 10607 input 10608 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 10609 * [35]14810 (c++) tree check failures with invalid code involving 10610 templates 10611 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 10612 cp/parser.c 10613 * [37]15044 (c++) ICE on syntax error, template header 10614 * [38]15057 (c++) Compiling of conditional value throw constructs 10615 cause a segmentation violation 10616 * [39]15064 (c++) typeid of template parameter gives ICE 10617 * [40]15142 (c++) ICE when passing a string where a char* is expected 10618 in a throw statement 10619 * [41]15159 ICE in rtl_verify_flow_info_1 10620 * [42]15165 (c++) ICE in instantiate_template 10621 * [43]15193 Unary minus using pointer to V4SF vector causes 10622 -fforce-mem to exhaust all memory 10623 * [44]15209 (c++) Runs out of memory with packed structs 10624 * [45]15227 (c++) Trouble with invalid function definition 10625 * [46]15285 (c++) instantiate_type ICE when forming pointer to 10626 template function 10627 * [47]15299 (c++) ICE in resolve_overloaded_unification 10628 * [48]15329 (c++) ICE on constructor of member template 10629 * [49]15550 ICE in extract_insn, in recog.c 10630 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 10631 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 10632 cp/name-lookup.c 10633 * [52]15666 [unit-at-a-time] Gcc abort on valid code 10634 * [53]15696 (c++) ICE with bad pointer-to-member code 10635 * [54]15701 (c++) ICE with friends and template template parameter 10636 * [55]15761 ICE in do_SUBST, in combine.c 10637 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 10638 10639 Ada 10640 10641 * [57]14538 All RTEMS targets broken for gnat 10642 10643 C front end 10644 10645 * [58]12391 missing warning about assigning to an incomplete type 10646 * [59]14649 atan(1.0) should not be a constant expression 10647 * [60]15004 [unit-at-a-time] no warning for unused paramater in 10648 static function 10649 * [61]15749 --pedantic-errors behaves differently from --pedantic 10650 with C-compiler on GNU/Linux 10651 10652 C++ compiler and library 10653 10654 * [62]10646 non-const reference is incorrectly matched in a "const T" 10655 partial specialization 10656 * [63]12077 wcin.rdbuf()->in_avail() return value too high 10657 * [64]13598 enc_filebuf doesn't work 10658 * [65]14211 const_cast returns lvalue but should be rvalue 10659 * [66]14220 num_put::do_put() undesired float/double behavior 10660 * [67]14245 problem with user-defined allocators in std::basic_string 10661 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 10662 const_iterator 10663 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 10664 FILE* 10665 * [70]14668 no warning anymore for reevaluation of declaration 10666 * [71]14775 LFS (large file support) tests missing 10667 * [72]14821 Duplicate namespace alias declaration should not conflict 10668 * [73]14930 Friend declaration ignored 10669 * [74]14932 cannot use offsetof to get offsets of array elements in 10670 g++ 3.4.0 10671 * [75]14950 [non unit-at-a-time] always_inline does not mix with 10672 templates and -O0 10673 * [76]14962 g++ ignores #pragma redefine_extname 10674 * [77]14975 Segfault on low-level write error during imbue 10675 * [78]15002 Linewise stream input is unusably slow (std::string slow) 10676 * [79]15025 compiler accepts redeclaration of template as 10677 non-template 10678 * [80]15046 [arm] Math functions misdetected by cross configuration 10679 * [81]15069 a bit test on a variable of enum type is miscompiled 10680 * [82]15074 g++ -lsupc++ still links against libstdc++ 10681 * [83]15083 spurious "statement has no effect" warning 10682 * [84]15096 parse error with templates and pointer to const member 10683 * [85]15287 combination of operator[] and operator .* fails in 10684 templates 10685 * [86]15317 __attribute__ unused in first parameter of constructor 10686 gives error 10687 * [87]15337 sizeof on incomplete type diagnostic 10688 * [88]15361 bitset<>::_Find_next fails 10689 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 10690 namespaces 10691 * [90]15427 valid code results in incomplete type error 10692 * [91]15471 Incorrect member pointer offsets in anonymous 10693 structs/unions 10694 * [92]15503 nested template problem 10695 * [93]15507 compiler hangs while laying out union 10696 * [94]15542 operator & and template definitions 10697 * [95]15565 SLES9: leading + sign for unsigned int with showpos 10698 * [96]15625 friend defined inside a template fails to find static 10699 function 10700 * [97]15629 Function templates, overloads, and friend name injection 10701 * [98]15742 'noreturn' attribute ignored in method of template 10702 functions. 10703 * [99]15775 Allocator::pointer consistently ignored 10704 * [100]15821 Duplicate namespace alias within namespace rejected 10705 * [101]15862 'enum yn' fails (confict with undeclared builtin) 10706 * [102]15875 rejects pointer to member in template 10707 * [103]15877 valid code using templates and anonymous enums is 10708 rejected 10709 * [104]15947 Puzzling error message for wrong destructor declaration 10710 in template class 10711 * [105]16020 cannot copy __gnu_debug::bitset 10712 * [106]16154 input iterator concept too restrictive 10713 * [107]16174 deducing top-level consts 10714 10715 Java 10716 10717 * [108]14315 Java compiler is not parallel make safe 10718 10719 Fortran 10720 10721 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 10722 10723 Objective-C 10724 10725 * [110]7993 private variables cannot be shadowed in subclasses 10726 10727 Optimization bugs 10728 10729 * [111]15228 useless copies of floating point operands 10730 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 10731 functions not optimized away 10732 * [113]15945 Incorrect floating point optimization 10733 * [114]15526 ftrapv aborts on 0 * (-1) 10734 * [115]14690 Miscompiled POOMA tests 10735 * [116]15112 GCC generates code to write to unchanging memory 10736 10737 Preprocessor 10738 10739 * [117]15067 Minor glitch in the source of cpp 10740 10741 Main driver program bugs 10742 10743 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 10744 ldstyle_liblookup 10745 10746 x86-specific (Intel/AMD) 10747 10748 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 10749 section} 10750 10751 HPPA-specific 10752 10753 * [120]14782 GCC produces an unaligned data access at -O2 10754 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 10755 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 10756 10757 IA64-specific 10758 10759 * [123]14610 __float80 constants incorrectly emitted 10760 * [124]14813 init_array sections are initialized in the wrong order 10761 * [125]14857 GCC segfault on duplicated asm statement 10762 * [126]15598 Gcc 3.4 ICE on valid code 10763 * [127]15653 Gcc 3.4 ICE on valid code 10764 10765 MIPS-specific 10766 10767 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 10768 -mno-split-addresses -mno-explicit-relocs 10769 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 10770 2.14.91 10771 * [130]16144 Bogus reference to __divdf3 when -O1 10772 * [131]16176 Miscompilation of unaligned data in MIPS backend 10773 10774 PowerPC-specific 10775 10776 * [132]11591 ICE in gcc.dg/altivec-5.c 10777 * [133]12028 powerpc-eabispe produces bad sCOND operation 10778 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 10779 * [135]14567 long double and va_arg complex args 10780 * [136]14715 Altivec stack layout may overlap gpr save with stack 10781 temps 10782 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 10783 option is used. 10784 * [138]14924 Compiler ICE on valid code 10785 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 10786 * [140]15106 vector varargs failure passing from altivec to 10787 non-altivec code for -m32 10788 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 10789 half-word operation 10790 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 10791 and stvx 10792 * [143]15662 Segmentation fault when an exception is thrown - even if 10793 try and catch are specified 10794 10795 s390-specific 10796 10797 * [144]15054 Bad code due to overlapping stack temporaries 10798 10799 SPARC-specific 10800 10801 * [145]15783 ICE with union assignment in 64-bit mode 10802 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 10803 R_SPARC_UA32" 10804 10805 x86-64-specific 10806 10807 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 10808 * [148]14723 Backported -march=nocona from mainline 10809 * [149]15290 __float128 failed to pass to function properly 10810 10811 Cygwin/Mingw32-specific 10812 10813 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 10814 conformant to MS layout 10815 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 10816 worker on windows32 targets 10817 10818 Bugs specific to embedded processors 10819 10820 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 10821 varaible on stack 10822 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 10823 gcc continues to use the register as if it was not clobbered 10824 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 10825 TARGET_COLDFIRE 10826 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 10827 * [156]14457 [m6811hc] ICE with simple c++ source 10828 * [157]14542 [m6811hc] ICE on simple source 10829 * [158]15100 [SH] cc1plus got hang-up on 10830 libstdc++-v3/testsuite/abi_check.cc 10831 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 10832 cris-* 10833 * [160]15396 [SH] ICE with -O2 -fPIC 10834 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 10835 ColdFire 10836 10837 Testsuite problems (compiler not affected) 10838 10839 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 10840 * [163]15488 (libstdc++) possibly insufficient file permissions for 10841 executing test suite 10842 * [164]15489 (libstdc++) testsuite_files determined incorrectly 10843 10844 Documentation bugs 10845 10846 * [165]13928 (libstdc++) no whatis info in some man pages generated 10847 by doxygen 10848 * [166]14150 Ada documentation out of date 10849 * [167]14949 (c++) Need to document method visibility changes 10850 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 10851 __________________________________________________________________ 10852 10853 GCC 3.4.2 10854 10855 Bug Fixes 10856 10857 This section lists the problem reports (PRs) from GCC's bug tracking 10858 system that are known to be fixed in the 3.4.2 release. This list might 10859 not be complete (that is, it is possible that some PRs that have been 10860 fixed are not listed here). 10861 10862 Bootstrap failures and issues 10863 10864 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 10865 libstdc++-v3/testsuite 10866 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 10867 profiledbootstrap does not work with the built compiler 10868 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 10869 10870 Multi-platform internal compiler errors (ICEs) 10871 10872 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 10873 cp_parser_class_specifier, in cp/parser.c 10874 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 10875 * [174]15461 (c++) ICE due to NRV and inlining 10876 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 10877 * [176]16180 ICE: segmentation fault in RTL optimization 10878 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 10879 * [178]16408 ICE: in delete_insn, in cfgrtl.c 10880 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 10881 the name of any other entity 10882 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 10883 * [181]16706 (c++) ICE in finish_member_declaration, in 10884 cp/semantics.c 10885 * [182]16810 (c++) Legal C++ program with cast gives ICE in 10886 build_ptrmemfunc 10887 * [183]16851 (c++) ICE when throwing a comma expression 10888 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 10889 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 10890 cp/typeck.c 10891 * [186]16905 (c++) ICE (segfault) with exceptions 10892 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 10893 redefinition 10894 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 10895 (identifier_node) in dependent_template_p, in cp/pt.c 10896 10897 Preprocessor bugs 10898 10899 * [189]16366 Preprocessor option -remap causes memory corruption 10900 10901 Optimization 10902 10903 * [190]15345 unreferenced nested inline functions not optimized away 10904 * [191]16590 Incorrect execution when compiling with -O2 10905 * [192]16693 Bitwise AND is lost when used within a cast to an enum 10906 of the same precision 10907 * [193]17078 Jump into if(0) substatement fails 10908 10909 Problems in generated debug information 10910 10911 * [194]13956 incorrect stabs for nested local variables 10912 10913 C front end bugs 10914 10915 * [195]16684 GCC should not warn about redundant redeclarations of 10916 built-ins 10917 10918 C++ compiler and library 10919 10920 * [196]12658 Thread safety problems in locale::global() and 10921 locale::locale() 10922 * [197]13092 g++ accepts invalid pointer-to-member conversion 10923 * [198]15320 Excessive memory consumption 10924 * [199]16246 Incorrect template argument deduction 10925 * [200]16273 Memory exhausted when using nested classes and virtual 10926 functions 10927 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 10928 * [202]16411 undefined reference to 10929 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 10930 >::file() 10931 * [203]16489 G++ incorrectly rejects use of a null constant integral 10932 expression as a null constant pointer 10933 * [204]16618 offsetof fails with constant member 10934 * [205]16637 syntax error reported for valid input code 10935 * [206]16717 __attribute__((constructor)) broken in C++ 10936 * [207]16813 compiler error in DEBUG version of range insertion 10937 std::map::insert 10938 * [208]16853 pointer-to-member initialization from incompatible one 10939 accepted 10940 * [209]16889 ambiguity is not detected 10941 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 10942 10943 Java compiler and library 10944 10945 * [211]7587 direct threaded interpreter not thread-safe 10946 * [212]16473 ServerSocket accept() leaks file descriptors 10947 * [213]16478 Hash synchronization deadlock with finalizers 10948 10949 Alpha-specific 10950 10951 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 10952 * [215]16974 could not split insn (ice in final_scan_insn, in 10953 final.c) 10954 10955 x86-specific 10956 10957 * [216]16298 ICE in output_operand 10958 * [217]17113 ICE with SSE2 intrinsics 10959 10960 x86-64 specific 10961 10962 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 10963 10964 MIPS-specific 10965 10966 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 10967 * [220]16325 [mips64] value profiling clobbers gp on mips 10968 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 10969 char[]s 10970 * [222]16380 [mips64] Use of uninitialised register after dbra 10971 conversion 10972 * [223]16407 [mips64] Unaligned access to local variables 10973 * [224]16643 [mips64] verify_local_live_at_start ICE after 10974 crossjumping & cfgcleanup 10975 10976 ARM-specific 10977 10978 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 10979 off by 1 10980 * [226]15948 THUMB: ICE with non-commutative cbranch 10981 * [227]17019 THUMB: bad switch statement in md code for 10982 addsi3_cbranch_scratch 10983 10984 IA64-specific 10985 10986 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 10987 (-mtune=merced) 10988 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 10989 (-mtune=itanium) 10990 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 10991 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 10992 result 10993 * [232]16445 ICE on valid code: don't count ignored insns 10994 * [233]16490 ICE (segfault) while compiling with -fprofile-use 10995 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 10996 10997 PowerPC-specific 10998 10999 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 11000 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 11001 issue) 11002 11003 SPARC-specific 11004 11005 * [237]16199 ICE while compiling apache 2.0.49 11006 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 11007 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 11008 11009 Bugs specific to embedded processors 11010 11011 * [240]16379 [m32r] can't output large model function call of memcpy 11012 * [241]17093 [m32r] ICE with -msdata=use -O0 11013 * [242]17119 [m32r] ICE at switch case 0x8000 11014 11015 DJGPP-specific 11016 11017 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 11018 11019 Alpha Tru64-specific 11020 11021 * [244]16210 libstdc++ gratuitously omits "long long" I/O 11022 11023 Testsuite, documentation issues (compiler is not affected): 11024 11025 * [245]15488 (libstdc++) possibly insufficient file permissions for 11026 executing test suite 11027 * [246]16250 ada/doctools runs makeinfo even in release tarball 11028 __________________________________________________________________ 11029 11030 GCC 3.4.3 11031 11032 This is the [247]list of problem reports (PRs) from GCC's bug tracking 11033 system that are known to be fixed in the 3.4.3 release. This list might 11034 not be complete (that is, it is possible that some PRs that have been 11035 fixed are not listed here). 11036 11037 Bootstrap failures 11038 11039 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 11040 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 11041 when undeclared 11042 11043 Internal compiler errors (ICEs) affecting multiple platforms 11044 11045 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 11046 .class files 11047 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 11048 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 11049 directive 11050 * [253]16566 ICE with flexible arrays 11051 * [254]17023 ICE with nested functions in parameter declaration 11052 * [255]17027 ICE with noreturn function in loop at -O2 11053 * [256]17524 ICE in grokdeclarator, in cp/decl.c 11054 * [257]17826 (c++) ICE in cp_tree_equal 11055 11056 C and optimization bugs 11057 11058 * [258]15526 -ftrapv aborts on 0 * (-1) 11059 * [259]16999 #ident stopped working 11060 * [260]17503 quadratic behaviour in invalid_mode_change_p 11061 * [261]17581 Long long arithmetic fails inside a switch/case 11062 statement when compiled with -O2 11063 * [262]18129 -fwritable-strings doesn't work 11064 11065 C++ compiler and library bugs 11066 11067 * [263]10975 incorrect initial ostringstream::tellp() 11068 * [264]11722 Unbuffered filebuf::sgetn is slow 11069 * [265]14534 Unrecognizing static function as a template parameter 11070 when its return value is also templated 11071 * [266]15172 Copy constructor optimization in aggregate 11072 initialization 11073 * [267]15786 Bad error message for frequently occuring error. 11074 * [268]16162 Rejects valid member-template-definition 11075 * [269]16612 empty basic_strings can't live in shared memory 11076 * [270]16715 std::basic_iostream is instantiated when used, even 11077 though instantiations are already contained in libstdc++ 11078 * [271]16848 code in /ext/demangle.h appears broken 11079 * [272]17132 GCC fails to eliminate function template specialization 11080 when argument deduction fails 11081 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 11082 in ropeimpl.h 11083 * [274]17327 use of `enumeral_type' in template type unification 11084 * [275]17393 "unused variable '._0'" warning with -Wall 11085 * [276]17501 Confusion with member templates 11086 * [277]17537 g++ not passing -lstdc++ to linker when all command line 11087 arguments are libraries 11088 * [278]17585 usage of unqualified name of static member from within 11089 class not allowed 11090 * [279]17821 Poor diagnostic for using "." instead of "->" 11091 * [280]17829 wrong error: call of overloaded function is ambiguous 11092 * [281]17851 Misleading diagnostic for invalid function declarations 11093 with undeclared types 11094 * [282]17976 Destructor is called twice 11095 * [283]18020 rejects valid definition of enum value in template 11096 * [284]18093 bogus conflict in namespace aliasing 11097 * [285]18140 C++ parser bug when using >> in templates 11098 11099 Fortran 11100 11101 * [286]17541 data statements with double precision constants fail 11102 11103 x86-specific 11104 11105 * [287]17853 -O2 ICE for MMX testcase 11106 11107 SPARC-specific 11108 11109 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 11110 11111 Darwin-specific 11112 11113 * [289]17167 FATAL:Symbol L_foo$stub already defined. 11114 11115 AIX-specific 11116 11117 * [290]17277 could not catch an exception when specified -maix64 11118 11119 Solaris-specific 11120 11121 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 11122 missing from system libraries 11123 11124 HP/UX specific: 11125 11126 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 11127 11128 ARM-specific 11129 11130 * [293]17384 ICE with mode attribute on structures 11131 11132 MIPS-specific 11133 11134 * [294]17770 No NOP after LWL with -mips1 11135 11136 Other embedded target specific 11137 11138 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 11139 * [296]14064 [avr-elf] -fdata-sections triggers ICE 11140 * [297]14678 [m68hc11-elf] gcc ICE 11141 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 11142 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 11143 target 11144 * [300]15886 [SH] Miscompilation with -O2 -fPIC 11145 * [301]16884 [avr-elf] [fweb related] bug while initializing 11146 variables 11147 11148 Bugs relating to debugger support 11149 11150 * [302]13841 missing debug info for _Complex function arguments 11151 * [303]15860 [big-endian targets] No DW_AT_location debug info is 11152 emitted for formal arguments to a function that uses "register" 11153 qualifiers 11154 11155 Testsuite issues (compiler not affected) 11156 11157 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 11158 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 11159 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 11160 testsuite 11161 11162 Documentation 11163 11164 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 11165 should be en_GB 11166 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 11167 document broken shell 11168 * [309]16406 USE_LD_AS_NEEDED undocumented 11169 __________________________________________________________________ 11170 11171 GCC 3.4.4 11172 11173 This is the [310]list of problem reports (PRs) from GCC's bug tracking 11174 system that are known to be fixed in the 3.4.4 release. This list might 11175 not be complete (that is, it is possible that some PRs that have been 11176 fixed are not listed here). 11177 __________________________________________________________________ 11178 11179 GCC 3.4.5 11180 11181 This is the [311]list of problem reports (PRs) from GCC's bug tracking 11182 system that are known to be fixed in the 3.4.5 release. This list might 11183 not be complete (that is, it is possible that some PRs that have been 11184 fixed are not listed here). 11185 11186 Bootstrap issues 11187 11188 * [312]24688 sco_math fixincl breaks math.h 11189 11190 C compiler bugs 11191 11192 * [313]17188 struct Foo { } redefinition 11193 * [314]20187 wrong code for ((unsigned char)(unsigned long 11194 long)((a?a:1)&(a*b)))?0:1) 11195 * [315]21873 infinite warning loop on bad array initializer 11196 * [316]21899 enum definition accepts values to be overriden 11197 * [317]22061 ICE in find_function_data, in function.c 11198 * [318]22308 Failure to diagnose violation of constraint 6.516p2 11199 * [319]22458 ICE on missing brace 11200 * [320]22589 ICE casting to long long 11201 * [321]24101 Segfault with preprocessed source 11202 11203 C++ compiler and library bugs 11204 11205 * [322]10611 operations on vector mode not recognized in C++ 11206 * [323]13377 unexpected behavior of namespace usage directive 11207 * [324]16002 Strange error message with new parser 11208 * [325]17413 local classes as template argument 11209 * [326]17609 spurious error message after using keyword 11210 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 11211 * [328]18124 ICE with invalid template template parameter 11212 * [329]18155 typedef in template declaration not rejected 11213 * [330]18177 ICE with const_cast for undeclared variable 11214 * [331]18368 C++ error message regression 11215 * [332]16378 ICE when returning a copy of a packed member 11216 * [333]18466 int ::i; accepted 11217 * [334]18512 ICE on invalid usage of template base class 11218 * [335]18454 ICE when returning undefined type 11219 * [336]18738 typename not allowed with non-dependent qualified name 11220 * [337]18803 rejects access to operator() in template 11221 * [338]19004 ICE in uses_template_parms, in cp/pt.c 11222 * [339]19208 Spurious error about variably modified type 11223 * [340]18253 bad error message / ICE for invalid template parameter 11224 * [341]19608 ICE after friend function definition in local class 11225 * [342]19884 ICE on explicit instantiation of a non-template 11226 constructor 11227 * [343]20153 ICE when C++ template function contains anonymous union 11228 * [344]20563 Infinite loop in diagnostic (and ice after error 11229 message) 11230 * [345]20789 ICE with incomplete type in template 11231 * [346]21336 Internal compiler error when using custom new operators 11232 * [347]21768 ICE in error message due to violation of coding 11233 conventions 11234 * [348]21853 constness of pointer to data member ignored 11235 * [349]21903 Default argument of template function causes a 11236 compile-time error 11237 * [350]21983 multiple diagnostics 11238 * [351]21987 New testsuite failure 11239 g++.dg/warn/conversion-function-1.C 11240 * [352]22153 ICE on invalid template specialization 11241 * [353]22172 Internal compiler error, seg fault. 11242 * [354]21286 filebuf::xsgetn vs pipes 11243 * [355]22233 ICE with wrong number of template parameters 11244 * [356]22508 ICE after invalid operator new 11245 * [357]22545 ICE with pointer to class member & user defined 11246 conversion operator 11247 * [358]23528 Wrong default allocator in ext/hash_map 11248 * [359]23550 char_traits requirements/1.cc test bad math 11249 * [360]23586 Bad diagnostic for invalid namespace-name 11250 * [361]23624 ICE in invert_truthvalue, in fold-const.c 11251 * [362]23639 Bad error message: not a member of '<declaration error>' 11252 * [363]23797 ICE on typename outside template 11253 * [364]23965 Bogus error message: no matching function for call to 11254 'foo(<type error>)' 11255 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 11256 error> 11257 * [366]24580 virtual base class cause exception not to be caught 11258 11259 Problems in generated debug information 11260 11261 * [367]24267 Bad DWARF for altivec vectors 11262 11263 Optimizations issues 11264 11265 * [368]17810 ICE in verify_local_live_at_start 11266 * [369]17860 Wrong generated code for loop with varying bound 11267 * [370]21709 ICE on compile-time complex NaN 11268 * [371]21964 broken tail call at -O2 or more 11269 * [372]22167 Strange optimization bug when using -Os 11270 * [373]22619 Compilation failure for real_const_1.f and 11271 real_const_2.f90 11272 * [374]23241 Invalid code generated for comparison of uchar to 255 11273 * [375]23478 Miscompilation due to reloading of a var that is also 11274 used in EH pad 11275 * [376]24470 segmentation fault in cc1plus when compiling with -O 11276 * [377]24950 ICE in operand_subword_force 11277 11278 Precompiled headers problems 11279 11280 * [378]14400 Cannot compile qt-x11-free-3.3.0 11281 * [379]14940 PCH largefile test fails on various platforms 11282 11283 Preprocessor bugs 11284 11285 * [380]20239 ICE on empty preprocessed input 11286 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 11287 source directory 11288 11289 Testsuite issues 11290 11291 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 11292 i686-pc-linux-gnu 11293 11294 Alpha specific 11295 11296 * [383]21888 bootstrap failure with linker relaxation enabled 11297 11298 ARM specific 11299 11300 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 11301 * [385]23985 Memory aliasing information incorrect in inlined memcpy 11302 11303 ColdFile specific 11304 11305 * [386]16719 Illegal move of byte into address register causes 11306 compiler to ICE 11307 11308 HPPA specific 11309 11310 * [387]21723 ICE while building libgfortran 11311 * [388]21841 -mhp-ld/-mgnu-ld documentation 11312 11313 IA-64 specific 11314 11315 * [389]23644 IA-64 hardware models and configuration options 11316 documentation error 11317 * [390]24718 Shared libgcc not used for linking by default 11318 11319 M68000 specific 11320 11321 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 11322 11323 MIPS specific 11324 11325 * [392]20621 ICE in change_address_1, in emit-rtl.c 11326 11327 PowerPC and PowerPC64 specific 11328 11329 * [393]18583 error on valid code: const 11330 __attribute__((altivec(vector__))) doesn't work in arrays 11331 * [394]20191 ICE in reload_cse_simplify_operands 11332 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 11333 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 11334 * [397]23404 gij trashes args of functions with more than 8 fp args 11335 * [398]23539 C & C++ compiler generating misaligned references 11336 regardless of compiler flags 11337 * [399]24102 floatdisf2_internal2 broken 11338 * [400]24465 -mminimal-toc miscompilation of __thread vars 11339 11340 Solaris specific 11341 11342 * [401]19933 Problem with define of HUGE_VAL in math_c99 11343 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 11344 symbols 11345 11346 SPARC specific 11347 11348 * [403]19300 PCH failures on sparc-linux 11349 * [404]20301 Assembler labels have a leading "-" 11350 * [405]20673 C PCH testsuite assembly comparison failure 11351 11352 x86 and x86_64 specific 11353 11354 * [406]18582 ICE with arrays of type V2DF 11355 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 11356 -fsched2-use-traces 11357 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 11358 * [409]24315 amd64 fails -fpeephole2 11359 __________________________________________________________________ 11360 11361 GCC 3.4.6 11362 11363 This is the [410]list of problem reports (PRs) from GCC's bug tracking 11364 system that are known to be fixed in the 3.4.6 release. This list might 11365 not be complete (that is, it is possible that some PRs that have been 11366 fixed are not listed here). 11367 11368 11369 For questions related to the use of GCC, please consult these web 11370 pages and the [411]GCC manuals. If that fails, the 11371 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11372 web pages and the development of GCC are welcome on our developer 11373 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 11374 archives. 11375 11376 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 11377 distribution of this entire article is permitted in any medium, 11378 provided this notice is preserved. 11379 11380 These pages are [416]maintained by the GCC team. Last modified 11381 2016-12-30[417]. 11382 11383 References 11384 11385 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 11386 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 11387 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 11388 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 11389 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 11390 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 11391 7. http://www.boost.org/ 11392 8. https://gcc.gnu.org/PR11953 11393 9. https://gcc.gnu.org/PR8361 11394 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 11395 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 11396 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 11397 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11398 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11399 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11400 16. http://www.gnu.org/software/classpath/ 11401 17. http://www.eclipse.org/ 11402 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 11403 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 11404 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 11405 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 11406 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 11407 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 11408 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 11409 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 11410 26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 11411 27. https://gcc.gnu.org/PR10129 11412 28. https://gcc.gnu.org/PR14576 11413 29. https://gcc.gnu.org/PR14760 11414 30. https://gcc.gnu.org/PR14671 11415 31. https://gcc.gnu.org/PR15093 11416 32. https://gcc.gnu.org/PR15178 11417 33. https://gcc.gnu.org/PR12753 11418 34. https://gcc.gnu.org/PR13985 11419 35. https://gcc.gnu.org/PR14810 11420 36. https://gcc.gnu.org/PR14883 11421 37. https://gcc.gnu.org/PR15044 11422 38. https://gcc.gnu.org/PR15057 11423 39. https://gcc.gnu.org/PR15064 11424 40. https://gcc.gnu.org/PR15142 11425 41. https://gcc.gnu.org/PR15159 11426 42. https://gcc.gnu.org/PR15165 11427 43. https://gcc.gnu.org/PR15193 11428 44. https://gcc.gnu.org/PR15209 11429 45. https://gcc.gnu.org/PR15227 11430 46. https://gcc.gnu.org/PR15285 11431 47. https://gcc.gnu.org/PR15299 11432 48. https://gcc.gnu.org/PR15329 11433 49. https://gcc.gnu.org/PR15550 11434 50. https://gcc.gnu.org/PR15554 11435 51. https://gcc.gnu.org/PR15640 11436 52. https://gcc.gnu.org/PR15666 11437 53. https://gcc.gnu.org/PR15696 11438 54. https://gcc.gnu.org/PR15701 11439 55. https://gcc.gnu.org/PR15761 11440 56. https://gcc.gnu.org/PR15829 11441 57. https://gcc.gnu.org/PR14538 11442 58. https://gcc.gnu.org/PR12391 11443 59. https://gcc.gnu.org/PR14649 11444 60. https://gcc.gnu.org/PR15004 11445 61. https://gcc.gnu.org/PR15749 11446 62. https://gcc.gnu.org/PR10646 11447 63. https://gcc.gnu.org/PR12077 11448 64. https://gcc.gnu.org/PR13598 11449 65. https://gcc.gnu.org/PR14211 11450 66. https://gcc.gnu.org/PR14220 11451 67. https://gcc.gnu.org/PR14245 11452 68. https://gcc.gnu.org/PR14340 11453 69. https://gcc.gnu.org/PR14600 11454 70. https://gcc.gnu.org/PR14668 11455 71. https://gcc.gnu.org/PR14775 11456 72. https://gcc.gnu.org/PR14821 11457 73. https://gcc.gnu.org/PR14930 11458 74. https://gcc.gnu.org/PR14932 11459 75. https://gcc.gnu.org/PR14950 11460 76. https://gcc.gnu.org/PR14962 11461 77. https://gcc.gnu.org/PR14975 11462 78. https://gcc.gnu.org/PR15002 11463 79. https://gcc.gnu.org/PR15025 11464 80. https://gcc.gnu.org/PR15046 11465 81. https://gcc.gnu.org/PR15069 11466 82. https://gcc.gnu.org/PR15074 11467 83. https://gcc.gnu.org/PR15083 11468 84. https://gcc.gnu.org/PR15096 11469 85. https://gcc.gnu.org/PR15287 11470 86. https://gcc.gnu.org/PR15317 11471 87. https://gcc.gnu.org/PR15337 11472 88. https://gcc.gnu.org/PR15361 11473 89. https://gcc.gnu.org/PR15412 11474 90. https://gcc.gnu.org/PR15427 11475 91. https://gcc.gnu.org/PR15471 11476 92. https://gcc.gnu.org/PR15503 11477 93. https://gcc.gnu.org/PR15507 11478 94. https://gcc.gnu.org/PR15542 11479 95. https://gcc.gnu.org/PR15565 11480 96. https://gcc.gnu.org/PR15625 11481 97. https://gcc.gnu.org/PR15629 11482 98. https://gcc.gnu.org/PR15742 11483 99. https://gcc.gnu.org/PR15775 11484 100. https://gcc.gnu.org/PR15821 11485 101. https://gcc.gnu.org/PR15862 11486 102. https://gcc.gnu.org/PR15875 11487 103. https://gcc.gnu.org/PR15877 11488 104. https://gcc.gnu.org/PR15947 11489 105. https://gcc.gnu.org/PR16020 11490 106. https://gcc.gnu.org/PR16154 11491 107. https://gcc.gnu.org/PR16174 11492 108. https://gcc.gnu.org/PR14315 11493 109. https://gcc.gnu.org/PR15151 11494 110. https://gcc.gnu.org/PR7993 11495 111. https://gcc.gnu.org/PR15228 11496 112. https://gcc.gnu.org/PR15345 11497 113. https://gcc.gnu.org/PR15945 11498 114. https://gcc.gnu.org/PR15526 11499 115. https://gcc.gnu.org/PR14690 11500 116. https://gcc.gnu.org/PR15112 11501 117. https://gcc.gnu.org/PR15067 11502 118. https://gcc.gnu.org/PR1963 11503 119. https://gcc.gnu.org/PR15717 11504 120. https://gcc.gnu.org/PR14782 11505 121. https://gcc.gnu.org/PR14828 11506 122. https://gcc.gnu.org/PR15202 11507 123. https://gcc.gnu.org/PR14610 11508 124. https://gcc.gnu.org/PR14813 11509 125. https://gcc.gnu.org/PR14857 11510 126. https://gcc.gnu.org/PR15598 11511 127. https://gcc.gnu.org/PR15653 11512 128. https://gcc.gnu.org/PR15189 11513 129. https://gcc.gnu.org/PR15331 11514 130. https://gcc.gnu.org/PR16144 11515 131. https://gcc.gnu.org/PR16176 11516 132. https://gcc.gnu.org/PR11591 11517 133. https://gcc.gnu.org/PR12028 11518 134. https://gcc.gnu.org/PR14478 11519 135. https://gcc.gnu.org/PR14567 11520 136. https://gcc.gnu.org/PR14715 11521 137. https://gcc.gnu.org/PR14902 11522 138. https://gcc.gnu.org/PR14924 11523 139. https://gcc.gnu.org/PR14960 11524 140. https://gcc.gnu.org/PR15106 11525 141. https://gcc.gnu.org/PR16026 11526 142. https://gcc.gnu.org/PR15191 11527 143. https://gcc.gnu.org/PR15662 11528 144. https://gcc.gnu.org/PR15054 11529 145. https://gcc.gnu.org/PR15783 11530 146. https://gcc.gnu.org/PR15626 11531 147. https://gcc.gnu.org/PR14326 11532 148. https://gcc.gnu.org/PR14723 11533 149. https://gcc.gnu.org/PR15290 11534 150. https://gcc.gnu.org/PR15250 11535 151. https://gcc.gnu.org/PR15551 11536 152. https://gcc.gnu.org/PR8309 11537 153. https://gcc.gnu.org/PR13250 11538 154. https://gcc.gnu.org/PR13803 11539 155. https://gcc.gnu.org/PR14093 11540 156. https://gcc.gnu.org/PR14457 11541 157. https://gcc.gnu.org/PR14542 11542 158. https://gcc.gnu.org/PR15100 11543 159. https://gcc.gnu.org/PR15296 11544 160. https://gcc.gnu.org/PR15396 11545 161. https://gcc.gnu.org/PR15782 11546 162. https://gcc.gnu.org/PR11610 11547 163. https://gcc.gnu.org/PR15488 11548 164. https://gcc.gnu.org/PR15489 11549 165. https://gcc.gnu.org/PR13928 11550 166. https://gcc.gnu.org/PR14150 11551 167. https://gcc.gnu.org/PR14949 11552 168. https://gcc.gnu.org/PR15123 11553 169. https://gcc.gnu.org/PR16469 11554 170. https://gcc.gnu.org/PR16344 11555 171. https://gcc.gnu.org/PR16842 11556 172. https://gcc.gnu.org/PR12608 11557 173. https://gcc.gnu.org/PR14492 11558 174. https://gcc.gnu.org/PR15461 11559 175. https://gcc.gnu.org/PR15890 11560 176. https://gcc.gnu.org/PR16180 11561 177. https://gcc.gnu.org/PR16224 11562 178. https://gcc.gnu.org/PR16408 11563 179. https://gcc.gnu.org/PR16529 11564 180. https://gcc.gnu.org/PR16698 11565 181. https://gcc.gnu.org/PR16706 11566 182. https://gcc.gnu.org/PR16810 11567 183. https://gcc.gnu.org/PR16851 11568 184. https://gcc.gnu.org/PR16870 11569 185. https://gcc.gnu.org/PR16904 11570 186. https://gcc.gnu.org/PR16905 11571 187. https://gcc.gnu.org/PR16964 11572 188. https://gcc.gnu.org/PR17068 11573 189. https://gcc.gnu.org/PR16366 11574 190. https://gcc.gnu.org/PR15345 11575 191. https://gcc.gnu.org/PR16590 11576 192. https://gcc.gnu.org/PR16693 11577 193. https://gcc.gnu.org/PR17078 11578 194. https://gcc.gnu.org/PR13956 11579 195. https://gcc.gnu.org/PR16684 11580 196. https://gcc.gnu.org/PR12658 11581 197. https://gcc.gnu.org/PR13092 11582 198. https://gcc.gnu.org/PR15320 11583 199. https://gcc.gnu.org/PR16246 11584 200. https://gcc.gnu.org/PR16273 11585 201. https://gcc.gnu.org/PR16401 11586 202. https://gcc.gnu.org/PR16411 11587 203. https://gcc.gnu.org/PR16489 11588 204. https://gcc.gnu.org/PR16618 11589 205. https://gcc.gnu.org/PR16637 11590 206. https://gcc.gnu.org/PR16717 11591 207. https://gcc.gnu.org/PR16813 11592 208. https://gcc.gnu.org/PR16853 11593 209. https://gcc.gnu.org/PR16889 11594 210. https://gcc.gnu.org/PR16959 11595 211. https://gcc.gnu.org/PR7587 11596 212. https://gcc.gnu.org/PR16473 11597 213. https://gcc.gnu.org/PR16478 11598 214. https://gcc.gnu.org/PR10695 11599 215. https://gcc.gnu.org/PR16974 11600 216. https://gcc.gnu.org/PR16298 11601 217. https://gcc.gnu.org/PR17113 11602 218. https://gcc.gnu.org/PR14697 11603 219. https://gcc.gnu.org/PR15869 11604 220. https://gcc.gnu.org/PR16325 11605 221. https://gcc.gnu.org/PR16357 11606 222. https://gcc.gnu.org/PR16380 11607 223. https://gcc.gnu.org/PR16407 11608 224. https://gcc.gnu.org/PR16643 11609 225. https://gcc.gnu.org/PR15927 11610 226. https://gcc.gnu.org/PR15948 11611 227. https://gcc.gnu.org/PR17019 11612 228. https://gcc.gnu.org/PR16130 11613 229. https://gcc.gnu.org/PR16142 11614 230. https://gcc.gnu.org/PR16278 11615 231. https://gcc.gnu.org/PR16414 11616 232. https://gcc.gnu.org/PR16445 11617 233. https://gcc.gnu.org/PR16490 11618 234. https://gcc.gnu.org/PR16683 11619 235. https://gcc.gnu.org/PR16195 11620 236. https://gcc.gnu.org/PR16239 11621 237. https://gcc.gnu.org/PR16199 11622 238. https://gcc.gnu.org/PR16416 11623 239. https://gcc.gnu.org/PR16430 11624 240. https://gcc.gnu.org/PR16379 11625 241. https://gcc.gnu.org/PR17093 11626 242. https://gcc.gnu.org/PR17119 11627 243. https://gcc.gnu.org/PR15928 11628 244. https://gcc.gnu.org/PR16210 11629 245. https://gcc.gnu.org/PR15488 11630 246. https://gcc.gnu.org/PR16250 11631 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 11632 248. https://gcc.gnu.org/PR17369 11633 249. https://gcc.gnu.org/PR17850 11634 250. https://gcc.gnu.org/PR13948 11635 251. https://gcc.gnu.org/PR14492 11636 252. https://gcc.gnu.org/PR16301 11637 253. https://gcc.gnu.org/PR16566 11638 254. https://gcc.gnu.org/PR17023 11639 255. https://gcc.gnu.org/PR17027 11640 256. https://gcc.gnu.org/PR17524 11641 257. https://gcc.gnu.org/PR17826 11642 258. https://gcc.gnu.org/PR15526 11643 259. https://gcc.gnu.org/PR16999 11644 260. https://gcc.gnu.org/PR17503 11645 261. https://gcc.gnu.org/PR17581 11646 262. https://gcc.gnu.org/PR18129 11647 263. https://gcc.gnu.org/PR10975 11648 264. https://gcc.gnu.org/PR11722 11649 265. https://gcc.gnu.org/PR14534 11650 266. https://gcc.gnu.org/PR15172 11651 267. https://gcc.gnu.org/PR15786 11652 268. https://gcc.gnu.org/PR16162 11653 269. https://gcc.gnu.org/PR16612 11654 270. https://gcc.gnu.org/PR16715 11655 271. https://gcc.gnu.org/PR16848 11656 272. https://gcc.gnu.org/PR17132 11657 273. https://gcc.gnu.org/PR17259 11658 274. https://gcc.gnu.org/PR17327 11659 275. https://gcc.gnu.org/PR17393 11660 276. https://gcc.gnu.org/PR17501 11661 277. https://gcc.gnu.org/PR17537 11662 278. https://gcc.gnu.org/PR17585 11663 279. https://gcc.gnu.org/PR17821 11664 280. https://gcc.gnu.org/PR17829 11665 281. https://gcc.gnu.org/PR17851 11666 282. https://gcc.gnu.org/PR17976 11667 283. https://gcc.gnu.org/PR18020 11668 284. https://gcc.gnu.org/PR18093 11669 285. https://gcc.gnu.org/PR18140 11670 286. https://gcc.gnu.org/PR17541 11671 287. https://gcc.gnu.org/PR17853 11672 288. https://gcc.gnu.org/PR17245 11673 289. https://gcc.gnu.org/PR17167 11674 290. https://gcc.gnu.org/PR17277 11675 291. https://gcc.gnu.org/PR17505 11676 292. https://gcc.gnu.org/PR17684 11677 293. https://gcc.gnu.org/PR17384 11678 294. https://gcc.gnu.org/PR17770 11679 295. https://gcc.gnu.org/PR11476 11680 296. https://gcc.gnu.org/PR14064 11681 297. https://gcc.gnu.org/PR14678 11682 298. https://gcc.gnu.org/PR15583 11683 299. https://gcc.gnu.org/PR15790 11684 300. https://gcc.gnu.org/PR15886 11685 301. https://gcc.gnu.org/PR16884 11686 302. https://gcc.gnu.org/PR13841 11687 303. https://gcc.gnu.org/PR15860 11688 304. https://gcc.gnu.org/PR17465 11689 305. https://gcc.gnu.org/PR17469 11690 306. https://gcc.gnu.org/PR18138 11691 307. https://gcc.gnu.org/PR15498 11692 308. https://gcc.gnu.org/PR15747 11693 309. https://gcc.gnu.org/PR16406 11694 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 11695 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 11696 312. https://gcc.gnu.org/PR24688 11697 313. https://gcc.gnu.org/PR17188 11698 314. https://gcc.gnu.org/PR20187 11699 315. https://gcc.gnu.org/PR21873 11700 316. https://gcc.gnu.org/PR21899 11701 317. https://gcc.gnu.org/PR22061 11702 318. https://gcc.gnu.org/PR22208 11703 319. https://gcc.gnu.org/PR22458 11704 320. https://gcc.gnu.org/PR22589 11705 321. https://gcc.gnu.org/PR24101 11706 322. https://gcc.gnu.org/PR10611 11707 323. https://gcc.gnu.org/PR13377 11708 324. https://gcc.gnu.org/PR16002 11709 325. https://gcc.gnu.org/PR17413 11710 326. https://gcc.gnu.org/PR17609 11711 327. https://gcc.gnu.org/PR17618 11712 328. https://gcc.gnu.org/PR18124 11713 329. https://gcc.gnu.org/PR18155 11714 330. https://gcc.gnu.org/PR18177 11715 331. https://gcc.gnu.org/PR18368 11716 332. https://gcc.gnu.org/PR18378 11717 333. https://gcc.gnu.org/PR18466 11718 334. https://gcc.gnu.org/PR18512 11719 335. https://gcc.gnu.org/PR18545 11720 336. https://gcc.gnu.org/PR18738 11721 337. https://gcc.gnu.org/PR18803 11722 338. https://gcc.gnu.org/PR19004 11723 339. https://gcc.gnu.org/PR19208 11724 340. https://gcc.gnu.org/PR19253 11725 341. https://gcc.gnu.org/PR19608 11726 342. https://gcc.gnu.org/PR19884 11727 343. https://gcc.gnu.org/PR20153 11728 344. https://gcc.gnu.org/PR20563 11729 345. https://gcc.gnu.org/PR20789 11730 346. https://gcc.gnu.org/PR21336 11731 347. https://gcc.gnu.org/PR21768 11732 348. https://gcc.gnu.org/PR21853 11733 349. https://gcc.gnu.org/PR21903 11734 350. https://gcc.gnu.org/PR21983 11735 351. https://gcc.gnu.org/PR21987 11736 352. https://gcc.gnu.org/PR22153 11737 353. https://gcc.gnu.org/PR22172 11738 354. https://gcc.gnu.org/PR21286 11739 355. https://gcc.gnu.org/PR22233 11740 356. https://gcc.gnu.org/PR22508 11741 357. https://gcc.gnu.org/PR22545 11742 358. https://gcc.gnu.org/PR23528 11743 359. https://gcc.gnu.org/PR23550 11744 360. https://gcc.gnu.org/PR23586 11745 361. https://gcc.gnu.org/PR23624 11746 362. https://gcc.gnu.org/PR23639 11747 363. https://gcc.gnu.org/PR23797 11748 364. https://gcc.gnu.org/PR23965 11749 365. https://gcc.gnu.org/PR24052 11750 366. https://gcc.gnu.org/PR24580 11751 367. https://gcc.gnu.org/PR24267 11752 368. https://gcc.gnu.org/PR17810 11753 369. https://gcc.gnu.org/PR17860 11754 370. https://gcc.gnu.org/PR21709 11755 371. https://gcc.gnu.org/PR21964 11756 372. https://gcc.gnu.org/PR22167 11757 373. https://gcc.gnu.org/PR22619 11758 374. https://gcc.gnu.org/PR23241 11759 375. https://gcc.gnu.org/PR23478 11760 376. https://gcc.gnu.org/PR24470 11761 377. https://gcc.gnu.org/PR24950 11762 378. https://gcc.gnu.org/PR14400 11763 379. https://gcc.gnu.org/PR14940 11764 380. https://gcc.gnu.org/PR20239 11765 381. https://gcc.gnu.org/PR15220 11766 382. https://gcc.gnu.org/PR19275 11767 383. https://gcc.gnu.org/PR21888 11768 384. https://gcc.gnu.org/PR15342 11769 385. https://gcc.gnu.org/PR23985 11770 386. https://gcc.gnu.org/PR16719 11771 387. https://gcc.gnu.org/PR21723 11772 388. https://gcc.gnu.org/PR21841 11773 389. https://gcc.gnu.org/PR23644 11774 390. https://gcc.gnu.org/PR24718 11775 391. https://gcc.gnu.org/PR18421 11776 392. https://gcc.gnu.org/PR20621 11777 393. https://gcc.gnu.org/PR18583 11778 394. https://gcc.gnu.org/PR20191 11779 395. https://gcc.gnu.org/PR22083 11780 396. https://gcc.gnu.org/PR23070 11781 397. https://gcc.gnu.org/PR23404 11782 398. https://gcc.gnu.org/PR23539 11783 399. https://gcc.gnu.org/PR24102 11784 400. https://gcc.gnu.org/PR24465 11785 401. https://gcc.gnu.org/PR19933 11786 402. https://gcc.gnu.org/PR21889 11787 403. https://gcc.gnu.org/PR19300 11788 404. https://gcc.gnu.org/PR20301 11789 405. https://gcc.gnu.org/PR20673 11790 406. https://gcc.gnu.org/PR18582 11791 407. https://gcc.gnu.org/PR19340 11792 408. https://gcc.gnu.org/PR21716 11793 409. https://gcc.gnu.org/PR24315 11794 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 11795 411. https://gcc.gnu.org/onlinedocs/ 11796 412. mailto:gcc-help (a] gcc.gnu.org 11797 413. mailto:gcc (a] gcc.gnu.org 11798 414. https://gcc.gnu.org/lists.html 11799 415. http://www.fsf.org/ 11800 416. https://gcc.gnu.org/about.html 11801 417. http://validator.w3.org/check/referer 11802 ====================================================================== 11803 http://gcc.gnu.org/gcc-3.3/index.html 11804 11805 GCC 3.3 Release Series 11806 11807 May 03, 2005 11808 11809 The [1]GNU project and the GCC developers are pleased to announce the 11810 release of GCC 3.3.6. 11811 11812 This release is a bug-fix release, containing fixes for regressions in 11813 GCC 3.3.5 relative to previous releases of GCC. 11814 11815 This release is the last of the series 3.3.x. 11816 11817 The GCC 3.3 release series includes numerous [2]new features, 11818 improvements, bug fixes, and other changes, thanks to an [3]amazing 11819 group of volunteers. 11820 11821 This release series is no longer maintained. 11822 11823 Release History 11824 11825 GCC 3.3.6 11826 May 3, 2005 ([4]changes) 11827 11828 GCC 3.3.5 11829 September 30, 2004 ([5]changes) 11830 11831 GCC 3.3.4 11832 May 31, 2004 ([6]changes) 11833 11834 GCC 3.3.3 11835 February 14, 2004 ([7]changes) 11836 11837 GCC 3.3.2 11838 October 16, 2003 ([8]changes) 11839 11840 GCC 3.3.1 11841 August 8, 2003 ([9]changes) 11842 11843 GCC 3.3 11844 May 14, 2003 ([10]changes) 11845 11846 References and Acknowledgements 11847 11848 GCC used to stand for the GNU C Compiler, but since the compiler 11849 supports several other languages aside from C, it now stands for the 11850 GNU Compiler Collection. 11851 11852 A list of [11]successful builds is updated as new information becomes 11853 available. 11854 11855 The GCC developers would like to thank the numerous people that have 11856 contributed new features, improvements, bug fixes, and other changes as 11857 well as test results to GCC. This [12]amazing group of volunteers is 11858 what makes GCC successful. 11859 11860 For additional information about GCC please refer to the [13]GCC 11861 project web site or contact the [14]GCC development mailing list. 11862 11863 To obtain GCC please use [15]our mirror sites, or our CVS server. 11864 11865 11866 For questions related to the use of GCC, please consult these web 11867 pages and the [16]GCC manuals. If that fails, the 11868 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11869 web pages and the development of GCC are welcome on our developer 11870 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 11871 archives. 11872 11873 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11874 distribution of this entire article is permitted in any medium, 11875 provided this notice is preserved. 11876 11877 These pages are [21]maintained by the GCC team. Last modified 11878 2016-09-30[22]. 11879 11880 References 11881 11882 1. http://www.gnu.org/ 11883 2. http://gcc.gnu.org/gcc-3.3/changes.html 11884 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11885 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 11886 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 11887 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 11888 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 11889 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 11890 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 11891 10. http://gcc.gnu.org/gcc-3.3/changes.html 11892 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 11893 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11894 13. http://gcc.gnu.org/index.html 11895 14. mailto:gcc (a] gcc.gnu.org 11896 15. http://gcc.gnu.org/mirrors.html 11897 16. https://gcc.gnu.org/onlinedocs/ 11898 17. mailto:gcc-help (a] gcc.gnu.org 11899 18. mailto:gcc (a] gcc.gnu.org 11900 19. https://gcc.gnu.org/lists.html 11901 20. http://www.fsf.org/ 11902 21. https://gcc.gnu.org/about.html 11903 22. http://validator.w3.org/check/referer 11904 ====================================================================== 11905 http://gcc.gnu.org/gcc-3.3/changes.html 11906 11907 GCC 3.3 Release Series 11908 Changes, New Features, and Fixes 11909 11910 The latest release in the 3.3 release series is [1]GCC 3.3.6. 11911 11912 Caveats 11913 11914 * The preprocessor no longer accepts multi-line string literals. They 11915 were deprecated in 3.0, 3.1, and 3.2. 11916 * The preprocessor no longer supports the -A- switch when appearing 11917 alone. -A- followed by an assertion is still supported. 11918 * Support for all the systems [2]obsoleted in GCC 3.1 has been 11919 removed from GCC 3.3. See below for a [3]list of systems which are 11920 obsoleted in this release. 11921 * Checking for null format arguments has been decoupled from the rest 11922 of the format checking mechanism. Programs which use the format 11923 attribute may regain this functionality by using the new [4]nonnull 11924 function attribute. Note that all functions for which GCC has a 11925 built-in format attribute, an appropriate built-in nonnull 11926 attribute is also applied. 11927 * The DWARF (version 1) debugging format has been deprecated and will 11928 be removed in a future version of GCC. Version 2 of the DWARF 11929 debugging format will continue to be supported for the foreseeable 11930 future. 11931 * The C and Objective-C compilers no longer accept the "Naming Types" 11932 extension (typedef foo = bar); it was already unavailable in C++. 11933 Code which uses it will need to be changed to use the "typeof" 11934 extension instead: typedef typeof(bar) foo. (We have removed this 11935 extension without a period of deprecation because it has caused the 11936 compiler to crash since version 3.0 and no one noticed until very 11937 recently. Thus we conclude it is not in widespread use.) 11938 * The -traditional C compiler option has been removed. It was 11939 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 11940 available.) The <varargs.h> header, used for writing variadic 11941 functions in traditional C, still exists but will produce an error 11942 message if used. 11943 * GCC 3.3.1 automatically places zero-initialized variables in the 11944 .bss section on some operating systems. Versions of GNU Emacs up to 11945 (and including) 21.3 will not work correctly when using this 11946 optimization; you can use -fno-zero-initialized-in-bss to disable 11947 it. 11948 11949 General Optimizer Improvements 11950 11951 * A new scheme for accurately describing processor pipelines, the 11952 [5]DFA scheduler, has been added. 11953 * Pavel Nejedly, Charles University Prague, has contributed new file 11954 format used by the edge coverage profiler (-fprofile-arcs). 11955 The new format is robust and diagnoses common mistakes where 11956 profiles from different versions (or compilations) of the program 11957 are combined resulting in nonsensical profiles and slow code to 11958 produced with profile feedback. Additionally this format allows 11959 extra data to be gathered. Currently, overall statistics are 11960 produced helping optimizers to identify hot spots of a program 11961 globally replacing the old intra-procedural scheme and resulting in 11962 better code. Note that the gcov tool from older GCC versions will 11963 not be able to parse the profiles generated by GCC 3.3 and vice 11964 versa. 11965 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 11966 pass enabled using -ftracer. This pass simplifies the control flow 11967 of functions allowing other optimizations to do better job. 11968 He also contributed the function reordering pass 11969 (-freorder-functions) to optimize function placement using profile 11970 feedback. 11971 11972 New Languages and Language specific improvements 11973 11974 C/ObjC/C++ 11975 11976 * The preprocessor now accepts directives within macro arguments. It 11977 processes them just as if they had not been within macro arguments. 11978 * The separate ISO and traditional preprocessors have been completely 11979 removed. The front end handles either type of preprocessed output 11980 if necessary. 11981 * In C99 mode preprocessor arithmetic is done in the precision of the 11982 target's intmax_t, as required by that standard. 11983 * The preprocessor can now copy comments inside macros to the output 11984 file when the macro is expanded. This feature, enabled using the 11985 -CC option, is intended for use by applications which place 11986 metadata or directives inside comments, such as lint. 11987 * The method of constructing the list of directories to be searched 11988 for header files has been revised. If a directory named by a -I 11989 option is a standard system include directory, the option is 11990 ignored to ensure that the default search order for system 11991 directories and the special treatment of system header files are 11992 not defeated. 11993 * A few more [6]ISO C99 features now work correctly. 11994 * A new function attribute, nonnull, has been added which allows 11995 pointer arguments to functions to be specified as requiring a 11996 non-null value. The compiler currently uses this information to 11997 issue a warning when it detects a null value passed in such an 11998 argument slot. 11999 * A new type attribute, may_alias, has been added. Accesses to 12000 objects with types with this attribute are not subjected to 12001 type-based alias analysis, but are instead assumed to be able to 12002 alias any other type of objects, just like the char type. 12003 12004 C++ 12005 12006 * Type based alias analysis has been implemented for C++ aggregate 12007 types. 12008 12009 Objective-C 12010 12011 * Generate an error if Objective-C objects are passed by value in 12012 function and method calls. 12013 * When -Wselector is used, check the whole list of selectors at the 12014 end of compilation, and emit a warning if a @selector() is not 12015 known. 12016 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 12017 * No longer need to include objc/objc-class.h to compile self calls 12018 in class methods (NeXT runtime only). 12019 * New -Wundeclared-selector option. 12020 * Removed selector bloating which was causing object files to be 10% 12021 bigger on average (GNU runtime only). 12022 * Using at run time @protocol() objects has been fixed in certain 12023 situations (GNU runtime only). 12024 * Type checking has been fixed and improved in many situations 12025 involving protocols. 12026 12027 Java 12028 12029 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 12030 1.4) API. 12031 * The JDK 1.4 assert facility has been implemented. 12032 * The bytecode interpreter is now direct threaded and thus faster. 12033 12034 Fortran 12035 12036 * Fortran improvements are listed in [7]the Fortran documentation. 12037 12038 Ada 12039 12040 * Ada tasking now works with glibc 2.3.x threading libraries. 12041 12042 New Targets and Target Specific Improvements 12043 12044 * The following changes have been made to the HP-PA port: 12045 + The port now defaults to scheduling for the PA8000 series of 12046 processors. 12047 + Scheduling support for the PA7300 processor has been added. 12048 + The 32-bit port now supports weak symbols under HP-UX 11. 12049 + The handling of initializers and finalizers has been improved 12050 under HP-UX 11. The 64-bit port no longer uses collect2. 12051 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 12052 + ABI fixes to correct the passing of small structures by value. 12053 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 12054 use the DFA processor pipeline description. 12055 * The following NetBSD configurations for the SuperH processor family 12056 have been added: 12057 + SH3, big-endian, sh-*-netbsdelf* 12058 + SH3, little-endian, shle-*-netbsdelf* 12059 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 12060 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 12061 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 12062 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 12063 * The following changes have been made to the IA-32/x86-64 port: 12064 + SSE2 and 3dNOW! intrinsics are now supported. 12065 + Support for thread local storage has been added to the IA-32 12066 and x86-64 ports. 12067 + The x86-64 port has been significantly improved. 12068 * The following changes have been made to the MIPS port: 12069 + All configurations now accept the -mabi switch. Note that you 12070 will need appropriate multilibs for this option to work 12071 properly. 12072 + ELF configurations will always pass an ABI flag to the 12073 assembler, except when the MIPS EABI is selected. 12074 + -mabi=64 no longer selects MIPS IV code. 12075 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 12076 been removed from this release. 12077 + -march now changes the core ISA level. In previous releases, 12078 it would change the use of processor-specific extensions, but 12079 would leave the core ISA unchanged. For example, mips64-elf 12080 -march=r8000 will now generate MIPS IV code. 12081 + Under most configurations, -mipsN now acts as a synonym for 12082 -march. 12083 + There are some new preprocessor macros to describe the -march 12084 and -mtune settings. See the documentation of those options 12085 for details. 12086 + Support for the NEC VR-Series processors has been added. This 12087 includes the 54xx, 5500, and 41xx series. 12088 + Support for the Sandcraft sr71k processor has been added. 12089 * The following changes have been made to the S/390 port: 12090 + Support to build the Java runtime libraries has been added. 12091 Java is now enabled by default on s390-*-linux* and 12092 s390x-*-linux* targets. 12093 + Multilib support for the s390x-*-linux* target has been added; 12094 this allows to build 31-bit binaries using the -m31 option. 12095 + Support for thread local storage has been added. 12096 + Inline assembler code may now use the 'Q' constraint to 12097 specify memory operands without index register. 12098 + Various platform-specific performance improvements have been 12099 implemented; in particular, the compiler now uses the BRANCH 12100 ON COUNT family of instructions and makes more frequent use of 12101 the TEST UNDER MASK family of instructions. 12102 * The following changes have been made to the PowerPC port: 12103 + Support for IBM Power4 processor added. 12104 + Support for Motorola e500 SPE added. 12105 + Support for AIX 5.2 added. 12106 + Function and Data sections now supported on AIX. 12107 + Sibcall optimizations added. 12108 * The support for H8 Tiny is added to the H8/300 port with -mn. 12109 12110 Obsolete Systems 12111 12112 Support for a number of older systems has been declared obsolete in GCC 12113 3.3. Unless there is activity to revive them, the next release of GCC 12114 will have their sources permanently removed. 12115 12116 All configurations of the following processor architectures have been 12117 declared obsolete: 12118 * Matsushita MN10200, mn10200-*-* 12119 * Motorola 88000, m88k-*-* 12120 * IBM ROMP, romp-*-* 12121 12122 Also, some individual systems have been obsoleted: 12123 * Alpha 12124 + Interix, alpha*-*-interix* 12125 + Linux libc1, alpha*-*-linux*libc1* 12126 + Linux ECOFF, alpha*-*-linux*ecoff* 12127 * ARM 12128 + Generic a.out, arm*-*-aout* 12129 + Conix, arm*-*-conix* 12130 + "Old ABI," arm*-*-oabi 12131 + StrongARM/COFF, strongarm-*-coff* 12132 * HPPA (PA-RISC) 12133 + Generic OSF, hppa1.0-*-osf* 12134 + Generic BSD, hppa1.0-*-bsd* 12135 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 12136 + HiUX, hppa*-*-hiux* 12137 + Mach Lites, hppa*-*-lites* 12138 * Intel 386 family 12139 + Windows NT 3.x, i?86-*-win32 12140 * MC68000 family 12141 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 12142 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 12143 m68k-sun-mach* 12144 + AT&T systems, m68000-att-sysv* 12145 + Atari systems, m68k-atari-sysv* 12146 + Motorola systems, m68k-motorola-sysv* 12147 + NCR systems, m68k-ncr-sysv* 12148 + Plexus systems, m68k-plexus-sysv* 12149 + Commodore systems, m68k-cbm-sysv* 12150 + Citicorp TTI, m68k-tti-* 12151 + Unos, m68k-crds-unos* 12152 + Concurrent RTU, m68k-ccur-rtu* 12153 + Linux a.out, m68k-*-linux*aout* 12154 + Linux libc1, m68k-*-linux*libc1* 12155 + pSOS, m68k-*-psos* 12156 * MIPS 12157 + Generic ECOFF, mips*-*-ecoff* 12158 + SINIX, mips-sni-sysv4 12159 + Orion RTEMS, mips64orion-*-rtems* 12160 * National Semiconductor 32000 12161 + OpenBSD, ns32k-*-openbsd* 12162 * POWER (aka RS/6000) and PowerPC 12163 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 12164 + Bull BOSX, rs6000-bull-bosx 12165 + Generic Mach, rs6000-*-mach* 12166 + Generic SysV, powerpc*-*-sysv* 12167 + Linux libc1, powerpc*-*-linux*libc1* 12168 * Sun SPARC 12169 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 12170 sparclite-*-aout*, and sparc86x-*-aout* 12171 + NetBSD a.out, sparc-*-netbsd*aout* 12172 + Generic BSD, sparc-*-bsd* 12173 + ChorusOS, sparc-*-chorusos* 12174 + Linux a.out, sparc-*-linux*aout* 12175 + Linux libc1, sparc-*-linux*libc1* 12176 + LynxOS, sparc-*-lynxos* 12177 + Solaris on HAL hardware, sparc-hal-solaris2* 12178 + SunOS versions 3 and 4, sparc-*-sunos[34]* 12179 * NEC V850 12180 + RTEMS, v850-*-rtems* 12181 * VAX 12182 + VMS, vax-*-vms* 12183 12184 Documentation improvements 12185 12186 Other significant improvements 12187 12188 * Almost all front-end dependencies in the compiler have been 12189 separated out into a set of language hooks. This should make adding 12190 a new front end clearer and easier. 12191 * One effect of removing the separate preprocessor is a small 12192 increase in the robustness of the compiler in general, and the 12193 maintainability of target descriptions. Previously target-specific 12194 built-in macros and others, such as __FAST_MATH__, had to be 12195 handled with so-called specs that were hard to maintain. Often they 12196 would fail to behave properly when conflicting options were 12197 supplied on the command line, and define macros in the user's 12198 namespace even when strict ISO compliance was requested. 12199 Integrating the preprocessor has cleanly solved these issues. 12200 * The Makefile suite now supports redirection of make install by 12201 means of the variable DESTDIR. 12202 __________________________________________________________________ 12203 12204 GCC 3.3 12205 12206 Detailed release notes for the GCC 3.3 release follow. 12207 12208 Bug Fixes 12209 12210 bootstrap failures 12211 12212 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 12213 [9]10198,[10]10338) 12214 12215 Internal compiler errors (multi-platform) 12216 12217 * [11]3581 large string causes segmentation fault in cc1 12218 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 12219 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 12220 init, invalid_op) 12221 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 12222 * [15]6412 (c++) ICE in retrieve_specialization 12223 * [16]6620 (c++) partial template specialization causes an ICE 12224 (segmentation fault) 12225 * [17]6663 (c++) ICE with attribute aligned 12226 * [18]7068 ICE with incomplete types 12227 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 12228 * [20]7647 (c++) ICE when data member has the name of the enclosing 12229 class 12230 * [21]7675 ICE in fixup_var_refs_1 12231 * [22]7718 'complex' template instantiation causes ICE 12232 * [23]8116 (c++) ICE in member template function 12233 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 12234 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 12235 fault 12236 * [26]8564 (c++) ICE in find_function_data, in function.c 12237 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 12238 * [28]8766 (c++) ICE after failed initialization of static template 12239 variable 12240 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 12241 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 12242 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 12243 definition 12244 * [32]9216 (c++) ICE on missing template parameter 12245 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 12246 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 12247 loop 12248 * [35]9429 (c++) ICE in template instantiation with a pointered new 12249 operator 12250 * [36]9516 Internal error when using a big array 12251 * [37]9600 (c++) ICE with typedefs in template class 12252 * [38]9629 (c++) virtual inheritance segfault 12253 * [39]9672 (c++) ICE: Error reporting routines re-entered 12254 * [40]9749 (c++) ICE in write_expression on invalid function 12255 prototype 12256 * [41]9794 (fortran) ICE: floating point exception during constant 12257 folding 12258 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 12259 * [43]9916 (c++) ICE with noreturn function in ?: statement 12260 * [44]9936 ICE with local function and variable-length 2d array 12261 * [45]10262 (c++) cc1plus crashes with large generated code 12262 * [46]10278 (c++) ICE in parser for invalid code 12263 * [47]10446 (c++) ICE on definition of nonexistent member function of 12264 nested class in a class template 12265 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 12266 declaration 12267 * [49]10506 (c++) ICE in build_new at cp/init.c with 12268 -fkeep-inline-functions and multiple inheritance 12269 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 12270 precision of the declared type 12271 12272 Optimization bugs 12273 12274 * [51]2001 Inordinately long compile times in reload CSE regs 12275 * [52]2391 Exponential compilation time explosion in combine 12276 * [53]2960 Duplicate loop conditions even with -Os 12277 * [54]4046 redundant conditional branch 12278 * [55]6405 Loop-unrolling related performance regressions 12279 * [56]6798 very long compile time with large case-statement 12280 * [57]6871 const objects shouldn't be moved to .bss 12281 * [58]6909 problem w/ -Os on modified loop-2c.c test case 12282 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 12283 non-void function'' warning 12284 * [60]7642 optimization problem with signbit() 12285 * [61]8634 incorrect code for inlining of memcpy under -O2 12286 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 12287 regular function call 12288 12289 C front end 12290 12291 * [63]2161 long if-else cascade overflows parser stack 12292 * [64]4319 short accepted on typedef'd char 12293 * [65]8602 incorrect line numbers in warning messages when using 12294 inline functions 12295 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 12296 AST nodes and breaks debugging dumps 12297 * [67]9853 miscompilation of non-constant structure initializer 12298 12299 c++ compiler and library 12300 12301 * [68]45 legal template specialization code is rejected (DUP: 12302 [69]3784) 12303 * [70]764 lookup failure: friend operator and dereferencing a pointer 12304 and templates (DUP: [71]5116) 12305 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 12306 2863) 12307 * [73]3663 G++ doesn't check access control during template 12308 instantiation 12309 * [74]3797 gcc fails to emit explicit specialization of a template 12310 member 12311 * [75]3948 Two destructors are called when no copy destructor is 12312 defined (ABI change) 12313 * [76]4137 Conversion operator within template is not accepted 12314 * [77]4361 bogus ambiguity taking the address of a member template 12315 * [78]4802 g++ accepts illegal template code (access to private 12316 member; DUP: [79]5837) 12317 * [80]4803 inline function is used but never defined, and g++ does 12318 not object 12319 * [81]5094 Partial specialization cannot be friend? 12320 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 12321 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 12322 time 12323 * [84]7015 certain __asm__ constructs rejected 12324 * [85]7086 compile time regression (quadratic behavior in 12325 fixup_var_refs) 12326 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 12327 std::abort 12328 * [87]7247 copy constructor missing when inlining enabled (invalid 12329 optimization?) 12330 * [88]7441 string array initialization compilation time regression 12331 from seconds to minutes 12332 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 12333 * [90]7804 bad printing of floating point constant in warning message 12334 * [91]8099 Friend classes and template specializations 12335 * [92]8117 member function pointers and multiple inheritance 12336 * [93]8205 using declaration and multiple inheritance 12337 * [94]8645 unnecessary non-zero checks in stl_tree.h 12338 * [95]8724 explicit destructor call for incomplete class allowed 12339 * [96]8805 compile time regression with many member variables 12340 * [97]8691 -O3 and -fno-implicit-templates are incompatible 12341 * [98]8700 unhelpful error message for binding temp to reference 12342 * [99]8724 explicit destructor call for incomplete class allowed 12343 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 12344 * [101]9016 Failure to consistently constant fold "constant" C++ 12345 objects 12346 * [102]9053 g++ confused about ambiguity of overloaded function 12347 templates 12348 * [103]9152 undefined virtual thunks 12349 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 12350 * [105]9297 data corruption due to codegen bug (when copying.) 12351 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 12352 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 12353 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 12354 local classes 12355 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 12356 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 12357 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 12358 and <iostream.h> 12359 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 12360 [114][DR 231] 12361 * [115]9555 ostream inserters fail to set badbit on exception 12362 * [116]9561 ostream inserters rethrow exception of wrong type 12363 * [117]9563 ostream::sentry returns true after a failed preparation 12364 * [118]9582 one-definition rule violation in std::allocator 12365 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 12366 * [120]9683 bug in initialization chains for static const variables 12367 from template classes 12368 * [121]9791 -Woverloaded-virtual reports hiding of destructor 12369 * [122]9817 collate::compare doesn't handle nul characters 12370 * [123]9825 filebuf::sputbackc breaks sbumpc 12371 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 12372 with custom traits 12373 * [125]9924 Multiple using statements for builtin functions not 12374 allowed 12375 * [126]9946 destructor is not called for temporary object 12376 * [127]9964 filebuf::close() sometimes fails to close file 12377 * [128]9988 filebuf::overflow writes EOF to file 12378 * [129]10033 optimization breaks polymorphic references w/ typeid 12379 operator 12380 * [130]10097 filebuf::underflow drops characters 12381 * [131]10132 filebuf destructor can throw exceptions 12382 * [132]10180 gcc fails to warn about non-inlined function 12383 * [133]10199 method parametrized by template does not work everywhere 12384 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 12385 * [135]10427 Stack corruption with variable-length automatic arrays 12386 and virtual destructors 12387 * [136]10503 Compilation never stops in fixed_type_or_null 12388 12389 Objective-C 12390 12391 * [137]5956 selectors aren't matched properly when added to the 12392 selector table 12393 12394 Fortran compiler and library 12395 12396 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 12397 detect 12398 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 12399 info requested 12400 * [140]5634 doc: explain that configure --prefix=~/... does not work 12401 * [141]6367 multiple repeat counts confuse namelist read into array 12402 * [142]6491 Logical operations error on logicals when using 12403 -fugly-logint 12404 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 12405 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 12406 on irix6.5 12407 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 12408 assume a direct access file 12409 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 12410 -fno-automatic) 12411 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 12412 * [148]7388 Incorrect output with 0-based array of characters 12413 * [149]8587 Double complex zero ** double precision number -> NaN 12414 instead of zero 12415 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 12416 unknown register name line-length-none 12417 * [151]10197 Direct access files not unformatted by default 12418 12419 Java compiler and library 12420 12421 * [152]6005 gcj fails to build rhug on alpha 12422 * [153]6389 System.getProperty("") should always throw an 12423 IllegalArgumentException 12424 * [154]6576 java.util.ResourceBundle.getResource ignores locale 12425 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 12426 * [156]7060 getMethod() doesn't search super interface 12427 * [157]7073 bytecode interpreter gives wrong answer for interface 12428 getSuperclass() 12429 * [158]7180 possible bug in 12430 javax.naming.spi.NamingManager.getPlusPath() 12431 * [159]7416 java.security startup refs "GNU libgcj.security" 12432 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 12433 env (DUP: [161]7578) 12434 * [162]7611 Internal error while compiling libjava with -O 12435 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 12436 * [164]7766 ZipInputStream.available returns 0 immediately after 12437 construction 12438 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 12439 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 12440 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 12441 * [168]8234 ZipInputStream chokes when InputStream.read() returns 12442 small chunks 12443 * [169]8415 reflection bug: exception info for Method 12444 * [170]8481 java.Random.nextInt(int) may return negative 12445 * [171]8593 Error reading GZIPped files with BufferedReader 12446 * [172]8759 java.beans.Introspector has no flushCaches() or 12447 flushFromCaches() methods 12448 * [173]8997 spin() calls Thread.sleep 12449 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 12450 instead of the root content of C: 12451 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 12452 wrong return codes 12453 * [176]9271 Severe bias in java.security.SecureRandom 12454 12455 Ada compiler and library 12456 12457 * [177]6767 make gnatlib-shared fails on -laddr2line 12458 * [178]9911 gnatmake fails to link when GCC configured with 12459 --with-sjlj-exceptions=yes 12460 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 12461 * [180]10546 Ada tasking not working on Red Hat 9 12462 12463 preprocessor 12464 12465 * [181]7029 preprocessor should ignore #warning with -M 12466 12467 ARM-specific 12468 12469 * [182]2903 [arm] Optimization bug with long long arithmetic 12470 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 12471 12472 FreeBSD-specific 12473 12474 * [184]7680 float functions undefined in math.h/cmath with #define 12475 _XOPEN_SOURCE 12476 12477 HP-UX or HP-PA-specific 12478 12479 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 12480 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 12481 fputc_unlocked 12482 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 12483 12484 m68hc11-specific 12485 12486 * [188]6744 Bad assembler code generated: reference to pseudo 12487 register z 12488 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 12489 in reload1.c 12490 12491 MIPS-specific 12492 12493 * [190]9496 [mips-linux] bug in optimizer? 12494 12495 PowerPC-specific 12496 12497 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 12498 space 12499 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 12500 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 12501 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 12502 12503 SPARC-specific 12504 12505 * [195]10267 (documentation) Wrong build instructions for 12506 *-*-solaris2* 12507 12508 x86-specific (Intel/AMD) 12509 12510 * [196]7916 ICE in instantiate_virtual_register_1 12511 * [197]7926 (c++) i486 instructions in header files make c++ programs 12512 crash on i386 12513 * [198]8555 ICE in gen_split_1231 12514 * [199]8994 ICE with -O -march=pentium4 12515 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 12516 * [201]9806 ICE in inline assembly with -fPIC flag 12517 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 12518 regs 12519 * [203]10233 64-bit comparison only comparing bottom 32-bits 12520 * [204]10286 type-punning doesn't work with __m64 and -O 12521 * [205]10308 [x86] ICE with -O -fgcse or -O2 12522 __________________________________________________________________ 12523 12524 GCC 3.3.1 12525 12526 Bug Fixes 12527 12528 This section lists the problem reports (PRs) from GCC's bug tracking 12529 system that are known to be fixed in the 3.3.1 release. This list might 12530 not be complete (that is, it is possible that some PRs that have been 12531 fixed are not listed here). 12532 12533 Bootstrap failures 12534 12535 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 12536 12537 Internal compiler errors (multi-platform) 12538 12539 * [207]5754 ICE on invalid nested template class 12540 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 12541 and --enable-checking 12542 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 12543 * [210]7053 (c++) ICE when declaring a function already defined as a 12544 friend method of a template class 12545 * [211]8164 (c++) ICE when using different const expressions as 12546 template parameter 12547 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 12548 * [213]9559 (c++) ICE with invalid initialization of a static const 12549 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 12550 when redeclaring a static member variable 12551 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 12552 dwarfout.c with -g -O -finline-functions 12553 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 12554 * [217]10475 ICE in subreg_highpart_offset for code with long long 12555 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 12556 from a void pointer 12557 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 12558 instantiating static member variables 12559 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 12560 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 12561 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 12562 MAX_INT_64BIT 12563 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 12564 sched.c 12565 * [224]10939 (c++) ICE with template code 12566 * [225]10956 (c++) ICE when specializing a template member function 12567 of a template class, in tsubst, in cp/pt.c 12568 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 12569 defined) 12570 * [227]11059 (c++) ICE with empty union 12571 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 12572 -O2 -fnon-call-exceptions 12573 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 12574 * [230]11149 (c++) ICE on error when instantiation with call function 12575 of a base type 12576 * [231]11228 (c++) ICE on new-expression using array operator new and 12577 default-initialization 12578 * [232]11282 (c++) Infinite memory usage after syntax error 12579 * [233]11301 (fortran) ICE with -fno-globals 12580 * [234]11308 (c++) ICE when using an enum type name as if it were a 12581 class or namespace 12582 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 12583 an empty struct 12584 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 12585 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 12586 template member functions 12587 12588 Optimization bugs 12589 12590 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 12591 problem) 12592 * [239]11304 Wrong code production with -fomit-frame-pointer 12593 * [240]11381 volatile memory access optimized away 12594 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 12595 * [242]11557 constant folding bug generates wrong code 12596 12597 C front end 12598 12599 * [243]5897 No warning for statement after return 12600 * [244]11279 DWARF-2 output mishandles large enums 12601 12602 Preprocessor bugs 12603 12604 * [245]11022 no warning for non-compatible macro redefinition 12605 12606 C++ compiler and library 12607 12608 * [246]2330 static_cast<>() to a private base is allowed 12609 * [247]5388 Incorrect message "operands to ?: have different types" 12610 * [248]5390 Libiberty fails to demangle multi-digit template 12611 parameters 12612 * [249]7877 Incorrect parameter passing to specializations of member 12613 function templates 12614 * [250]9393 Anonymous namespaces and compiling the same file twice 12615 * [251]10032 -pedantic converts some errors to warnings 12616 * [252]10468 const typeof(x) is non-const, but only in templates 12617 * [253]10527 confused error message with "new int()" parameter 12618 initializer 12619 * [254]10679 parameter MIN_INLINE_INSNS is not honored 12620 * [255]10682 gcc chokes on a typedef for an enum inside a class 12621 template 12622 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 12623 0. 12624 * [257]10845 template member function (with nested template as 12625 parameter) cannot be called anymore if another unrelated template 12626 member function is defined 12627 * [258]10849 Cannot define an out-of-class specialization of a 12628 private nested template class 12629 * [259]10888 Suppress -Winline warnings for system headers 12630 * [260]10929 -Winline warns about functions for which no definition 12631 is visible 12632 * [261]10931 valid conversion static_cast<const unsigned 12633 int&>(lvalue-of-type-int) is rejected 12634 * [262]10940 Bad code with explicit specialization 12635 * [263]10968 If member function implicitly instantiated, explicit 12636 instantiation of class fails to instantiate it 12637 * [264]10990 Cannot convert with dynamic_cast<> to a private base 12638 class from within a member function 12639 * [265]11039 Bad interaction between implicit typename deprecation 12640 and friendship 12641 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 12642 "__unused__" instead 12643 * [267]11095 C++ iostream manipulator causes segfault when called 12644 with negative argument 12645 * [268]11098 g++ doesn't emit complete debugging information for 12646 local variables in destructors 12647 * [269]11137 GNU/Linux shared library constructors not called unless 12648 there's one global object 12649 * [270]11154 spurious ambiguity report for template class 12650 specialization 12651 * [271]11329 Compiler cannot find user defined implicit typecast 12652 * [272]11332 Spurious error with casts in ?: expression 12653 * [273]11431 static_cast behavior with subclasses when default 12654 constructor available 12655 * [274]11528 money_get facet does not accept "$.00" as valid 12656 * [275]11546 Type lookup problems in out-of-line definition of a 12657 class doubly nested from a template class 12658 * [276]11567 C++ code containing templated member function with same 12659 name as pure virtual member function results in linking failure 12660 * [277]11645 Failure to deal with using and private inheritance 12661 12662 Java compiler and library 12663 12664 * [278]5179 Qualified static field access doesn't initialize its 12665 class 12666 * [279]8204 gcj -O2 to native reorders certain instructions 12667 improperly 12668 * [280]10838 java.io.ObjectInputStream syntax error 12669 * [281]10886 The RMI registry that comes with GCJ does not work 12670 correctly 12671 * [282]11349 JNDI URL context factories not located correctly 12672 12673 x86-specific (Intel/AMD) 12674 12675 * [283]4823 ICE on inline assembly code 12676 * [284]8878 miscompilation with -O and SSE 12677 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 12678 -masm=intel 12679 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 12680 in reload1.c 12681 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 12682 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 12683 * [289]11044 [x86] out of range loop instructions for FP code on K6 12684 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 12685 built-ins 12686 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 12687 is used 12688 12689 SPARC- or Solaris- specific 12690 12691 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 12692 * [293]10142 [SPARC64] gcc produces wrong code when passing 12693 structures by value 12694 * [294]10663 New configure check aborts with Sun tools. 12695 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 12696 * [296]10876 ICE in calculate_giv_inc when building KDE 12697 * [297]10955 wrong code at -O3 for structure argument in context of 12698 structure return 12699 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 12700 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 12701 Linux kernel 12702 12703 ia64 specific 12704 12705 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 12706 * [301]11320 scheduler bug (in machine depended reorganization pass) 12707 * [302]11599 bug with conditional and __builtin_prefetch 12708 12709 PowerPC specific 12710 12711 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 12712 during loop) 12713 * [304]10871 error in rs6000_stack_info save_size computation 12714 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 12715 cures it 12716 12717 m68k-specific 12718 12719 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 12720 * [307]10557 [m68k] ICE in subreg_offset_representable_p 12721 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 12722 12723 ARM-specific 12724 12725 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 12726 functions with __attribute__ ((interrupt ("IRQ"))) 12727 * [310]10842 [arm] Clobbered link register is copied to pc under 12728 certain circumstances 12729 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 12730 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 12731 (3.4) 12732 12733 MIPS-specific 12734 12735 * [313]11084 ICE in propagate_one_insn, in flow.c 12736 12737 SH-specific 12738 12739 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 12740 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 12741 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 12742 C++ files 12743 12744 GNU/Linux (or Hurd?) specific 12745 12746 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 12747 12748 UnixWare specific 12749 12750 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 12751 7.1.1 12752 12753 Cygwin (or mingw) specific 12754 12755 * [319]5287 ICE with dllimport attribute 12756 * [320]10148 [MingW/CygWin] Compiler dumps core 12757 12758 DJGPP specific 12759 12760 * [321]8787 GCC fails to emit .intel_syntax when invoked with 12761 -masm=intel on DJGPP 12762 12763 Darwin (and MacOS X) specific 12764 12765 * [322]10900 trampolines crash 12766 12767 Documentation 12768 12769 * [323]1607 (c++) Format attributes on methods undocumented 12770 * [324]4252 Invalid option `-fdump-translation-unit' 12771 * [325]4490 Clarify restrictions on -m96bit-long-double, 12772 -m128bit-long-double 12773 * [326]10355 document an issue with regparm attribute on some systems 12774 (e.g. Solaris) 12775 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 12776 (Unix)" is wrong 12777 * [328]10805 document bug in old version of Sun assembler 12778 * [329]10815 warn against GNU binutils on AIX 12779 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 12780 * [331]11280 Manual incorrect with respect to -freorder-blocks 12781 * [332]11466 Document -mlittle-endian and its restrictions for the 12782 sparc64 port 12783 12784 Testsuite bugs (compiler itself is not affected) 12785 12786 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 12787 report failure 12788 * [334]10810 gcc-3.3 fails make check: buffer overrun in 12789 test_demangle.c 12790 __________________________________________________________________ 12791 12792 GCC 3.3.2 12793 12794 Bug Fixes 12795 12796 This section lists the problem reports (PRs) from GCC's bug tracker 12797 that are known to be fixed in the 3.3.2 release. This list might not be 12798 complete (that is, it is possible that some PRs that have been fixed 12799 are not listed here). 12800 12801 Bootstrap failures and problems 12802 12803 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 12804 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 12805 --enable-threads=posix 12806 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 12807 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 12808 7.1.1) 12809 * [339]11687 xstormy16-elf build fails in libf2c 12810 * [340]12263 [SGI IRIX] bootstrap fails during compile of 12811 libf2c/libI77/backspace.c 12812 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 12813 fix-header processing) 12814 12815 Internal compiler errors (multi-platform) 12816 12817 * [342]7277 Casting integers to vector types causes ICE 12818 * [343]7939 (c++) ICE on invalid function template specialization 12819 * [344]11063 (c++) ICE on parsing initialization list of const array 12820 member 12821 * [345]11207 ICE with negative index in array element designator 12822 * [346]11522 (fortran) g77 dwarf-2 ICE in 12823 add_abstract_origin_attribute 12824 * [347]11595 (c++) ICE on duplicate label definition 12825 * [348]11646 (c++) ICE in commit_one_edge_insertion with 12826 -fnon-call-exceptions -fgcse -O 12827 * [349]11665 ICE in struct initializer when taking address 12828 * [350]11852 (c++) ICE with bad struct initializer. 12829 * [351]11878 (c++) ICE in cp_expr_size 12830 * [352]11883 ICE with any -O on mercury-generated C code 12831 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 12832 cp/typeck2.c when applying typeid operator to template template 12833 parameter 12834 * [354]12146 ICE in lookup_template_function, in cp/pt.c 12835 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 12836 -fno-gcse -O2 12837 * [356]12369 (c++) ICE with templates and friends 12838 * [357]12446 ICE in emit_move_insn on complicated array reference 12839 * [358]12510 ICE in final_scan_insn 12840 * [359]12544 ICE with large parameters used in nested functions 12841 12842 C and optimization bugs 12843 12844 * [360]9862 spurious warnings with -W -finline-functions 12845 * [361]10962 lookup_field is a linear search on a linked list (can be 12846 slow if large struct) 12847 * [362]11370 -Wunreachable-code gives false complaints 12848 * [363]11637 invalid assembly with -fnon-call-exceptions 12849 * [364]11885 Problem with bitfields in packed structs 12850 * [365]12082 Inappropriate unreachable code warnings 12851 * [366]12180 Inline optimization fails for variadic function 12852 * [367]12340 loop unroller + gcse produces wrong code 12853 12854 C++ compiler and library 12855 12856 * [368]3907 nested template parameter collides with member name 12857 * [369]5293 confusing message when binding a temporary to a reference 12858 * [370]5296 [DR115] Pointers to functions and to template functions 12859 behave differently in deduction 12860 * [371]7939 ICE on function template specialization 12861 * [372]8656 Unable to assign function with __attribute__ and pointer 12862 return type to an appropriate variable 12863 * [373]10147 Confusing error message for invalid template function 12864 argument 12865 * [374]11400 std::search_n() makes assumptions about Size parameter 12866 * [375]11409 issues with using declarations, overloading, and 12867 built-in functions 12868 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 12869 multiple bits in mask 12870 * [377]11786 operator() call on variable in other namespace not 12871 recognized 12872 * [378]11867 static_cast ignores ambiguity 12873 * [379]11928 bug with conversion operators that are typedefs 12874 * [380]12114 Uninitialized memory accessed in dtor 12875 * [381]12163 static_cast + explicit constructor regression 12876 * [382]12181 Wrong code with comma operator and c++ 12877 * [383]12236 regparm and fastcall messes up parameters 12878 * [384]12266 incorrect instantiation of unneeded template during 12879 overload resolution 12880 * [385]12296 istream::peek() doesn't set eofbit 12881 * [386]12298 [sjlj exceptions] Stack unwind destroys 12882 not-yet-constructed object 12883 * [387]12369 ICE with templates and friends 12884 * [388]12337 apparently infinite loop in g++ 12885 * [389]12344 stdcall attribute ignored if function returns a pointer 12886 * [390]12451 missing(late) class forward declaration in cxxabi.h 12887 * [391]12486 g++ accepts invalid use of a qualified name 12888 12889 x86 specific (Intel/AMD) 12890 12891 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 12892 builtins 12893 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 12894 -O2 12895 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 12896 * [395]12116 [k6] Invalid assembly output values with X-MAME code 12897 * [396]12070 ICE converting between double and long double with 12898 -msoft-float 12899 12900 ia64-specific 12901 12902 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 12903 * [398]11535 __builtin_return_address may not work on ia64 12904 * [399]11693 [ia64] ICE in gen_nop_type 12905 * [400]12224 [ia64] Thread-local storage doesn't work 12906 12907 PowerPC-specific 12908 12909 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 12910 kernel 12911 * [402]11319 loop miscompiled on ppc32 12912 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 12913 12914 SPARC-specific 12915 12916 * [404]11662 wrong code for expr. with cast to long long and 12917 exclusive or 12918 * [405]11965 invalid assembler code for a shift < 32 operation 12919 * [406]12301 (c++) stack corruption when a returned expression throws 12920 an exception 12921 12922 Alpha-specific 12923 12924 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 12925 kernel 2.4.22-pre8 12926 12927 HPUX-specific 12928 12929 * [408]11313 problem with #pragma weak and static inline functions 12930 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 12931 12932 Solaris specific 12933 12934 * [410]12166 Profiled programs crash if PROFDIR is set 12935 12936 Solaris-x86 specific 12937 12938 * [411]12101 i386 Solaris no longer works with GNU as? 12939 12940 Miscellaneous embedded target-specific bugs 12941 12942 * [412]10988 [m32r-elf] wrong blockmove code with -O3 12943 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 12944 -O2 12945 * [414]11902 [sh4] spec file improperly inserts rpath even when none 12946 needed 12947 * [415]11903 [sh4] -pthread fails to link due to error in spec file 12948 on sh4 12949 __________________________________________________________________ 12950 12951 GCC 3.3.3 12952 12953 Minor features 12954 12955 In addition to the bug fixes documented below, this release contains 12956 few minor features such as: 12957 * Support for --with-sysroot 12958 * Support for automatic detection of executable stacks 12959 * Support for SSE3 instructions 12960 * Support for thread local storage debugging under GDB on S390 12961 12962 Bug Fixes 12963 12964 This section lists the problem reports (PRs) from GCC's bug tracker 12965 that are known to be fixed in the 3.3.3 release. This list might not be 12966 complete (that is, it is possible that some PRs that have been fixed 12967 are not listed here). 12968 12969 Bootstrap failures and issues 12970 12971 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 12972 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 12973 unable to infer tagged configuration 12974 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 12975 subdirectories properly 12976 12977 Internal compiler errors (multi-platform) 12978 12979 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 12980 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 12981 * [420]10555 (c++) ICE on undefined template argument 12982 * [421]10706 (c++) ICE in mangle_class_name_for_template 12983 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 12984 active 12985 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 12986 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 12987 * [425]12632 (fortran) -fbounds-check ICE 12988 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 12989 3.3.2 12990 * [427]12726 (c++) ICE (segfault) on trivial code 12991 * [428]12890 (c++) ICE on compilation of class with throwing method 12992 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 12993 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 12994 correct code with -O2 -fno-force-mem 12995 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 12996 template 12997 * [432]13318 ICE: floating point exception in the loop optimizer 12998 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 12999 except.c 13000 * [434]13574 (c++) invalid array default initializer in class lets 13001 gcc consume all memory and die 13002 * [435]13475 ICE on SIMD variables with partial value initialization 13003 * [436]13797 (c++) ICE on invalid template parameter 13004 * [437]13824 (java) gcj SEGV with simple .java program 13005 13006 C and optimization bugs 13007 13008 * [438]8776 loop invariants are not removed (most likely) 13009 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 13010 strncmp by memcmp 13011 * [440]11350 undefined labels with -Os -fPIC 13012 * [441]12826 Optimizer removes reference through volatile pointer 13013 * [442]12500 stabs debug info: void no longer a predefined / builtin 13014 type 13015 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 13016 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 13017 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 13018 * [446]13507 spurious printf format warning 13019 * [447]13382 Type information for const pointer disappears during 13020 optimization. 13021 * [448]13394 noreturn attribute ignored on recursive invokation 13022 * [449]13400 Compiled code crashes storing to read-only location 13023 * [450]13521 Endless loop in calculate_global_regs_live 13024 13025 C++ compiler and library 13026 13027 Some of the bug fixes in this list were made to implement decisions 13028 that the ISO C++ standards committee has made concerning several defect 13029 reports (DRs). Links in the list below point to detailed discussion of 13030 the relevant defect report. 13031 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 13032 unification 13033 * [452]2294 using declaration confusion 13034 * [453]5050 template instantiation depth exceeds limit: recursion 13035 problem? 13036 * [454]9371 Bad exception handling in 13037 i/ostream::operator>>/<<(streambuf*) 13038 * [455]9546 bad exception handling in ostream members 13039 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 13040 face of unknown locales 13041 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 13042 * [459]10095 istream::operator>>(int&) sets ios::badbit when 13043 ios::failbit is set. 13044 * [460]11554 Warning about reordering of initializers doesn't mention 13045 location of constructor 13046 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 13047 * [462]12352 Exception safety problems in src/localename.cc 13048 * [463]12438 Memory leak in locale::combine() 13049 * [464]12540 Memory leak in locale::locale(const char*) 13050 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 13051 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 13052 * [470]12696 memory eating infinite loop in diagnostics (error 13053 recovery problem) 13054 * [471]12815 Code compiled with optimization behaves unexpectedly 13055 * [472]12862 Conflicts between typedefs/enums and namespace member 13056 declarations 13057 * [473]12926 Wrong value after assignment in initialize list using 13058 bit-fields 13059 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 13060 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 13061 * [478]13007 basic_streambuf::pubimbue, imbue wrong 13062 * [479]13009 Implicitly-defined assignment operator writes to wrong 13063 memory 13064 * [480]13057 regparm attribute not applied to destructor 13065 * [481]13070 -Wformat option ignored in g++ 13066 * [482]13081 forward template declarations in <complex> let inlining 13067 fail 13068 * [483]13239 Assertion does not seem to work correctly anymore 13069 * [484]13262 "xxx is private within this context" when initializing a 13070 self-contained template class 13071 * [485]13290 simple typo in concept checking for std::generate_n 13072 * [486]13323 Template code does not compile in presence of typedef 13073 * [487]13369 __verify_grouping (and __add_grouping?) not correct 13074 * [488]13371 infinite loop with packed struct and inlining 13075 * [489]13445 Template argument replacement "dereferences" a typedef 13076 * [490]13461 Fails to access protected-ctor from public constant 13077 * [491]13462 Non-standard-conforming type set::pointer 13078 * [492]13478 gcc uses wrong constructor to initialize a const 13079 reference 13080 * [493]13544 "conflicting types" for enums in different scopes 13081 * [494]13650 string::compare should not (always) use 13082 traits_type::length() 13083 * [495]13683 bogus warning about passing non-PODs through ellipsis 13084 * [496]13688 Derived class is denied access to protected base class 13085 member class 13086 * [497]13774 Member variable cleared in virtual multiple inheritance 13087 class 13088 * [498]13884 Protect sstream.tcc from extern template use 13089 13090 Java compiler and library 13091 13092 * [499]10746 [win32] garbage collection crash in GCJ 13093 13094 Objective-C compiler and library 13095 13096 * [500]11433 Crash due to dereferencing null pointer when querying 13097 protocol 13098 13099 Fortran compiler and library 13100 13101 * [501]12633 logical expression gives incorrect result with 13102 -fugly-logint option 13103 * [502]13037 [gcse-lm] g77 generates incorrect code 13104 * [503]13213 Hex constant problem when compiling with -fugly-logint 13105 and -ftypeless-boz 13106 13107 x86-specific (Intel/AMD) 13108 13109 * [504]4490 ICE with -m128bit-long-double 13110 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 13111 `reg' in make_field_assignment, in combine.c 13112 * [506]12441 ICE: can't find a register to spill 13113 * [507]12943 array static-init failure under -fpic, -fPIC 13114 * [508]13608 Incorrect code with -O3 -ffast-math 13115 13116 PowerPC-specific 13117 13118 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 13119 __attribute__((aligned(16))) 13120 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 13121 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 13122 altivec.md) 13123 * [512]12537 g++ generates writeable text sections 13124 13125 SPARC-specific 13126 13127 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 13128 -m64 13129 * [514]12865 mprotect call to make trampoline executable may fail 13130 * [515]13354 ICE in sparc_emit_set_const32 13131 13132 ARM-specific 13133 13134 * [516]10467 [arm] ICE in pre_insert_copy_insn, 13135 13136 ia64-specific 13137 13138 * [517]11226 ICE passing struct arg with two floats 13139 * [518]11227 ICE for _Complex float, _Complex long double args 13140 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 13141 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 13142 * Various fixes for libunwind 13143 13144 Alpha-specific 13145 13146 * [521]12654 Incorrect comparison code generated for Alpha 13147 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 13148 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 13149 13150 HPPA-specific 13151 13152 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 13153 * [525]12158 [hppa] compilation does not terminate at -O1 13154 13155 S390-specific 13156 13157 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 13158 (1<<24)-1 possible for CLCL-Instruction 13159 13160 SH-specific 13161 13162 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 13163 * [528]10392 optimizer generates faulty array indexing 13164 * [529]11322 SH profiler outputs multiple definitions of symbol 13165 * [530]13069 gcc/config/sh/rtems.h broken 13166 * [531]13302 Putting a va_list in a struct causes seg fault 13167 * [532]13585 Incorrect optimization of call to sfunc 13168 * Fix inappropriately exported libgcc functions from the shared 13169 library 13170 13171 Other embedded target specific 13172 13173 * [533]8916 [mcore] unsigned char assign gets hosed. 13174 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 13175 * [535]13122 [h8300] local variable gets corrupted by function call 13176 when -fomit-frame-pointer is given 13177 * [536]13256 [cris] strict_low_part mistreated in delay slots 13178 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 13179 -fexpensive-optimizations produces wrong code on mcore 13180 13181 GNU HURD-specific 13182 13183 * [538]12561 gcc/config/t-gnu needs updating to work with 13184 --with-sysroot 13185 13186 Tru64 Unix specific 13187 13188 * [539]6243 testsuite fails almost all tests due to no libintl in 13189 LD_LIBRARY_PATH during test. 13190 * [540]11397 weak aliases broken on Tru64 UNIX 13191 13192 AIX-specific 13193 13194 * [541]12505 build failure due to defines of uchar in cpphash.h and 13195 sys/types.h 13196 * [542]13150 WEAK symbols not exported by collect2 13197 13198 IRIX-specific 13199 13200 * [543]12666 fixincludes problem on IRIX 6.5.19m 13201 13202 Solaris-specific 13203 13204 * [544]12969 Including sys/byteorder.h breaks configure checks 13205 13206 Testsuite problems (compiler is not affected) 13207 13208 * [545]10819 testsuite creates CR+LF on compiler version lines in 13209 test summary files 13210 * [546]11612 abi_check not finding correct libgcc_s.so.1 13211 13212 Miscellaneous 13213 13214 * [547]13211 using -###, incorrect warnings about unused linker file 13215 are produced 13216 __________________________________________________________________ 13217 13218 GCC 3.3.4 13219 13220 This is the [548]list of problem reports (PRs) from GCC's bug tracking 13221 system that are known to be fixed in the 3.3.4 release. This list might 13222 not be complete (that is, it is possible that some PRs that have been 13223 fixed are not listed here). 13224 __________________________________________________________________ 13225 13226 GCC 3.3.5 13227 13228 This is the [549]list of problem reports (PRs) from GCC's bug tracking 13229 system that are known to be fixed in the 3.3.5 release. This list might 13230 not be complete (that is, it is possible that some PRs that have been 13231 fixed are not listed here). 13232 __________________________________________________________________ 13233 13234 GCC 3.3.6 13235 13236 This is the [550]list of problem reports (PRs) from GCC's bug tracking 13237 system that are known to be fixed in the 3.3.6 release. This list might 13238 not be complete (that is, it is possible that some PRs that have been 13239 fixed are not listed here). 13240 13241 13242 For questions related to the use of GCC, please consult these web 13243 pages and the [551]GCC manuals. If that fails, the 13244 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13245 web pages and the development of GCC are welcome on our developer 13246 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 13247 archives. 13248 13249 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 13250 distribution of this entire article is permitted in any medium, 13251 provided this notice is preserved. 13252 13253 These pages are [556]maintained by the GCC team. Last modified 13254 2016-01-30[557]. 13255 13256 References 13257 13258 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 13259 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 13260 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 13261 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 13262 5. http://gcc.gnu.org/news/dfa.html 13263 6. http://gcc.gnu.org/c99status.html 13264 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 13265 8. https://gcc.gnu.org/PR10140 13266 9. https://gcc.gnu.org/PR10198 13267 10. https://gcc.gnu.org/PR10338 13268 11. https://gcc.gnu.org/PR3581 13269 12. https://gcc.gnu.org/PR4382 13270 13. https://gcc.gnu.org/PR5533 13271 14. https://gcc.gnu.org/PR6387 13272 15. https://gcc.gnu.org/PR6412 13273 16. https://gcc.gnu.org/PR6620 13274 17. https://gcc.gnu.org/PR6663 13275 18. https://gcc.gnu.org/PR7068 13276 19. https://gcc.gnu.org/PR7083 13277 20. https://gcc.gnu.org/PR7647 13278 21. https://gcc.gnu.org/PR7675 13279 22. https://gcc.gnu.org/PR7718 13280 23. https://gcc.gnu.org/PR8116 13281 24. https://gcc.gnu.org/PR8358 13282 25. https://gcc.gnu.org/PR8511 13283 26. https://gcc.gnu.org/PR8564 13284 27. https://gcc.gnu.org/PR8660 13285 28. https://gcc.gnu.org/PR8766 13286 29. https://gcc.gnu.org/PR8803 13287 30. https://gcc.gnu.org/PR8846 13288 31. https://gcc.gnu.org/PR8906 13289 32. https://gcc.gnu.org/PR9216 13290 33. https://gcc.gnu.org/PR9261 13291 34. https://gcc.gnu.org/PR9263 13292 35. https://gcc.gnu.org/PR9429 13293 36. https://gcc.gnu.org/PR9516 13294 37. https://gcc.gnu.org/PR9600 13295 38. https://gcc.gnu.org/PR9629 13296 39. https://gcc.gnu.org/PR9672 13297 40. https://gcc.gnu.org/PR9749 13298 41. https://gcc.gnu.org/PR9794 13299 42. https://gcc.gnu.org/PR9829 13300 43. https://gcc.gnu.org/PR9916 13301 44. https://gcc.gnu.org/PR9936 13302 45. https://gcc.gnu.org/PR10262 13303 46. https://gcc.gnu.org/PR10278 13304 47. https://gcc.gnu.org/PR10446 13305 48. https://gcc.gnu.org/PR10451 13306 49. https://gcc.gnu.org/PR10506 13307 50. https://gcc.gnu.org/PR10549 13308 51. https://gcc.gnu.org/PR2001 13309 52. https://gcc.gnu.org/PR2391 13310 53. https://gcc.gnu.org/PR2960 13311 54. https://gcc.gnu.org/PR4046 13312 55. https://gcc.gnu.org/PR6405 13313 56. https://gcc.gnu.org/PR6798 13314 57. https://gcc.gnu.org/PR6871 13315 58. https://gcc.gnu.org/PR6909 13316 59. https://gcc.gnu.org/PR7189 13317 60. https://gcc.gnu.org/PR7642 13318 61. https://gcc.gnu.org/PR8634 13319 62. https://gcc.gnu.org/PR8750 13320 63. https://gcc.gnu.org/PR2161 13321 64. https://gcc.gnu.org/PR4319 13322 65. https://gcc.gnu.org/PR8602 13323 66. https://gcc.gnu.org/PR9177 13324 67. https://gcc.gnu.org/PR9853 13325 68. https://gcc.gnu.org/PR45 13326 69. https://gcc.gnu.org/PR3784 13327 70. https://gcc.gnu.org/PR764 13328 71. https://gcc.gnu.org/PR5116 13329 72. https://gcc.gnu.org/PR2862 13330 73. https://gcc.gnu.org/PR3663 13331 74. https://gcc.gnu.org/PR3797 13332 75. https://gcc.gnu.org/PR3948 13333 76. https://gcc.gnu.org/PR4137 13334 77. https://gcc.gnu.org/PR4361 13335 78. https://gcc.gnu.org/PR4802 13336 79. https://gcc.gnu.org/PR5837 13337 80. https://gcc.gnu.org/PR4803 13338 81. https://gcc.gnu.org/PR5094 13339 82. https://gcc.gnu.org/PR5730 13340 83. https://gcc.gnu.org/PR6713 13341 84. https://gcc.gnu.org/PR7015 13342 85. https://gcc.gnu.org/PR7086 13343 86. https://gcc.gnu.org/PR7099 13344 87. https://gcc.gnu.org/PR7247 13345 88. https://gcc.gnu.org/PR7441 13346 89. https://gcc.gnu.org/PR7768 13347 90. https://gcc.gnu.org/PR7804 13348 91. https://gcc.gnu.org/PR8099 13349 92. https://gcc.gnu.org/PR8117 13350 93. https://gcc.gnu.org/PR8205 13351 94. https://gcc.gnu.org/PR8645 13352 95. https://gcc.gnu.org/PR8724 13353 96. https://gcc.gnu.org/PR8805 13354 97. https://gcc.gnu.org/PR8691 13355 98. https://gcc.gnu.org/PR8700 13356 99. https://gcc.gnu.org/PR8724 13357 100. https://gcc.gnu.org/PR8949 13358 101. https://gcc.gnu.org/PR9016 13359 102. https://gcc.gnu.org/PR9053 13360 103. https://gcc.gnu.org/PR9152 13361 104. https://gcc.gnu.org/PR9182 13362 105. https://gcc.gnu.org/PR9297 13363 106. https://gcc.gnu.org/PR9318 13364 107. https://gcc.gnu.org/PR9320 13365 108. https://gcc.gnu.org/PR9400 13366 109. https://gcc.gnu.org/PR9424 13367 110. https://gcc.gnu.org/PR9425 13368 111. https://gcc.gnu.org/PR9439 13369 112. https://gcc.gnu.org/PR9474 13370 113. https://gcc.gnu.org/PR9548 13371 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 13372 115. https://gcc.gnu.org/PR9555 13373 116. https://gcc.gnu.org/PR9561 13374 117. https://gcc.gnu.org/PR9563 13375 118. https://gcc.gnu.org/PR9582 13376 119. https://gcc.gnu.org/PR9622 13377 120. https://gcc.gnu.org/PR9683 13378 121. https://gcc.gnu.org/PR9791 13379 122. https://gcc.gnu.org/PR9817 13380 123. https://gcc.gnu.org/PR9825 13381 124. https://gcc.gnu.org/PR9826 13382 125. https://gcc.gnu.org/PR9924 13383 126. https://gcc.gnu.org/PR9946 13384 127. https://gcc.gnu.org/PR9964 13385 128. https://gcc.gnu.org/PR9988 13386 129. https://gcc.gnu.org/PR10033 13387 130. https://gcc.gnu.org/PR10097 13388 131. https://gcc.gnu.org/PR10132 13389 132. https://gcc.gnu.org/PR10180 13390 133. https://gcc.gnu.org/PR10199 13391 134. https://gcc.gnu.org/PR10300 13392 135. https://gcc.gnu.org/PR10427 13393 136. https://gcc.gnu.org/PR10503 13394 137. https://gcc.gnu.org/PR5956 13395 138. https://gcc.gnu.org/PR1832 13396 139. https://gcc.gnu.org/PR3924 13397 140. https://gcc.gnu.org/PR5634 13398 141. https://gcc.gnu.org/PR6367 13399 142. https://gcc.gnu.org/PR6491 13400 143. https://gcc.gnu.org/PR6742 13401 144. https://gcc.gnu.org/PR7113 13402 145. https://gcc.gnu.org/PR7236 13403 146. https://gcc.gnu.org/PR7278 13404 147. https://gcc.gnu.org/PR7384 13405 148. https://gcc.gnu.org/PR7388 13406 149. https://gcc.gnu.org/PR8587 13407 150. https://gcc.gnu.org/PR9038 13408 151. https://gcc.gnu.org/PR10197 13409 152. https://gcc.gnu.org/PR6005 13410 153. https://gcc.gnu.org/PR6389 13411 154. https://gcc.gnu.org/PR6576 13412 155. https://gcc.gnu.org/PR6652 13413 156. https://gcc.gnu.org/PR7060 13414 157. https://gcc.gnu.org/PR7073 13415 158. https://gcc.gnu.org/PR7180 13416 159. https://gcc.gnu.org/PR7416 13417 160. https://gcc.gnu.org/PR7570 13418 161. https://gcc.gnu.org/PR7578 13419 162. https://gcc.gnu.org/PR7611 13420 163. https://gcc.gnu.org/PR7709 13421 164. https://gcc.gnu.org/PR7766 13422 165. https://gcc.gnu.org/PR7785 13423 166. https://gcc.gnu.org/PR7786 13424 167. https://gcc.gnu.org/PR8142 13425 168. https://gcc.gnu.org/PR8234 13426 169. https://gcc.gnu.org/PR8415 13427 170. https://gcc.gnu.org/PR8481 13428 171. https://gcc.gnu.org/PR8593 13429 172. https://gcc.gnu.org/PR8759 13430 173. https://gcc.gnu.org/PR8997 13431 174. https://gcc.gnu.org/PR9253 13432 175. https://gcc.gnu.org/PR9254 13433 176. https://gcc.gnu.org/PR9271 13434 177. https://gcc.gnu.org/PR6767 13435 178. https://gcc.gnu.org/PR9911 13436 179. https://gcc.gnu.org/PR10020 13437 180. https://gcc.gnu.org/PR10546 13438 181. https://gcc.gnu.org/PR7029 13439 182. https://gcc.gnu.org/PR2903 13440 183. https://gcc.gnu.org/PR7873 13441 184. https://gcc.gnu.org/PR7680 13442 185. https://gcc.gnu.org/PR8705 13443 186. https://gcc.gnu.org/PR9986 13444 187. https://gcc.gnu.org/PR10056 13445 188. https://gcc.gnu.org/PR6744 13446 189. https://gcc.gnu.org/PR7361 13447 190. https://gcc.gnu.org/PR9496 13448 191. https://gcc.gnu.org/PR7067 13449 192. https://gcc.gnu.org/PR8480 13450 193. https://gcc.gnu.org/PR8784 13451 194. https://gcc.gnu.org/PR10315 13452 195. https://gcc.gnu.org/PR10267 13453 196. https://gcc.gnu.org/PR7916 13454 197. https://gcc.gnu.org/PR7926 13455 198. https://gcc.gnu.org/PR8555 13456 199. https://gcc.gnu.org/PR8994 13457 200. https://gcc.gnu.org/PR9426 13458 201. https://gcc.gnu.org/PR9806 13459 202. https://gcc.gnu.org/PR10077 13460 203. https://gcc.gnu.org/PR10233 13461 204. https://gcc.gnu.org/PR10286 13462 205. https://gcc.gnu.org/PR10308 13463 206. https://gcc.gnu.org/PR11272 13464 207. https://gcc.gnu.org/PR5754 13465 208. https://gcc.gnu.org/PR6597 13466 209. https://gcc.gnu.org/PR6949 13467 210. https://gcc.gnu.org/PR7053 13468 211. https://gcc.gnu.org/PR8164 13469 212. https://gcc.gnu.org/PR8384 13470 213. https://gcc.gnu.org/PR9559 13471 214. https://gcc.gnu.org/PR9649 13472 215. https://gcc.gnu.org/PR9864 13473 216. https://gcc.gnu.org/PR10432 13474 217. https://gcc.gnu.org/PR10475 13475 218. https://gcc.gnu.org/PR10635 13476 219. https://gcc.gnu.org/PR10661 13477 220. https://gcc.gnu.org/PR10700 13478 221. https://gcc.gnu.org/PR10712 13479 222. https://gcc.gnu.org/PR10796 13480 223. https://gcc.gnu.org/PR10890 13481 224. https://gcc.gnu.org/PR10939 13482 225. https://gcc.gnu.org/PR10956 13483 226. https://gcc.gnu.org/PR11041 13484 227. https://gcc.gnu.org/PR11059 13485 228. https://gcc.gnu.org/PR11083 13486 229. https://gcc.gnu.org/PR11105 13487 230. https://gcc.gnu.org/PR11149 13488 231. https://gcc.gnu.org/PR11228 13489 232. https://gcc.gnu.org/PR11282 13490 233. https://gcc.gnu.org/PR11301 13491 234. https://gcc.gnu.org/PR11308 13492 235. https://gcc.gnu.org/PR11473 13493 236. https://gcc.gnu.org/PR11503 13494 237. https://gcc.gnu.org/PR11513 13495 238. https://gcc.gnu.org/PR11198 13496 239. https://gcc.gnu.org/PR11304 13497 240. https://gcc.gnu.org/PR11381 13498 241. https://gcc.gnu.org/PR11536 13499 242. https://gcc.gnu.org/PR11557 13500 243. https://gcc.gnu.org/PR5897 13501 244. https://gcc.gnu.org/PR11279 13502 245. https://gcc.gnu.org/PR11022 13503 246. https://gcc.gnu.org/PR2330 13504 247. https://gcc.gnu.org/PR5388 13505 248. https://gcc.gnu.org/PR5390 13506 249. https://gcc.gnu.org/PR7877 13507 250. https://gcc.gnu.org/PR9393 13508 251. https://gcc.gnu.org/PR10032 13509 252. https://gcc.gnu.org/PR10468 13510 253. https://gcc.gnu.org/PR10527 13511 254. https://gcc.gnu.org/PR10679 13512 255. https://gcc.gnu.org/PR10682 13513 256. https://gcc.gnu.org/PR10689 13514 257. https://gcc.gnu.org/PR10845 13515 258. https://gcc.gnu.org/PR10849 13516 259. https://gcc.gnu.org/PR10888 13517 260. https://gcc.gnu.org/PR10929 13518 261. https://gcc.gnu.org/PR10931 13519 262. https://gcc.gnu.org/PR10940 13520 263. https://gcc.gnu.org/PR10968 13521 264. https://gcc.gnu.org/PR10990 13522 265. https://gcc.gnu.org/PR11039 13523 266. https://gcc.gnu.org/PR11062 13524 267. https://gcc.gnu.org/PR11095 13525 268. https://gcc.gnu.org/PR11098 13526 269. https://gcc.gnu.org/PR11137 13527 270. https://gcc.gnu.org/PR11154 13528 271. https://gcc.gnu.org/PR11329 13529 272. https://gcc.gnu.org/PR11332 13530 273. https://gcc.gnu.org/PR11431 13531 274. https://gcc.gnu.org/PR11528 13532 275. https://gcc.gnu.org/PR11546 13533 276. https://gcc.gnu.org/PR11567 13534 277. https://gcc.gnu.org/PR11645 13535 278. https://gcc.gnu.org/PR5179 13536 279. https://gcc.gnu.org/PR8204 13537 280. https://gcc.gnu.org/PR10838 13538 281. https://gcc.gnu.org/PR10886 13539 282. https://gcc.gnu.org/PR11349 13540 283. https://gcc.gnu.org/PR4823 13541 284. https://gcc.gnu.org/PR8878 13542 285. https://gcc.gnu.org/PR9815 13543 286. https://gcc.gnu.org/PR10402 13544 287. https://gcc.gnu.org/PR10504 13545 288. https://gcc.gnu.org/PR10673 13546 289. https://gcc.gnu.org/PR11044 13547 290. https://gcc.gnu.org/PR11089 13548 291. https://gcc.gnu.org/PR11420 13549 292. https://gcc.gnu.org/PR9362 13550 293. https://gcc.gnu.org/PR10142 13551 294. https://gcc.gnu.org/PR10663 13552 295. https://gcc.gnu.org/PR10835 13553 296. https://gcc.gnu.org/PR10876 13554 297. https://gcc.gnu.org/PR10955 13555 298. https://gcc.gnu.org/PR11018 13556 299. https://gcc.gnu.org/PR11556 13557 300. https://gcc.gnu.org/PR10907 13558 301. https://gcc.gnu.org/PR11320 13559 302. https://gcc.gnu.org/PR11599 13560 303. https://gcc.gnu.org/PR9745 13561 304. https://gcc.gnu.org/PR10871 13562 305. https://gcc.gnu.org/PR11440 13563 306. https://gcc.gnu.org/PR7594 13564 307. https://gcc.gnu.org/PR10557 13565 308. https://gcc.gnu.org/PR11054 13566 309. https://gcc.gnu.org/PR10834 13567 310. https://gcc.gnu.org/PR10842 13568 311. https://gcc.gnu.org/PR11052 13569 312. https://gcc.gnu.org/PR11183 13570 313. https://gcc.gnu.org/PR11084 13571 314. https://gcc.gnu.org/PR10331 13572 315. https://gcc.gnu.org/PR10413 13573 316. https://gcc.gnu.org/PR11096 13574 317. https://gcc.gnu.org/PR2873 13575 318. https://gcc.gnu.org/PR3163 13576 319. https://gcc.gnu.org/PR5287 13577 320. https://gcc.gnu.org/PR10148 13578 321. https://gcc.gnu.org/PR8787 13579 322. https://gcc.gnu.org/PR10900 13580 323. https://gcc.gnu.org/PR1607 13581 324. https://gcc.gnu.org/PR4252 13582 325. https://gcc.gnu.org/PR4490 13583 326. https://gcc.gnu.org/PR10355 13584 327. https://gcc.gnu.org/PR10726 13585 328. https://gcc.gnu.org/PR10805 13586 329. https://gcc.gnu.org/PR10815 13587 330. https://gcc.gnu.org/PR10877 13588 331. https://gcc.gnu.org/PR11280 13589 332. https://gcc.gnu.org/PR11466 13590 333. https://gcc.gnu.org/PR10737 13591 334. https://gcc.gnu.org/PR10810 13592 335. https://gcc.gnu.org/PR8336 13593 336. https://gcc.gnu.org/PR9330 13594 337. https://gcc.gnu.org/PR9631 13595 338. https://gcc.gnu.org/PR9877 13596 339. https://gcc.gnu.org/PR11687 13597 340. https://gcc.gnu.org/PR12263 13598 341. https://gcc.gnu.org/PR12490 13599 342. https://gcc.gnu.org/PR7277 13600 343. https://gcc.gnu.org/PR7939 13601 344. https://gcc.gnu.org/PR11063 13602 345. https://gcc.gnu.org/PR11207 13603 346. https://gcc.gnu.org/PR11522 13604 347. https://gcc.gnu.org/PR11595 13605 348. https://gcc.gnu.org/PR11646 13606 349. https://gcc.gnu.org/PR11665 13607 350. https://gcc.gnu.org/PR11852 13608 351. https://gcc.gnu.org/PR11878 13609 352. https://gcc.gnu.org/PR11883 13610 353. https://gcc.gnu.org/PR11991 13611 354. https://gcc.gnu.org/PR12146 13612 355. https://gcc.gnu.org/PR12215 13613 356. https://gcc.gnu.org/PR12369 13614 357. https://gcc.gnu.org/PR12446 13615 358. https://gcc.gnu.org/PR12510 13616 359. https://gcc.gnu.org/PR12544 13617 360. https://gcc.gnu.org/PR9862 13618 361. https://gcc.gnu.org/PR10962 13619 362. https://gcc.gnu.org/PR11370 13620 363. https://gcc.gnu.org/PR11637 13621 364. https://gcc.gnu.org/PR11885 13622 365. https://gcc.gnu.org/PR12082 13623 366. https://gcc.gnu.org/PR12180 13624 367. https://gcc.gnu.org/PR12340 13625 368. https://gcc.gnu.org/PR3907 13626 369. https://gcc.gnu.org/PR5293 13627 370. https://gcc.gnu.org/PR5296 13628 371. https://gcc.gnu.org/PR7939 13629 372. https://gcc.gnu.org/PR8656 13630 373. https://gcc.gnu.org/PR10147 13631 374. https://gcc.gnu.org/PR11400 13632 375. https://gcc.gnu.org/PR11409 13633 376. https://gcc.gnu.org/PR11740 13634 377. https://gcc.gnu.org/PR11786 13635 378. https://gcc.gnu.org/PR11867 13636 379. https://gcc.gnu.org/PR11928 13637 380. https://gcc.gnu.org/PR12114 13638 381. https://gcc.gnu.org/PR12163 13639 382. https://gcc.gnu.org/PR12181 13640 383. https://gcc.gnu.org/PR12236 13641 384. https://gcc.gnu.org/PR12266 13642 385. https://gcc.gnu.org/PR12296 13643 386. https://gcc.gnu.org/PR12298 13644 387. https://gcc.gnu.org/PR12369 13645 388. https://gcc.gnu.org/PR12337 13646 389. https://gcc.gnu.org/PR12344 13647 390. https://gcc.gnu.org/PR12451 13648 391. https://gcc.gnu.org/PR12486 13649 392. https://gcc.gnu.org/PR8869 13650 393. https://gcc.gnu.org/PR9786 13651 394. https://gcc.gnu.org/PR11689 13652 395. https://gcc.gnu.org/PR12116 13653 396. https://gcc.gnu.org/PR12070 13654 397. https://gcc.gnu.org/PR11184 13655 398. https://gcc.gnu.org/PR11535 13656 399. https://gcc.gnu.org/PR11693 13657 400. https://gcc.gnu.org/PR12224 13658 401. https://gcc.gnu.org/PR11087 13659 402. https://gcc.gnu.org/PR11319 13660 403. https://gcc.gnu.org/PR11949 13661 404. https://gcc.gnu.org/PR11662 13662 405. https://gcc.gnu.org/PR11965 13663 406. https://gcc.gnu.org/PR12301 13664 407. https://gcc.gnu.org/PR11717 13665 408. https://gcc.gnu.org/PR11313 13666 409. https://gcc.gnu.org/PR11712 13667 410. https://gcc.gnu.org/PR12166 13668 411. https://gcc.gnu.org/PR12101 13669 412. https://gcc.gnu.org/PR10988 13670 413. https://gcc.gnu.org/PR11805 13671 414. https://gcc.gnu.org/PR11902 13672 415. https://gcc.gnu.org/PR11903 13673 416. https://gcc.gnu.org/PR11890 13674 417. https://gcc.gnu.org/PR12399 13675 418. https://gcc.gnu.org/PR13068 13676 419. https://gcc.gnu.org/PR10060 13677 420. https://gcc.gnu.org/PR10555 13678 421. https://gcc.gnu.org/PR10706 13679 422. https://gcc.gnu.org/PR11496 13680 423. https://gcc.gnu.org/PR11741 13681 424. https://gcc.gnu.org/PR12440 13682 425. https://gcc.gnu.org/PR12632 13683 426. https://gcc.gnu.org/PR12712 13684 427. https://gcc.gnu.org/PR12726 13685 428. https://gcc.gnu.org/PR12890 13686 429. https://gcc.gnu.org/PR12900 13687 430. https://gcc.gnu.org/PR13060 13688 431. https://gcc.gnu.org/PR13289 13689 432. https://gcc.gnu.org/PR13318 13690 433. https://gcc.gnu.org/PR13392 13691 434. https://gcc.gnu.org/PR13574 13692 435. https://gcc.gnu.org/PR13475 13693 436. https://gcc.gnu.org/PR13797 13694 437. https://gcc.gnu.org/PR13824 13695 438. https://gcc.gnu.org/PR8776 13696 439. https://gcc.gnu.org/PR10339 13697 440. https://gcc.gnu.org/PR11350 13698 441. https://gcc.gnu.org/PR12826 13699 442. https://gcc.gnu.org/PR12500 13700 443. https://gcc.gnu.org/PR12941 13701 444. https://gcc.gnu.org/PR12953 13702 445. https://gcc.gnu.org/PR13041 13703 446. https://gcc.gnu.org/PR13507 13704 447. https://gcc.gnu.org/PR13382 13705 448. https://gcc.gnu.org/PR13394 13706 449. https://gcc.gnu.org/PR13400 13707 450. https://gcc.gnu.org/PR13521 13708 451. https://gcc.gnu.org/PR2094 13709 452. https://gcc.gnu.org/PR2294 13710 453. https://gcc.gnu.org/PR5050 13711 454. https://gcc.gnu.org/PR9371 13712 455. https://gcc.gnu.org/PR9546 13713 456. https://gcc.gnu.org/PR10081 13714 457. https://gcc.gnu.org/PR10093 13715 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 13716 459. https://gcc.gnu.org/PR10095 13717 460. https://gcc.gnu.org/PR11554 13718 461. https://gcc.gnu.org/PR12297 13719 462. https://gcc.gnu.org/PR12352 13720 463. https://gcc.gnu.org/PR12438 13721 464. https://gcc.gnu.org/PR12540 13722 465. https://gcc.gnu.org/PR12594 13723 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 13724 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 13725 468. https://gcc.gnu.org/PR12657 13726 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 13727 470. https://gcc.gnu.org/PR12696 13728 471. https://gcc.gnu.org/PR12815 13729 472. https://gcc.gnu.org/PR12862 13730 473. https://gcc.gnu.org/PR12926 13731 474. https://gcc.gnu.org/PR12967 13732 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 13733 476. https://gcc.gnu.org/PR12971 13734 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 13735 478. https://gcc.gnu.org/PR13007 13736 479. https://gcc.gnu.org/PR13009 13737 480. https://gcc.gnu.org/PR13057 13738 481. https://gcc.gnu.org/PR13070 13739 482. https://gcc.gnu.org/PR13081 13740 483. https://gcc.gnu.org/PR13239 13741 484. https://gcc.gnu.org/PR13262 13742 485. https://gcc.gnu.org/PR13290 13743 486. https://gcc.gnu.org/PR13323 13744 487. https://gcc.gnu.org/PR13369 13745 488. https://gcc.gnu.org/PR13371 13746 489. https://gcc.gnu.org/PR13445 13747 490. https://gcc.gnu.org/PR13461 13748 491. https://gcc.gnu.org/PR13462 13749 492. https://gcc.gnu.org/PR13478 13750 493. https://gcc.gnu.org/PR13544 13751 494. https://gcc.gnu.org/PR13650 13752 495. https://gcc.gnu.org/PR13683 13753 496. https://gcc.gnu.org/PR13688 13754 497. https://gcc.gnu.org/PR13774 13755 498. https://gcc.gnu.org/PR13884 13756 499. https://gcc.gnu.org/PR10746 13757 500. https://gcc.gnu.org/PR11433 13758 501. https://gcc.gnu.org/PR12633 13759 502. https://gcc.gnu.org/PR13037 13760 503. https://gcc.gnu.org/PR13213 13761 504. https://gcc.gnu.org/PR4490 13762 505. https://gcc.gnu.org/PR12292 13763 506. https://gcc.gnu.org/PR12441 13764 507. https://gcc.gnu.org/PR12943 13765 508. https://gcc.gnu.org/PR13608 13766 509. https://gcc.gnu.org/PR11598 13767 510. https://gcc.gnu.org/PR11793 13768 511. https://gcc.gnu.org/PR12467 13769 512. https://gcc.gnu.org/PR12537 13770 513. https://gcc.gnu.org/PR12496 13771 514. https://gcc.gnu.org/PR12865 13772 515. https://gcc.gnu.org/PR13354 13773 516. https://gcc.gnu.org/PR10467 13774 517. https://gcc.gnu.org/PR11226 13775 518. https://gcc.gnu.org/PR11227 13776 519. https://gcc.gnu.org/PR12644 13777 520. https://gcc.gnu.org/PR13149 13778 521. https://gcc.gnu.org/PR12654 13779 522. https://gcc.gnu.org/PR12965 13780 523. https://gcc.gnu.org/PR13031 13781 524. https://gcc.gnu.org/PR11634 13782 525. https://gcc.gnu.org/PR12158 13783 526. https://gcc.gnu.org/PR11992 13784 527. https://gcc.gnu.org/PR9365 13785 528. https://gcc.gnu.org/PR10392 13786 529. https://gcc.gnu.org/PR11322 13787 530. https://gcc.gnu.org/PR13069 13788 531. https://gcc.gnu.org/PR13302 13789 532. https://gcc.gnu.org/PR13585 13790 533. https://gcc.gnu.org/PR8916 13791 534. https://gcc.gnu.org/PR11576 13792 535. https://gcc.gnu.org/PR13122 13793 536. https://gcc.gnu.org/PR13256 13794 537. https://gcc.gnu.org/PR13373 13795 538. https://gcc.gnu.org/PR12561 13796 539. https://gcc.gnu.org/PR6243 13797 540. https://gcc.gnu.org/PR11397 13798 541. https://gcc.gnu.org/PR12505 13799 542. https://gcc.gnu.org/PR13150 13800 543. https://gcc.gnu.org/PR12666 13801 544. https://gcc.gnu.org/PR12969 13802 545. https://gcc.gnu.org/PR10819 13803 546. https://gcc.gnu.org/PR11612 13804 547. https://gcc.gnu.org/PR13211 13805 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 13806 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 13807 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 13808 551. https://gcc.gnu.org/onlinedocs/ 13809 552. mailto:gcc-help (a] gcc.gnu.org 13810 553. mailto:gcc (a] gcc.gnu.org 13811 554. https://gcc.gnu.org/lists.html 13812 555. http://www.fsf.org/ 13813 556. https://gcc.gnu.org/about.html 13814 557. http://validator.w3.org/check/referer 13815 ====================================================================== 13816 http://gcc.gnu.org/gcc-3.2/index.html 13817 13818 GCC 3.2 Release Series 13819 13820 April 25, 2003 13821 13822 The [1]GNU project and the GCC developers are pleased to announce the 13823 release of GCC 3.2.3. 13824 13825 The purpose of the GCC 3.2 release series is to provide a stable 13826 platform for OS distributors to use building their next releases. A 13827 primary objective was to stabilize the C++ ABI; we believe that the 13828 interface to the compiler and the C++ standard library are now 13829 relatively stable. 13830 13831 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 13832 interoperate with code compiled by GCC 3.1.1 or earlier. 13833 13834 Please refer to our [2]detailed list of news, caveats, and bug-fixes 13835 for further information. 13836 13837 This release series is no longer maintained. 13838 13839 Release History 13840 13841 GCC 3.2.3 13842 April 25, 2003 ([3]changes) 13843 13844 GCC 3.2.2 13845 February 5, 2003 ([4]changes) 13846 13847 GCC 3.2.1 13848 November 19, 2002 ([5]changes) 13849 13850 GCC 3.2 13851 August 14, 2002 ([6]changes) 13852 13853 References and Acknowledgements 13854 13855 GCC used to stand for the GNU C Compiler, but since the compiler 13856 supports several other languages aside from C, it now stands for the 13857 GNU Compiler Collection. 13858 13859 A list of [7]successful builds is updated as new information becomes 13860 available. 13861 13862 The GCC developers would like to thank the numerous people that have 13863 contributed new features, improvements, bug fixes, and other changes as 13864 well as test results to GCC. This [8]amazing group of volunteers is 13865 what makes GCC successful. 13866 13867 For additional information about GCC please refer to the [9]GCC project 13868 web site or contact the [10]GCC development mailing list. 13869 13870 To obtain GCC please use [11]our mirror sites, or our CVS server. 13871 13872 13873 For questions related to the use of GCC, please consult these web 13874 pages and the [12]GCC manuals. If that fails, the 13875 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13876 web pages and the development of GCC are welcome on our developer 13877 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 13878 archives. 13879 13880 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 13881 distribution of this entire article is permitted in any medium, 13882 provided this notice is preserved. 13883 13884 These pages are [17]maintained by the GCC team. Last modified 13885 2016-09-30[18]. 13886 13887 References 13888 13889 1. http://www.gnu.org/ 13890 2. http://gcc.gnu.org/gcc-3.2/changes.html 13891 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 13892 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 13893 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 13894 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 13895 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 13896 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13897 9. http://gcc.gnu.org/index.html 13898 10. mailto:gcc (a] gcc.gnu.org 13899 11. http://gcc.gnu.org/mirrors.html 13900 12. https://gcc.gnu.org/onlinedocs/ 13901 13. mailto:gcc-help (a] gcc.gnu.org 13902 14. mailto:gcc (a] gcc.gnu.org 13903 15. https://gcc.gnu.org/lists.html 13904 16. http://www.fsf.org/ 13905 17. https://gcc.gnu.org/about.html 13906 18. http://validator.w3.org/check/referer 13907 ====================================================================== 13908 http://gcc.gnu.org/gcc-3.2/changes.html 13909 13910 GCC 3.2 Release Series 13911 Changes, New Features, and Fixes 13912 13913 The latest release in the 3.2 release series is [1]GCC 3.2.3. 13914 13915 Caveats and New Features 13916 13917 Caveats 13918 13919 * The C++ compiler does not correctly zero-initialize 13920 pointers-to-data members. You must explicitly initialize them. For 13921 example: int S::*m(0); will work, but depending on 13922 default-initialization to zero will not work. This bug cannot be 13923 fixed in GCC 3.2 without inducing unacceptable risks. It will be 13924 fixed in GCC 3.3. 13925 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 13926 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 13927 a number of C++ ABI fixes which make its C++ compiler generate 13928 binary code which is incompatible with the C++ compilers found in 13929 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 13930 13931 Frontend Enhancements 13932 13933 C/C++/Objective-C 13934 13935 * The method of constructing the list of directories to be searched 13936 for header files has been revised. If a directory named by a -I 13937 option is a standard system include directory, the option is 13938 ignored to ensure that the default search order for system 13939 directories and the special treatment of system header files are 13940 not defeated. 13941 * The C and Objective-C compilers no longer accept the "Naming Types" 13942 extension (typedef foo = bar); it was already unavailable in C++. 13943 Code which uses it will need to be changed to use the "typeof" 13944 extension instead: typedef typeof(bar) foo. (We have removed this 13945 extension without a period of deprecation because it has caused the 13946 compiler to crash since version 3.0 and no one noticed until very 13947 recently. Thus we conclude it is not in widespread use.) 13948 13949 C++ 13950 13951 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 13952 in GCC and the multi-vendor standard, but more have been found 13953 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 13954 code which is affected by these bugs. We will fix these bugs in 13955 some future release, once we are confident that all have been 13956 found; until then, it is our intention to make changes to the ABI 13957 only if they are necessary for correct compilation of C++, as 13958 opposed to conformance to the ABI documents. 13959 * For details on how to build an ABI compliant compiler for GNU/Linux 13960 systems, check the [3]common C++ ABI page. 13961 13962 New Targets and Target Specific Improvements 13963 13964 IA-32 13965 13966 * Fixed a number of bugs in SSE and MMX intrinsics. 13967 * Fixed common compiler crashes with SSE instruction set enabled 13968 (implied by -march=pentium3, pentium4, athlon-xp) 13969 * __m128 and __m128i is not 128bit aligned when used in structures. 13970 13971 x86-64 13972 13973 * A bug whereby the compiler could generate bad code for bzero has 13974 been fixed. 13975 * ABI fixes (implying ABI incompatibilities with previous version in 13976 some corner cases) 13977 * Fixed prefetch code generation 13978 __________________________________________________________________ 13979 13980 GCC 3.2.3 13981 13982 3.2.3 is a bug fix release only; there are no new features that were 13983 not present in GCC 3.2.2. 13984 13985 Bug Fixes 13986 13987 This section lists the problem reports (PRs) from GCC's bug tracking 13988 system that are known to be fixed in the 3.2.3 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), and some of the titles have been changed to 13991 make them more clear. 13992 13993 Internal Compiler Errors (multi-platform) 13994 13995 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 13996 cc1plus 13997 * [5]6440: (c++) template specializations cause ICE 13998 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 13999 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 14000 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 14001 * [9]8068: exceedingly high (infinite) memory usage 14002 * [10]8178: ICE with __builtin_ffs 14003 * [11]8396: ICE in copy_to_mode_reg, in explow.c 14004 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 14005 * [13]9768: ICE when optimizing inline code at -O2 14006 * [14]9798: (c++) Infinite recursion (segfault) in 14007 cp/decl.c:push_using_directive with recursive using directives 14008 * [15]9799: mismatching structure initializer with nested flexible 14009 array member: ICE 14010 * [16]9928: ICE on duplicate enum declaration 14011 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 14012 sparc, alpha) 14013 * [18]10352: ICE in find_reloads_toplev 14014 * [19]10336: ICE with -Wunreachable-code 14015 14016 C/optimizer bugs: 14017 14018 * [20]8224: Incorrect joining of signed and unsigned division 14019 * [21]8613: -O2 produces wrong code with builtin strlen and 14020 postincrements 14021 * [22]8828: gcc reports some code is unreachable when it is not 14022 * [23]9226: GCSE breaking argument passing 14023 * [24]9853: miscompilation of non-constant structure initializer 14024 * [25]9797: C99-style struct initializers are miscompiled 14025 * [26]9967: Some standard C function calls should not be replaced 14026 when optimizing for size 14027 * [27]10116: ce2: invalid merge of join_bb in the context of switch 14028 statements 14029 * [28]10171: wrong code for inlined function 14030 * [29]10175: -Wunreachable-code doesn't work for single lines 14031 14032 C++ compiler and library: 14033 14034 * [30]8316: Confusing diagnostic for code that misuses conversion 14035 operators 14036 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 14037 * [32]9420: incomplete type incorrectly reported 14038 * [33]9459: typeof in return type specification of template not 14039 supported 14040 * [34]9507: filebuf::open handles ios_base::ate incorrectly 14041 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 14042 * [36]9602: Total confusion about template/friend/virtual/abstract 14043 * [37]9993: destructor not called for local object created within and 14044 returned from infinite loop 14045 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 14046 system 14047 14048 Java compiler and library: 14049 14050 * [39]9652: libgcj build fails on irix6.5.1[78] 14051 * [40]10144: gas on solaris complains about bad .stabs lines for 14052 java, native as unaffected 14053 14054 x86-specific (Intel/AMD): 14055 14056 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 14057 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 14058 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 14059 failed 14060 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 14061 failed 14062 14063 SPARC-specific: 14064 14065 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 14066 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 14067 unroll.c 14068 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 14069 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 14070 execute/loop-2d.c 14071 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 14072 * [50]9414: Scheduling bug on Ultrasparc 14073 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 14074 14075 m68k-specific: 14076 14077 * [52]7248: broken "inclusive or" code 14078 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 14079 14080 PowerPC-specific: 14081 14082 * [54]9732: Wrong code with -O2 -fPIC 14083 * [55]10073: ICE: powerpc cannot split insn 14084 14085 Alpha-specific: 14086 14087 * [56]7702: optimization problem on a DEC alpha under OSF1 14088 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 14089 14090 HP-specific: 14091 14092 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 14093 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 14094 (missing symbol) 14095 * [60]10271: Floating point args don't get reloaded across function 14096 calls with -O2 14097 14098 MIPS specific: 14099 14100 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 14101 compile/920501-4.c 14102 14103 CRIS specific: 14104 14105 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 14106 14107 Miscellaneous and minor bugs: 14108 14109 * [63]6955: collect2 says "core dumped" when there is no core 14110 __________________________________________________________________ 14111 14112 GCC 3.2.2 14113 14114 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 14115 install by means of the DESTDIR variable. Parts of the GCC tree have 14116 featured that support long before, but now it is available even from 14117 the top level. 14118 14119 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 14120 features that were not present in GCC 3.2.1. 14121 14122 Bug Fixes 14123 14124 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 14125 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 14126 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 14127 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 14128 change, and thus restores ABI-compatibility with previous releases 14129 (except GCC 3.2.1) on these platforms. 14130 14131 This section lists the problem reports (PRs) from GCC's bug tracking 14132 system that are known to be fixed in the 3.2.2 release. This list might 14133 not be complete (that is, it is possible that some PRs that have been 14134 fixed are not listed here) and some of the titles have been changed to 14135 make them more clear. 14136 14137 Internal Compiler Errors (multi-platform) 14138 14139 * [64]5919: (c++) ICE when passing variable array to template 14140 function 14141 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 14142 * [66]7507: ICE with -O2 when address of called function is a 14143 complicated expression 14144 * [67]7622: ICE with nested inline functions if function's address is 14145 taken 14146 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 14147 [69]9258) 14148 * [70]8031: (c++) ICE in code comparing typeids and casting from 14149 virtual base 14150 * [71]8275: ICE in simplify_subreg 14151 * [72]8332: (c++) builtin strlen/template interaction causes ICE 14152 * [73]8372: (c++) ICE on explicit call of destructor 14153 * [74]8439: (c, not c++) empty struct causes ICE 14154 * [75]8442: (c++) ICE with nested template classes 14155 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 14156 * [77]8615: (c++) ICE with out-of-range character constant template 14157 argument 14158 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 14159 * [79]8799: (c++) ICE: error reporting routines re-entered 14160 * [80]9328: (c++) ICE with typeof(X) for overloaded X 14161 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 14162 14163 C++ (compiler and library) bugs 14164 14165 * [82]47: scoping in nested classes is broken 14166 * [83]6745: problems with iostream rdbuf() member function 14167 * [84]8214: conversion from const char* const to char* sometimes 14168 accepted illegally 14169 * [85]8493: builtin strlen and overload resolution (same bug as 14170 [86]8332) 14171 * [87]8503: strange behaviour of function types 14172 * [88]8727: compiler confused by inheritance from an anonymous struct 14173 * [89]7445: poor performance of std::locale::classic() in 14174 multi-threaded applications 14175 * [90]8230: mishandling of overflow in vector<T>::resize 14176 * [91]8399: sync_with_stdio(false) breaks unformatted input 14177 * [92]8662: illegal access of private member of unnamed class is 14178 accepted 14179 * [93]8707: "make distclean" fails in libstdc++-v3 directory 14180 * [94]8708: __USE_MALLOC doesn't work 14181 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 14182 * [96]8887: Bug in date formats with --enable-clocale=generic 14183 * [97]9076: Call Frame Instructions are not handled correctly during 14184 unwind operation 14185 * [98]9151: std::setprecision limited to 16 digits when outputting a 14186 double to a stream 14187 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 14188 * [100]9269: libstdc++ headers: explicit specialization of function 14189 must precede its first use 14190 * [101]9322: return value of basic_streambuf<>::getloc affected by 14191 locale::global 14192 * [102]9433: segfault in runtime support for dynamic_cast 14193 14194 C and optimizer bugs 14195 14196 * [103]8032: GCC incorrectly initializes static structs that have 14197 flexible arrays 14198 * [104]8639: simple arithmetic expression broken 14199 * [105]8794: optimization improperly eliminates certain expressions 14200 * [106]8832: traditional "asm volatile" code is illegally optimized 14201 * [107]8988: loop optimizer bug: with -O2, code is generated that 14202 segfaults (found on i386, bug present for all platforms) 14203 * [108]9492: structure copy clobbers subsequent stores to structure 14204 14205 Objective-C bugs 14206 14207 * [109]9267: Objective-C parser won't build with newer bison versions 14208 (e.g. 1.875) 14209 14210 Ada bugs 14211 14212 * [110]8344: Ada build problem due to conflict between gcc/final.o, 14213 gcc/ada/final.o 14214 14215 Preprocessor bugs 14216 14217 * [111]8524: _Pragma within macros is improperly expanded 14218 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 14219 -fshort-wchar 14220 14221 ARM-specific 14222 14223 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 14224 14225 x86-specific (Intel/AMD) 14226 14227 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 14228 * [115]8599: loop unroll bug with -march=k6-3 14229 * [116]9506: ABI breakage in structure return (affects BSD and 14230 Cygwin, but not GNU/Linux) 14231 14232 FreeBSD 5.0 specific 14233 14234 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 14235 14236 RTEMS-specific 14237 14238 * [118]9292: hppa1.1-rtems configurery problems 14239 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 14240 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 14241 * [121]9296: gthr-rtems regression 14242 * [122]9316: powerpc-rtems: extending multilibs 14243 14244 HP-PA specific 14245 14246 * [123]9493: ICE with -O2 when building a simple function 14247 14248 Documentation 14249 14250 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 14251 * [125]8947: Please add a warning about "-malign-double" in docs 14252 * [126]7448, [127]8882: typo cleanups 14253 __________________________________________________________________ 14254 14255 GCC 3.2.1 14256 14257 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 14258 generates code that is known not to be binary-compatible with the 14259 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 14260 in the distribution, for details. 14261 14262 This release also removes an old GCC extension, "naming types", and the 14263 documentation now directs users to use a different GCC extension, 14264 __typeof__, instead. The feature had evidently been broken for a while. 14265 14266 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 14267 the new warning there are no new features that were not present in GCC 14268 3.2. 14269 14270 In addition, the previous fix for [128]PR 7445 (poor performance of 14271 std::locale::classic() in multi-threaded applications) was reverted 14272 ("unfixed"), because the "fix" was not thread-safe. 14273 14274 Bug Fixes 14275 14276 This section lists the problem reports (PRs) from GCC's bug tracking 14277 system that are known to be fixed in the 3.2.1 release. This list might 14278 not be complete (that is, it is possible that some PRs that have been 14279 fixed are not listed here). As you can see, the number of bug fixes is 14280 quite large, so it is strongly recommended that users of earlier GCC 14281 3.x releases upgrade to GCC 3.2.1. 14282 14283 Internal Compiler Errors (multi-platform) 14284 14285 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 14286 * [130]5661: (c++) ICE instantiating template on array of unknown 14287 size (bad code) 14288 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 14289 64-bit platforms 14290 * [132]6994: (c++) ICE in find_function_data 14291 * [133]7150: preprocessor: GCC -dM -E gives an ICE 14292 * [134]7160: ICE when optimizing branches without a return value 14293 * [135]7228: (c++) ICE when using member template and template 14294 function 14295 * [136]7266: (c++) ICE with -pedantic on missing typename 14296 * [137]7353: ICE from use of "Naming Types" extension, see above 14297 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 14298 * [139]7478: (c++) ICE on static_cast inside template 14299 * [140]7526: preprocessor core dump when _Pragma implies #pragma 14300 dependency 14301 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 14302 is a duplicate) 14303 * [143]7754: (c++) ICE on union with template parameter 14304 * [144]7788: (c++) redeclaring a definition as an incomplete class 14305 causes ICE 14306 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 14307 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 14308 kernel 14309 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 14310 variables 14311 * [148]8134: (c++) ICE in force_store_init_value on legal code 14312 * [149]8149: (c++) ICE on incomplete type 14313 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 14314 initialization 14315 14316 C++ (compiler and library) bugs 14317 14318 * [151]5607: No pointer adjustment in covariant return types 14319 * [152]6579: Infinite loop with statement expressions in member 14320 initialization 14321 * [153]6803: Default copy constructor bug in GCC 3.1 14322 * [154]7176: g++ confused by friend and static member with same name 14323 * [155]7188: Segfault with template class and recursive (incorrect) 14324 initializer list 14325 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 14326 inheritance if a method has a variable number of arguments 14327 * [157]7461: ctype<char>::classic_table() returns offset array on 14328 Cygwin 14329 * [158]7524: f(const float arg[3]) fails 14330 * [159]7584: Erroneous ambiguous base error on using declaration 14331 * [160]7676: Member template overloading problem 14332 * [161]7679: infinite loop when a right parenthesis is missing 14333 * [162]7811: default locale not taken from environment 14334 * [163]7961: compare( char *) implemented incorrectly in 14335 basic_string<> 14336 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 14337 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 14338 [166]6745) 14339 * [167]8096: deque::at() throws std::range_error instead of 14340 std::out_of_range 14341 * [168]8127: cout << cin.rdbuf() infinite loop 14342 * [169]8218: Excessively large memory consumed for classes with large 14343 array members 14344 * [170]8287: GCC 3.2: Destructor called for non-constructed local 14345 object 14346 * [171]8347: empty vector range used in string construction causes 14347 core dump 14348 * [172]8348: fail() flag is set in istringstream when eof() flag is 14349 set 14350 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 14351 14352 C and optimizer bugs 14353 14354 * [174]6627: -fno-align-functions doesn't seem to disable function 14355 alignment 14356 * [175]6631: life_analysis misoptimizes code to initialize fields of 14357 a structure 14358 * [176]7102: unsigned char division results in floating exception 14359 * [177]7120: Run once loop should *always* be unrolled 14360 (pessimization) 14361 * [178]7209: Bug involving array referencing and ?: operator 14362 * [179]7515: invalid inlining of global function with -O3 14363 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 14364 * [181]8467: bug in sibling call optimization 14365 14366 Preprocessor bugs 14367 14368 * [182]4890: incorrect line markers from the traditional preprocessor 14369 * [183]7357: -M option omits system headers files (making it the same 14370 as -MM) 14371 * [184]7358: Changes to Sun's make Dependencies 14372 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 14373 C headers 14374 * [186]7862: Interrupting GCC -MD removes .d file but not .o 14375 * [187]8190: Failed compilation deletes -MD dependency file 14376 * [188]8524: _Pragma within macro is improperly expanded 14377 14378 x86 specific (Intel/AMD) 14379 14380 * [189]5351: (i686-only) function pass-by-value structure copy 14381 corrupts stack ([190]7591 is a duplicate) 14382 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 14383 -march=pentium3/pentium2/athlon (these are all the same underlying 14384 bug, in MMX register use) 14385 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 14386 as above?) 14387 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 14388 * [199]6981: wrong code in 64-bit manipulation on x86 14389 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 14390 macro 14391 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 14392 intrinsics are broken 14393 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 14394 -march=pentium4 14395 * [203]7693: Typo in i386 mmintrin.h header 14396 * [204]7723: ICE - Pentium3 sse - GCC 3.2 14397 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 14398 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 14399 14400 PowerPC specific 14401 14402 * [207]5967: GCC bug when profiling nested functions on powerpc 14403 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 14404 loop on PowerPC 14405 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 14406 * [210]7130: miscompiled code for GCC-3.1 on 14407 powerpc-unknown-linux-gnu with -funroll-all-loops 14408 * [211]7133: PowerPC ICE: unrecognizable insn 14409 * [212]7380: ICE in extract_insn, at recog.c:2148 14410 * [213]8252: ICE on Altivec code with optimization turned on 14411 * [214]8451: Altivec ICE in GCC 3.2 14412 14413 HP/PA specific 14414 14415 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 14416 14417 SPARC specific 14418 14419 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 14420 in the wrong place on sparc-solaris 14421 * [217]7151: ICE when compiling for UltraSPARC 14422 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 14423 double and -O1 14424 * [219]7842: [REGRESSION] SPARC code gen bug 14425 14426 ARM specific 14427 14428 * [220]7856: [arm] invalid offset in constant pool reference 14429 * [221]7967: optimization produces wrong code (ARM) 14430 14431 Alpha specific 14432 14433 * [222]7374: __builtin_fabsl broken on alpha 14434 14435 IBM s390 specific 14436 14437 * [223]7370: ICE in fixup_var_refs_1 on s390x 14438 * [224]7409: loop optimization bug on s390x-linux-gnu 14439 * [225]8232: s390x: ICE when using bcmp with int length argument 14440 14441 SCO specific 14442 14443 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 14444 symbol: BITS_PER_UNIT 14445 14446 m68k/Coldfire specific 14447 14448 * [227]8314: crtbegin, crtend need to be multilib'ed for this 14449 platform 14450 14451 Documentation 14452 14453 * [228]761: Document some undocumented options 14454 * [229]5610: Fix documentation about invoking SSE instructions 14455 (-mfpmath=sse) 14456 * [230]7484: List -Wmissing-declarations as C-only option 14457 * [231]7531: -mcmodel not documented for x86-64 14458 * [232]8120: Update documentation of bad use of ## 14459 __________________________________________________________________ 14460 14461 GCC 3.2 14462 14463 3.2 is a small bug fix release, but there is a change to the 14464 application binary interface (ABI), hence the change to the second part 14465 of the version number. 14466 14467 The main purpose of the 3.2 release is to correct a couple of problems 14468 in the C++ ABI, with the intention of providing a stable interface 14469 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 14470 14471 Bug Fixes 14472 14473 C++ 14474 14475 * [233]7320: g++ 3.2 relocation problem 14476 * [234]7470: vtable: virtual function pointers not in declaration 14477 order 14478 14479 libstdc++ 14480 14481 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 14482 * [236]6503, [237]6642, [238]7186: Problems with comparing or 14483 subtracting various types of const and non-const iterators 14484 * [239]7216: ambiguity with basic_iostream::traits_type 14485 * [240]7220: problem with basic_istream::ignore(0,delimiter) 14486 * [241]7222: locale::operator==() doesn't work on std::locale("") 14487 * [242]7286: placement operator delete issue 14488 * [243]7442: cxxabi.h does not match the C++ ABI 14489 * [244]7445: poor performance of std::locale::classic() in 14490 multi-threaded applications 14491 14492 x86-64 specific 14493 14494 * [245]7291: off-by-one in generated inline bzero code for x86-64 14495 14496 14497 For questions related to the use of GCC, please consult these web 14498 pages and the [246]GCC manuals. If that fails, the 14499 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14500 web pages and the development of GCC are welcome on our developer 14501 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 14502 archives. 14503 14504 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 14505 distribution of this entire article is permitted in any medium, 14506 provided this notice is preserved. 14507 14508 These pages are [251]maintained by the GCC team. Last modified 14509 2016-01-30[252]. 14510 14511 References 14512 14513 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 14514 2. http://gcc.gnu.org/gcc-3.1/changes.html 14515 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 14516 4. https://gcc.gnu.org/PR3782 14517 5. https://gcc.gnu.org/PR6440 14518 6. https://gcc.gnu.org/PR7050 14519 7. https://gcc.gnu.org/PR7741 14520 8. https://gcc.gnu.org/PR7982 14521 9. https://gcc.gnu.org/PR8068 14522 10. https://gcc.gnu.org/PR8178 14523 11. https://gcc.gnu.org/PR8396 14524 12. https://gcc.gnu.org/PR8674 14525 13. https://gcc.gnu.org/PR9768 14526 14. https://gcc.gnu.org/PR9798 14527 15. https://gcc.gnu.org/PR9799 14528 16. https://gcc.gnu.org/PR9928 14529 17. https://gcc.gnu.org/PR10114 14530 18. https://gcc.gnu.org/PR10352 14531 19. https://gcc.gnu.org/PR10336 14532 20. https://gcc.gnu.org/PR8224 14533 21. https://gcc.gnu.org/PR8613 14534 22. https://gcc.gnu.org/PR8828 14535 23. https://gcc.gnu.org/PR9226 14536 24. https://gcc.gnu.org/PR9853 14537 25. https://gcc.gnu.org/PR9797 14538 26. https://gcc.gnu.org/PR9967 14539 27. https://gcc.gnu.org/PR10116 14540 28. https://gcc.gnu.org/PR10171 14541 29. https://gcc.gnu.org/PR10175 14542 30. https://gcc.gnu.org/PR8316 14543 31. https://gcc.gnu.org/PR9169 14544 32. https://gcc.gnu.org/PR9420 14545 33. https://gcc.gnu.org/PR9459 14546 34. https://gcc.gnu.org/PR9507 14547 35. https://gcc.gnu.org/PR9538 14548 36. https://gcc.gnu.org/PR9602 14549 37. https://gcc.gnu.org/PR9993 14550 38. https://gcc.gnu.org/PR10167 14551 39. https://gcc.gnu.org/PR9652 14552 40. https://gcc.gnu.org/PR10144 14553 41. https://gcc.gnu.org/PR8746 14554 42. https://gcc.gnu.org/PR9888 14555 43. https://gcc.gnu.org/PR9638 14556 44. https://gcc.gnu.org/PR9954 14557 45. https://gcc.gnu.org/PR7784 14558 46. https://gcc.gnu.org/PR7796 14559 47. https://gcc.gnu.org/PR8281 14560 48. https://gcc.gnu.org/PR8366 14561 49. https://gcc.gnu.org/PR8726 14562 50. https://gcc.gnu.org/PR9414 14563 51. https://gcc.gnu.org/PR10067 14564 52. https://gcc.gnu.org/PR7248 14565 53. https://gcc.gnu.org/PR8343 14566 54. https://gcc.gnu.org/PR9732 14567 55. https://gcc.gnu.org/PR10073 14568 56. https://gcc.gnu.org/PR7702 14569 57. https://gcc.gnu.org/PR9671 14570 58. https://gcc.gnu.org/PR8694 14571 59. https://gcc.gnu.org/PR9953 14572 60. https://gcc.gnu.org/PR10271 14573 61. https://gcc.gnu.org/PR6362 14574 62. https://gcc.gnu.org/PR10377 14575 63. https://gcc.gnu.org/PR6955 14576 64. https://gcc.gnu.org/PR5919 14577 65. https://gcc.gnu.org/PR7129 14578 66. https://gcc.gnu.org/PR7507 14579 67. https://gcc.gnu.org/PR7622 14580 68. https://gcc.gnu.org/PR7681 14581 69. https://gcc.gnu.org/PR9528 14582 70. https://gcc.gnu.org/PR8031 14583 71. https://gcc.gnu.org/PR8275 14584 72. https://gcc.gnu.org/PR8332 14585 73. https://gcc.gnu.org/PR8372 14586 74. https://gcc.gnu.org/PR8439 14587 75. https://gcc.gnu.org/PR8442 14588 76. https://gcc.gnu.org/PR8518 14589 77. https://gcc.gnu.org/PR8615 14590 78. https://gcc.gnu.org/PR8663 14591 79. https://gcc.gnu.org/PR8799 14592 80. https://gcc.gnu.org/PR9328 14593 81. https://gcc.gnu.org/PR9465 14594 82. https://gcc.gnu.org/PR47 14595 83. https://gcc.gnu.org/PR6745 14596 84. https://gcc.gnu.org/PR8214 14597 85. https://gcc.gnu.org/PR8493 14598 86. https://gcc.gnu.org/PR8332 14599 87. https://gcc.gnu.org/PR8503 14600 88. https://gcc.gnu.org/PR8727 14601 89. https://gcc.gnu.org/PR7445 14602 90. https://gcc.gnu.org/PR8230 14603 91. https://gcc.gnu.org/PR8399 14604 92. https://gcc.gnu.org/PR8662 14605 93. https://gcc.gnu.org/PR8707 14606 94. https://gcc.gnu.org/PR8708 14607 95. https://gcc.gnu.org/PR8790 14608 96. https://gcc.gnu.org/PR8887 14609 97. https://gcc.gnu.org/PR9076 14610 98. https://gcc.gnu.org/PR9151 14611 99. https://gcc.gnu.org/PR9168 14612 100. https://gcc.gnu.org/PR9269 14613 101. https://gcc.gnu.org/PR9322 14614 102. https://gcc.gnu.org/PR9433 14615 103. https://gcc.gnu.org/PR8032 14616 104. https://gcc.gnu.org/PR8639 14617 105. https://gcc.gnu.org/PR8794 14618 106. https://gcc.gnu.org/PR8832 14619 107. https://gcc.gnu.org/PR8988 14620 108. https://gcc.gnu.org/PR9492 14621 109. https://gcc.gnu.org/PR9267 14622 110. https://gcc.gnu.org/PR8344 14623 111. https://gcc.gnu.org/PR8524 14624 112. https://gcc.gnu.org/PR8880 14625 113. https://gcc.gnu.org/PR9090 14626 114. https://gcc.gnu.org/PR8588 14627 115. https://gcc.gnu.org/PR8599 14628 116. https://gcc.gnu.org/PR9506 14629 117. https://gcc.gnu.org/PR9484 14630 118. https://gcc.gnu.org/PR9292 14631 119. https://gcc.gnu.org/PR9293 14632 120. https://gcc.gnu.org/PR9295 14633 121. https://gcc.gnu.org/PR9296 14634 122. https://gcc.gnu.org/PR9316 14635 123. https://gcc.gnu.org/PR9493 14636 124. https://gcc.gnu.org/PR7341 14637 125. https://gcc.gnu.org/PR8947 14638 126. https://gcc.gnu.org/PR7448 14639 127. https://gcc.gnu.org/PR8882 14640 128. https://gcc.gnu.org/PR7445 14641 129. https://gcc.gnu.org/PR2521 14642 130. https://gcc.gnu.org/PR5661 14643 131. https://gcc.gnu.org/PR6419 14644 132. https://gcc.gnu.org/PR6994 14645 133. https://gcc.gnu.org/PR7150 14646 134. https://gcc.gnu.org/PR7160 14647 135. https://gcc.gnu.org/PR7228 14648 136. https://gcc.gnu.org/PR7266 14649 137. https://gcc.gnu.org/PR7353 14650 138. https://gcc.gnu.org/PR7411 14651 139. https://gcc.gnu.org/PR7478 14652 140. https://gcc.gnu.org/PR7526 14653 141. https://gcc.gnu.org/PR7721 14654 142. https://gcc.gnu.org/PR7803 14655 143. https://gcc.gnu.org/PR7754 14656 144. https://gcc.gnu.org/PR7788 14657 145. https://gcc.gnu.org/PR8031 14658 146. https://gcc.gnu.org/PR8055 14659 147. https://gcc.gnu.org/PR8067 14660 148. https://gcc.gnu.org/PR8134 14661 149. https://gcc.gnu.org/PR8149 14662 150. https://gcc.gnu.org/PR8160 14663 151. https://gcc.gnu.org/PR5607 14664 152. https://gcc.gnu.org/PR6579 14665 153. https://gcc.gnu.org/PR6803 14666 154. https://gcc.gnu.org/PR7176 14667 155. https://gcc.gnu.org/PR7188 14668 156. https://gcc.gnu.org/PR7306 14669 157. https://gcc.gnu.org/PR7461 14670 158. https://gcc.gnu.org/PR7524 14671 159. https://gcc.gnu.org/PR7584 14672 160. https://gcc.gnu.org/PR7676 14673 161. https://gcc.gnu.org/PR7679 14674 162. https://gcc.gnu.org/PR7811 14675 163. https://gcc.gnu.org/PR7961 14676 164. https://gcc.gnu.org/PR8071 14677 165. https://gcc.gnu.org/PR8127 14678 166. https://gcc.gnu.org/PR6745 14679 167. https://gcc.gnu.org/PR8096 14680 168. https://gcc.gnu.org/PR8127 14681 169. https://gcc.gnu.org/PR8218 14682 170. https://gcc.gnu.org/PR8287 14683 171. https://gcc.gnu.org/PR8347 14684 172. https://gcc.gnu.org/PR8348 14685 173. https://gcc.gnu.org/PR8391 14686 174. https://gcc.gnu.org/PR6627 14687 175. https://gcc.gnu.org/PR6631 14688 176. https://gcc.gnu.org/PR7102 14689 177. https://gcc.gnu.org/PR7120 14690 178. https://gcc.gnu.org/PR7209 14691 179. https://gcc.gnu.org/PR7515 14692 180. https://gcc.gnu.org/PR7814 14693 181. https://gcc.gnu.org/PR8467 14694 182. https://gcc.gnu.org/PR4890 14695 183. https://gcc.gnu.org/PR7357 14696 184. https://gcc.gnu.org/PR7358 14697 185. https://gcc.gnu.org/PR7602 14698 186. https://gcc.gnu.org/PR7862 14699 187. https://gcc.gnu.org/PR8190 14700 188. https://gcc.gnu.org/PR8524 14701 189. https://gcc.gnu.org/PR5351 14702 190. https://gcc.gnu.org/PR7591 14703 191. https://gcc.gnu.org/PR6845 14704 192. https://gcc.gnu.org/PR7034 14705 193. https://gcc.gnu.org/PR7124 14706 194. https://gcc.gnu.org/PR7174 14707 195. https://gcc.gnu.org/PR7134 14708 196. https://gcc.gnu.org/PR7375 14709 197. https://gcc.gnu.org/PR7390 14710 198. https://gcc.gnu.org/PR6890 14711 199. https://gcc.gnu.org/PR6981 14712 200. https://gcc.gnu.org/PR7242 14713 201. https://gcc.gnu.org/PR7396 14714 202. https://gcc.gnu.org/PR7630 14715 203. https://gcc.gnu.org/PR7693 14716 204. https://gcc.gnu.org/PR7723 14717 205. https://gcc.gnu.org/PR7951 14718 206. https://gcc.gnu.org/PR8146 14719 207. https://gcc.gnu.org/PR5967 14720 208. https://gcc.gnu.org/PR6984 14721 209. https://gcc.gnu.org/PR7114 14722 210. https://gcc.gnu.org/PR7130 14723 211. https://gcc.gnu.org/PR7133 14724 212. https://gcc.gnu.org/PR7380 14725 213. https://gcc.gnu.org/PR8252 14726 214. https://gcc.gnu.org/PR8451 14727 215. https://gcc.gnu.org/PR7250 14728 216. https://gcc.gnu.org/PR6668 14729 217. https://gcc.gnu.org/PR7151 14730 218. https://gcc.gnu.org/PR7335 14731 219. https://gcc.gnu.org/PR7842 14732 220. https://gcc.gnu.org/PR7856 14733 221. https://gcc.gnu.org/PR7967 14734 222. https://gcc.gnu.org/PR7374 14735 223. https://gcc.gnu.org/PR7370 14736 224. https://gcc.gnu.org/PR7409 14737 225. https://gcc.gnu.org/PR8232 14738 226. https://gcc.gnu.org/PR7623 14739 227. https://gcc.gnu.org/PR8314 14740 228. https://gcc.gnu.org/PR761 14741 229. https://gcc.gnu.org/PR5610 14742 230. https://gcc.gnu.org/PR7484 14743 231. https://gcc.gnu.org/PR7531 14744 232. https://gcc.gnu.org/PR8120 14745 233. https://gcc.gnu.org/PR7320 14746 234. https://gcc.gnu.org/PR7470 14747 235. https://gcc.gnu.org/PR6410 14748 236. https://gcc.gnu.org/PR6503 14749 237. https://gcc.gnu.org/PR6642 14750 238. https://gcc.gnu.org/PR7186 14751 239. https://gcc.gnu.org/PR7216 14752 240. https://gcc.gnu.org/PR7220 14753 241. https://gcc.gnu.org/PR7222 14754 242. https://gcc.gnu.org/PR7286 14755 243. https://gcc.gnu.org/PR7442 14756 244. https://gcc.gnu.org/PR7445 14757 245. https://gcc.gnu.org/PR7291 14758 246. https://gcc.gnu.org/onlinedocs/ 14759 247. mailto:gcc-help (a] gcc.gnu.org 14760 248. mailto:gcc (a] gcc.gnu.org 14761 249. https://gcc.gnu.org/lists.html 14762 250. http://www.fsf.org/ 14763 251. https://gcc.gnu.org/about.html 14764 252. http://validator.w3.org/check/referer 14765 ====================================================================== 14766 http://gcc.gnu.org/gcc-3.1/index.html 14767 14768 GCC 3.1 14769 14770 July 27, 2002 14771 14772 The [1]GNU project and the GCC developers are pleased to announce the 14773 release of GCC 3.1.1. 14774 14775 The links below still apply to GCC 3.1.1. 14776 14777 This release series is no longer maintained. 14778 14779 May 15, 2002 14780 14781 The [2]GNU project and the GCC developers are pleased to announce the 14782 release of GCC 3.1. 14783 14784 GCC used to stand for the GNU C Compiler, but since the compiler 14785 supports several other languages aside from C, it now stands for the 14786 GNU Compiler Collection. 14787 14788 A list of [3]successful builds is updated as new information becomes 14789 available. 14790 14791 The GCC developers would like to thank the numerous people that have 14792 contributed [4]new features, improvements, bug fixes, and other changes 14793 as well as test results to GCC. This [5]amazing group of volunteers is 14794 what makes GCC successful. 14795 14796 For additional information about GCC please refer to the [6]GCC project 14797 web site or contact the [7]GCC development mailing list. 14798 14799 To obtain GCC please use [8]our mirror sites, or our CVS server. 14800 __________________________________________________________________ 14801 14802 14803 For questions related to the use of GCC, please consult these web 14804 pages and the [9]GCC manuals. If that fails, the 14805 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14806 web pages and the development of GCC are welcome on our developer 14807 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 14808 archives. 14809 14810 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 14811 distribution of this entire article is permitted in any medium, 14812 provided this notice is preserved. 14813 14814 These pages are [14]maintained by the GCC team. Last modified 14815 2016-09-30[15]. 14816 14817 References 14818 14819 1. http://www.gnu.org/ 14820 2. http://www.gnu.org/ 14821 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 14822 4. http://gcc.gnu.org/gcc-3.1/changes.html 14823 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14824 6. http://gcc.gnu.org/index.html 14825 7. mailto:gcc (a] gcc.gnu.org 14826 8. http://gcc.gnu.org/mirrors.html 14827 9. https://gcc.gnu.org/onlinedocs/ 14828 10. mailto:gcc-help (a] gcc.gnu.org 14829 11. mailto:gcc (a] gcc.gnu.org 14830 12. https://gcc.gnu.org/lists.html 14831 13. http://www.fsf.org/ 14832 14. https://gcc.gnu.org/about.html 14833 15. http://validator.w3.org/check/referer 14834 ====================================================================== 14835 http://gcc.gnu.org/gcc-3.1/changes.html 14836 14837 GCC 3.1 Release Series 14838 Changes, New Features, and Fixes 14839 14840 Additional changes in GCC 3.1.1 14841 14842 * A bug related to how structures and unions are returned has been 14843 fixed for powerpc-*-netbsd*. 14844 * An important bug in the implementation of -fprefetch-loop-arrays 14845 has been fixed. Previously the optimization prefetched random 14846 blocks of memory for most targets except for i386. 14847 * The Java compiler now compiles Java programs much faster and also 14848 works with parallel make. 14849 * Nested functions have been fixed for mips*-*-netbsd*. 14850 * Some missing floating point support routines have beed added for 14851 mips*-*-netbsd*. 14852 * This [1]message gives additional information about the bugs fixed 14853 in this release. 14854 14855 Caveats 14856 14857 * The -traditional C compiler option has been deprecated and will be 14858 removed in GCC 3.3. (It remains possible to preprocess non-C code 14859 with the traditional preprocessor.) 14860 * The default debugging format for most ELF platforms (including 14861 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 14862 from stabs to DWARF2. This requires GDB 5.1.1 or later. 14863 14864 General Optimizer Improvements 14865 14866 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 14867 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 14868 for profile driven optimizations. 14869 Options -fprofile-arcs and -fbranch-probabilities can now be used 14870 to improve speed of the generated code by profiling the actual 14871 program behaviour on typical runs. In the absence of profile info 14872 the compiler attempts to guess the profile statically. 14873 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 14874 monitor performance of the generated code. 14875 According to the SPECInt2000 results on an AMD Athlon CPU, the code 14876 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 14877 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 14878 is about 2.1% faster compared to 2.95.3. Tests were done using the 14879 -O2 -march=athlon command-line options. 14880 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 14881 infrastructure developed by CodeSourcery, LLC for the C++ front 14882 end, so that it is now used in the C front end too. Inlining 14883 functions as trees exposes them earlier to the compiler, giving it 14884 more opportunities for optimization. 14885 * Support for data prefetching instructions has been added to the GCC 14886 back end and several targets. A new __builtin_prefetch intrinsic is 14887 available to explicitly insert prefetch instructions and 14888 experimental support for loop array prefetching has been added (see 14889 -fprefetch-loop-array documentation). 14890 * Support for emitting debugging information for macros has been 14891 added for DWARF2. It is activated using -g3. 14892 14893 New Languages and Language specific improvements 14894 14895 C/C++ 14896 14897 * A few more [4]ISO C99 features. 14898 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 14899 * The preprocessor's symbol table has been merged with the symbol 14900 table of the C, C++ and Objective-C front ends. 14901 * The preprocessor consumes less memory than the preprocessor in GCC 14902 3.0, often significantly so. On normal input files, it typically 14903 consumes less memory than pre-3.0 cccp-based GCC, too. 14904 14905 C++ 14906 14907 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 14908 was a workaround to allow std compliant code to work with the 14909 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 14910 * The C++ ABI has been fixed so that void (A::*)() const is mangled 14911 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 14912 affects pointer to cv-qualified member function types. 14913 * The C++ ABI has been changed to correctly handle this code: 14914 struct A { 14915 void operator delete[] (void *, size_t); 14916 }; 14917 14918 struct B : public A { 14919 }; 14920 14921 new B[10]; 14922 14923 The amount of storage allocated for the array will be greater than 14924 it was in 3.0, in order to store the number of elements in the 14925 array, so that the correct size can be passed to operator delete[] 14926 when the array is deleted. Previously, the value passed to operator 14927 delete[] was unpredictable. 14928 This change will only affect code that declares a two-argument 14929 operator delete[] with a second parameter of type size_t in a base 14930 class, and does not override that definition in a derived class. 14931 * The C++ ABI has been changed so that: 14932 struct A { 14933 void operator delete[] (void *, size_t); 14934 void operator delete[] (void *); 14935 }; 14936 14937 does not cause unnecessary storage to be allocated when an array of 14938 A objects is allocated. 14939 This change will only affect code that declares both of these forms 14940 of operator delete[], and declared the two-argument form before the 14941 one-argument form. 14942 * The C++ ABI has been changed so that when a parameter is passed by 14943 value, any cleanup for that parameter is performed in the caller, 14944 as specified by the ia64 C++ ABI, rather than the called function 14945 as before. As a result, classes with a non-trivial destructor but a 14946 trivial copy constructor will be passed and returned by invisible 14947 reference, rather than by bitwise copy as before. 14948 * G++ now supports the "named return value optimization": for code 14949 like 14950 A f () { 14951 A a; 14952 ... 14953 return a; 14954 } 14955 14956 G++ will allocate a in the return value slot, so that the return 14957 becomes a no-op. For this to work, all return statements in the 14958 function must return the same variable. 14959 * Improvements to the C++ library are listed in [5]the libstdc++-v3 14960 FAQ. 14961 14962 Objective-C 14963 14964 * Annoying linker warnings (due to incorrect code being generated) 14965 have been fixed. 14966 * If a class method cannot be found, the compiler no longer issues a 14967 warning if a corresponding instance method exists in the root 14968 class. 14969 * Forward @protocol declarations have been fixed. 14970 * Loading of categories has been fixed in certain situations (GNU run 14971 time only). 14972 * The class lookup in the run-time library has been rewritten so that 14973 class method dispatch is more than twice as fast as it used to be 14974 (GNU run time only). 14975 14976 Java 14977 14978 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 14979 javax.transaction. 14980 * Property files and other system resources can be compiled into 14981 executables which use libgcj using the new gcj --resource feature. 14982 * libgcj has been ported to more platforms. In particular there is 14983 now a mostly-functional mingw32 (Windows) target port. 14984 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 14985 Java code can now be called from a C/C++ application. 14986 * gcj can now use builtin functions for certain known methods, for 14987 instance Math.cos. 14988 * gcj can now automatically remove redundant array-store checks in 14989 some common cases. 14990 * The --no-store-checks optimization option was added. This can be 14991 used to omit runtime store checks for code which is known not to 14992 throw ArrayStoreException 14993 * The following third party interface standards were added to libgcj: 14994 org.w3c.dom and org.xml.sax. 14995 * java.security has been merged with GNU Classpath. The new package 14996 is now JDK 1.2 compliant, and much more complete. 14997 * A bytecode verifier was added to the libgcj interpreter. 14998 * java.lang.Character was rewritten to comply with the Unicode 3.0 14999 standard, and improve performance. 15000 * Partial support for many more locales was added to libgcj. 15001 * Socket timeouts have been implemented. 15002 * libgcj has been merged into a single shared library. There are no 15003 longer separate shared libraries for the garbage collector and 15004 zlib. 15005 * Several performance improvements were made to gcj and libgcj: 15006 + Hash synchronization (thin locks) 15007 + A special allocation path for finalizer-free objects 15008 + Thread-local allocation 15009 + Parallel GC, and other GC tweaks 15010 15011 Fortran 15012 15013 Fortran improvements are listed in [6]the Fortran documentation. 15014 15015 Ada 15016 15017 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front 15018 end and associated tools. The GNAT compiler fully implements the Ada 15019 language as defined by the ISO/IEC 8652 standard. 15020 15021 Please note that the integration of the Ada front end is still work in 15022 progress. 15023 15024 New Targets and Target Specific Improvements 15025 15026 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU 15027 architecture used in new editions of Donald E. Knuth's The Art of 15028 Computer Programming. 15029 * Axis Communications has contributed its port to the CRIS CPU 15030 architecture, used in the ETRAX system-on-a-chip series. 15031 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 15032 SH5 64-bit RISC microprocessor architecture, extending the existing 15033 SH port. 15034 * UltraSPARC is fully supported in 64-bit mode. The option -m64 15035 enables it. 15036 * For compatibility with the Sun compiler #pragma redefine_extname 15037 has been implemented on Solaris. 15038 * The x86 back end has had some noticeable work done to it. 15039 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 15040 Jaeger have contributed a port to the AMD x86-64 architecture. 15041 For more information on x86-64 see http://www.x86-64.org. 15042 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 15043 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 15044 enable the respective instruction sets. Intel C++ compatible 15045 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 15046 will be added in next major release. 15047 + Following those improvements, targets for Pentium MMX, K6-2, 15048 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 15049 added. Refer to the documentation on -march= and -mcpu= 15050 options for details. 15051 + For those targets that support it, -mfpmath=sse will cause the 15052 compiler to generate SSE/SSE2 instructions for floating point 15053 math instead of x87 instructions. Usually, this will lead to 15054 quicker code -- especially on the Pentium 4. Note that only 15055 scalar floating point instructions are used and GCC does not 15056 exploit SIMD features yet. 15057 + Prefetch support has been added to the Pentium III, Pentium 4, 15058 K6-2, K6-3, and Athlon series. 15059 + Code generated for floating point to integer conversions has 15060 been improved leading to better performance of many 3D 15061 applications. 15062 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 15063 * C++ support for AIX has been improved. 15064 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 15065 PowerPC port supporting the AltiVec programming model (SIMD). The 15066 support, though presently useful, is experimental and is expected 15067 to stabilize for 3.2. The support is written to conform to 15068 Motorola's AltiVec specs. See -maltivec. 15069 15070 Obsolete Systems 15071 15072 Support for a number of older systems has been declared obsolete in GCC 15073 3.1. Unless there is activity to revive them, the next release of GCC 15074 will have their sources permanently removed. 15075 15076 All configurations of the following processor architectures have been 15077 declared obsolete: 15078 * MIL-STD-1750A, 1750a-*-* 15079 * AMD A29k, a29k-*-* 15080 * Convex, c*-convex-* 15081 * Clipper, clipper-*-* 15082 * Elxsi, elxsi-*-* 15083 * Intel i860, i860-*-* 15084 * Sun picoJava, pj-*-* and pjl-*-* 15085 * Western Electric 32000, we32k-*-* 15086 15087 Most configurations of the following processor architectures have been 15088 declared obsolete, but we are preserving a few systems which may have 15089 active developers. It is unlikely that the remaining systems will 15090 survive much longer unless we see definite signs of port activity. 15091 * Motorola 88000 except 15092 + Generic a.out, m88k-*-aout* 15093 + Generic SVR4, m88k-*-sysv4 15094 + OpenBSD, m88k-*-openbsd* 15095 * NS32k except 15096 + NetBSD, ns32k-*-netbsd* 15097 + OpenBSD, ns32k-*-openbsd*. 15098 * ROMP except 15099 + OpenBSD, romp-*-openbsd*. 15100 15101 Finally, only some configurations of these processor architectures are 15102 being obsoleted. 15103 * Alpha: 15104 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 15105 alpha*-*-osf[45], are still supported.) 15106 * ARM: 15107 + RISCiX, arm-*-riscix*. 15108 * i386: 15109 + 386BSD, i?86-*-bsd* 15110 + Chorus, i?86-*-chorusos* 15111 + DG/UX, i?86-*-dgux* 15112 + FreeBSD 1.x, i?86-*-freebsd1.* 15113 + IBM AIX, i?86-*-aix* 15114 + ISC UNIX, i?86-*-isc* 15115 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 15116 + NEXTstep, i?86-next-* 15117 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 15118 + RTEMS/coff, i?86-*-rtemscoff* 15119 + RTEMS/go32, i?86-go32-rtems* 15120 + Sequent/BSD, i?86-sequent-bsd* 15121 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 15122 i?86-sequent-sysv3* 15123 + SunOS, i?86-*-sunos* 15124 * Motorola 68000: 15125 + Altos, m68[k0]*-altos-* 15126 + Apollo, m68[k0]*-apollo-* 15127 + Apple A/UX, m68[k0]*-apple-* 15128 + Bull, m68[k0]*-bull-* 15129 + Convergent, m68[k0]*-convergent-* 15130 + Generic SVR3, m68[k0]*-*-sysv3* 15131 + ISI, m68[k0]*-isi-* 15132 + LynxOS, m68[k0]*-*-lynxos* 15133 + NEXT, m68[k0]*-next-* 15134 + RTEMS/coff, m68[k0]*-*-rtemscoff* 15135 + Sony, m68[k0]*-sony-* 15136 * MIPS: 15137 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 15138 + Generic BSD, mips-*-bsd* 15139 + Generic System V, mips-*-sysv* 15140 + IRIX before version 5, mips-sgi-irix[1234]* 15141 + RiscOS, mips-*-riscos* 15142 + Sony, mips-sony-* 15143 + Tandem, mips-tandem-* 15144 * SPARC: 15145 + RTEMS/a.out, sparc-*-rtemsaout*. 15146 15147 Documentation improvements 15148 15149 * The old manual ("Using and Porting the GNU Compiler Collection") 15150 has been replaced by a users manual ("Using the GNU Compiler 15151 Collection") and a separate internals reference manual ("GNU 15152 Compiler Collection Internals"). 15153 * More complete and much improved documentation about GCC's internal 15154 representation used by the C and C++ front ends. 15155 * Many cleanups and improvements in general. 15156 15157 15158 For questions related to the use of GCC, please consult these web 15159 pages and the [9]GCC manuals. If that fails, the 15160 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15161 web pages and the development of GCC are welcome on our developer 15162 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 15163 archives. 15164 15165 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 15166 distribution of this entire article is permitted in any medium, 15167 provided this notice is preserved. 15168 15169 These pages are [14]maintained by the GCC team. Last modified 15170 2017-01-29[15]. 15171 15172 References 15173 15174 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 15175 2. http://gcc.gnu.org/news/profiledriven.html 15176 3. http://gcc.gnu.org/benchmarks/ 15177 4. http://gcc.gnu.org/c99status.html 15178 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 15179 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 15180 7. http://www.adacore.com/ 15181 8. http://www-cs-faculty.stanford.edu/~uno/mmix.html 15182 9. https://gcc.gnu.org/onlinedocs/ 15183 10. mailto:gcc-help (a] gcc.gnu.org 15184 11. mailto:gcc (a] gcc.gnu.org 15185 12. https://gcc.gnu.org/lists.html 15186 13. http://www.fsf.org/ 15187 14. https://gcc.gnu.org/about.html 15188 15. http://validator.w3.org/check/referer 15189 ====================================================================== 15190 http://gcc.gnu.org/gcc-3.0/index.html 15191 15192 GCC 3.0.4 15193 15194 February 20, 2002 15195 15196 The [1]GNU project and the GCC developers are pleased to announce the 15197 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 15198 series. 15199 15200 This release series is no longer maintained. 15201 15202 GCC used to stand for the GNU C Compiler, but since the compiler 15203 supports several other languages aside from C, it now stands for the 15204 GNU Compiler Collection. 15205 15206 GCC 3.0.x has several new optimizations, new targets, new languages and 15207 many other new features, relative to GCC 2.95.x. See the [2]new 15208 features page for a more complete list. 15209 15210 A list of [3]successful builds is updated as new information becomes 15211 available. 15212 15213 The GCC developers would like to thank the numerous people that have 15214 contributed new features, test results, bug fixes, etc to GCC. This 15215 [4]amazing group of volunteers is what makes GCC successful. 15216 15217 And finally, we can't in good conscience fail to mention some 15218 [5]caveats to using GCC 3.0.x. 15219 15220 For additional information about GCC please refer to the [6]GCC project 15221 web site or contact the [7]GCC development mailing list. 15222 15223 To obtain GCC please use [8]our mirror sites, or our CVS server. 15224 __________________________________________________________________ 15225 15226 Previous 3.0.x Releases 15227 15228 December 20, 2001: GCC 3.0.3 has been released. 15229 October 25, 2001: GCC 3.0.2 has been released. 15230 August 20, 2001: GCC 3.0.1 has been released. 15231 June 18, 2001: GCC 3.0 has been released. 15232 15233 15234 For questions related to the use of GCC, please consult these web 15235 pages and the [9]GCC manuals. If that fails, the 15236 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15237 web pages and the development of GCC are welcome on our developer 15238 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 15239 archives. 15240 15241 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 15242 distribution of this entire article is permitted in any medium, 15243 provided this notice is preserved. 15244 15245 These pages are [14]maintained by the GCC team. Last modified 15246 2016-09-30[15]. 15247 15248 References 15249 15250 1. http://www.gnu.org/ 15251 2. http://gcc.gnu.org/gcc-3.0/features.html 15252 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 15253 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15254 5. http://gcc.gnu.org/gcc-3.0/caveats.html 15255 6. http://gcc.gnu.org/index.html 15256 7. mailto:gcc (a] gcc.gnu.org 15257 8. http://gcc.gnu.org/mirrors.html 15258 9. https://gcc.gnu.org/onlinedocs/ 15259 10. mailto:gcc-help (a] gcc.gnu.org 15260 11. mailto:gcc (a] gcc.gnu.org 15261 12. https://gcc.gnu.org/lists.html 15262 13. http://www.fsf.org/ 15263 14. https://gcc.gnu.org/about.html 15264 15. http://validator.w3.org/check/referer 15265 ====================================================================== 15266 http://gcc.gnu.org/gcc-3.0/features.html 15267 15268 GCC 3.0 New Features 15269 15270 Additional changes in GCC 3.0.4 15271 15272 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 15273 system, which use the ELF object file format, on x86 processors. 15274 * Correct debugging information is generated from functions that have 15275 lines from multiple files (e.g. yacc output). 15276 * A fix for whitespace handling in the -traditional preprocessor, 15277 which can affect Fortran. 15278 * Fixes to the exception handling runtime. 15279 * More fixes for bad code generation in C++. 15280 * A fix for shared library generation under AIX 4.3. 15281 * Documentation updates. 15282 * Port of GCC to Tensilica's Xtensa processor contributed. 15283 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 15284 15285 Additional changes in GCC 3.0.3 15286 15287 * A fix to correct an accidental change to the PowerPC ABI. 15288 * Fixes for bad code generation on a variety of architectures. 15289 * Improvements to the debugging information generated for C++ 15290 classes. 15291 * Fixes for bad code generation in C++. 15292 * A fix to avoid crashes in the C++ demangler. 15293 * A fix to the C++ standard library to avoid buffer overflows. 15294 * Miscellaneous improvements for a variety of architectures. 15295 15296 Additional changes in GCC 3.0.2 15297 15298 * Fixes for bad code generation during loop unrolling. 15299 * Fixes for bad code generation by the sibling call optimization. 15300 * Minor improvements to x86 code generation. 15301 * Implementation of function descriptors in C++ vtables for IA64. 15302 * Numerous minor bug-fixes. 15303 15304 Additional changes in GCC 3.0.1 15305 15306 * C++ fixes for incorrect code-generation. 15307 * Improved cross-compiling support for the C++ standard library. 15308 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 15309 in GCC 3.0. 15310 * Fixes for various exception-handling bugs. 15311 * A port to the S/390 architecture. 15312 15313 General Optimizer Improvements 15314 15315 * [2]Basic block reordering pass. 15316 * New if-conversion pass with support for conditional (predicated) 15317 execution. 15318 * New tail call and sibling call elimination optimizations. 15319 * New register renaming pass. 15320 * New (experimental) [3]static single assignment (SSA) representation 15321 support. 15322 * New dead-code elimination pass implemented using the SSA 15323 representation. 15324 * [4]Global null pointer test elimination. 15325 * [5]Global code hoisting/unification. 15326 * More builtins and optimizations for stdio.h, string.h and old BSD 15327 functions, as well as for ISO C99 functions. 15328 * New builtin __builtin_expect for giving hints to the branch 15329 predictor. 15330 15331 New Languages and Language specific improvements 15332 15333 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 15334 and supported, including the run-time library containing most 15335 common non-GUI Java classes, a bytecode interpreter, and the Boehm 15336 conservative garbage collector. Many bugs have been fixed. GCJ can 15337 compile Java source or Java bytecodes to either native code or Java 15338 class files, and supports native methods written in either the 15339 standard JNI or the more efficient and convenient CNI. 15340 * Here is a [6]partial list of C++ improvements, both new features 15341 and those no longer supported. 15342 * New C++ ABI. On the IA-64 platform GCC is capable of 15343 inter-operating with other IA-64 compilers. 15344 * The new ABI also significantly reduces the size of symbol and debug 15345 information. 15346 * New C++ support library and many C++ bug fixes, vastly improving 15347 our conformance to the ISO C++ standard. 15348 * New [7]inliner for C++. 15349 * Rewritten C preprocessor, integrated into the C, C++ and Objective 15350 C compilers, with very many improvements including ISO C99 support 15351 and [8]improvements to dependency generation. 15352 * Support for more [9]ISO C99 features. 15353 * Many improvements to support for checking calls to format functions 15354 such as printf and scanf, including support for ISO C99 format 15355 features, extensions from the Single Unix Specification and GNU 15356 libc 2.2, checking of strfmon formats and features to assist in 15357 auditing for format string security bugs. 15358 * New warnings for C code that may have undefined semantics because 15359 of violations of sequence point rules in the C standard (such as a 15360 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 15361 * Additional warning option -Wfloat-equal. 15362 * Improvements to -Wtraditional. 15363 * Fortran improvements are listed in [10]the Fortran documentation. 15364 15365 New Targets and Target Specific Improvements 15366 15367 * New x86 back end, generating much improved code. 15368 * Support for a generic i386-elf target contributed. 15369 * New option to emit x86 assembly code using Intel style syntax 15370 (-mintel-syntax). 15371 * HPUX 11 support contributed. 15372 * Improved PowerPC code generation, including scheduled prologue and 15373 epilogue. 15374 * Port of GCC to Intel's IA-64 processor contributed. 15375 * Port of GCC to Motorola's MCore 210 and 340 contributed. 15376 * New unified back-end for Arm, Thumb and StrongArm contributed. 15377 * Port of GCC to Intel's XScale processor contributed. 15378 * Port of GCC to Atmel's AVR microcontrollers contributed. 15379 * Port of GCC to Mitsubishi's D30V processor contributed. 15380 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 15381 processor family) contributed. 15382 * Port of GCC to Fujitsu's FR30 processor contributed. 15383 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 15384 * Port of GCC to Sun's picoJava processor core contributed. 15385 15386 Documentation improvements 15387 15388 * Substantially rewritten and improved C preprocessor manual. 15389 * Many improvements to other documentation. 15390 * Manpages for gcc, cpp and gcov are now generated automatically from 15391 the master Texinfo manual, eliminating the problem of manpages 15392 being out of date. (The generated manpages are only extracts from 15393 the full manual, which is provided in Texinfo form, from which 15394 info, HTML, other formats and a printed manual can be generated.) 15395 * Generated info files are included in the release tarballs alongside 15396 their Texinfo sources, avoiding problems on some platforms with 15397 building makeinfo as part of the GCC distribution. 15398 15399 Other significant improvements 15400 15401 * Garbage collection used internally by the compiler for most memory 15402 allocation instead of obstacks. 15403 * Lengauer and Tarjan algorithm used for computing dominators in the 15404 CFG. This algorithm can be significantly faster and more space 15405 efficient than our older algorithm. 15406 * gccbug script provided to assist in submitting bug reports to our 15407 bug tracking system. (Bug reports previously submitted directly to 15408 our mailing lists, for which you received no bug tracking number, 15409 should be submitted again using gccbug if you can reproduce the 15410 problem with GCC 3.0.) 15411 * The internal libgcc library is [11]built as a shared library on 15412 systems that support it. 15413 * Extensive testsuite included with GCC, with many new tests. In 15414 addition to tests for GCC bugs that have been fixed, many tests 15415 have been added for language features, compiler warnings and 15416 builtin functions. 15417 * Additional language-independent warning options -Wpacked, -Wpadded, 15418 -Wunreachable-code and -Wdisabled-optimization. 15419 * Target-independent options -falign-functions, -falign-loops and 15420 -falign-jumps. 15421 15422 Plus a great many bug fixes and almost all the [12]features found in 15423 GCC 2.95. 15424 15425 15426 For questions related to the use of GCC, please consult these web 15427 pages and the [13]GCC manuals. If that fails, the 15428 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15429 web pages and the development of GCC are welcome on our developer 15430 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 15431 archives. 15432 15433 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 15434 distribution of this entire article is permitted in any medium, 15435 provided this notice is preserved. 15436 15437 These pages are [18]maintained by the GCC team. Last modified 15438 2016-08-23[19]. 15439 15440 References 15441 15442 1. http://www.netbsd.org/ 15443 2. http://gcc.gnu.org/news/reorder.html 15444 3. http://gcc.gnu.org/news/ssa.html 15445 4. http://gcc.gnu.org/news/null.html 15446 5. http://gcc.gnu.org/news/unify.html 15447 6. http://gcc.gnu.org/gcc-3.0/c++features.html 15448 7. http://gcc.gnu.org/news/inlining.html 15449 8. http://gcc.gnu.org/news/dependencies.html 15450 9. http://gcc.gnu.org/c99status.html 15451 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 15452 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 15453 12. http://gcc.gnu.org/gcc-2.95/features.html 15454 13. https://gcc.gnu.org/onlinedocs/ 15455 14. mailto:gcc-help (a] gcc.gnu.org 15456 15. mailto:gcc (a] gcc.gnu.org 15457 16. https://gcc.gnu.org/lists.html 15458 17. http://www.fsf.org/ 15459 18. https://gcc.gnu.org/about.html 15460 19. http://validator.w3.org/check/referer 15461 ====================================================================== 15462 http://gcc.gnu.org/gcc-3.0/caveats.html 15463 15464 GCC 3.0 Caveats 15465 15466 * -fstrict-aliasing is now part of -O2 and higher optimization 15467 levels. This allows the compiler to assume the strictest aliasing 15468 rules applicable to the language being compiled. For C and C++, 15469 this activates optimizations based on the type of expressions. This 15470 optimization may thus break old, non-compliant code. 15471 * Enumerations are now properly promoted to int in function 15472 parameters and function returns. Normally this change is not 15473 visible, but when using -fshort-enums this is an ABI change. 15474 * The undocumented extension that allowed C programs to have a label 15475 at the end of a compound statement has been deprecated and may be 15476 removed in a future version. Programs that now generate a warning 15477 about this may be fixed by adding a null statement (a single 15478 semicolon) after the label. 15479 * The poorly documented extension that allowed string constants in C, 15480 C++ and Objective C to contain unescaped newlines has been 15481 deprecated and may be removed in a future version. Programs using 15482 this extension may be fixed in several ways: the bare newline may 15483 be replaced by \n, or preceded by \n\, or string concatenation may 15484 be used with the bare newline preceded by \n" and " placed at the 15485 start of the next line. 15486 * The Chill compiler is not included in GCC 3.0, because of the lack 15487 of a volunteer to convert it to use garbage collection. 15488 * Certain non-standard iostream methods from earlier versions of 15489 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 15490 ostream::form, and istream::gets. 15491 * The new C++ ABI is not yet fully supported by current (as of 15492 2001-07-01) releases and development versions of GDB, or any 15493 earlier versions. There is a problem setting breakpoints by line 15494 number, and other related issues that have been fixed in GCC 3.0 15495 but not yet handled in GDB: 15496 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 15497 15498 15499 For questions related to the use of GCC, please consult these web 15500 pages and the [2]GCC manuals. If that fails, the 15501 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15502 web pages and the development of GCC are welcome on our developer 15503 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 15504 15505 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15506 distribution of this entire article is permitted in any medium, 15507 provided this notice is preserved. 15508 15509 These pages are [7]maintained by the GCC team. Last modified 15510 2016-01-30[8]. 15511 15512 References 15513 15514 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 15515 2. https://gcc.gnu.org/onlinedocs/ 15516 3. mailto:gcc-help (a] gcc.gnu.org 15517 4. mailto:gcc (a] gcc.gnu.org 15518 5. https://gcc.gnu.org/lists.html 15519 6. http://www.fsf.org/ 15520 7. https://gcc.gnu.org/about.html 15521 8. http://validator.w3.org/check/referer 15522 ====================================================================== 15523 http://gcc.gnu.org/gcc-2.95/index.html 15524 15525 GCC 2.95 15526 15527 March 16, 2001: The GNU project and the GCC developers are pleased to 15528 announce the release of GCC version 2.95.3. 15529 15530 This release is no longer maintained. 15531 15532 Release History 15533 15534 GCC 2.95.3 15535 March 16, 2001 15536 15537 GCC 2.95.2 15538 October 27, 1999 15539 15540 GCC 2.95.1 15541 August 19, 1999 15542 15543 GCC 2.95 15544 July 31, 1999. This is the first release of GCC since the April 15545 1999 GCC/EGCS reunification and includes nearly a year's worth 15546 of new development and bugfixes. 15547 15548 References and Acknowledgements 15549 15550 GCC used to stand for the GNU C Compiler, but since the compiler 15551 supports several other languages aside from C, it now stands for the 15552 GNU Compiler Collection. 15553 15554 The whole suite has been extensively [1]regression tested and 15555 [2]package tested. It should be reliable and suitable for widespread 15556 use. 15557 15558 The compiler has several new optimizations, new targets, new languages 15559 and other new features. See the [3]new features page for a more 15560 complete list of new features found in the GCC 2.95 releases. 15561 15562 The sources include installation instructions in both HTML and 15563 plaintext forms in the install directory in the distribution. However, 15564 the most up to date installation instructions and [4]build/test status 15565 are on the web pages. We will update those pages as new information 15566 becomes available. 15567 15568 The GCC developers would like to thank the numerous people that have 15569 contributed new features, test results, bugfixes, etc to GCC. This 15570 [5]amazing group of volunteers is what makes GCC successful. 15571 15572 And finally, we can't in good conscience fail to mention some 15573 [6]caveats to using GCC 2.95. 15574 15575 Download GCC 2.95 from one of our many [7]mirror sites. 15576 15577 For additional information about GCC please see the [8]GCC project web 15578 server or contact the [9]GCC development mailing list. 15579 15580 15581 For questions related to the use of GCC, please consult these web 15582 pages and the [10]GCC manuals. If that fails, the 15583 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15584 web pages and the development of GCC are welcome on our developer 15585 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 15586 archives. 15587 15588 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 15589 distribution of this entire article is permitted in any medium, 15590 provided this notice is preserved. 15591 15592 These pages are [15]maintained by the GCC team. Last modified 15593 2016-09-30[16]. 15594 15595 References 15596 15597 1. http://gcc.gnu.org/gcc-2.95/regress.html 15598 2. http://gcc.gnu.org/gcc-2.95/othertest.html 15599 3. http://gcc.gnu.org/gcc-2.95/features.html 15600 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 15601 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15602 6. http://gcc.gnu.org/gcc-2.95/caveats.html 15603 7. http://gcc.gnu.org/mirrors.html 15604 8. http://gcc.gnu.org/index.html 15605 9. mailto:gcc (a] gcc.gnu.org 15606 10. https://gcc.gnu.org/onlinedocs/ 15607 11. mailto:gcc-help (a] gcc.gnu.org 15608 12. mailto:gcc (a] gcc.gnu.org 15609 13. https://gcc.gnu.org/lists.html 15610 14. http://www.fsf.org/ 15611 15. https://gcc.gnu.org/about.html 15612 16. http://validator.w3.org/check/referer 15613 ====================================================================== 15614 http://gcc.gnu.org/gcc-2.95/features.html 15615 15616 GCC 2.95 New Features 15617 15618 * General Optimizer Improvements: 15619 + [1]Localized register spilling to improve speed and code 15620 density especially on small register class machines. 15621 + [2]Global CSE using lazy code motion algorithms. 15622 + [3]Improved global constant/copy propagation. 15623 + [4]Improved control flow graph analysis and manipulation. 15624 + [5]Local dead store elimination. 15625 + [6]Memory Load hoisting/store sinking in loops. 15626 + [7]Type based alias analysis is enabled by default. Note this 15627 feature will expose bugs in the Linux kernel. Please refer to 15628 the FAQ (as shipped with GCC 2.95) for additional information 15629 on this issue. 15630 + Major revamp of GIV detection, combination and simplification 15631 to improve loop performance. 15632 + Major improvements to register allocation and reloading. 15633 * New Languages and Language specific improvements 15634 + [8]Many C++ improvements. 15635 + [9]Many Fortran improvements. 15636 + [10]Java front-end has been integrated. A [11]runtime library 15637 is available separately. 15638 + [12]ISO C99 support 15639 + [13]Chill front-end and runtime has been integrated. 15640 + Boehm garbage collector support in libobjc. 15641 + More support for various pragmas which appear in vendor 15642 include files 15643 * New Targets and Target Specific Improvements 15644 + [14]SPARC backend rewrite. 15645 + -mschedule=8000 will optimize code for PA8000 class 15646 processors; -mpa-risc-2-0 will generate code for PA2.0 15647 processors 15648 + Various micro-optimizations for the ia32 port. K6 15649 optimizations 15650 + Compiler will attempt to align doubles in the stack on the 15651 ia32 port 15652 + Alpha EV6 support 15653 + PowerPC 750 15654 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 15655 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 15656 + c3x, c4x 15657 + HyperSPARC 15658 + SparcLite86x 15659 + sh4 15660 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 15661 arm-linux) 15662 + vxWorks targets include support for vxWorks threads 15663 + StrongARM 110 and ARM9 support added. ARM Scheduling 15664 parameters rewritten. 15665 + Various changes to the MIPS port to avoid assembler macros, 15666 which in turn improves performance 15667 + Various performance improvements to the i960 port. 15668 + Major rewrite of ns32k port 15669 * Other significant improvements 15670 + [15]Ability to dump cfg information and display it using vcg. 15671 + The new faster scheme for fixing vendor header files is 15672 enabled by default. 15673 + Experimental internationalization support. 15674 + multibyte character support 15675 + Some compile-time speedups for pathological problems 15676 + Better support for complex types 15677 * Plus the usual mountain of bugfixes 15678 * Core compiler is based on the gcc2 development tree from Sept 30, 15679 1998, so we have all of the [16]features found in GCC 2.8. 15680 15681 Additional Changes in GCC 2.95.1 15682 15683 * Generic bugfixes and improvements 15684 + Various documentation fixes related to the GCC/EGCS merger. 15685 + Fix memory management bug which could lead to spurious aborts, 15686 core dumps or random parsing errors in the compiler. 15687 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 15688 support. 15689 + Fix infinite loop in the CSE optimizer. 15690 + Avoid undefined behavior in compiler FP emulation code 15691 + Fix install problem when prefix is overridden on the make 15692 install command. 15693 + Fix problem with unwanted installation of assert.h on some 15694 systems. 15695 + Fix problem with finding the wrong assembler in a single tree 15696 build. 15697 + Avoid increasing the known alignment of a register that is 15698 already known to be a pointer. 15699 * Platform specific bugfixes and improvements 15700 + Codegen bugfix for prologue/epilogue for cpu32 target. 15701 + Fix long long code generation bug for the Coldfire target. 15702 + Fix various aborts in the SH compiler. 15703 + Fix bugs in libgcc support library for the SH. 15704 + Fix alpha ev6 code generation bug. 15705 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 15706 AIX platforms. 15707 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 15708 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 15709 targets. 15710 + Fix weak symbol handling for rs6000/ppc svr4 targets. 15711 + Fix various problems with 64bit code generation for the 15712 rs6000/ppc port. 15713 + Fix codegen bug which caused tetex to be mis-compiled on the 15714 x86. 15715 + Fix compiler abort in new cfg code exposed by x86 port. 15716 + Fix out of range array reference in code convert flat 15717 registers to the x87 stacked FP register file. 15718 + Fix minor vxworks configuration bug. 15719 + Fix return type of bsearch for SunOS 4.x. 15720 * Language & Runtime specific fixes. 15721 + The G++ signature extension has been deprecated. It will be 15722 removed in the next major release of G++. Use of signatures 15723 will result in a warning from the compiler. 15724 + Several bugs relating to templates and namespaces were fixed. 15725 + A bug that caused crashes when combining templates with -g on 15726 DWARF1 platforms was fixed. 15727 + Pointers-to-members, virtual functions, and multiple 15728 inheritance should now work together correctly. 15729 + Some code-generation bugs relating to function try blocks were 15730 fixed. 15731 + G++ is a little bit more lenient with certain archaic 15732 constructs than in GCC 2.95. 15733 + Fix to prevent shared library version #s from bring truncated 15734 to 1 digit 15735 + Fix missing std:: in the libstdc++ library. 15736 + Fix stream locking problems in libio. 15737 + Fix problem in java compiler driver. 15738 15739 Additional Changes in GCC 2.95.2 15740 15741 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 15742 the optimizations performed by -fstrict-aliasing are valid according to 15743 the C and C++ standards, the optimization have caused some problems, 15744 particularly with old non-conforming code. 15745 15746 The GCC developers are experimenting with ways to warn users about code 15747 which violates the C/C++ standards, but those warnings are not ready 15748 for widespread use at this time. Rather than wait for those warnings 15749 the GCC developers have chosen to disable -fstrict-aliasing by default 15750 for the GCC 2.95.2 release. 15751 15752 We strongly encourage developers to find and fix code which violates 15753 the C/C++ standards as -fstrict-aliasing may be enabled by default in 15754 future releases. Use the option -fstrict-aliasing to re-enable these 15755 optimizations. 15756 * Generic bugfixes and improvements 15757 + Fix incorrectly optimized memory reference in global common 15758 subexpression elimination (GCSE) optimization pass. 15759 + Fix code generation bug in regmove.c in which it could 15760 incorrectly change a "const" value. 15761 + Fix bug in optimization of conditionals involving volatile 15762 memory references. 15763 + Avoid over-allocation of stack space for some procedures. 15764 + Fixed bug in the compiler which caused incorrect optimization 15765 of an obscure series of bit manipulations, shifts and 15766 arithmetic. 15767 + Fixed register allocator bug which caused teTeX to be 15768 mis-compiled on SPARC targets. 15769 + Avoid incorrect optimization of degenerate case statements for 15770 certain targets such as the ARM. 15771 + Fix out of range memory reference in the jump optimizer. 15772 + Avoid dereferencing null pointer in fix-header. 15773 + Fix test for GCC specific features so that it is possible to 15774 bootstrap with gcc-2.6.2 and older versions of GCC. 15775 + Fix typo in scheduler which could potentially cause out of 15776 range memory accesses. 15777 + Avoid incorrect loop reversal which caused incorrect code for 15778 certain loops on PowerPC targets. 15779 + Avoid incorrect optimization of switch statements on certain 15780 targets (for example the ARM). 15781 * Platform specific bugfixes and improvements 15782 + Work around bug in Sun V5.0 compilers which caused bootstrap 15783 comparison failures on SPARC targets. 15784 + Fix SPARC backend bug which caused aborts in final.c. 15785 + Fix sparc-hal-solaris2* configuration fragments. 15786 + Fix bug in sparc block profiling. 15787 + Fix obscure code generation bug for the PARISC targets. 15788 + Define __STDC_EXT__ for HPUX configurations. 15789 + Various POWERPC64 code generation bugfixes. 15790 + Fix abort for PPC targets using ELF (ex GNU/Linux). 15791 + Fix collect2 problems for AIX targets. 15792 + Correct handling of .file directive for PPC targets. 15793 + Fix bug in fix_trunc x86 patterns. 15794 + Fix x86 port to correctly pop the FP stack for functions that 15795 return structures in memory. 15796 + Fix minor bug in strlen x86 pattern. 15797 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 15798 + Fix template repository code to handle leading underscore in 15799 mangled names. 15800 + Fix weak/weak alias support for OpenBSD. 15801 + GNU/Linux for the ARM has C++ compatible include files. 15802 * Language & Runtime specific fixes. 15803 + Fix handling of constructor attribute in the C front-end which 15804 caused problems building the Chill runtime library on some 15805 targets. 15806 + Fix minor problem merging type qualifiers in the C front-end. 15807 + Fix aliasing bug for pointers and references (C/C++). 15808 + Fix incorrect "non-constant initializer bug" when -traditional 15809 or -fwritable-strings is enabled. 15810 + Fix build error for Chill front-end on SunOS. 15811 + Do not complain about duplicate instantiations when using 15812 -frepo (C++). 15813 + Fix array bounds handling in C++ front-end which caused 15814 problems with dwarf debugging information in some 15815 circumstances. 15816 + Fix minor namespace problem. 15817 + Fix problem linking java programs. 15818 15819 Additional Changes in GCC 2.95.3 15820 15821 * Generic bugfixes and improvements 15822 + Fix numerous problems that caused incorrect optimization in 15823 the register reloading code. 15824 + Fix numerous problems that caused incorrect optimization in 15825 the loop optimizer. 15826 + Fix aborts in the functions build_insn_chain and scan_loops 15827 under some circumstances. 15828 + Fix an alias analysis bug. 15829 + Fix an infinite compilation bug in the combiner. 15830 + A few problems with complex number support have been fixed. 15831 + It is no longer possible for gcc to act as a fork bomb when 15832 installed incorrectly. 15833 + The -fpack-struct option should be recognized now. 15834 + Fixed a bug that caused incorrect code to be generated due to 15835 a lost stack adjustment. 15836 * Platform specific bugfixes and improvements 15837 + Support building ARM toolchains hosted on Windows. 15838 + Fix attribute calculations in ARM toolchains. 15839 + arm-linux support has been improved. 15840 + Fix a PIC failure on sparc targets. 15841 + On ix86 targets, the regparm attribute should now work 15842 reliably. 15843 + Several updates for the h8300 port. 15844 + Fix problem building libio with glibc 2.2. 15845 15846 15847 For questions related to the use of GCC, please consult these web 15848 pages and the [17]GCC manuals. If that fails, the 15849 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15850 web pages and the development of GCC are welcome on our developer 15851 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 15852 archives. 15853 15854 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 15855 distribution of this entire article is permitted in any medium, 15856 provided this notice is preserved. 15857 15858 These pages are [22]maintained by the GCC team. Last modified 15859 2016-12-29[23]. 15860 15861 References 15862 15863 1. http://gcc.gnu.org/news/spill.html 15864 2. http://gcc.gnu.org/news/lcm.html 15865 3. http://gcc.gnu.org/news/cprop.html 15866 4. http://gcc.gnu.org/news/cfg.html 15867 5. http://gcc.gnu.org/news/dse.html 15868 6. http://gcc.gnu.org/news/hoist.html 15869 7. http://gcc.gnu.org/news/alias.html 15870 8. http://gcc.gnu.org/gcc-2.95/c++features.html 15871 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 15872 10. http://gcc.gnu.org/news/gcj-announce.txt 15873 11. http://gcc.gnu.org/news/javaannounce.html 15874 12. http://gcc.gnu.org/c99status.html 15875 13. http://gcc.gnu.org/news/chill.html 15876 14. http://gcc.gnu.org/news/sparc.html 15877 15. http://gcc.gnu.org/news/egcs-vcg.html 15878 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15879 17. https://gcc.gnu.org/onlinedocs/ 15880 18. mailto:gcc-help (a] gcc.gnu.org 15881 19. mailto:gcc (a] gcc.gnu.org 15882 20. https://gcc.gnu.org/lists.html 15883 21. http://www.fsf.org/ 15884 22. https://gcc.gnu.org/about.html 15885 23. http://validator.w3.org/check/referer 15886 ====================================================================== 15887 http://gcc.gnu.org/gcc-2.95/caveats.html 15888 15889 GCC 2.95 Caveats 15890 15891 * GCC 2.95 will issue an error for invalid asm statements that had 15892 been silently accepted by earlier versions of the compiler. This is 15893 particularly noticeable when compiling older versions of the Linux 15894 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 15895 for more information on this issue. 15896 * GCC 2.95 implements type based alias analysis to disambiguate 15897 memory references. Some programs, particularly the Linux kernel 15898 violate ANSI/ISO aliasing rules and therefore may not operate 15899 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 15900 shipped with GCC 2.95) for more information on this issue. 15901 * GCC 2.95 has a known bug in its handling of complex variables for 15902 64bit targets. Instead of silently generating incorrect code, GCC 15903 2.95 will issue a fatal error for situations it can not handle. 15904 This primarily affects the Fortran community as Fortran makes more 15905 use of complex variables than C or C++. 15906 * GCC 2.95 has an integrated libstdc++, but does not have an 15907 integrated libg++. Furthermore old libg++ releases will not work 15908 with GCC 2.95. You can retrieve a recent copy of libg++ from the 15909 [1]GCC ftp server. 15910 Note most C++ programs only need libstdc++. 15911 * Exception handling may not work with shared libraries, particularly 15912 on alphas, hppas, rs6000/powerpc and mips based platforms. 15913 Exception handling is known to work on x86 GNU/Linux platforms with 15914 shared libraries. 15915 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 15916 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 15917 or EGCS 1.1. As a result it may be necessary to fix C++ code before 15918 it will compile with GCC 2.95. 15919 * G++ is also converting toward the ISO C++ standard; as a result 15920 code which was previously valid (and thus accepted by other 15921 compilers and older versions of g++) may no longer be accepted. The 15922 flag -fpermissive may allow some non-conforming code to compile 15923 with GCC 2.95. 15924 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 15925 1.1.x, EGCS 1.0.x or GCC 2.8.x. 15926 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 15927 between Sept 30, 1998 and April 30, 1999 (the official end of the 15928 GCC 2.8 project). Future GCC releases will include all the changes 15929 from the defunct GCC 2.8 sources. 15930 15931 15932 For questions related to the use of GCC, please consult these web 15933 pages and the [2]GCC manuals. If that fails, the 15934 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15935 web pages and the development of GCC are welcome on our developer 15936 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 15937 15938 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15939 distribution of this entire article is permitted in any medium, 15940 provided this notice is preserved. 15941 15942 These pages are [7]maintained by the GCC team. Last modified 15943 2016-01-30[8]. 15944 15945 References 15946 15947 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 15948 2. https://gcc.gnu.org/onlinedocs/ 15949 3. mailto:gcc-help (a] gcc.gnu.org 15950 4. mailto:gcc (a] gcc.gnu.org 15951 5. https://gcc.gnu.org/lists.html 15952 6. http://www.fsf.org/ 15953 7. https://gcc.gnu.org/about.html 15954 8. http://validator.w3.org/check/referer 15955 ====================================================================== 15956 http://gcc.gnu.org/egcs-1.1/index.html 15957 15958 EGCS 1.1 15959 15960 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 15961 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 15962 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 15963 15964 EGCS is a free software project to further the development of the GNU 15965 compilers using an open development environment. 15966 15967 EGCS 1.1 is a major new release of the EGCS compiler system. It has 15968 been [1]extensively tested and is believed to be stable and suitable 15969 for widespread use. 15970 15971 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 15972 development sources; it contains all of the new features found in GCC 15973 2.8.1 as well as all new development from GCC up to June 6, 1998. 15974 15975 EGCS 1.1 also contains many improvements and features not found in GCC 15976 or in older versions of EGCS: 15977 * Global common subexpression elimination and global constant/copy 15978 propagation (aka [2]gcse) 15979 * Ongoing improvements to the [3]alias analysis support to allow for 15980 better optimizations throughout the compiler. 15981 * Vastly improved [4]C++ compiler and integrated C++ runtime 15982 libraries. 15983 * Fixes for the /tmp symlink race security problems. 15984 * New targets including mips16, arm-thumb and 64 bit PowerPC. 15985 * Improvements to GNU Fortran (g77) compiler and runtime library made 15986 since g77 version 0.5.23. 15987 15988 See the [5]new features page for a more complete list of new features 15989 found in EGCS 1.1 releases. 15990 15991 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 15992 1.1: 15993 * General improvements and fixes 15994 + Avoid some stack overflows when compiling large functions. 15995 + Avoid incorrect loop invariant code motions. 15996 + Fix some core dumps on Linux kernel code. 15997 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 15998 + Fix code generation problem in gcse. 15999 + Various documentation related fixes. 16000 * g++/libstdc++ improvements and fixes 16001 + MT safe EH fix for setjmp/longjmp based exception handling. 16002 + Fix a few bad interactions between optimization and exception 16003 handling. 16004 + Fixes for demangling of template names starting with "__". 16005 + Fix a bug that would fail to run destructors in some cases 16006 with -O2. 16007 + Fix 'new' of classes with virtual bases. 16008 + Fix crash building Qt on the Alpha. 16009 + Fix failure compiling WIFEXITED macro on GNU/Linux. 16010 + Fix some -frepo failures. 16011 * g77 and libf2c improvements and fixes 16012 + Various documentation fixes. 16013 + Avoid compiler crash on RAND intrinsic. 16014 + Fix minor bugs in makefiles exposed by BSD make programs. 16015 + Define _XOPEN_SOURCE for libI77 build to avoid potential 16016 problems on some 64-bit systems. 16017 + Fix problem with implicit endfile on rewind. 16018 + Fix spurious recursive I/O errors. 16019 * platform specific improvements and fixes 16020 + Match all versions of UnixWare7. 16021 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 16022 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 16023 from unsigned ints to double precision floats. 16024 + Fix ARM ABI issue with NetBSD. 16025 + Fix a few arm code generation bugs. 16026 + Fixincludes will fix additional broken SCO OpenServer header 16027 files. 16028 + Fix a m68k backend bug which caused invalid offsets in reg+d 16029 addresses. 16030 + Fix problems with 64bit AIX 4.3 support. 16031 + Fix handling of long longs for varargs/stdarg functions on the 16032 ppc. 16033 + Minor fixes to CPP predefines for Windows. 16034 + Fix code generation problems with gpr<->fpr copies for 64bit 16035 ppc. 16036 + Fix a few coldfire code generation bugs. 16037 + Fix some more header file problems on SunOS 4.x. 16038 + Fix assert.h handling for RTEMS. 16039 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 16040 + Fix x86 compiler abort in reg-stack pass. 16041 + Fix cygwin/windows problem with section attributes. 16042 + Fix Alpha code generation problem exposed by SMP Linux 16043 kernels. 16044 + Fix typo in m68k 32->64bit integer conversion. 16045 + Make sure target libraries build with -fPIC for PPC & Alpha 16046 targets. 16047 16048 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 16049 1.1.1: 16050 * General improvements and fixes 16051 + Fix bug in loop optimizer which caused the SPARC (and 16052 potentially other) ports to segfault. 16053 + Fix infinite recursion in alias analysis and combiner code. 16054 + Fix bug in regclass preferencing. 16055 + Fix incorrect loop reversal which caused incorrect code to be 16056 generated for several targets. 16057 + Fix return value for builtin memcpy. 16058 + Reduce compile time for certain loops which exposed quadratic 16059 behavior in the loop optimizer. 16060 + Fix bug which caused volatile memory to be written multiple 16061 times when only one write was needed/desired. 16062 + Fix compiler abort in caller-save.c 16063 + Fix combiner bug which caused incorrect code generation for 16064 certain division by constant operations. 16065 + Fix incorrect code generation due to a bug in range check 16066 optimizations. 16067 + Fix incorrect code generation due to mis-handling of clobbered 16068 values in CSE. 16069 + Fix compiler abort/segfault due to incorrect register 16070 splitting when unrolling loops. 16071 + Fix code generation involving autoincremented addresses with 16072 ternary operators. 16073 + Work around bug in the scheduler which caused qt to be 16074 mis-compiled on some platforms. 16075 + Fix code generation problems with -fshort-enums. 16076 + Tighten security for temporary files. 16077 + Improve compile time for codes which make heavy use of 16078 overloaded functions. 16079 + Fix multiply defined constructor/destructor symbol problems. 16080 + Avoid setting bogus RPATH environment variable during 16081 bootstrap. 16082 + Avoid GNU-make dependencies in the texinfo subdir. 16083 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 16084 --enable-cpp=<dirname> can be used to specify an additional 16085 install directory for the cpp wrapper script. 16086 + Fix CSE bug which caused incorrect label-label refs to appear 16087 on some platforms. 16088 + Avoid linking in EH routines from libgcc if they are not 16089 needed. 16090 + Avoid obscure bug in aliasing code. 16091 + Fix bug in weak symbol handling. 16092 * Platform-specific improvements and fixes 16093 + Fix detection of PPro/PII on Unixware 7. 16094 + Fix compiler segfault when building spec99 and other programs 16095 for SPARC targets. 16096 + Fix code-generation bugs for integer and floating point 16097 conditional move instructions on the PPro/PII. 16098 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 16099 + Fix build failure for the arc port. 16100 + Fix floating point format configuration for i?86-gnu port. 16101 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 16102 threads are enabled. 16103 + Fix coldfire code generation bugs. 16104 + Fix "unrecognized insn" problems for Alpha and PPC ports. 16105 + Fix h8/300 code generation problem with floating point values 16106 in memory. 16107 + Fix unrecognized insn problems for the m68k port. 16108 + Fix namespace-pollution problem for the x86 port. 16109 + Fix problems with old assembler on x86 NeXT systems. 16110 + Fix PIC code-generation problems for the SPARC port. 16111 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 16112 + Fix minor ISO namespace violation in Alpha varargs/stdarg 16113 support. 16114 + Fix incorrect "braf" instruction usage for the SH port. 16115 + Fix minor bug in va-sh which prevented its use with -ansi. 16116 + Fix problems recognizing and supporting FreeBSD. 16117 + Handle OpenBSD systems correctly. 16118 + Minor fixincludes fix for Digital UNIX 4.0B. 16119 + Fix problems with ctors/dtors in SCO shared libraries. 16120 + Abort instead of generating incorrect code for PPro/PII 16121 floating point conditional moves. 16122 + Avoid multiply defined symbols on GNU/Linux systems using 16123 libc-5.4.xx. 16124 + Fix abort in alpha compiler. 16125 * Fortran-specific fixes 16126 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 16127 is in the documented, non-Y2K-compliant range of 0-99, instead 16128 of being returned as 100 in the year 2000. 16129 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 16130 milliseconds value properly in Values(8). 16131 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 16132 information properly in SArray(7). 16133 16134 Each release includes installation instructions in both HTML and 16135 plaintext forms (see the INSTALL directory in the toplevel directory of 16136 the distribution). However, we also keep the most up to date 16137 installation instructions and [6]build/test status on our web page. We 16138 will update those pages as new information becomes available. 16139 16140 The EGCS project would like to thank the numerous people that have 16141 contributed new features, test results, bugfixes, etc. This [7]amazing 16142 group of volunteers is what makes EGCS successful. 16143 16144 And finally, we can't in good conscience fail to mention some 16145 [8]caveats to using EGCS 1.1. 16146 16147 Download EGCS from egcs.cygnus.com (USA California). 16148 16149 The EGCS 1.1 release is also available on many mirror sites. 16150 [9]Goto mirror list to find a closer site. 16151 16152 16153 For questions related to the use of GCC, please consult these web 16154 pages and the [10]GCC manuals. If that fails, the 16155 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16156 web pages and the development of GCC are welcome on our developer 16157 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 16158 archives. 16159 16160 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 16161 distribution of this entire article is permitted in any medium, 16162 provided this notice is preserved. 16163 16164 These pages are [15]maintained by the GCC team. Last modified 16165 2016-01-30[16]. 16166 16167 References 16168 16169 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 16170 2. http://gcc.gnu.org/news/gcse.html 16171 3. http://gcc.gnu.org/news/alias.html 16172 4. http://gcc.gnu.org/egcs-1.1/c++features.html 16173 5. http://gcc.gnu.org/egcs-1.1/features.html 16174 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 16175 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16176 8. http://gcc.gnu.org/egcs-1.1/caveats.html 16177 9. http://gcc.gnu.org/mirrors.html 16178 10. https://gcc.gnu.org/onlinedocs/ 16179 11. mailto:gcc-help (a] gcc.gnu.org 16180 12. mailto:gcc (a] gcc.gnu.org 16181 13. https://gcc.gnu.org/lists.html 16182 14. http://www.fsf.org/ 16183 15. https://gcc.gnu.org/about.html 16184 16. http://validator.w3.org/check/referer 16185 ====================================================================== 16186 http://gcc.gnu.org/egcs-1.1/features.html 16187 16188 EGCS 1.1 new features 16189 16190 * Integrated GNU Fortran (g77) compiler and runtime library with 16191 improvements, based on g77 version 0.5.23. 16192 * Vast improvements in the C++ compiler; so many they have [1]page of 16193 their own! 16194 * Compiler implements [2]global common subexpression elimination and 16195 global copy/constant propagation. 16196 * More major improvements in the [3]alias analysis code. 16197 * More major improvements in the exception handling code to improve 16198 performance, lower static overhead and provide the infrastructure 16199 for future improvements. 16200 * The infamous /tmp symlink race security problems have been fixed. 16201 * The regmove optimization pass has been nearly completely rewritten 16202 to improve performance of generated code. 16203 * The compiler now recomputes register usage information before local 16204 register allocation. By providing more accurate information to the 16205 priority based allocator, we get better register allocation. 16206 * The register reloading phase of the compiler optimizes spill code 16207 much better than in previous releases. 16208 * Some bad interactions between the register allocator and 16209 instruction scheduler have been fixed, resulting in much better 16210 code for certain programs. Additionally, we have tuned the 16211 scheduler in various ways to improve performance of generated code 16212 for some architectures. 16213 * The compiler's branch shortening algorithms have been significantly 16214 improved to work better on targets which align jump targets. 16215 * The compiler now supports -Os to prefer optimizing for code space 16216 over optimizing for code speed. 16217 * The compiler will now totally eliminate library calls which compute 16218 constant values. This primarily helps targets with no integer 16219 div/mul support and targets without floating point support. 16220 * The compiler now supports an extensive "--help" option. 16221 * cpplib has been greatly improved and may be suitable for limited 16222 use. 16223 * Memory footprint for the compiler has been significantly reduced 16224 for some pathological cases. 16225 * The time to build EGCS has been improved for certain targets 16226 (particularly the alpha and mips platforms). 16227 * Many infrastructure improvements throughout the compiler, plus the 16228 usual mountain of bugfixes and minor improvements. 16229 * Target dependent improvements: 16230 + SPARC port now includes V8 plus and V9 support as well as 16231 performance tuning for Ultra class machines. The SPARC port 16232 now uses the Haifa scheduler. 16233 + Alpha port has been tuned for the EV6 processor and has an 16234 optimized expansion of memcpy/bzero. The Alpha port now uses 16235 the Haifa scheduler. 16236 + RS6000/PowerPC: support for the Power64 architecture and AIX 16237 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 16238 + x86: Alignment of static store data and jump targets is per 16239 Intel recommendations now. Various improvements throughout the 16240 x86 port to improve performance on Pentium processors 16241 (including improved epilogue sequences for Pentium chips and 16242 backend improvements which should help register allocation on 16243 all x86 variants. Conditional move support has been fixed and 16244 enabled for PPro processors. The x86 port also better supports 16245 64bit operations now. Unixware 7, a System V Release 5 target, 16246 is now supported and SCO OpenServer targets can support GAS. 16247 + MIPS has improved multiply/multiply-add support and now 16248 includes mips16 ISA support. 16249 + M68k has many micro-optimizations and Coldfire fixes. 16250 * Core compiler is based on the GCC development tree from June 9, 16251 1998, so we have all of the [4]features found in GCC 2.8. 16252 16253 16254 For questions related to the use of GCC, please consult these web 16255 pages and the [5]GCC manuals. If that fails, the 16256 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16257 web pages and the development of GCC are welcome on our developer 16258 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 16259 16260 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 16261 distribution of this entire article is permitted in any medium, 16262 provided this notice is preserved. 16263 16264 These pages are [10]maintained by the GCC team. Last modified 16265 2016-01-30[11]. 16266 16267 References 16268 16269 1. http://gcc.gnu.org/egcs-1.1/c++features.html 16270 2. http://gcc.gnu.org/news/gcse.html 16271 3. http://gcc.gnu.org/news/alias.html 16272 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 16273 5. https://gcc.gnu.org/onlinedocs/ 16274 6. mailto:gcc-help (a] gcc.gnu.org 16275 7. mailto:gcc (a] gcc.gnu.org 16276 8. https://gcc.gnu.org/lists.html 16277 9. http://www.fsf.org/ 16278 10. https://gcc.gnu.org/about.html 16279 11. http://validator.w3.org/check/referer 16280 ====================================================================== 16281 http://gcc.gnu.org/egcs-1.1/caveats.html 16282 16283 EGCS 1.1 Caveats 16284 16285 * EGCS has an integrated libstdc++, but does not have an integrated 16286 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 16287 Lu has made a libg++-2.8.1.2 snapshot available which may work with 16288 EGCS. 16289 Note most C++ programs only need libstdc++. 16290 * Exception handling may not work with shared libraries, particularly 16291 on alphas, hppas, rs6000/powerpc and mips based platforms. 16292 Exception handling is known to work on x86-linux platforms with 16293 shared libraries. 16294 * Some versions of the Linux kernel have bugs which prevent them from 16295 being compiled or from running when compiled by EGCS. See the FAQ 16296 (as shipped with EGCS 1.1) for additional information. 16297 * In general, EGCS is more rigorous about rejecting invalid C++ code 16298 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 16299 a result it may be necessary to fix C++ code before it will compile 16300 with EGCS. 16301 * G++ is also converting toward the ISO C++ standard; as a result 16302 code which was previously valid (and thus accepted by other 16303 compilers and older versions of g++) may no longer be accepted. 16304 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 16305 or GCC 2.8.x due to changes necessary to support thread safe 16306 exception handling. 16307 16308 16309 For questions related to the use of GCC, please consult these web 16310 pages and the [1]GCC manuals. If that fails, the 16311 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16312 web pages and the development of GCC are welcome on our developer 16313 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 16314 16315 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 16316 distribution of this entire article is permitted in any medium, 16317 provided this notice is preserved. 16318 16319 These pages are [6]maintained by the GCC team. Last modified 16320 2016-01-30[7]. 16321 16322 References 16323 16324 1. https://gcc.gnu.org/onlinedocs/ 16325 2. mailto:gcc-help (a] gcc.gnu.org 16326 3. mailto:gcc (a] gcc.gnu.org 16327 4. https://gcc.gnu.org/lists.html 16328 5. http://www.fsf.org/ 16329 6. https://gcc.gnu.org/about.html 16330 7. http://validator.w3.org/check/referer 16331 ====================================================================== 16332 http://gcc.gnu.org/egcs-1.0/index.html 16333 16334 EGCS 1.0 16335 16336 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 16337 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 16338 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 16339 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 16340 16341 EGCS is a collaborative effort involving several groups of hackers 16342 using an open development model to accelerate development and testing 16343 of GNU compilers and runtime libraries. 16344 16345 An important goal of EGCS is to allow wide scale testing of 16346 experimental features and optimizations; therefore, EGCS contains some 16347 features and optimizations which are still under development. However, 16348 EGCS has been carefully tested and should be comparable in quality to 16349 most GCC releases. 16350 16351 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 16352 development sources; it contains nearly all of the new features found 16353 in GCC 2.8. 16354 16355 EGCS 1.0 also contains many improvements and features not found in GCC 16356 2.7 and even the GCC 2.8 series (which was released after the original 16357 EGCS 1.0 release). 16358 * Integrated C++ runtime libraries, including support for most major 16359 GNU/Linux systems! 16360 * The integrated libstdc++ library includes a verbatim copy of SGI's 16361 STL release. 16362 * Integrated GNU Fortran compiler. 16363 * New instruction scheduler. 16364 * New alias analysis code. 16365 16366 See the [1]new features page for a more complete list of new features. 16367 16368 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 16369 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 16370 EGCS 1.0 release: 16371 * Add support for Red Hat 5.0 Linux and better support for Linux 16372 systems using glibc2. 16373 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 16374 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 16375 fix these problems. 16376 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 16377 handling interfaces. 16378 To avoid future compatibility problems, we strongly urge anyone who 16379 is planning on distributing shared libraries that contain C++ code 16380 to upgrade to EGCS 1.0.1 first. 16381 Soon after EGCS 1.0 was released, the GCC developers made some 16382 incompatible changes in libgcc's exception handling interfaces. 16383 These changes were needed to solve problems on some platforms. This 16384 means that GCC 2.8.0, when released, will not be seamlessly 16385 compatible with shared libraries built by EGCS 1.0. The reason is 16386 that the libgcc.a in GCC 2.8.0 will not contain a function needed 16387 by the old interface. 16388 The result of this is that there may be compatibility problems with 16389 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 16390 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 16391 and libgcc.a has the support routines for both the old and the new 16392 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 16393 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 16394 The maintainers of GCC 2.x have decided against including seamless 16395 support for the old interface in 2.8.0, since it was never 16396 "official", so to avoid future compatibility problems we recommend 16397 against distributing any shared libraries built by EGCS 1.0 that 16398 contain C++ code (upgrade to 1.0.1 and use that). 16399 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 16400 The x86 changes fix code generation errors exposed when building 16401 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 16402 The hppa change fixes a compiler abort when configured for use with 16403 RTEMS. 16404 The MIPS changes fix problems with the definition of LONG_MAX on 16405 newer systems, allow for command line selection of the target ABI, 16406 and fix one code generation problem. 16407 The rs6000/ppc change fixes some problems with passing structures 16408 to varargs/stdarg functions. 16409 * A few machine independent bugfixes, mostly to fix code generation 16410 errors when building Linux kernels or glibc. 16411 * Fix a few critical exception handling and template bugs in the C++ 16412 compiler. 16413 * Fix Fortran namelist bug on alphas. 16414 * Fix build problems on x86-solaris systems. 16415 16416 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 16417 serious problems in EGCS 1.0.1. 16418 * General improvements and fixes 16419 + Memory consumption significantly reduced, especially for 16420 templates and inline functions. 16421 + Fix various problems with glibc2.1. 16422 + Fix loop optimization bug exposed by rs6000/ppc port. 16423 + Fix to avoid potential code generation problems in jump.c. 16424 + Fix some undefined symbol problems in dwarf1 debug support. 16425 * g++/libstdc++ improvements and fixes 16426 + libstdc++ in the EGCS release has been updated and should be 16427 link compatible with libstdc++-2.8. 16428 + Various fixes in libio/libstdc++ to work better on GNU/Linux 16429 systems. 16430 + Fix problems with duplicate symbols on systems that do not 16431 support weak symbols. 16432 + Memory corruption bug and undefined symbols in bastring have 16433 been fixed. 16434 + Various exception handling fixes. 16435 + Fix compiler abort for very long thunk names. 16436 * g77 improvements and fixes 16437 + Fix compiler crash for omitted bound in Fortran CASE 16438 statement. 16439 + Add missing entries to g77 lang-options. 16440 + Fix problem with -fpedantic in the g77 compiler. 16441 + Fix "backspace" problem with g77 on alphas. 16442 + Fix x86 backend problem with Fortran literals and -fpic. 16443 + Fix some of the problems with negative subscripts for g77 on 16444 alphas. 16445 + Fixes for Fortran builds on cygwin32/mingw32. 16446 * platform specific improvements and fixes 16447 + Fix long double problems on x86 (exposed by glibc). 16448 + x86 ports define i386 again to keep imake happy. 16449 + Fix exception handling support on NetBSD ports. 16450 + Several changes to collect2 to fix many problems with AIX. 16451 + Define __ELF__ for GNU/Linux on rs6000. 16452 + Fix -mcall-linux problem on GNU/Linux on rs6000. 16453 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 16454 + Allow autoconf to select a proper install problem on AIX 3.1. 16455 + m68k port support includes -mcpu32 option as well as cpu32 16456 multilibs. 16457 + Fix stdarg bug for irix6. 16458 + Allow EGCS to build on irix5 without the gnu assembler. 16459 + Fix problem with static linking on sco5. 16460 + Fix bootstrap on sco5 with native compiler. 16461 + Fix for abort building newlib on H8 target. 16462 + Fix fixincludes handling of math.h on SunOS. 16463 + Minor fix for Motorola 3300 m68k systems. 16464 16465 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 16466 problems reported by Red Hat for builds of Red Hat 5.1. 16467 * Generic bugfixes: 16468 + Fix a typo in the libio library which resulted in incorrect 16469 behavior of istream::get. 16470 + Fix the Fortran negative array index problem. 16471 + Fix a major problem with the ObjC runtime thread support 16472 exposed by glibc2. 16473 + Reduce memory consumption of the Haifa scheduler. 16474 * Target specific bugfixes: 16475 + Fix one x86 floating point code generation bug exposed by 16476 glibc2 builds. 16477 + Fix one x86 internal compiler error exposed by glibc2 builds. 16478 + Fix profiling bugs on the Alpha. 16479 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 16480 + Fix rs6000/ppc bug when converting values from integer types 16481 to floating point types. 16482 16483 The EGCS 1.0 releases include installation instructions in both HTML 16484 and plaintext forms (see the INSTALL directory in the toplevel 16485 directory of the distribution). However, we also keep the most up to 16486 date installation instructions and [2]build/test status on our web 16487 page. We will update those pages as new information becomes available. 16488 16489 And, we can't in good conscience fail to mention some [3]caveats to 16490 using EGCS. 16491 16492 Update: Big thanks to Stanford for providing a high speed link for 16493 downloading EGCS (go.cygnus.com)! 16494 16495 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 16496 (USA California -- High speed link provided by Stanford). 16497 16498 The EGCS 1.0 release is also available many mirror sites. 16499 [4]Goto mirror list to find a closer site 16500 16501 We'd like to thank the numerous people that have contributed new 16502 features, test results, bugfixes, etc. Unfortunately, they're far too 16503 numerous to mention by name. 16504 16505 16506 For questions related to the use of GCC, please consult these web 16507 pages and the [5]GCC manuals. If that fails, the 16508 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16509 web pages and the development of GCC are welcome on our developer 16510 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 16511 16512 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 16513 distribution of this entire article is permitted in any medium, 16514 provided this notice is preserved. 16515 16516 These pages are [10]maintained by the GCC team. Last modified 16517 2016-01-30[11]. 16518 16519 References 16520 16521 1. http://gcc.gnu.org/egcs-1.0/features.html 16522 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 16523 3. http://gcc.gnu.org/egcs-1.0/caveats.html 16524 4. http://gcc.gnu.org/mirrors.html 16525 5. https://gcc.gnu.org/onlinedocs/ 16526 6. mailto:gcc-help (a] gcc.gnu.org 16527 7. mailto:gcc (a] gcc.gnu.org 16528 8. https://gcc.gnu.org/lists.html 16529 9. http://www.fsf.org/ 16530 10. https://gcc.gnu.org/about.html 16531 11. http://validator.w3.org/check/referer 16532 ====================================================================== 16533 http://gcc.gnu.org/egcs-1.0/features.html 16534 16535 EGCS 1.0 features 16536 16537 * Core compiler is based on the gcc2 development tree from Aug 2, 16538 1997, so we have most of the [1]features found in GCC 2.8. 16539 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 16540 * Vast improvements in the C++ compiler; so many they have [2]page of 16541 their own! 16542 * Integrated C++ runtime libraries, including support for most major 16543 GNU/Linux systems! 16544 * New instruction scheduler from IBM Haifa which includes support for 16545 function wide instruction scheduling as well as superscalar 16546 scheduling. 16547 * Significantly improved alias analysis code. 16548 * Improved register allocation for two address machines. 16549 * Significant code generation improvements for Fortran code on 16550 Alphas. 16551 * Various optimizations from the g77 project as well as improved loop 16552 optimizations. 16553 * Dwarf2 debug format support for some targets. 16554 * egcs libstdc++ includes the SGI STL implementation without changes. 16555 * As a result of these and other changes, egcs libstc++ is not binary 16556 compatible with previous releases of libstdc++. 16557 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 16558 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 16559 1.1), Support for RTEMS on several embedded targets, Support for 16560 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 16561 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 16562 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 16563 * RS6000/PowerPC ports generate code which can run on all 16564 RS6000/PowerPC variants by default. 16565 * -mcpu= and -march= switches for the x86 port to allow better 16566 control over how the x86 port generates code. 16567 * Includes the template repository patch (aka repo patch); note the 16568 new template code makes repo obsolete for ELF systems using gnu-ld 16569 such as GNU/Linux. 16570 * Plus the usual assortment of bugfixes and improvements. 16571 16572 16573 For questions related to the use of GCC, please consult these web 16574 pages and the [3]GCC manuals. If that fails, the 16575 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16576 web pages and the development of GCC are welcome on our developer 16577 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 16578 16579 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 16580 distribution of this entire article is permitted in any medium, 16581 provided this notice is preserved. 16582 16583 These pages are [8]maintained by the GCC team. Last modified 16584 2016-01-30[9]. 16585 16586 References 16587 16588 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 16589 2. http://gcc.gnu.org/egcs-1.0/c++features.html 16590 3. https://gcc.gnu.org/onlinedocs/ 16591 4. mailto:gcc-help (a] gcc.gnu.org 16592 5. mailto:gcc (a] gcc.gnu.org 16593 6. https://gcc.gnu.org/lists.html 16594 7. http://www.fsf.org/ 16595 8. https://gcc.gnu.org/about.html 16596 9. http://validator.w3.org/check/referer 16597 ====================================================================== 16598 http://gcc.gnu.org/egcs-1.0/caveats.html 16599 16600 EGCS 1.0 Caveats 16601 16602 * EGCS has an integrated libstdc++, but does not have an integrated 16603 libg++. Furthermore old libg++ releases will not work with egc; HJ 16604 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 16605 Note most C++ programs only need libstdc++. 16606 * Note that using -pedantic or -Wreturn-type can cause an explosion 16607 in the amount of memory needed for template-heavy C++ code, such as 16608 code that uses STL. Also note that -Wall includes -Wreturn-type, so 16609 if you use -Wall you will need to specify -Wno-return-type to turn 16610 it off. 16611 * Exception handling may not work with shared libraries, particularly 16612 on alphas, hppas, and mips based platforms. Exception handling is 16613 known to work on x86-linux platforms with shared libraries. 16614 * Some versions of the Linux kernel have bugs which prevent them from 16615 being compiled or from running when compiled by EGCS. See the FAQ 16616 (as shipped with EGCS 1.0) for additional information. 16617 * In general, EGCS is more rigorous about rejecting invalid C++ code 16618 or deprecated C++ constructs than G++ 2.7. As a result it may be 16619 necessary to fix C++ code before it will compile with EGCS. 16620 * G++ is also aggressively tracking the C++ standard; as a result 16621 code which was previously valid (and thus accepted by other 16622 compilers and older versions of G++) may no longer be accepted. 16623 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 16624 1.0.x and later releases should work with Red Hat Linux 5.0. 16625 16626 16627 For questions related to the use of GCC, please consult these web 16628 pages and the [1]GCC manuals. If that fails, the 16629 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16630 web pages and the development of GCC are welcome on our developer 16631 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 16632 16633 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 16634 distribution of this entire article is permitted in any medium, 16635 provided this notice is preserved. 16636 16637 These pages are [6]maintained by the GCC team. Last modified 16638 2016-01-30[7]. 16639 16640 References 16641 16642 1. https://gcc.gnu.org/onlinedocs/ 16643 2. mailto:gcc-help (a] gcc.gnu.org 16644 3. mailto:gcc (a] gcc.gnu.org 16645 4. https://gcc.gnu.org/lists.html 16646 5. http://www.fsf.org/ 16647 6. https://gcc.gnu.org/about.html 16648 7. http://validator.w3.org/check/referer 16649 ====================================================================== 16650
Indexes created Fri Jun 26 00:24:39 UTC 2026